1. Introduction

1.1 Motivation for developing rare earth doped mid-infrared fiber lasers

The mid-infrared (IR), defined as the spectral region 3-25 µm, covers the important atmospheric windows, of 3-5 µm and 8-12 µm, and the molecular fingerprints of numerous gases, liquids and solids as diverse as: greenhouse gases; ground, water and air pollutants; pharmaceuticals; toxic agents; oil, oil products and plastics; biological tissue and so on. In this spectral range, molecular species exhibit fundamental vibrational absorption bands with large extinction coefficients, hence mid-IR spectroscopy potentially provides extremely sensitive chemical analysis. Therefore, both discrete and broadband mid-IR sources are of great interest for remote, real-time spectroscopy for a broad range of potential applications. These applications include: real-time monitoring of manufacturing processes (thereby enabling control); ‘biophotonics’ such as spectral tissue mapping for medical diagnosis (thereby hastening medical decision-taking and treatment planning); sensing, for instance of explosives and narcotics, for security purposes; sensing of pollutants for on-line monitoring of environmental-quality e.g. exhaust gases from industrial plant, cars and jet-engines and mid-IR ‘radar’ such as for collision avoidance. To exploit the mid-IR region effectively, requires the development of a raft of new passive and active mid-infrared fiber, and waveguide, narrow and broadband sources, sensors and components, including mid-IR fiber lasers.

Rare-earth (RE)-ion-doped mid-IR fiber lasers, in addition, have the potential for direct mid-infrared imaging. Mid-IR laser reflection of plant leaves was shown to image not only the surface cellular structure but also the sub-surface cellular structure [1]. Analogous mid-IR imaging of tissue during medical endoscopic surgical procedures may allow, for instance, cancer tumour margins to be examined in real-time.

New wavelengths for laser machining and welding of materials are of interest. Mid-IR fiber laser machining and welding of polymers will be facilitated by resonance between the laser light frequency and the vibrational absorption of the polymer chemical bonding. Likewise human tissue, for which new wavelengths for laser medical surgery are of international interest [2] and new mid-IR fiber laser wavelengths would give surgeons more choice in the mid-IR than the current common option of the CO₂ laser, with the aim of tailoring laser scalpels for purpose and minimising collateral damage.

1.2 Background

For the near-IR, a silica glass host allows efficient lasing for instance of erbium and neodymium ions doped into the lattice and a glass fiber cavity can be formed. Near-IR silica glass fiber lasers are undergoing a massive development and offer high power, efficiency and excellent beam quality, in a compact package.

However, silica glass rapidly becomes opaque beyond 2 μm wavelength. On the other hand, RE-ions are known to offer numerous transitions from 3 to 10 μm for potential exploitation in a glass fiber laser format. Mid-IR fiber laser operation up to 3 μm wavelength has been demonstrated in fluoride glass hosts. But beyond 3 μm the multiphonon relaxation rates of fluoride glasses compete with the sharp-line luminescence. So other vitreous or crystalline materials with lower phonon energies must be sought as the host materials.

The chalcogenide glasses are based on the chalcogen group (Group XVI) of the Periodic Table comprising: sulfur, selenium and tellurium usually formulated with other elements to build up robust glass matrices. It is helpful to classify the chalcogenide glasses in terms of the relative ionicity of the chemical bonds constituting the glass matrix as follows: (i) the more ionic gallium-lanthanum-chalcogen glasses and (ii) the covalently-bonded glasses based on the Periodic Table ‘p block’ elements such as germanium-arsenic-chalcogen or germanium-antimony-chalcogen glasses, which can be more covalent than silica glass. Chalcogenide glasses offer favourable properties for RE-doped fiber lasing such as high refractive indices for high absorption and emission cross-sections and generally low phonon energies for efficient radiative processes of doped RE-ions. Chalcogenide glasses are chemically durable in liquid water, and the open atmosphere, and are mechanically robust. Chalcogenide glasses have not yet been developed as mid-IR, RE-ion doped glass fiber lasers and to date there are no mid-IR fiber lasers available for use beyond 3 μm wavelength.

Below the current challenges in fabricating RE-ion-doped chalcogenide-glass fibers for developing mid-IR fiber lasers are reviewed. For the first time a coherent explanation is forwarded for the failure to date to develop a gallium-lanthanum-sulfide glass mid-IR fiber laser. For the more covalent chalcogenide glasses, the importance of optimizing the glass host and glass processing routes in order to minimize non-radiative decay and to avoid rare earth ion clustering and glass devitrification is discussed. For the first time a new idea is explored to explain an additional method of non-radiative depopulation of the excited state in the mid-IR that has not been properly recognized before: that of impurity multiphonon relaxation. Practical characterization of candidate selenide glasses is presented using X-ray diffraction, Fourier transform infrared spectroscopy and viscosity/temperature relations. Progress in modeling mid-IR fiber lasers is summarised and measured mid-IR fluorescence lifetimes and quantum efficiencies are collated.

2. Chalcogenide-glasses as host lattices for rare earth ions

Jorgensen and Reisfeld noted that the possibility of stimulated light emission was first discussed by Einstein in 1917, eight years before the quantum mechanical description of energy levels of many-electron systems [3]. The spectral lines that are obtained both in absorption and emission from the lanthanide ions (also known as the RE-ions) although formally forbidden transitions, and hence relatively weak, are almost as sharp as for gaseous elements. For practical, convenient lasing devices, the RE-ions are supported in a solid host. The rather sharp spectral lines of the RE-ions are due to f-f transitions, within the RE-ion 4f shell, which are shielded from the surrounding host lattice by the electrostatic screening afforded by the outer-lying, closed 5p shell of the RE-ion. Hence the center-line wavelength of the light absorbed or emitted by the RE ions is not much affected by the chemical nature of the lattice. Although, due to the nephel-auxetic effect (‘electron cloud expanding’), RE-ion radiative absorption and emission tend to red-shift slightly with increasing covalency of the lattice. In contrast, the probability of the radiative emission per unit time (W_r) of the doped RE-ions is dependent on the chemical nature of the lattice. Lattice phonons (vibrational quanta) can couple to the RE-ions to provide a non-radiative multiphonon decay route down to the next lowest lying manifold of Stark levels. The overall emission probability from the initial state (W_tot) (the inverse of the lifetime of the initial state (τ)) is a combination of W_r and the probability of non-radiative emission per unit time (W_nr):

Thus, for RE emission where the electronic energy gap may be bridged by a few phonons, the non-radiative multiphonon decay competes successfully with RE-ion fluorescence and laser action is prevented. Therefore it is important to select the host glass lattice carefully.

2.1 Glass lattice multiphonon relaxation in rare-earth-ion-doped chalcogenide glasses

The lattice vibration of maximum phonon energy is usually taken to be the most active in multiphonon decay, although lower energy phonons cannot be ruled out from participating. The maximum phonon energy has generally been taken as the highest energy absorption band occurring in the host glass Raman spectrum. This is usually the symmetric stretching vibration; this vibration undergoes no dipole change and hence is infrared-inactive.

The maximum phonon energy of silica glass is ~1000 cm⁻¹. For RE-ions doped in a silica glass host, RE-ion fluorescence occurs for energy gaps larger than ~4600 cm⁻¹ (~< 2.2 μm), therefore 5 phonons can effectively bridge the energy gap [6]. For fluoride glasses, the maximum phonon energy is 525 cm⁻¹ and fluorescence occurs for energy gaps greater than ~3100 cm⁻¹ (~<3.2 μm), and hence up to 6 phonons bridge the gap to prevent fluorescence. For the sulfide glasses the maximum phonon energy is generally accepted to be 425 cm⁻¹ [7] and for the selenide and selenide/telluride based glasses it is ~230-300 cm⁻¹ [8]. Assuming conservatively that energy gaps for the sulfide, selenide and the mixed-chalcogen selenide/telluride glasses can be bridged effectively by 4 phonons, this suggests a potential window for RE-ion laser operation above ~1700 cm⁻¹ (≤ 6 μm) for the sulfide glasses and above ~900-1200 cm⁻¹ (~≤ 8 μm - ≤ 11 μm) for the selenide and selenide/telluride glasses. Neglecting any other considerations, the lattice multiphonon non-radiative decay rate depends on the maximum lattice phonon energy.

In 1996, a near-IR fiber laser based on Nd³⁺-doped gallium-lanthanum-sulfide with output at 1.08 μm was reported [9]. However, this remains the only demonstration of a chalcogenide glass fiber laser to date. It appears that the main challenge in achieving mid-IR fiber lasers in RE-doped chalcogenide glasses is to improve sufficiently the glass quality of the fibers. Glass quality is key to the more ionic and more covalent types of chalcogenide glass matrices.

To understand why fiber lasers have been demonstrated in the near-IR in RE-doped gallium-lanthanum–sulfide-based glasses, yet not in the mid-IR, it is necessary to consider both the glass quality and the lattice multiphonon relaxation rates.

Regarding glass quality: in general, glass optical fibers may be fabricated by either: (i) melt-drawing, which means shaping the fiber whilst cooling the glass-melt down from above the liquidus temperature (T_L), or (ii) reheating a solid glass preform from room temperature to above the glass transition (T_g) and below T_L, shaping the supercooled liquid to form fiber and cooling the fiber to below T_g to form glass again. RE-doped gallium-lanthanum–sulfide-based glass fibers are made by preform-drawing. A simple measure of stability towards crystallization on reheating a glass to above T_g for shaping, like preform-drawing, is given by the Hruby parameter [10]: [T_x-T_g] (where: T_x is the onset temperature of crystallization [11]). For fiber-drawing, an appropriate viscosity (~10^4-5Pa s) must be accessible within the [T_x-T_g] gap in order to avoid any crystallization occurring during the fiber-drawing process. Large [T_x-T_g] gaps offer more chance of successfully avoiding crystallization.

Gallium-lanthanum-sulfide glasses have been formulated typically as GLS for low oxide glasses e.g. 65Ga₂S₃:30La₂S₃:5La₂O₃ (molar (mol)%) and GLSO for higher oxide glasses e.g. 78Ga₂S₃:22La₂O₃ [12]. Figure 1 shows how the Hruby parameter varies with oxide content of GLS [13,14] and that the oxide content plays a pivotal role in glass stability towards crystallization above T_g during fiber-drawing.

 

Fig. 1 [T_x-T_g] as a function of oxide content of Ga₂S₃-La₂S₃-La₂O₃ glasses; T_g is the extrapolated onset glass transition and T_x the extrapolated onset temperature of crystallization (both measured ( ± 0.5°C) using differential thermal analysis). For all glasses: n = 0.70 (where n = atomic ratio = (Ga/(Ga + La)). (Compositions with ≥0.49 wt%[O] had 0.1 & 1.07 wt%[O] contamination in the supplied Ga₂S₃ & La₂S₃, respectively. Compositions with ≤ 0.38wt%[O] had 0.12 & 0.14 wt%[O] contamination in the Ga₂S₃ & La₂S₃, respectively.) Figure shows a maximum in the [T_x-T_g] gap at 0.49 wt%[O] and that the [T_x-T_g] gap increases > 2.12 wt%[O]. For ≤ 0.13 wt%[O] there was no glass formation. (Figure adapted from [13,14].)

Download Full Size | PDF

Large oxide additions are known to be necessary to increase gallium-lanthanum-sulfide glass stability enough (enlarge the Hruby parameter) to avoid crystallization during fiber drawing [12,15] and typically up to 30 mol% La₂O₃ is added to achieve crystal-free fiber. Thus, at the higher oxide levels required for successful fiber drawing, the host lattice has become an oxysulfide matrix rather than a simple sulfide matrix.

Figure 2 shows the increasing vibrational absorption at 8.6 μm (1163 cm⁻¹) on adding La₂O₃ to gallium-lanthanum-sulfide glasses [14]; this vibrational absorption is thought to be due to a mixed sulfur-oxygen species, such as sulfate anions [16]. The presence of sulfate ions is feasible due to the ionic nature of the glass lattice. The maximum phonon energy of the oxysulfide glass is now close to 8.6 μm (1163 cm⁻¹).

 

Fig. 2 Infrared absorption spectra for the Ga₂S₃-La₂S₃-La₂O₃ series of glass compositions with ≥0.49wt%[O], see Fig. 1. (Reproduced with kind permission from [14].)

Download Full Size | PDF

Multiphonon processes in RE-ion-doped crystals or glasses arise from the same general physics [4]. In a crystal the RE-ion has nearest neighbour anions and it is the vibration of these and more distant anions that contribute to the fluctuating Stark field which induces the non-radiative transitions. The aperiodic structure of a glass prevents such concerted lattice vibration. The high energy modes that couple to the RE-ion in a glass matrix are best approximated by independent oscillators [4]. Doped active rare earth ions usually have high coordination numbers of ~6-12. In the GLSO oxysulfide lattice the active RE-ions may, in extreme cases, be coordinated solely by oxygen or solely by sulfur. More likely, the first coordination sphere will be composed of proportions of oxygen anions, sulfur anions and sulfate.

The lowest energy gap across which fluorescence has been measured is ~2000 cm⁻¹ (~5 μm) in Nd³⁺-doped-GLS glasses but not in GLSO glasses [17]. Taking the maximum phonon energy in GLSO glasses as ~that of the S-O lattice mode at 1163 cm⁻¹ (8.6 μm) (Fig. 2) and assuming conservatively that an energy gap of the RE-ion which may be bridged by ≤ 4 phonons will undergo fluorescence, then potential useful laser operation in a GLSO host is predicted only above ~4760 cm⁻¹ (< ~2.1 μm), as seen in practice. In summary, in the past a too low phonon energy has been used for GLS glasses as the premise of their ability to lase in the mid-IR. When a more realistic phonon energy is used, as proposed here for the first time, then it is a straightforward matter to understand why it is that, over many years of investigation, GLS glasses have failed to lase in the mid-IR.

The rest of this paper will concentrate on the more covalent ‘p-block’ chalcogenide glasses.

2.2 Impurity multiphonon relaxation in rare-earth-ion-doped chalcogenide glasses

The situation is rather different for the ‘p-block’ chalcogenide sulfide, selenide and selenide/telluride chalcogenide glass hosts, for instance: Ge-Sb-S, Ge-As-Se and Ge-As-Se-Te glass hosts. Firstly, for these more covalent chalcogenide glasses, oxide exists as an impurity manifesting extrinsic absorption bands in the infrared spectra of glasses rather than contributing to the multiphonon edge. Moreover, we have found that too much oxide in selenide glasses actually prevents glass formation rather than aiding it [15,18]. Finally, oxides are the most electronegative anions and the RE dopants are the most electropositive (metallic) species in these covalent chalcogenide glass melts and so the rare earth ions tend to scavenge oxygen in the melt, forming strong oxide bonds [19]. In a glass host containing oxide impurity, for instance based on Ge-Ga-As-Se, the first coordination shell of a proportion of the doped RE-ions will contain oxygen, bonded as RE-O-As, RE-O-Ge, RE-O-Ga or RE-O-RE. RE-O-RE may contribute to ion-to-ion relaxation (see section 2.4); [As-O] and [Ge-O] impurity exhibit high-energy, extrinsic-impurity, vibrational absorption bands. These broad bands are centred: ([As-O], [Ge-O]) at ~1266 cm⁻¹ (~7.9 μm) and [As-O] at ~770 cm⁻¹ (~13 μm) [20,21].

Assuming similar physics to that under-lying intrinsic lattice multiphonon relaxation, then extrinsic impurities local to the RE-ion in a glass host lattice cause impurity multiphonon relaxation of the RE-ion excited states in an analogous way by means of local independent oscillators. For instance, Pr³⁺-doped Ge-Ga-As-Se glass has a potential fluorescent emission at ~2857-1818 cm⁻¹ (~3.5-5.5 μm) [22], the electronic excitation energy ΔE can be released by non-radiative emission to the local impurity of [RE-O-As] which accepts 3 phonons (3ħω_imp ≈2307 cm⁻¹) and RE-O-Ge which accepts 2 phonons (2ħω_imp ≈2532 cm⁻¹); moreover for [RE-Se-H], the Se-H vibration [21] performs a resonant absorption (ħω_imp ≈2174 cm⁻¹). At room temperature vibrating species, such as the impurity independent oscillators, are normally in the ground state [23] and 1, 2 or 3 vibrational quanta may be readily accepted (Fig. 3 ). These are usually referred to as the fundamental vibrational, 1st overtone and 2nd overtone absorptions, respectively (Fig. 3). The vibrational energy level separation tends to decrease slightly at high energies.

 

Fig. 3 (Adapted from [23]. The allowed vibrational energy levels and some transitions between them for a diatomic molecule undergoing anharmonic oscillations.) It is suggested that impurity species such as [As-O], coordinating RE-ions doped into chalcogenide glass hosts, accept energy non-radiatively from the excited RE-ions and in turn undergo vibrational excitation.

Download Full Size | PDF

It is important to note that a key difference between lattice multiphonon decay and impurity multiphonon decay is the inability of the latter to act on all RE-ion oscillators. This is because of the rather low concentration of extrinsic impurities. The maximum phonon energy of the extrinsic impurity oxide in a selenide host is at a far higher energy than that of the lattice. So, taking into account the principle that the higher the maximum phonon energy, the higher the non-radiative decay rate, then any doped RE-ions located closely to extrinsic impurity oxide, which is a small population of the whole RE-ion population, will have a higher probability of non-radiative emission by impurity multiphonon decay (W_imp) than by lattice multiphonon decay (following the multiphonon-energy gap law). The remaining doped RE-ions are exposed to the lattice multiphonon decay W_mp.

Considering RE-doped “p-block” sulfide, selenide and selenide-telluride glasses it is now clear that their cut-off wavelength for fluorescence is no longer governed solely by the maximum lattice phonon energy but also by the maximum impurity phonon energy of each impurity and by the amount of that impurity present. So a more realistic expression for equation [2] is:

To summarise, there are different populations of RE-ions in a chalcogenide glass host, as follows. There is a ‘well behaved’ population whose radiative behavior is governed by the maximum lattice phonon energy, which is low for instance in selenide glasses. But there are other smaller populations of RE-ions whose radiative behavior is dominated instead by the maximum phonon energies of local impurities which are usually each much higher than the maximum lattice phonon energy and which can lead to fast non-radiative decay for some fraction of the RE-ion population.

In support of these ideas, Quimby and Aitken [24] have re-evaluated the α, Β parameters (Eq. (3) of the multiphonon energy gap law for rare earth doped sulfide glasses. The temperature dependence of the RE-dopant fluorescent lifetime was measured in order to separate out true multiphonon decay from other non-radiative processes. It was shown for the host: Ge₂₅As_8.33Ga_1.67S₆₅ (at%) that for energy gaps > 2500 cm⁻¹ (i.e. emission < 4 μm), non-radiative processes other than lattice multiphonon decay systematically dominated. These other non-radiative decay processes were proposed to be energy transfer to vibrational impurities like O-H and S-H. The mean levels of these impurities in the glasses were ~400 ppmw [SH] and ~25 ppmw [OH]. More studies like this are needed to understand better the role of W_imp in non-radiative decay. Furthermore, Moizan et al. [25]. have shown near-infrared, radiative-lifetimes of Er³⁺ doped in sulfide glasses (Er³⁺-Ge-Ga-Sb-S) become lower beyond a threshold of [S-H] impurity of ~95 ppm.

For selenide glasses, the [H-Se] impurity absorption band overlaps Er³⁺, Dy³⁺, Pr³⁺ and Tb³⁺ emission bands in the 4-5 μm wavelength region (see Table 1 ). Thus, only one [H-Se] impurity phonon is required to de-activate the RE-ion excited state. On the other hand, for the selenide glass Ge-Ga-As-Se, high quantum efficiencies have been measured for several mid-IR transitions of doped Dy³⁺, Pr³⁺ and Tb³⁺ ions in the bulk glasses [26] (see Table 1). However a measured high quantum efficiency may be misleading if the measured fluorescent intensities are in fact low and fast non-radiative decay due to impurity multiphonon relaxation has already occurred. EXAFS (extended X-ray absorption spectra fine structure spectroscopy) has shown in a study of Tm³⁺-doped Ge₂₅Ga₁₀S₆₅, that Tm³⁺ was coordinated with 7 [S]; but when 10 mol% CsBr was included in the formulation, keeping the Ge:Ga:S atomic ratio the same, the Tm³⁺ was instead coordinated by 6 [Br] [27]. More of these types of study should again be carried out to understand better the local environment of the RE-ions in chalcogenide glasses.

Table 1. Collation of the mid-IR emission of RE-ion doped bulk chalcogenide glasses, and fiber, at ≥ 3 μm wavelength. Glasses are in blocks according to their ionic or covalent behavior and the host-glass chalcogen. References are in date-order in each block. Key: - means data are not available.

View Table

Overall, the optical loss spectrum of a chalcogenide glass fiber not only gives information about the likely intrinsic and extrinsic optical loss in the fiber laser cavity but also indicates the potential for unwanted multiphonon fast relaxation from the excited state by impurities.

2.3 Effect of the electronic absorption edge in rare-earth-ion-doped chalcogenide glasses

The high energy electronic absorption edge of chalcogenide glasses, the Urbach edge, corresponds to an exponential increase of the glass absorption coefficient to ~10⁴ cm⁻¹ (see later Fig. 9 ). The weak absorption tail (WAT) emanates from the low energy side of the Urbach edge and has a smaller-slope, exponential-decrease with decreasing photon energy. WAT loss is attributed to impurities and localized native defects in the chalcogenide glass; the native defects are considered to be bonding departures from the usual valence requirements. Bishop et al. [28] have demonstrated that, on broad-band pumping into the host-glass optical band-gap, resonant energy transfer can occur from the host to doped RE-ions which exhibit f-f transitions overlapping the WAT. This energy transfer is proposed to be mediated by defects in the glass. But, Harada and Tanaka [29] have pointed out that on pumping the optical band-gap, photodarkening eventually occurs and the luminescent intensity reduces with exposure time. No work has been done on investigating the impact on mid-IR RE-ion transitions of possible energy transfer from the host. There is some evidence that the energy transfer is weaker for lower-energy RE-ion transitions. Similarly, it is not known if RE-ions that are excited through direct population of f-f levels can dissipate energy to the host via a reverse process so as to cause non-radiative decay of excited RE-ion states.

 

Fig. 9 Attenuation versus wavelength plot for Teflon-clad As₄₀S₅₅Se₅ glass fiber fabricated using: (A) as-received chemicals and (B) purified chemicals. Also shown are the Urbach and multiphonon absorption edges and the weak absorption tail (WAT) and the calculated Rayleigh scattering loss. (Reproduced with kind permission from [ 40 ].)

Download Full Size | PDF

2.4 Clustering of rare-earth-ions in chalcogenide glasses

When like RE-ions in a lattice are too close to one another then ion-ion interactions can occur which depopulate the excited state and contribute to W_nr. This phenomenon is often termed RE-clustering and tends to quench luminescence, which is known as concentration quenching. The term clustering has become synonymous with the RE-ions being too close together but the term is often applied without knowledge of the local chemical environment of the RE-ion. For instance, clusters have been imaged using transmission electron microscopy as physical discontinuities in the homogenous glass matrix in over-doped RE-ion-doped silica-based glasses [30] and in Er³⁺-doped nano-glass-ceramics [31]. It is important to understand the RE-ion environment in order to try to overcome the problem of concentration quenching. There are two extremes. For homogeneously distributed ions in a glass lattice then, as the RE-ion concentration is increased, the concentration eventually is so high that the mean RE-ion separation becomes small enough for spatial migration of the excitation energy from one RE-ion to a neighboring RE-ion to become possible. At the other extreme, the RE-ion concentration may exceed the equilibrium solid-solid solubility of the RE-ions in the glass matrix and the RE-ions may precipitate out (or never dissolve) and, in aggregating, lead to too small ion-to-ion distances. Alternatively, the excess RE-ions may precipitate out either on glass-melt-cooling or on glass-reheating during fiber-drawing and this may be accompanied by phase separation of the host matrix itself. RE-ions may nucleate the separating phase. The separating phase in the glass may be crystalline or amorphous. If amorphous the phase separation will have originated through liquid-liquid phase separation which either occurred above the liquidus, during glass melting, or occurred during glass-melt-cooling (for instance on melt-drawing fiber) as metastable, sub-liquidus, liquid-liquid phase separation. On cooling below Tg the separated liquid phase cools as a separated glassy phase distributed throughout the parent glass. The problem of RE-ion clustering and concentration quenching is intrinsically linked to phase separation because the RE dopant may concentrate in the separated phase or, even worse, concentrate there together with impurities.

High concentrations of RE-ions are possible where a substantial level of lanthanide ions constitute the glass matrix itself, as in the fluoride glasses [32] and GLS and GLSO. For instance, 1 mol% Er₂S₃ was homogeneously distributed throughout a 70Ga₂S₃-23La₂S-6La₂O₃ glass which was then used as the target for sputtered waveguides [33]. The waveguides were measured to have 0.5 at% Er³⁺ (oxygen content was ignored) and demonstrated to show near-IR gain. For compositions like this, the solubility of the active Er³⁺ ion is encouraged due to its similar ionic radius and charge to the La³⁺ which is a component of the glass host matrix.

The very low solubility of RE-ions in silica glass is partially overcome by adding the co-dopant Al₂O₃ [34]. Similarly, for the “p-block” chalcogenide glass compositions like Ge-Sb-S and Ge-As-Se, the intrinsic RE-ion solubility is extremely low [35]. Addition of Ga^III to the host in a concentration ratio of at least 10:1 = [Ga^III]: [RE-ion] can increase RE emission by an order of magnitude for the same level of RE-dopant (Fig. 4 ).

 

Fig. 4 ¹G₄ → ³H⁵ emission spectra of Pr³⁺-doped GeAs sulfide glasses with (open circles) and without (solid diamonds) Ga^III co-dopant. (Reproduced with permission from [ 35 ].)

Download Full Size | PDF

Ga^III adopts tetrahedral coordination in the glass matrix, like Ge^IV, but Ga^III is trivalent and charge deficient, so the addition of Ga^III is thought to encourage complexing of the RE-ions at the Ga^III sites. Prior to addition of Ga^III the RE-ions are described as clustered [35,36] but their behavior may instead be due to a lack of solubility and random precipitation [19].

We have shown recently that addition of too much Ga^III can lead to phase separation of Ga₂Se₃ in RE-doped Ge-Ga-As-Se glasses [19]. Ležal [37] has concluded that the homogeneity of RE-ion-doped sulfide, selenide and selenide-telluride glasses can depend on the chemical form of the RE-ion added. Moreover, for Pr³⁺-doped Ge-Ga-S glasses he reported an increase in number-density of the clusters, which were imaged and appeared like crystalline inclusions ~1 μm diameter, with increasing [O-H] impurity concentration in the glass.

3. Rare earth doped chalcogenide glasses: glass quality

As mentioned above, chalcogenide glass fibers may be drawn either by: (i) shaping during cooling of a glass melt, held close to or above the liquidus where crystallization is not thermodynamically possible or (ii) solid glass preforms may be made and drawn to fiber by reheating to access the supercooled liquid temperature range above T_g. Both processes are preceded by glass-melting and silica–glass containment is used generally for the ‘p-block’ chalcogenide glasses. A chemical vapor deposition (CVD) route to unary chalcogenide glass thin films has been reported [38] but CVD routes to the rather complex multicomponent chalcogenide glasses required to solubilise rare earths are unlikely to be easily developed.

In recent work, we have shown the importance of optimising the balance between the RE-dopant concentration and Ga^III concentration in Ge-Ga-As-Se glasses. As discussed earlier (section 2.4) not enough Ga^III means that the RE-dopant is not dispersed homogeneously in the glass [35,36]. But if too much Ga^III is added, the RE-ions tend to nucleate out Ga₂Se₃ on melt-cooling of the glass as shown for the series 0-2000 ppmw (by weight) of Dy³⁺ dopant added to the host glass Ge_16.5Ga₁₀As₉Se_64.5 [19].

New results, presented for the first time here, are for a lower level of added gallium of Ga₃ rather than Ga₁₀ (host glass composition: Ge_16.5Ga₃As₁₆Se_64.5). A series of glasses was synthesised [method as in [19]] with increasing levels of Dy³⁺ doping from 0 to 3000 ppmw. Powder X-ray diffraction (XRD) (collected in the open atmosphere using a Siemens Krystalloflex 810, with CuK_α radiation from 10 to 70 °2θ in steps of 0.2°2θ) indicated that all were amorphous (see Fig. 5 patterns a – g); a crystallized Ga₁₀ glass is shown for comparison also (in pattern h) [17]).

 

Fig. 5 Powder X-ray diffraction patterns of: a – g: 0, 500, 800, 1000, 1500, 2000 and 3000 ppmw Dy³⁺ doped Ge_16.5Ga₃As₁₆Se_64.5. The XRD patterns indicate that the as-prepared glasses are amorphous. h: For comparison, a 1000 ppm Dy³⁺ doped Ge_16.5Ga₁₀As₉Se_64.5 glass showing the onset of bulk crystallization to Ga₂Se₃ [19].

Download Full Size | PDF

However, Fourier transform infrared (FTIR) spectroscopy (Fig. 6 , collected for disc samples polished to a 1 μm finish under non-aqueous media and scanned with a W lamp source, KBr beamsplitter and DTGS detector in a Bruker IFS 66/S spectrometer after purging the sample chamber for 1h (Parker Filtration purge gas to remove H₂O and CO₂)) reveals that the higher doping levels of 1500-3000 ppmw show scattering in the near- and mid-IR region indicating scattering particles may be present of a size of the order of the wavelength of the near-IR light used (wavelength-independent scattering is show by the highest doped glasses at smaller wavelengths than shown in Fig. 6). For Ga₁₀ glasses such scattering was found to be evident when there was precipitated Ga₂Se₃ inside the bulk of the glass [19].

 

Fig. 6 Near- and mid-infrared absorption spectra of as-prepared Dy³⁺-doped Ge_16.5 Ga₃As₁₆Se_64.5 glasses. The 1500 ppmw, 2000 ppmw and 3000 ppmw doped glasses exhibit scattering in the near- to mid-infrared region.

Download Full Size | PDF

Finally, Fig. 7 shows that at the higher level of dopant not only are scattering centers present but also the absorption band at 1.3 μm (⁶H_15/2→ (⁶H_9/2, ⁶F_11/2) of the Dy³⁺ starts to show structuring, indicative of a more ordered local environment. All of this information was hidden on the amorphous XRD patterns (Fig. 5). We suggest this is because of the low specific volume occupied by any crystallites present making them below the XRD detection limit (~2 vol% crystallinity). Scattering may also come from silica particles; we know that [Si] and [O] can be incorporated in the chalcogenide glass from the silica containment [19]. The large refractive index difference between silica and chalcogenide would mean that very small (<< 1 μm) silica particles could exert strong scattering. However, we have never imaged silica particles in the chalcogenide glasses under transmission electron microscope imaging.

 

Fig. 7 As-prepared 800 ppmw Dy³⁺-doped Ge_16.5 Ga₃As₁₆Se_64.5 glass exhibits a smooth absorption band at 1.3 μm due to the (⁶H_15/2→ (⁶H_9/2, ⁶F_11/2) absorption of the Dy³⁺ indicating that the local environment of the Dy³⁺ ions is amorphous. At the higher doping level of 1500 ppmw Dy³⁺, the 1.3 μm absorption band shows structuring, indicative of a more ordered local environment and indicating that the glass has undergone unwanted crystallization during the melt-quenching procedure to make the glass.

Download Full Size | PDF

A useful approach for anticipating the ability of glasses to withstand the process of reheating to draw fiber is viscosity/temperature measurement. We have used the parallel plate method to access fiber-drawing viscosities (see [11] for experimental details). From Fig. 8 it may be seen that the viscosity/temperature curve of the Ga₁₀ glass doped with 800 ppm Dy³⁺ smoothly decreased from 10^7.5 Pas to 10^5.8 Pas with increasing temperature, there then followed a rise in viscosity with increasing temperature before decreasing once more. This viscosity rise is consistent with the glass-forming liquid above T_g (note that viscosity at T_g is ~10^12.5 Pas) devitrifying on heating [39]. The crystals in the supercooled liquid provide a temporary resistance to flow. The residual, super-cooled liquid exhibits viscous flow as the temperature is increased. In marked contrast, the Ga₃ glass doped with 800 ppm Dy³⁺ showed a smooth response to the measurement and appeared to allow access to the fiber-drawing viscosity, within the time-frame of the experiment.

 

Fig. 8 Viscosity/temperature curves of: (a) 800 ppmw Dy³⁺ doped Ga₁₀ glass (Ge_16.5Ga₁₀As₉Se₆₄) showing the smooth fall in viscosity with temperature is arrested at 10^5.8 Pas due to unwanted crystal growth and (b) in contrast, the viscosity of the 800 ppmw Dy³⁺ doped Ga₃ glass (Ge_16.5Ga₃As₁₆Se₆₄) falls smoothly over the temperature range studied.

Download Full Size | PDF

4. Optical loss of chalcogenide glass fibers

For constructing the fiber laser cavity it is important to have low optical loss both at the pump wavelength and at the emission wavelength. Figure 9 shows optical loss plots constructed by Sanghera et al. [40] for a Teflon-coated As₄₀S₅₅Se₅ glass fiber fabricated using as-received chemicals and purified chemicals.

Although > 5x9s purity chemicals are readily bought ‘off-the-shelf’, unfortunately only the cationic impurities are specified and unknown amounts of hydride and oxide impurities are present which affect the mid-IR transmission window of the chalcogenide glasses and can cause impurity multiphonon relaxation of the RE-ion excited state (see section 2.2). Also shown in Fig. 9 is the calculated theoretical minimum optical loss curve based on the measured Urbach and multiphonon absorption edges, the weak absorption tail (WAT) and the calculated Rayleigh scattering loss.

Champion fiber losses for the simple chalcogenide glass compositions, like As-S, have been achieved in exemplary work by the group of Churbanov [41]. They have also shown that for ‘selenium-tellurium’ glasses, melt-drawn fiber exhibited lower background scattering loss than did the preform-drawn fiber although absorption coefficients of extrinsic vibrational impurities such as Se-H appeared the same for both types of fiber drawing [42]. More work is now needed in addressing achieving low loss in the more complex compositions required to solubilise the RE-dopants.

Our recent work has shown the ability of high levels of rare-earth-ion dopant, in the company of high levels of Ga^III, to solubilise the silica of the containment during chalcogenide glass melting thereby increasing Si-O extrinsic impurities as well as potentially contributing to scattering loss [19].

5. Modeling performance of mid-infrared lasers

RE-ion emission in the mid-IR usually involves lasing transitions not down to the ground state thus bottle-neck populations may occur in the lower lasing level. Quimby et al. [43] have modeled a cascade pumping scheme for Dy³⁺-doped selenide glasses, to offset this problem, involving multiple Bragg gratings within the fiber laser to build up the excited state population of the required upper lasing level. Their numerical simulations predict high efficiency and power scalability around 4.5 μm output with a key requirement being the need to reduce fiber loss to the 1-3 dBm⁻¹ range. On the other hand, we have proposed a far simpler scheme, obviating the need for Bragg gratings, based upon the innate high Fresnel reflection (due to the high refractive indices of the Se and Se/Te glasses) at the fiber end-faces, augmented by abutted or evaporated mirrors [44]. It is important to note that both models predict efficient fiber lasing of Dy³⁺-doped selenide glass fibers.

A conservative estimate of achievable overall efficiency of the RE-doped, mid-IR fiber lasers is 2%, using a 3-4 W pump to deliver 100 mW uncooled laser output. Note that Quimby et al. [43] have modeled 33% quantum efficiency for Dy³⁺ output at 4.5 µm with a quantum slope efficiency of ~16% - typically 80 mW output is expected for losses of ~1 dBm⁻¹. We have predicted 100 mW output for a 5 W near-IR pump and 3x10¹⁹ RE-ions cm⁻³ with losses of 1 dBm⁻¹.

Prudenzano et al. [45] have modeled a photonic crystal fiber laser of a 10,000 ppm Er³⁺-doped GeGaSb-sulfide glass operating at 4.5 μm output, with pumping at 806 nm. The design was based on measured absorption and emission cross-sections and was for a large-mode-area core, thereby preserving monomode operation and offsetting nonlinear effects. For an output power of 100 mW, the predicted change across the wavelength range of 4.45-4.65 μm was ~11 mW.

6. Towards RE-doped chalcogenide glass mid-infrared fiber lasers: results to date

Table 1 collates absorption and emission behavior of RE-ion doped bulk chalcogenide glasses, and fiber, strictly in the mid-infrared ≥ 3μm wavelength range; near-infrared emissions are not included. In general, quantum efficiency has been calculated by dividing the reciprocal of the measured (experimental) lifetime of the RE-ion radiative state by the theoretical probability of radiative emission as calculated using Judd-Ofelt modeling [46,47]. Mid-IR fluorescence, but not lasing, has been demonstrated in rare earth doped chalcogenide fibers [22,25]. Calculated quantum efficiencies decrease from selenide-based host glasses to sulfide-based host glasses, and from the more covalent ‘p block’ sulfide glass hosts to the more ionic sulfide glasses.

7. Conclusions

Mid-IR rare-earth-doped chalcogenide glass fiber lasers have great potential application across diverse fields. The fiber laser configuration offers better beam quality than competing technologies, for instance, than quantum cascade lasers and transition-metal-doped selenides, and more versatility for pulsed operation. To reach the development stage of mid-IR fiber lasers, a better understanding is required of the local RE-ion environment to obviate unwanted non-radiative decay. Lower optical loss fiber of the complex Ga^III-containing chalcogenide glass compositions is required. A better understanding of the impact of host chalcogenide glass native defects on the doped RE-ion fluorescence is also needed.