1. Introduction and background

1.1. Overview

Over the last decade, there has been a strong growth in the number of advances in fiber-based sources that emit MIR radiation, and in the range of applications [1] that employ these unique sources. Although the start of the MIR spectral range has been defined very loosely by researchers working on MIR technologies, with “the MIR spectral range” being defined as broadly as wavelengths between 2 µm to 50 µm, and as narrowly as 8 µm to 14 µm [2], the international (ISO) definition [3] has the short wavelength limit of MIR as starting at 3 µm. However, we will relax this definition—consistent with the broader usage of “MIR” by numerous practitioners and authors of publications on MIR detectors and lasers, and define the short wavelength limit of MIR as 2.5 µm, to enable discussion of a plethora of fiber lasers—and other MIR sources—that have been developed in the past three decades at wavelengths in the 2.5 µm to 12 µm spectral range. Given the fact that for wavelengths shorter than 2.5 µm, silicate glasses are the dominant materials in fiber optics, we feel that extending the definition to 2.5 µm allows consideration of a practical optical materials-based wavelength “boundary” between silica fibers and the soft glass fibers used at MIR wavelengths. Note that the nonradiative lifetimes and radiative quantum efficiencies of electronic transitions (as described in Section 2.3.1 below) using alternative oxide glasses—such as germanate and tellurite glasses—are also extremely small to enable efficient laser action at wavelengths > 2.5 µm.

The recent growth of activity on MIR fiber lasers has arisen due to several factors. First of all, the variety and quality of optical fibers has reached a point where numerous solid core MIR transparent fibers with losses of <10 dB/km and hollow core silica-based fibers exhibiting optical losses of <100 dB/km have now been demonstrated at wavelengths near the short wavelength edge of the MIR spectrum [4]. These loss values allow lower MIR laser thresholds and a wider variety of lasing transitions to be demonstrated and higher average output powers and pulse energies both from ionic transitions and at other wavelengths via nonlinear interactions. The lowering of optical losses in optical fibers in the MIR is by far the most important parameter relevant to the efficient generation of coherent MIR radiation with fiber-based sources. Likewise, lower costs for fibers have opened the field up to substantial research and commercial opportunities.

The number of physical processes that have been demonstrated to create coherent MIR emissions have also grown. This variety extends from the new electronic transitions in rare earth doped glass to the extensive range of nonlinear processes that have been used in MIR fibers, both in solid-core fibers and in gases inserted into hollow-core fibers. Although reasonably broad tunability has also been demonstrated in the inhomogeneously broadened (due to local field variations in ionic sites in amorphous media such as glasses) electronic transitions such as the ⁴I_11/2 → ⁴I_13/2 transition in the erbium ion [5], a broad range of MIR frequencies are also enabled via supercontinuum generation, in which numerous nonlinear processes, notably stimulated Raman scattering (SRS), self-phase modulation (SPM), four wave mixing (FWM) and cross-phase modulation (XPM) contribute significantly to the near-continuous filling of the spectral range over a wide range of MIR frequencies. In addition, the number of MIR fiber sources—particularly fiber lasers that are relatively easy to build and that provide powerful and efficient high-beam quality laser outputs—continues to grow, as do the number of available pump sources that are relatively easy to incorporate for pumping of MIR transitions, either in single wavelength or multi-wavelength [6] pumping configurations.

1.2. Need for advanced sources of coherent MIR radiation

The need to create high power MIR coherent radiation is driven by the vast number of applications that are currently benefit or that could benefit from operation in the MIR. It is widely known that nearly all biological and organic chemical compounds, many of which are very important to our health, security and the environment have strong fundamental absorption features at MIR wavelengths [7–9]. Due to the strong absorption band in water at 3 µm, and the consequent strong absorption of radiation at this wavelength in body tissues, precision surgery was one of the early motivators for developing fiber-based MIR sources of high average power and high pulse energy fiber lasers [10–12]. For example, it has been shown that 3-µm radiation from a cw fiber laser [13] creates a better excision boundary in various soft tissues (i.e., less carbonisation and minimal cell death) than 2-µm radiation [14] because of the much larger absorption coefficient (and the consequent shorter absorption depth). The wide atmospheric transmission in the 3 µm to 5 µm and the 8 µm to 13 µm bands also offer the opportunity for high bandwidth line-of-sight free-space communications (at least on clear days) and specialized MIR-Longwave IR (LIR) laser weapons and infrared countermeasures (IRCMs); however, reported MIR communications systems to date are reliant on complex optical arrangements involving nonlinear frequency conversion architectures [15,16]. In addition, cutting or advanced processing of materials that have fundamental absorption features in the MIR allows sculpting of precision micron-sized features to create more cost-effective devices and instruments with exceptional operational characteristics. Of recent interest has been the demonstration of accurate ablation of polymers using fiber laser emission at one of the fundamental absorption wavelengths [17]. In defense, optical blocking of infrared signatures and blinding of the “eyes” of incoming missiles via IRCMs increases the survivability of military aircraft [18], and inevitably also of commercial aircraft. The growing field of high-performance MIR sensing has also seen many important developments [19] and requires a whole suite of MIR sources to cover the wide variety of anticipated applications.

In summary, it is becoming increasingly important to provide more efficient and reliable sources of MIR light, preferably with one or more of the following specific characteristics: high beam qualities, high average powers, ultrashort pulsewidths, high peak powers, specific wavelengths of operation, large wavelength tunabilities and narrow linewidths. Achieving specific targets for one or more of these characteristics are the major drivers for new MIR fiber lasers and other advanced sources of coherent MIR radiation. In this review, because of the enormous growth in fiber-based source technologies and of the broad range of devices and applications possible with such fiber-based sources of coherent MIR radiation, we have primarily restricted our discussion to MIR fiber technologies and to some strongly related natural “add-ons”. We examine the key parameters that have resulted in record performance achievements, especially recently, in terms of bandwidths, efficiencies, output powers and emitted wavelengths, and also discuss a few specific anticipated future needs and potentially disruptive approaches to future MIR fiber source developments.

1.3. Advantages of—and specific issues related to—fiber-based sources of coherent MIR radiation

Fiber lasers have established themselves as outstanding sources of high-brightness coherent radiation, particularly at near infrared (NIR) wavelengths. Fiber lasers can generally be fabricated into relatively compact and robust arrangements because the light stays confined in a waveguide (or core), which can be designed to enable light propagation in a single spatial mode, allowing high optical intensities over long propagation lengths in compact arrangements. The light in the core is tightly confined over long lengths, leading to higher gain-length products and lower pump power thresholds for especially “near-ground state” terminating transitions compared to bulk solid-state systems (which also often involve much more complex cavity arrangements). Optical fiber lasers are also much more amenable to working at very high average optical power levels because of the relative ease of cooling the long thin geometry of a fiber. This attribute is particularly important for industrial applications not only because of the high threshold powers needed for the onset of certain processes (such as welding or the cutting of a thick steel plate) but also because of the proportionality between high average powers and the high yields (that are critical for commercial viability of several industrial applications). Fibers also enable very good beam pointing stability because the light is confined to the core and does not rely on relatively sensitive alignment of mirrors. Notably, in a feature unique to optical fibers, the overall performance reliability and robustness is vastly improved when the fibers and devices comprising the laser and delivery systems are spliced together forming nearly integrated “all-fiber” arrangements.

As may be expected, the primary difference between the development of solid-core fiber-based coherent light sources in the MIR and NIR wavelengths are the materials used in the fibers themselves. In the NIR, very robust oxide-based materials dominate, with the silicate glass family by far the most important. Extension of the use of these robust glass materials systems to MIR wavelengths, however, is prohibited by the transparency of the silicate glasses. MIR emitting solid-core optical fiber-based light sources are based on a range of low phonon energy “soft glass” materials, primarily from the fluoride and chalcogenide glass families, although tellurites provide interesting options. Unlike the NIR, which is based more or less entirely on the silicates, in part because of their robustness and transparency, and also because non-radiative decay is not an issue at large transition energies (elucidated in Section 2.3.1 below), it has become clear that the creation of MIR light from an optical fiber platform will not be dependent on just a single “glass family”. As such, a large variety of “soft glass” families are currently used; and in the future, high quality single mode fibers drawn from advanced transparent glass ceramics will most likely be needed, particularly at wavelengths >4 µm, as elaborated in Sections 5 and 6.

There are several parameters that facilitate high power lasers in the MIR because of the specific optical factors that become more favourable at increasing wavelengths. The mode area of the lowest order mode propagating in the core of the fiber, in the weakly guiding approximation, scales as λ², where λ is the laser wavelength. This has a highly beneficial effect on the power scaling opportunities for coherent MIR fiber sources. For example, some losses—and other detrimental effects such as unwanted phase modulation—due to optical nonlinearities decrease with decreasing light intensities, which are obviously reduced (for the same power transported in the fiber) when the mode area is increased. Moreover, nonlinear losses from stimulated Brillouin scattering (SBS) and SRS are smaller—due to reduced SBS and SRS gains—at longer wavelengths; also of significant benefit is the fact that the threshold for catastrophic self-focusing in an optical fiber varies as the square of the wavelength and that the power thresholds for laser induced breakdown increases at longer wavelengths because more photons are required to bridge the bandgap. These favourable parameters drive the creation of high-power light in the MIR, but in practice other issues such as lower Raman gains (when exploited for wavelength shifting), problems with incorporating dopants at sufficiently high concentrations, comparatively low softening and melting points, lower damage thresholds, and lower thermal conductivities of the soft glasses [20,21] used to create the fibers, and inevitable differences (quantum defect) between convenient excitation sources and the emitted wavelengths present new major challenges that can inhibit device or system optimization.

In the NIR, the ability to create rugged fiber Bragg gratings (FBGs), reliable fiber splices, couplers and beam combiners have resulted in robust, alignment-free high-power fiber lasers, offering many unique features, for instance, numerous choices of high power and high beam quality narrow linewidth [22] or ultrashort pulsed [23] laser outputs. Keeping the light confined in the core within all components of the NIR laser, and from the pump source to the end application provides a high degree of robustness and efficiency, which have made fiber lasers an inherently disruptive technology. In contrast, widespread adoption of coherent MIR fiber sources has been delayed by the unfortunate lack of similar fiber component technologies for the MIR. This problem originates in part due to limitations in the glasses used for fiber-based MIR systems. The most successfully—and widely used—MIR-transparent glass to date, particularly for the creation of ionic transition-based MIR single mode laser radiation is a fluoride glass (e.g., a heavy metal fluoride glass composed of ZrF₄, BaF₂, LaF₃, AlF₃ and NaF commonly abbreviated to ZBLAN, described in further detail in Section 2.3 below). The fundamental material properties of ZBLAN are significantly different than those of the silicates i.e., it is less stable environmentally, has weaker thermomechanical properties, and is more difficult to cleave and handle. Despite these challenges, major technological advances have been made steadily with singlemode ZBLAN (and other closely related fluoride) fibers, including demonstrations of FBGs written into doped fluoride fibers [24,25] and splicing special end caps to fluoride fibers to isolate the doped fiber from the environment (and from adverse effects of humidity) and to reduce heating of the end faces when high pump powers are used. This latter procedure has proven to be particularly important when the emitting wavelength is attenuated by small quantities of water that may get absorbed by the fiber [26,27], and has enabled demonstrations of numerous all-fiber coherent MIR sources, such as those that emit tens of watts of output power [28,29] and octave-spanning (2.4 µm to 5.4 µm) MIR supercontinuum sources [30].

2. Generation of MIR radiation via ionic transitions in fibers: choice of ions and host glasses

2.1. Introduction

Coherent radiation is most easily generated by storage of energy in specific excited levels in atoms, ions, or molecules, via a population inversion between targeted energy states, and subsequent stimulated emission between these states. In solid core fibers, this is most commonly achieved via the incorporation of appropriate rare earth ions and transition metal ions in the glass host comprising the fiber core. Note that although such population inversion has been achieved relatively easily in rare-earth doped glass fibers in both NIR and MIR wavelengths, population inversion in transition metal dopants has been much more difficult due to the sensitivity of the MIR transitions to the local field environments and related symmetry requirements for “acceptable” ionic sites, as discussed in Section 2.2.1 below. Nevertheless, lasing has been achieved at NIR wavelengths in compositionally-embedded bismuth [31] fibers, and more recently, preliminary indications of optical gain and fiber laser emission have been demonstrated in the shortwave-MIR (2.3 µm) spectral region [32] with the use of chromium (a transition metal) ions doped in ZnSe crystal-based glass fibers [33]. The Cr²⁺-doped crystalline ZnSe optical fibers used by [32] for their 2.3 µm fiber lasers comprised of a near-crystalline ZnSe semiconductor core embedded in a silica-clad fiber via high pressure chemical vapor deposition processes [34] and are still in a relative stage of infancy, exhibiting intolerably high scattering losses.

2.2. Choices of specific ions and rransitions

2.2.1. Rare earth and transition metals Ions: influence of local field environments on oscillator strengths (absorption and emission cross sections)

The optical transitions of rare-earth ions (RE³⁺) in the NIR and MIR regions all occur within the 4f subshell, which is shielded by the outer 5s and 5p subshells; as such, the influence of the host material is relatively small compared to the situation with the 3d transitions in transition-metal ions which are not shielded and are thus heavily influenced by the host (resulting in major spectral perturbations and significantly shortened lifetimes). In crystalline environments, because of the locations of ions in a narrow range of choices (often just one) of crystalline sites (see Fig. 1) (usually determined by the valence and size of the dopant ions and the host substitutional sites), ionic transitions in crystals are usually well-defined in frequency, depending on the sites that they occupy. This feature results in relatively narrow absorption and emission lines relative to amorphous hosts, such as optical glasses, in which the rare earth ions occupy a broad range of sites, with significantly different local fields, leading to locally perturbed transitions and significant inhomogeneous broadening of the absorption and emission spectra.

Figure 1 shows examples of specific sites occupied by RE³⁺ “dopants” (Er³⁺ in this specific example) in NaYF₄ crystal lattices, and Fig. 2 shows an example of sites occupied typically by transition metal ion (TM²⁺) dopants (such as Cr²⁺ ions) in a “zinc-blende” crystal lattice structure. Note that—in contrast with these well-defined sites in crystalline lattices—ions in glasses occupy a very large range of local field environments, and a near continuum of Stark shifts.

 

Fig. 1. Schematic depiction of Er³⁺ ions doped in (a) hexagonal (b) and cubic phase NaYF₄ crystal lattices. In (a) the hexagonal NaYF₄, an ordered array of F⁻ offers two types of cation sites: one occupied by Na⁺ ions, and the other occupied by Y³⁺ and Na⁺ ions, the Y³⁺ is located in 9-coordination sites with low symmetry, whereas in (b) the cubic phase NaYF₄, Y³⁺ occupies the 8-coordination cubic symmetry site. As expected, for both of these lattices, the Er³⁺ ions are located in sites that would be nominally occupied by a trivalent Y ion. Adapted from [35].

Download Full Size | PDF

 

Fig. 2. Crystal structure of Cr²⁺-doped ZnS crystal, and electronic states responsible for the MIR emission [36].

Download Full Size | PDF

Unlike the luminescence spectra from ions in well-defined crystal sites, single crystal-field transitions between two spin-orbit multiplets cannot be distinguished in glasses, at least at ambient temperatures, because inhomogeneous spectral-line broadening occurs due to the local variation of the electric field experienced by the ion. Homogeneous broadening mechanisms are also relevant in a number of glasses, but inhomogeneous broadening dominates.

A key point to note about RE³⁺ ions is that—because of shielding by the 5s and 5p subshells—the oscillator strengths and emission cross-sections are relatively insensitive to local field environments, and the key differences in the absorption and emission spectra of the same transitions in different glasses is the precise location of the transitions and the absorption and emission profiles and bandwidths. Relative to crystalline hosts, the broadening of rare-earth transitions in glasses is actually an advantage when broadband, continuous wavelength tuning is required; nevertheless, the broadening results in lower absorption and emission cross-sections compared to crystalline hosts. These reduced cross sections can lead to low gains and higher pump power needs for high gains per unit length, but the high light confinement in optical fibers—and consequent high pump intensities—over long interaction lengths more than makes up for the reduced absorption and emission cross-sections in glasses (versus crystals), as long as background losses at the target lasing wavelengths are not a serious issue. As such, this feature—of high pump intensities and long interaction lengths—usually results in much lower pump thresholds and increased conversion efficiencies in fiber lasers relative to their bulk laser counterparts when compared to high-quality crystalline gain media.

However, in sharp contrast with the behaviour exhibited by RE³⁺ ions, TM²⁺ ions, such as the Cr²⁺ ions depicted in Fig. 2 are extremely sensitive to their local field environments, and their oscillator strengths and emission cross sections can vary by several orders of magnitude depending on these local field environments [37,38]. A case in point is the fact that, unlike rare-earth ions, TM²⁺ ions exhibit strong oscillator strengths and correspondingly strong absorption and the emission bands in noncentrosymmetric tetrahedral sites (i.e. sites lacking inversion symmetry) in “zinc-blende” crystal structures (such as ZnS and ZnSe), but these transitions – and corresponding absorption and emission cross-sections—are relatively weak in near-centrosymmetric amorphous and other crystalline sites (such as octahedral sites). Thus, these absorption and emission bands are virtually “non-existent” in amorphous glass-like structures such as ZBLAN or chalcogenide “host glasses” (discussed in Section 2.3 below). The broadband MIR emission spectrum measured from Cr²⁺ ions in ZnS [36] is depicted in Fig. 3; note that despite the fact that this emission is from a crystalline host, unlike the case of RE³⁺ ionic emissions, the emission bandwidth is extremely broad, and even broader than MIR emissions typically seen from various RE³⁺ ions (e.g., Tm³⁺, Ho³⁺, Er³⁺) in amorphous glasses (e.g., ZBLAN), whose spectral locations and approximate lineshapes are also depicted in Fig. 3 for perspective (ignore vertical scales).

 

Fig. 3. Broadband MIR emission spectrum from Cr²⁺ ions in ZnS; note the large bandwidth of the Cr:ZnS emission relative to Tm³⁺, Ho³⁺, Er³⁺ ions in ZBLAN [36].

Download Full Size | PDF

2.2.2. Energy level structures

The dominant rare earth and transition metal ion lasing transitions used for the creation of stimulated-emission-based coherent NIR and MIR light sources are shown in Figs. 4 and 5 respectively. Most of the NIR—and several MIR—fiber lasers have been developed using the “near-ground state” terminating transitions of the RE³⁺ ions that are optically pumped by diode lasers (although the use of NIR fiber lasers as laser pumps at 1, 1.5 and 2 µm wavelengths is becoming increasingly important for MIR fiber lasers, because of their high efficiencies, high beam qualities and specific emission wavelengths [6]). At NIR wavelengths, the near-ground state terminating transitions in Ho³⁺, Tm³⁺, Er³⁺, Yb³⁺, and Nd³⁺ are the dominant transitions, and these NIR lasers are based on optical fibers made from silicate glasses. The transitions used for MIR emissions, however, are a mixture of “near ground state” and excited state terminating transitions, of which only a fraction can be directly diode pumped. In many cases [see for instance, 25,26,39], it has also been necessary to increase the doping of RE³⁺ ions and to exploit favourable energy transfer (ET) processes—as described in further detail in the following section—in order to enhance the emission of MIR light. As stated earlier, none of the transition metals have been made to lase at MIR wavelengths in glass fibers yet, but these transitions hold significant promise for future fiber lasers based on MIR luminescent transparent glass ceramics, as elaborated in Section 5 below.

 

Fig. 4. (a) Energy level diagrams of common triply ionized rare earth ions (RE³⁺). The most common pump and emission wavelengths are indicated by blue and red arrows, respectively. (b) Typical luminescence spectra or emission bands for different RE^3+­­­­ [40].

Download Full Size | PDF

 

Fig. 5. (a) Energy level diagrams of several common doubly ionized transition metal ions (TM²⁺). The pump and emission wavelengths are indicated by blue and red arrows, respectively. (b) Typical mid-IR emission bands for several TM²⁺ in ZnSe crystals [41].

Download Full Size | PDF

2.2.3. Ion-ion interactions: cross relaxation and other energy transfer processes

The simple process of stimulated emission invoked by absorption of pump light and the emission of MIR laser radiation in low-to-moderately-doped glass fibers has many benefits. The design of a fiber laser is usually much simpler, since the dopant concentration can be tailored for optimal absorption at the pump wavelength, including the use of moderate doping in long fibers for control of the heat load. Moreover, the dopant densities and longitudinal distribution can be adjusted to create gain in a manner where the background loss becomes an insignificant issue. Thus, relatively low dopant concentrations might be preferable for high power systems, but many systems benefit when the dopant concentrations are extremely high (in part because optimal inversion and gain creation is often dependent on high doping densities, as discussed below for the case of the 2.7 µm transition in erbium). Most notably, in double clad fibers, the dopant concentration is usually higher because the absorption coefficient of the fiber scales with core to cladding area ratio, leading to a strong reduction in the effective absorption coefficient and the amount of absorbed pump heat load per unit length. Also, for some MIR transitions, the lower laser level has a longer lifetime than the upper level (see Figs. 6(a) and (b) below) and a quenching process for the lower level is essential for efficient lasing, especially in the cw regime (as elaborated later for the case of the Er³⁺ and Ho³⁺ ions, elucidated by energy level diagrams in Figs. 8 and 9). Lastly, a sensitizer ion that absorbs the pump light and transfers that excitation to the lasing ion can be exploited for more efficient optical pumping of the upper laser level (as described later for the Co-Fe TM²⁺- ZnSe system in Fig. 17). To date for MIR fiber laser systems, which are currently all based on rare earth ions to excite the upper laser level, there has been little need for adding sensitising ions and it has been the need to quench the lower laser level that has required significant energy transfer (ET) considerations. We will discuss the most important examples below, namely the well-known energy recycling process relevant to the ⁴I_11/2 → ⁴I_13/2 transition of Er³⁺ [42,43] and the co-doping of Ho³⁺ with Pr³⁺ ions. Note that MIR Er³⁺ fiber lasers have has also been co-doped with Pr³⁺ ions leading to the first demostration of >1 W output from a MIR fiber laser [44], but this approach has been abandoned in favour of the more efficient energy recycling process [45,46].

 

Fig. 6. Simplified energy level diagram of (a) the Er³⁺ ion showing the primary MIR 2700 nm – 2800 nm laser transition and the primary pump transition which is relatively easily addressed via high power 975 nm to 980 nm diode laser pump radiation [39]; (b) the Ho³⁺ ion showing the primary MIR laser transition (2840 nm – 2950 nm) and the primary pump transition which is also normally addressed using 1150 nm diode lasers [52], and (c) the Dy³⁺ ion showing the primary near-ground-state terminating MIR laser transition at 3000 nm, and the numerous choices of primary pump transitions which are normally addressed using fiber laser pumping [53].

Download Full Size | PDF

As seen in Fig. 4(b) above, there are three separate RE³⁺ ions that yield MIR emission at wavelengths around 3 µm, and each of these has been used successfully for strong lasing action in fibers. Figures 6 and 7 display the energy level diagrams and the measured emission cross sections for each of these transitions in additional detail, along with information on nominal radiative lifetimes of each level and the most common pump wavelength options (determined largely by readily available pump sources). We focus first on the “grand-daddy” of MIR fiber laser systems, namely that based on the 3 µm transition of the Er³⁺ ion. Starting with the simplified energy level diagram for the Er³⁺ ion in Fig. 6(a), Fig. 8 schematically depicts an energy transfer upconversion (ETU) process [43] that occurs quite readily between two Er³⁺ ions at small interion spacings, (i.e., at high doping densities, typically >20,000 ppm) and high pump intensities (typically >1 Watt of pump power). This ETU process not only quenches the lower laser level but also helps “recycle” energy from the lower laser level to the upper laser level. In other words, quenching of the “potential bottleneck” at the ⁴I_13/2 level of the Er³⁺ ion is achieved via the ETU process ⁴I_13/2, ⁴I_13/2 → ⁴I_9/2, ⁴I_15/2, which—as shown very clearly in Fig. 8 below—populates the ⁴I_9/2 level after energy from one excited Er³⁺ ion transfers energy to an adjacent excited Er³⁺ ion. This mechanism was analysed carefully via a comprehensive rate equation model by Pollnau and Jackson [47], and predicted to create overall lasing efficiencies that are significantly higher (i.e., >50%) than simply quenching the lower laser level [40,44] using ET to a desensitising co-dopant such as Pr³⁺. Experimentally, however, this has been difficult to prove with only one demonstration to date [48] of conversion efficiencies (i.e., 39.5%) beyond the 34.3% Stokes limited efficiency for this NIR pumped MIR laser system. It has been postulated [49] that the rate parameters for ETU are much lower in the optical fiber compared to the bulk material [50] for reasons that are still not clear, perhaps in part due to the heavy clustering of Er³⁺ ions that has been observed to occur at higher ionic densities [46], and in part because of the introduction of non-radiative decay from drawing-induced defects. Recent time-resolved spectroscopy measurements suggest that depending on the fiber manufacturer, the rate parameters for ET may occur somewhere between the values relating to the strongly interacting and weakly-interacting parameters [51].

 

Fig. 7. Measured emission cross sections of the 3 µm class transitions of the Ho³⁺, Er³⁺ and Dy³⁺ ions extracted from the fluorescence spectra measured when these ions are doped into ZBLAN glass [54].

Download Full Size | PDF

 

Fig. 8. Energy transfer upconversion (ETU) process between two proximally located ⁴I_13/2 excited Er³⁺ ions, preferably at interion distances of <20 nm from one another [43,46]. In this process, energy transfer between two Er³⁺ ions initially in the first excited state (⁴I_13/2) results in a recycling of excitation back to the upper laser level (⁴I_11/2) that creates a “two-for-one” (two lasing photons per pump photon) mechanism that boosts the slope efficiency. The competing “gain lowering ETU process”: ⁴I_11/2, ⁴I_11/2 → ⁴F_7/2, ⁴I_15/2 is much weaker than the desirable “gain increasing ETU process”: ⁴I_13/2, ⁴I_13/2 → ⁴I_9/2, ⁴I_15/2. The small interion distances are facilitated by using extremely high concentrations of Er³⁺ ions (up to 10 mol.%, limited only by solvation issues) in double clad fluoride fiber geometries.

Download Full Size | PDF

The other significant ion-ion interaction process that assists with the emission of MIR laser radiation from a fiber laser relates to the Ho³⁺ ion (see Fig. 6(b)). Like the Er³⁺ ion, the ⁵I₇ lower laser level of Ho³⁺, has a longer lifetime (∼11 ms) compared to the upper laser level lifetime (∼3.5 ms), and a worse lifetime ratio, leading to accumulation of population at the ⁵I₇ lower laser level, which can easily create a “bottleneck” and cause this lasing transition to self-saturate. Figure 9 depicts a unique process [55,56], namely ET out of the lower laser level, by which this self-saturation of the output has been mitigated for the Ho³⁺ ion. In the first demonstration of MIR fibre laser emission using Ho³⁺, Pr³⁺-co-doped ZBLAN fiber [55] a Ho³⁺:Pr³⁺ concentration ratio was 1:20 and direct core pumping at a wavelength of 1100 nm was used. The measured slope efficiency of 3.2% was hampered by pump excited state absorption (ESA) and probably a non-optimal RE³⁺ ion concentration ratio. The ET process Ho³⁺(⁵I₇) → Pr³⁺(³F₂, ³H₆) is resonant, making it fast and effective. The post-ET excited Pr³⁺ excited states (³F₂, ³H₆) relax very quickly (<10 µs) via multiphonon emission to the ground state. There are also no Pr³⁺ ion energy levels that are resonant (or energetically proximate) with the upper laser level of Ho³⁺, unlike the case of the Er³⁺-Pr³⁺ co-doped system, making Pr³⁺ co-doping a very effective method for quenching the lower laser level of Ho³⁺, and thereby providing optical gain for this MIR transition in holmium. Also, unlike for the case of the Er³⁺ ion, ETU will not assist in recycling the energy [56]; the rate parameter for ETU initiating from the upper (⁵I₆) laser level of the Ho³⁺ ion, i.e., the ETU process: ⁵I₆, ⁵I₆ → ⁵I₈, ⁵F₅ is larger by an order of magnitude compared to the rate parameter for ETU that starts from the lower (⁵I₇) laser level i.e., the ETU process: ⁵I₇, ⁵I₇ → ⁵I₈, ⁵I₆. Therefore, in contrast with Er³⁺, ETU for the Ho³⁺ transition reduces the population inversion and co-doping Ho³⁺ with Pr³⁺ ion has thus been the preferred approach. Experimentally, Pr³⁺ ion concentrations in the range 2000 to 3000 ppm, and Ho³⁺:Pr³⁺ concentration ratios of approximately10:1 are sufficient to quench the lower laser level.

 

Fig. 9. Energy transfer between excited Ho³⁺ ions and ground state Pr³⁺ ions. The energy transfer process is highly resonant meaning relatively low concentrations of Pr³⁺ are required for effective quenching of the lower laser level of the Ho³⁺ transition [55].

Download Full Size | PDF

2.3. Host (glass) issues

2.3.1. Phonon energy requirements and multiphonon processes

A radiative transition from an excited initial state (energy level) i to a lower-lying state j is characterized by the radiative rate A_ij which has units of s⁻¹. If a number of decay processes occur from this initial level i to several lower states, the overall radiative rate is given by $A_{i} = \sum_{j}^{n} A_{ij}$, i.e., the sum of all individual rate constants. The branching ratio of each radiative transition is defined as β_ij = A_ij/A_i [57]. Radiative decay processes of any excited state however compete with nonradiative relaxation of that level i to the various lower levels j, as may occur by simultaneous transfer of energy from level i to optical phonons in the host. The rate of nonradiative relaxation (i.e., multiphonon decay), W_i, decreases exponentially with the number of phonons that need to be emitted to bridge the energy gap from level i to the lower levels, but if the lower levels are widely spaced in energy, the decay to the lower energy level that is closest in energy to the upper level j is the most dominant. This is largely due to the inverse exponential dependence of the nonradiative decay rates on these energy gaps; as a “rule of thumb”, the nonradiative multiphonon relaxation rate is usually negligible relative to the radiative decay rates of each level if the energy gap between that level and the adjacent lower level is over 4 times the highest optical phonon energy in the material [58]. The rate of multiphonon relaxation also increases with temperature; as such, for several early demonstrations of MIR lasing from a fiber (and other solid state MIR lasers), the host material was cooled to cryogenic temperatures, especially if the highest phonon energies in the host glass are comparable to the MIR transition energy [38,59]. Note that laser efficiencies of many traditional MIR lasers can still be enhanced significantly with the lowering of temperatures of the host because of this issue. The final luminescence lifetime, τ_i, of an excited energy level i is the inverse of the sum of the radiative rate constant A_i and the rate constant of multiphonon relaxation, W_i. Note also that the radiative quantum efficiency is usually defined as QE = A_i / (A_i + W_i).

The contribution of the multiphonon relaxation process to the overall relaxation rate of any given energy level is much stronger in oxide glasses (e.g., the borates, phosphates, silicates and germanates) compared to fluorides [58] because of the larger values of the highest phonon energies in these glasses (comprised of lighter constituent atoms), which imply the need for fewer phonons to be simultaneously generated to enable this nonradiative decay process [58]. This factor is much more obvious when the oxide glass families are compared to the “heavy atom” glass families—such as the chalcogenides—which have even lower phonon energies. Table 1 lists “typical” reported values of several important physical parameters of major host glass families used for MIR fiber lasers, along with comparable “reference” values of silicate glasses. As stated above, the higher phonon energies in silicate glasses imply that the nonradiative multiphonon relaxation rates [58] are much higher in these glasses, and they are thus unsuitable for MIR lasing transitions, as is more obvious from quantitative estimates of these relaxation rates for specific energy gaps or transition energies in Fig. 10 below. A related factor of great significance for the choices of glasses for MIR fibers is the relationship between the MIR transparency and the phonon energies due to the role of multi-phonon absorption, which once again favors the choice of low phonon energy glasses (such as the chalcogenides) for MIR fibers. Last, but not the least, heavy metal glasses have weaker bond strengths and consequently lower softening and melting points (hence their labeling as “soft glasses”), which favor several practical fabrication issues, but limit the “ruggedness” of these glasses and glass fibers (relative to their silica counterparts).

 

Fig. 10. Plots of theoretically estimated room temperature nonradiative decay rates as a function of the energy gap between an energy state and its adjacent lower level (typically “upper” and “lower” energy levels of an ionic transition) for several glass families, based on assumed values of highest phonon energies [58].

Download Full Size | PDF

Table 1. Reported “typical” values of maximum phonon energies and several other key physical parameters for major glass material families used to make MIR optical fibers

View Table | View all tables in this article

As stated above, as a simple “rule of thumb”, nonradiative decay can dominate over radiative decay (with typical radiative lifetimes of the order of ∼3–10 ms) if four (or less) phonons are required to bridge the energy gap. This is illustrated quantitatively in Fig. 10 (adapted from [58]) via plots of the nonradiative decay rates (W_nr), which also indicate the decay rates for a couple of important ionic rare earth transitions (see Figs. 4 and 5 above) for a large number of glass families as a function of the energy gap between that level and the closest or adjacent level below the level “i” under consideration. Note that these plots only provide an approximate indication for the nonradiative decay rates or nonradiative lifetimes of a specific energy level in a chosen “isolated” ion inserted in an appropriate glass family, based on maximum phonon energy assumptions for each glass family in the plot; a more serious reader should get actual values of the energy gaps and phonon energies for their specific ion transition-glass combination, and recalculate the values of W_nr and A_nr based on [58] or other related theoretical models. As an example, the nonradiative decay rate for the ⁴I_11/2 transition in Er³⁺ is estimated to be ∼100 s⁻¹ in ZBLAN, a fluorozirconate glass, corresponding to a nonradiative lifetime of ∼10 ms, which is comparable to the radiative lifetime for this level, and does not have a devastating impact on its energy storage properties and the achievement of population inversion, whereas achievement of population inversion of this transition in germanate, tellurite, silicate, phosphate and borate glasses becomes increasingly more difficult due to the increase in phonon energies in these glasses—and consequent increases in the nonradiative decay for this level—totally inhibiting energy storage and bypassing the radiative decay channel, making population inversion extremely difficult no matter how hard one pumps energy into this state in such glass hosts. Note also that the plots show a characteristic logarithmic dependence of the decay rates, but the slopes of these plots strongly depend on the host [58] and are best estimated empirically.

Last, but not the least, ignoring other issues related to fiberizability of the glass or crystal that is to be used, this plot elucidates the difficulty or near-impossibility of achieving efficient population inversion at wavelengths longer than 4 µm (i.e., an energy gap of 2500 cm⁻¹) in currently developed MIR transparent glass fibers. As an example, the 5 µm ⁷F₅ → ⁷F₆ transition in terbium has an anticipated nonradiative decay rate of over 3000 s⁻¹ (implying a nonradiative lifetime of approximately 300 µsec) in the lowest phonon energy chalcogenide glasses, whose fiberizability into low loss single mode fibers is still relatively poor. As such, novel solutions, such as the use of transition-metal-doped ZnSe glass fibers [36] and the use of advanced transparent glass ceramics containing ion-doped PbF₂ LaF₃ or ZnSe crystallites in appropriately-fabricated TGCs (transparent glass ceramics), ideally with suitably index-matched glasses [67], as discussed in Section 4 below, are anticipated to be critical for efficient inversion-based MIR coherent sources at wavelengths longer than 4 µm. Nevertheless, the need for longer-wavelength coherent MIR sources can be easily addressed by the use of nonlinear conversion methods such as SRS, Raman fiber lasers (RFLs) and optical parametric oscillators (OPOs), the latter invoking phase-matched FWM [68] in appropriately-designed MIR transparent fibers with appropriately large nonlinearities and optical properties (such as birefringence and modal group velocities in multimode fibers) that facilitate efficient phase matching.

As mentioned earlier, NIR fiber lasers using silicate glass fibers are a very mature technology that forms the backbone of a large range of commercial fiber laser products and the multibillion-dollar fiber amplifier products used in the fiber optic telecommunications industry; however, solid core silicate glass fiber lasers are limited to emission wavelengths shorter than about 2.2 µm [69] because of multiphonon absorption and relaxation processes. In addition, heavy metal oxide glasses based on tellurium and germanium usually contain unacceptably high precursor impurities which represent an additional source of high loss; as such, the quantum efficiency of MIR transitions in heavy metal oxide glasses is practically negligible [70]. Fluoride or fluorozirconate glass fibers based on the ZBLAN compositions are the most advanced MIR glasses, but other compositions, most notably glasses based around InF₃ precursors appear to present excellent alternative glass choices for low-loss MIR fibers in the 2.5 µm to 4 µm spectral range. In addition, fluoride glasses demonstrate very high rare-earth solubilities; as such, fluoride glasses can be doped with very high rare earth concentrations (of the order of 100,000 ppm in Er³⁺:ZBLAN), and this property, combined with the ultralow singlemode fiber losses demonstrated in the wavelength region between 2.5 µm and 4 µm have made fluorozirconate (notably ZBLAN) fibers the best candidates for the development of practical, efficient and powerful 3-µm-class (and potentially 4-µm-class) rare earth doped fiber coherent sources.

Figure 11 displays typical MIR loss spectra for state-of-the-art single mode fluoride optical fibers along with an “solid core” chalcogenide fiber with one of the lowest reported losses to date [71]. Note that lower losses (<50 dB/km) have been demonstrated in microstructured chalcogenide fibers [66]. The chalcogenide glass families display unsurpassed transparencies at wavelengths longer than 5 µm (because of the much lower phonon energies of these glasses, and consequently reduced multiphonon absorption at longer MIR wavelengths) and high optical nonlinearities (related to the high refractive indices in glasses based on heavy atomic constituents), but are generally very difficult to dope with rare earth cations without detrimental effects to fiber loss. Nevertheless, a recent study [72] reporting long luminescence lifetimes and relatively low fiber background loss illustrates the potential of GeAsSe chalcogenide glasses fiber for direct ion doping with moderately high concentrations. Also, as discussed in Section 5.3.1.3 in considerable detail, ion doped TGCs have exhibited longer luminescence lifetimes and relatively bright broadband luminescence, but the drawing of such TGCs into ultralow loss singlemode optical fibers is still very much in a relative stage of infancy.

 

Fig. 11. Measured attenuation spectrum of commercial passive (undoped) fluoride optical fibers. Fluoride data supplied by Le Verre Fluoré (Brittany, France) and GeAsSe data from [71]. Note the large peak at ∼4.5 µm (that resulted in off-scale absorption) relates to a fundamental resonance with the Se-H bond.

Download Full Size | PDF

2.3.2. Physical issues: fiber draw process and related glass properties

The creation of “usable” optical fibers involves a number of important fabrication steps. As opposed to silicate glass fibers, which are normally fabricated using gaseous phase precursor materials and vapour phase deposition processes with ultrahigh purity gas precursors, soft glass fibers involve solid-state precursors which are much less pure than gases. For low loss fibers, the precursor materials usually require further purification because even small (<1 ppm) concentrations of impurities can result in unacceptably large fiber loss values [73,74]. The initial preform fabrication process involves batching, melting, casting and usually annealing. Such “melt-cast” glasses form within a region of the compositional phase diagram with a resultant large variation in bond angles and bond lengths. Note also that these glasses require fabrication in a controlled environment (typically in a glove box) in order to prevent further contamination. Despite the disadvantage of low overall purities, and the related impurity losses, a major advantage of this method of glass formation is its relatively high simplicity and extreme versatility, allowing an almost limitless number of glass combinations; the latter factor explains the large range of soft glass studies in the literature, including a large number of spectroscopic studies of RE³⁺ ion doped MIR (or “soft”) glasses.

Preforms can be made from such glasses by a large variety of methods, including a standard rod-in-tube or the double crucible methods [75] or by extrusion processes [76]; an optical fiber preform is then placed in the draw tower and optical fibers are drawn under controlled conditions of temperature and draw rate, which are significantly different than those in silicate fibers because of the much lower melting points of the soft glasses used for all MIR optical fibers (i.e., between 350 and 700 °C, for soft glass fibers versus the 1750 to 1950 °C draw temperatures used for silicate glass fibers). The temperature of the furnace in the draw tower clearly depends on the softening temperature of the glass; many soft glasses usually have T_g < 500 °C. As is typical of all drawn fibers, the geometry of the optical fiber is established in the preform and is maintained in the fiber. Of the glass parameters, the differential thermal expansion coefficients of the various parts of the preform are particularly important since they control the degree of the stress that is imparted to the fiber. (In polarizing fiber, for example, the differential expansion coefficients can be taken advantage of in order to create large in-fiber stresses and hence strong birefringence.) The thermal conductivity of the glass is important for many applications of the optical fiber, especially when cooling is a requirement or an advantage; the larger output powers produced from silicate-based fiber lasers are in part related to their relatively higher thermal conductivities relative to the soft glasses, see Table 1.

3. Generation of MIR radiation in fibers via nonlinear conversion methods

3.1. Overview of nonlinear conversion processes in fibers

As indicated earlier, because of the high intensities achievable over extremely long propagation lengths in optical fibers, it is relatively easy to observe extremely large manifestations of a large range of nonlinear optical effects, such as SBS, SRS, FWM, SPM, and XPM, in optical fibers even when the input optical powers are relatively modest [77–79]. Although many of these processes are highly detrimental for simple optical transmission applications, and can have disastrous effects on fiber optical communications systems [for instance, see 79], these effects can be exploited for very efficient wavelength conversion of coherent NIR and MIR optical sources to generate wavelengths that are not easily achievable by ionic, atomic, or molecular transitions. Stimulated Brillouin scattering exhibits very low threshold powers (<10 mW) for sub-gigahertz optical inputs and is potentially very useful for fine tuning frequency shifts of such narrow linewidth MIR sources by amounts of the order of 10 GHz. However, it has seldom been used for MIR applications to date and will not be discussed any further in this review.

Frequency conversion of high average power cw or long-pulsed sources is best done via SRS or FWM, both of which have proven to be extremely effective sources of frequency conversion in optical fibers. However, since phase matching of the FWM interaction is a much more complex process in non-birefringent singlemode fibers, SRS is generally the simpler and more practical process for the attainment of new MIR wavelengths, especially for cases that involve down conversion or long-wavelength shifting of the input radiation to MIR wavelengths beyond 4 µm. The use of SRS to make RFLs is discussed in some detail in the following section. When ultrashort pulses inputs are used, SPM and XPM, along with FWM can work in tandem with SRS to continuously broaden the bandwidth of the radiation through a series of cascaded steps to generate a broad continuum of wavelengths; this multistep complex nonlinear optical interaction is simply designated as “continuum generation”.

3.2. Stimulated Raman scattering and Raman fiber lasers

As discussed above, RFLs are very attractive sources of laser radiation, particularly at wavelengths at which there are no “invertible” atomic, ionic, or molecular transitions [78–80], as is frequently true at MIR wavelengths. The attractiveness of RFLs stems largely from the fact that SRS with extremely high optical gain coefficients (termed “Raman gains”) can be easily achieved in low-loss optical fibers at nearly arbitrary target wavelengths λ_R in the transparency windows of the fibers simply by using moderate pump powers at pump wavelengths λ_P. The choice of lasing wavelengths, pump wavelengths, and pump bandwidths is not only enabled by the broad range of phonon energies available in glasses (because of their amorphous nature) leading to very broad Raman gain bandwidths, but also by the feasibility of using multi-Stokes or “cascading” processes—along with a cascaded series of FBGs—because of the extremely high gains and high conversion efficiencies achievable in each Stokes order [81–87]. As such, CW tunable Raman fiber lasers have been demonstrated over the last few decades (since the 1970’s) by numerous researchers [78,81,83,88,89] as highly efficient sources of broadly-tunable coherent radiation at many wavelengths ranging from the visible to the MIR. The high efficiency of RFLs—limited only by the quantum defect between the pump and the Stokes wavelengths—is not only enabled by the low fiber losses and the high gains due to the possibility of confining high pump intensities over long interaction lengths, but also due to the general absence of other superfluous loss mechanisms in “intra-fiber” laser cavities, especially when all needed component fibers are fusion spliced to each other.

Figure 12 and Table 2 summarize the Raman gain properties of key glasses of interest here. Figure 12 shows the normalized Raman gain spectra, while Table 2 shows (a) the estimated peak Raman gain coefficients [90–95] for representative members of each “family” of glasses, (b) the approximate Stokes shifts (in cm⁻¹) at the peak Raman gain energies (corresponding to the largest phonon densities), (c) the approximate Raman gain bandwidths (in cm⁻¹), and (d) the nominal spectral “transparency windows” for several of the commonly used MIR glass families namely tellurites, chalcogenides, and fluorides (represented best in the present context by ZBLAN). The data for silica glass—which is the most extensively studied and used glass for Raman fiber glass—has been added to Table 2 for reference as a comparison baseline or “standard”, in part to give perspective on the relative magnitude of the Raman gains, Stokes shifts, and bandwidths in the various MIR glasses. Note that most of the Raman gain measurements were performed at NIR wavelengths and—except for silica—the peak Raman gain coefficients in the MIR glasses shown in Table 2 have been scaled to a wavelength of 3 µm using the inverse wavelength dependence [77,91] of the Raman gain coefficient on the excitation wavelength, λ_ex (G_R ∝ 1/λ_ex). As such, the peak Raman gain in the relatively “popular” ZBLAN glass used extensively for rare-earth-doped MIR fiber lasers is comparable to that in silica. The peak Raman gains in several tellurite (e.g., 78%TeO₂-%Bi₂O₃-%ZnO-12%Na₂O, abbreviated TBZN) and chalcogenide (i.e., As₂S₃, As₂Se₃) glasses on the other hand are nearly two to three orders of magnitude larger than those in silica, indicating their strong promise for low-pump-threshold RFLs even with fiber lengths of less than a meter.

 

Fig. 12. Normalized Raman gain spectra for arsenic selenide (As₂Se₃), arsenic sulfide (As₂S₃) and tellurite (TBZN) glasses [90].

Download Full Size | PDF

Table 2. Raman gain characteristics of a few key NIR and MIR glasses along with their transparency windows

View Table | View all tables in this article

Note also that the chosen values of Raman gain bandwidths in this table correspond to the 50% gain point in the representative Raman gain curves in Fig. 12. At higher pump powers, much larger gain bandwidths (and a much larger range of corresponding lasing wavelengths of the RFL) should be possible. As an example, for the tellurite glass TBZN, if the pump powers were high enough to enable attainment of lasing threshold at the 15% gain point in the Raman gain curve (see Fig. 12), a Raman laser tuning range of 750 cm⁻¹ (corresponding to Stokes shifts from 75 cm⁻¹ to 825 cm⁻¹), i.e., over 5 times larger than the 140 cm⁻¹ gain bandwidth value stated in Table 2 (for the 50% gain point) should be readily achievable.

3.3. Continuum generation

As stated in the Overview section above (Section 3.1), when ultrashort pulses are used as the input “pump lasers”, the nonlinear processes of SPM and XPM, along with FWM can work in tandem with SRS to continuously broaden the bandwidth of the radiation in a series of cascaded steps to generate a broad continuum of wavelengths usually called continuum generation. Since the discovery of supercontinuum generation in bulk glasses in 1970 [96,97], there has been a significant increase in the number of demonstrations of supercontinuum sources [98], particularly in optical fibers, with a large number of these experiments focussed on demonstrating ultrabroadband MIR sources. These demonstrations are founded on advances in the materials used to make optical fibers, the control of the dispersion properties of the fibers, and the specific laser sources used to excite the supercontinua. Key achievements and recent advances on MIR continuum sources are elaborated on in Section 4.5 below.

4. Key scientific and technological accomplishments to date

4.1. Overview of key achievements with fiber-based coherent sources

As stated earlier, advances in low-loss glasses and fibers and of the related fiber components and technologies—such as the development of high-quality fusion splices and spliced endcaps, thermally-conductive coatings for efficient fiber cooling, and development of very sophisticated intrafiber reflectors or FBGs have led to many significant technologically feasible ideas and experimental advances in fiber-based coherent sources of MIR radiation in the last decade, particularly in the past few years. This section describes several of the key achievements in fusion splicing and FBG inscription in soft glass fibers, and their use for developing high average power MIR fiber lasers, ultrashort pulse sources, RFLs and high average power sources of continuum radiation.

4.2. Post-processing of fibers: issues and key achievements to date

Advanced post processing technologies for optical fibers —to improve its properties or to enable fusion splices, transparent end caps at the end of doped fibers, and unique and sophisticated intra-fiber reflectors (i.e., FBGs)—are essential requirements for the maturation of the any fiber-based device, including coherent sources of light. These critical milestone technologies were developed quite thoroughly for silicate glass fibers between the seventies and the nineties, but development of these critical post-processing technologies and related components occurred much later for soft glass fibers. The first reports of significant progress in these technologies occurred in the mid-nineties and continue to be developed and refined.

4.2.1. Coupling of radiation into fibers, fusion splicers, and end caps

Splicing soft glass optical fiber keeps the light “in fiber” which makes the overall optical system more robust. Splicing can even be carried out between fibers made from different glasses. Srinivasan et. al. [99] demonstrated rugged fusion splices between similar 15 µm core diameter single mode ZBLAN fibers with record values as low as 0.08 dB, and average splice losses of 0.3 dB but splices attempted between fluoride and silica fibers were fragile despite occasional demonstrations of splice losses below 0.2 dB. More recently, Pei et. al. [100] demonstrated fusion splicing between ZBLAN and silica fibers with different mode diameters (7.8 and 4.3 µm, respectively) using an arc splicer, with a splice loss of 0.14 dB, followed by strengthening the fragile fusion-spliced joint with optical glue. Subsequently, Okamoto et. al. [101] demonstrated a mechanical splice between single mode silica and Pr³⁺-doped ZBLAN fibers with a transmission loss of 0.5 dB. Thermal splicing methods, with splice losses of <0.6 dB between fluoride and silica fibers were also demonstrated by Mahrous et. al. [102] and by Nauriyal et. al. [103], followed by a much more successful demonstration of <0.3 dB reproducible splice losses by Huang et. al. [104] between single mode ZBLAN and silica fibers by setting the filament position closer to the silica fiber to facilitate increased softening of the silica fiber. Finally, Carbonnier et. al. [105] demonstrated average transmission losses of 0.225 dB between the ZBLAN and silica fibers with the use of CO₂ laser glass-processing station and a novel encapsulation technique. Chalcogenide to silica fiber fusion splices have also been demonstrated [106], but apparently have not yet been used as extensively as the fluoride to silica splices in advanced fiber laser systems.

Fusion splicing has also been used to make “end caps”, i.e. short sections of undoped ZBLAN fibers at the end of heavily doped gain fibers. For instance, undoped germanate-based fibers have been used as end caps for the delivery of high power 3 µm radiation from fiber lasers [26], and more recently Uehara et. al. [107] used a CaF₂ end cap for demonstrating a power-scaled ultrastable 30 W amplified Er³⁺-ZBLAN fiber laser.

4.2.2. Intrafiber reflectors: fiber Bragg gratings

As stated above, FBGs function very well as intra-fiber reflectors at specific target wavelengths and bandwidths. More specifically, FBGs can be designed either as ultra-narrowband reflectors or filters within the fiber or as reflectors over a broad range of wavelengths, limited only by the length of the photosensitive fiber, the range of frequency chirp imparted on the FBGs, and practical setup limitations [98] with arbitrary reflectivity designs and controlled amount of reflectivities at all wavelengths within the target bandwidth [108]. In the past two decades, several developments, including (1) materials modifications to enhance photosensitivity [109] of MIR core glass materials and (2) new photo-inscription techniques—including creation of localized phase transitions and other methods of permanent index modifications via multiphoton processes induced by high intensity ultrashort pulse lasers [110–116]—have been developed to “write” optimally designed FBGs in soft glass fibers (chalcogenides, tellurites, and fluorides), leading to unique intracavity high reflectivity (HR) mirrors and output couplers (OCs) that enable control of lasing bandwidths and optimized output coupling from such MIR fiber lasers. For instance, by appropriate adjustment of the photo-inscription techniques, these FBGs may be uniform (a simple sinusoidal index pattern) or “chirped” to enable larger reflection bandwidths [108], or contain highly-localized discrete phase shifts (to enable ultra-narrow band transmission capabilities as with π-shifted gratings), or be apodized or superstructured for specific fiber laser architectures and spectral requirements. Moreover, most of these specialized FBGs can be written either via holographic or phase-mask interference methods or via point-by-point writing, the latter usually being done by ultrahigh intensity laser writing methods [114–116].

A key parameter relevant to the design of FBGs is the maximum value of the refractive index change or “modulation”, Δn, that can be achieved for FBGs inscribed in the cores of the fibers, which in turn determines the coupling coefficient, κ, and thus the reflectivity [108] that may be achievable in appropriate fibers made from such materials. Table 3 below depicts information on previously demonstrated values of the index modulation depth (Δn) and the corresponding coupling coefficients, along with the fabrication methods (illumination wavelengths and writing conditions) chosen for FBGs written in silica (the standard “reference” material) and in various MIR glasses [109–113,115–120].

Table 3. Values of refractive index modulation (Δn) demonstrated in various representative glasses, the corresponding coupling coefficients (κ) and the illumination wavelengths and writing conditions used to achieve these values

View Table | View all tables in this article

Table 3 not only shows that that much higher coupling coefficients are readily achievable in FBGs written in As₂Se₃ fibers compared to those written in silica fibers, but that for the most relevant chalcogenide glasses, these FBGs can be written quite easily with relatively simple low power laser sources. On the other hand, if larger Stokes shifts are desired, particularly with the generation of longer MIR wavelengths with relatively short wavelength MIR pump sources, it is usually preferable to use tellurite glasses similar to TBZN (including 75%TeO₂-%ZnO-5%Na₂O glass abbreviated TZN and 73%TeO₂-%ZnO-5%Na₂O-2%La₂O₃ glass abbreviated TZNL), in which the FBG lithography is not as easy as in the chalcogenides; nevertheless, FBG lithography in most MIR fibers still compares very favorably to that in silica, both in terms of ease of writing and the magnitudes of the coupling coefficients that are achievable. One should also note that, fabrication of FBGs and of creation of localized phase shifts (such as the π phase shift element) at MIR wavelengths (relative to the situation at NIR wavelengths) requires lower writing resolutions, simplifying the task of writing such FBGs, particularly for the much longer MIR wavelength sources discussed in the cascaded-RFL-pumped narrow linewidth (NLW) sources described in Section 4.4.1.2 below.

Based on the above discussion, it should be clear that the choice of glass to be used for a specific Raman fiber laser application is based in part on the availability of robust near single-mode low-loss fibers—preferably with very small mode areas—based on the above-described choices of soft glasses, but also on the magnitude of the peak Raman gain coefficients, the desired values and ranges of the Raman shifts, and the ease with which the desired FBGs can be written in these fibers with appropriately large coupling coefficients. In this regard, selenium-based chalcogenides and TBZN glasses appear particularly promising [90], assuming that MIR fibers of sufficiently low mode areas and ultralow losses can be achieved in fibers made from these glasses.

4.3. MIR fiber lasers and fiber amplifiers based on ionic transitions

This section consists of 4 subsections, all covering key accomplishments to date on coherent MIR sources, (MIR fiber lasers or fibre amplifiers) based on ionic transitions. The first subsection (4.3.1) focuses on high average power broadband cw MIR fiber lasers and power scaling issues, the second (4.3.2) discusses wavelength tunable MIR fiber lasers, the third (4.3.3) covers long pulsed (nanosecond to millisecond duration) sources based on Q-switching, gain switching and gain modulation, and the fourth describes advances in ultrashort pulse duration (100 picoseconds and shorter) modelocked MIR fiber lasers.

4.3.1. High power MIR fiber lasers and power scaling issues

The initial demonstration of lasing, either by cw or pulsed pumping, of any laser device is considered to be a critical step for viability of a new laser transition, including the choice of the glass host and the pumping process used. This is particularly important for the MIR fiber lasers based on RE³⁺ ion transitions because of the relatively large number of choices of transitions that can emit at MIR wavelengths. Unfortunately, the necessary pump wavelengths needed to excite the upper laser level and create a population inversion at some of these transitions do not always match with the availability of efficient, high power laser diodes. In addition, many of the MIR transitions are far above the ground state of the ion, resulting in an apparent need for short wavelength pump sources unless multiphoton pumping or fortuitous energy transfer processes can be used. Without the use of suitable energy transfer processes (as are feasible with Er³⁺ and Tm³⁺ ions at 3 µm and 2 µm respectively), short wavelength pumps can lead to extremely low overall efficiencies and significant internal heating because of the inevitable large quantum defect. As a result, it has been the demonstration of cw—or long pulse (nanoseconds or longer durations)—MIR laser operation with readily available (and powerful) NIR pump sources that has usually determined the “technological fate” or practical commercial utility of a MIR fiber laser. Recently however, with the growing number of efficient and now moderately powerful MIR fiber sources available, these sources are themselves being used for the generation of alternative sources of cw and pulsed MIR laser radiation, both as optical pumps for ionic transitions [121] and for nonlinear wavelength shifters.

Figure 13 depicts a popular plot [122] of the reported maximum output powers versus wavelength, from past publications on various fiber lasers, and includes updates on more recent reported maximum powers and wavelengths. Over the last few years, there has been a steady increase in the output power demonstrated by MIR fiber lasers emitting in the MIR, in particular for the 2.8 µm [29] and 3.5 µm [123] transitions of Er³⁺, the 3.2 µm transition of Dy³⁺ [124], and the 3.9 µm transition of Ho³⁺ [125]. We note that all demonstrations of these high-power outputs have involved a fluoride glass host. The nearly exponential decline in power with increasing wavelength seen in this figure is due in part to the steady decline in the quantum defect between the preferred diode laser pump source (usually a 975–980 nm NIR diode laser) and the MIR emission wavelength, but is also due to the limited power handling capabilities of fluoride glass fibers. It is almost always preferable to use diode lasers directly as excitation sources for lasers, but the lack of commercial high-power MIR diode lasers, and the consequent need to use high power NIR diode lasers poses a serious problem; not only is the overall conversion efficiency seriously limited, but the large difference between the pump and laser energies (i.e., quantum defect) gets dissipated in the form of high amounts of heat. This is a significant problem for MIR fiber lasers because the materials used necessarily have weaker thermo-mechanical properties—and lower softening and melting points—than materials used for the high power (say 1 µm or 1.5 µm) NIR fiber lasers, which use much lighter metal oxides, viz., the silicates.

 

Fig. 13. Reported maximum output power emitted as a function of the emission wavelength for published cw fiber laser systems. Note that in the calculation of the line of best fit, the result relating to emission at 3.22 µm from the Ho³⁺ ion was not considered.

Download Full Size | PDF

It is thus becoming clear that to create high average power MIR laser light from ionic emitter sources (such as RE³⁺ and TM²⁺ ions), it is better to “spread out” the heat burden due to this quantum defect loss, perhaps in a cascade of stages, so that the final pump wavelength is close to the final laser wavelength. The longer the emitted wavelength required, therefore, the longer the pump wavelength needed. In the case of fiber lasers, this has resulted in the concepts of all-NIR inband pumping for 2 µm-class systems [126] in which Tm³⁺-doped silicate fiber lasers are used to pump Ho³⁺-doped silicate fiber lasers, and all-MIR inband pumping when Er³⁺-doped ZBLAN fiber lasers are used to pump Dy³⁺-doped ZBLAN fibers for 3 µm-class systems [121]. These demonstrations also illustrate the importance of using low dopant concentrations which enable a much more manageable heat load per unit length of fiber.

As seen in Fig. 13, which plots the highest reported output powers emitted from cw fiber lasers as a function of emission wavelengths, the highest-power MIR fiber lasers—and choices of dopant ions (Er³⁺, Ho³⁺, Dy³⁺) and bandwidths—occur at wavelengths near 3 µm. These near-3 µm lasers, first described by the ionic transitions and emission cross sections discussions related to Figs. 6 and 7, are now occasionally labeled collectively as “the 3-µm-class” of fiber lasers; the evolution of the maximum cw output powers over the last three decades from this class of lasers is depicted in Fig. 14. Since the first demonstration of mW power level 2.8 µm emission back in 1988 [127] to the 40-Watt-plus power levels demonstrated in recent years (see Fig. 14, [29]) from the ⁴I­_11/2 → ⁴I_13/2 transition of Er³⁺-doped fluoride fibers (that are now typically diode pumped at 976 nm), near-3 µm emissions have been demonstrated from the slightly longer-wavelength emission from the ⁵I­₆ → ⁵I₇ transition in Ho³⁺-doped fluoride fiber, and the ⁶H­_13/2 → ⁶H_15/2 transition in Dy³⁺-doped fluoride fibers; a significant part of recent research has concentrated on the latter (dysprosium) 3-µm system. Note that the ⁵S­₂ → ⁵F₅ transition of the Ho³⁺ ion, with an emission wavelength of 3.22 µm [128], can also be included in the 3-µm-class of fiber lasers. Unfortunately, the location of this transition “high up in the energy level structure” of Ho³⁺ requires short wavelength excitation, with an inevitable reduction in efficiency due to the large quantum deficit, as discussed above. In addition, since the emission wavelength from Dy³⁺ is relatively close to the 3.22 µm Ho³⁺ emission wavelength, relatively little work has been reported on the further development of the 3.22 µm Ho³⁺ fiber laser. Of course, with the increased availability of low-cost blue emitting GaN diode lasers, and the potential for dual wavelength pumping using NIR sources, more effective methods of pumping this transition are expected in the future. Table 4 lists the characteristics of cw fibre lasers based on the significant MIR transitions of the RE³⁺ ions.

 

Fig. 14. Published cw output power from the “3-µm-class” fiber lasers as a function of time since the first demonstration in 1988 [127].

Download Full Size | PDF

There are a number of trends that one can note from an examination of Fig. 14. Firstly, the output power emitted from fiber lasers employing the Er³⁺ ion have tended to stabilise to approximately the 40 W level. This apparent power limit relates to the ingress of water vapor from the surrounding environment into the tips of the fluoride glass fiber, as mentioned briefly in Section 1.3. The 2.8 µm emission wavelength from the Er³⁺-doped ZBLAN fiber laser overlaps strongly with the O-H stretching vibration that normally causes thermal runaway and ultimately failure of the fiber tip, an effect first observed and analysed in detail in 2012 [27]. End-capping the Er³⁺-doped ZBLAN fiber with undoped AlF₃-based glass fibre [28] and more recently CaF₂ [107] has lessened the problem to some degree, allowing the extension of the maximum power level to this apparent 40 W limit. Further power scaling was prohibited by subsequent failure of the end cap itself prompting a recent investigation [26] into the use of other materials for end capping including non-fluoride materials e.g., silicate glass. With further advances in the end-capping technology, a breakthrough output power of 100 W—a benchmark in terms of impact—is certainly possible from a single fiber laser system.

Table 4. Characteristics of reported highest power cw MIR fiber lasers

View Table | View all tables in this article

On the other hand, MIR fiber lasers based on the 3 µm transition of the Ho³⁺ ion have not reached an apparent saturation in the emitted output power even though they are currently limited by insufficient pump power at the required pump wavelength (i.e., 1150 nm). It is conceivable that future power scaling experiments of Ho³⁺-doped ZBLAN fiber lasers will also be beset with the same problems associated with fibre tip failure, but there are other transitions that offer possibly better opportunities. The Dy³⁺ ion has received much less attention in the literature, but it has the potential to be one of the dominant ions amongst the 3 µm class of fiber lasers. One traditional problem with the Dy³⁺ ion is that its absorption features do not overlap strongly with the light emission from standard commercial high-power diode lasers, making excitation of the Dy³⁺ ion usually more difficult. Recent results involving excitation using a Ti:Sapphire laser [130] however, suggest that high power diode pumping at 803 nm has real potential prompting more work in the future. As observed in Fig. 6(c), many of the energy levels of the Dy³⁺ ion are relatively closely spaced, meaning that Dy³⁺ exhibits many emission lines in the MIR when a (very) low phonon host is used. The 3 µm transition of Dy³⁺ was first operated continuous wave (cw) by doping it into a ZBLAN fiber and pumping it at 1.1 µm using an Yb^3+­­-doped silicate fiber laser [131]; a higher 23% slope efficiency (i.e., 62% of the Stokes limit) was achieved when a higher reflectivity outcoupling mirror was used [132]. Despite these encouraging results, these efforts were hindered by excessive levels of pump ESA, excitation at a wavelength that involved a large quantum defect and a single fiber design. Directly exciting the upper laser level with 2.8 µm from Er³⁺-doped ZBLAN fiber lasers was a breakthrough [121], allowing recently demonstrated slope efficiencies of 91% [133] which is essentially the Stokes limit for the pump wavelength of 2.825 µm.

It has been shown [134] that the rate parameters for energy transfer between the Dy³⁺ ions (in a ZBLAN host) that are excited to the ⁶H­_13/2 level (i.e., the upper laser level for the 3 µm transition) are quite small and that there are no significant concentration quenching problems for the 3 µm transition for Dy³⁺ concentrations up to 4 mol.%. Thus, when moderate doping levels are used, essentially no energy transfer processes are involved. This allows one to design optical fibres with some degree of flexibility i.e., when core is directly pumped, low concentrations are used and when the cladding is pumped, higher concentrations are required. To add to this, the emission wavelengths from Dy³⁺-based MIR fiber lasers are typically between 3.0 µm and 3.2 µm which are further away from the absorption peak of liquid water (i.e., at 2.94 µm) and much further away from water vapor absorption (i.e., at 2.65 µm). Thus, with the combination of robust end caps, multiplexed excitation from a number of Er³⁺-doped ZBLAN lasers and the introduction of all-fiber arrangements, demonstration of a 100 W level Dy³⁺-doped fiber laser is expected in the foreseeable future.

4.3.2. Wavelength tunable MIR fiber lasers and applications

For a number of spectroscopic and sensing applications, a tunable light source is needed. Fiber lasers operating in the MIR offer a reliable and efficient alterative to OPOs particularly if the application utilises only a portion of the spectrum. Primarily as a result of its dominance in the early years of MIR fiber laser research and hence is higher level of maturity, the 2.8 µm transition of the Er³⁺ ion was the first system to reach watt-level output whilst being tunable [135,136]. This was followed in 2010 by power levels reaching 10 W and a tuning range of 110 nm [137]. The 2.9 µm transition of the Ho³⁺ ion gained recognition as an alternative source of tunable emission in this spectral region when 6.5 W was demonstrated over a tuning range of 150 nm [129]. More recently, the 3.5 µm transition of the Er³⁺ ion was used to establish a then record tuning range of 450 nm and tuning up to 3.78 µm [138] but the ground state terminating 3 µm transition of the Dy³⁺ ion once again proved its potential after the widest tuning range of any RE³⁺-based system, fiber or bulk was demonstrated in 2018 [139]. A total tuning range of 573 nm albeit at power levels <200 mW was measured. Dysprosium also offers opportunities for broadly tunable pulsed output with reported pulsewidths ranging from the ps level [140] to the µs level [141], as described in Sections 4.3.3 and 4.3.4 below.

4.3.3. Long pulse (nanosecond to millisecond duration) MIR fiber lasers based on Q-switching and gain switching

MIR fiber lasers have been operated in the pulsed mode via Q-switching, gain switching, and gain modulation by numerous researchers since the late nineties. One of the earliest demonstrations of pulsed operation from a 3-µm class fiber laser [142] involved a rather simple Er³⁺, Pr³⁺-doped ZBLAN fiber laser arrangement in which the Q factor for the cavity was modulated with a rotating shutter. Since the “on-off” time of a rotating shutter is quite slow, the resulting pulses approached time durations in the µs regime, which are fortunately well-suited to a number of medical applications [10–12]. A more traditional approach to Q-switching, however, involves the use of acousto-optic modulators (AOMs), which have a much faster “on-off” time and enable the creation of much shorter pulses i.e., typically in the ns regime. In 2011 [143], large pulse energies and a high average output power of 12 W were demonstrated using an AOM in conjunction with a laser cavity that was sealed-off from the environment in order to avoid water ingress as discussed above. Extrapolation of the Q-switching technique to the 2.9 µm transition of the Ho³⁺ ion [144] was a logical extension and provided pulsed emission at wavelengths much closer to the peak of liquid water absorption, as is highly desirable for the development of practical ablation tools for medicine.

Recently, there has been strong interest in the development of Q-switching MIR fiber lasers using passive, rather than active modulation techniques. Of particular importance is the development of topological or two-dimensional nanomaterials as saturable absorbers, the most important being graphene which was employed in one of the early demonstrations of passive Q-switching of MIR fibre lasers [145]. Notably in this demonstration, the emission wavelength was 2.935 µm which aligns strongly with water absorption and the pulses were of µs-long durations. One of the most important saturable absorber materials involves semiconductors that can be layered to form grating structures for cavity mirrors. These are known as Semiconductor Saturable Absorber Mirrors (SESAM) and one of the first demonstrations involved the 3 µm transition of Ho³⁺ and pulses again of µs-long durations [146]. Interestingly, the 2.9 µm transition of the Ho³⁺ ion can be made to co-lase with the adjacent ground-state-terminating transition, with emission at a wavelength 2.1 µm see Fig. 6(b), thereby forming a “cascade”. In a recent demonstration [147], the SESAM Q-switched 2.9 µm transition forced the 2.1 µm transition to gain switch producing two sets of pulses with an adjustable relative repetition rate dependent on the pump power. Another important saturable absorber material is black phosphorous which is a two-dimensional nanomaterial that is structurally similar to graphene and with a bandgap that is narrow for a monolayer and widens with the number of additional layers that are deposited. Black phosphorous essentially bridges the gap between graphene which does not have a bandgap and the semiconducting transition metal dichalcogenides which have wider bandgaps. The first demonstration of Q-switching a MIR fiber laser using black phosphorous [148] produced an average output power of 485 mW, with pulse energies of 7.7 µJ and pulse widths of 1.18 µs

Rather than modulating the loss, an alternative method for producing pulses from a laser involves modulating the gain by pumping the fiber laser with optical pulses in a so-called “gain-switching” process. One of the first demonstrations involved once again the 2.8 µm transition of the Er³⁺ ion [149]. A key advantage of switching the gain (and not the loss) is that less components are used and some complexity is therefore removed from the MIR laser cavity. This means that more robust “all fiber” arrangements can be more readily created. As a result, average power levels of 10 W and pulse energies up to 80 µJ have been demonstrated [150]. Extending the gain switching approach to other RE³⁺ MIR transitions has been shown for the 3.5 µm transition of Er³⁺ [151,152], the 2.9 µm transition of the Ho³⁺ ion [153] and the 3 µm transition of Dy³⁺ which can also be broadly tunable over a 300 nm emission bandwidth [141,154].

4.3.4. Ultrashort pulse MIR fiber lasers based on mode locking

Because of the large bandwidths of the RE³⁺ transitions in glasses discussed in Section 2.2.1, and the high gains that are readily attainable, fiber lasers have become crucial sources of ultrashort pulses in both the NIR and MIR. These sources are the basis for the creation of on-chip supercontinua [155,156] and tools for spectroscopy and sensing [157–160], nonlinear optics, and ultraprecise modification and machining of materials. Besides the above applications, there are several military applications for high peak power ultracompact ultrashort pulse MIR wavelengths, such as for complex infrared counter-measure scenarios and annulment of incoming hostile weaponry in aerial situations and surveillance and communication devices in land-based and underwater scenarios. In particular, ultrashort pulse MIR laser sources are particularly promising for high peak power and high peak intensity directed energy weapon systems because of their potential for propagation over relatively long distances relative to their NIR counterparts and the well-known wavelength-squared ( λ²) dependence of the critical power limit for atmospheric transmission [161]. Ultrafast MIR optics has traditionally relied on OPOs and amplifiers (OPAs) and ultrashort pulse Ti:Sapphire laser pumps for the generation of short-pulse MIR light. These devices are remarkably useful but several important issues, including the beam quality of the MIR (usually the idler) emission, overall ruggedness, financial cost and the maximum average output power are constraints that limit their potential to meet applications more broadly. There is now a growing need to develop all fiber sources of ultrashort pulses in the MIR that are more powerful, more compact, and reliable.

Two groups [162,163] simultaneously reported the successful demonstration of sub-picosecond ultrashort-pulse mode-locked 3-µm-class fiber lasers using Er³⁺-doped ZBLAN fibers in a ring cavity arrangement. These demonstrations utilised the well-known nonlinear polarization rotation (NPR) technique [164] for mode-locking. Pulse durations of 207 fs [163] and 497 fs [162] were demonstrated with peak powers of 3.5 kW and 6.4 kW, respectively. The key to these demonstrations was the commercial availability of MIR transparent Faraday rotators and waveplates. Translating this technology to industry has been boosted by the incorporation of end caps that assist in maintaining reliability [165]. In contrast with NIR fiber systems, the lack of a full range of off-the-shelf fully fiberized optical components is a problem for fiber-based MIR oscillators. This has meant that all demonstrations have involved a combination of fiber and bulk optics. The presence of bulk optical components in a set up creates air gaps between them, unless the entire laser cavity is evacuated or inserted in a vacuum chamber, which makes the overall laser system quite cumbersome. It is known that atmospheric water vapour absorption within the cavity must be avoided with dry gas purging or a liquid fluorocarbon (which also cools the fiber) [166] because it can lead to fiber tip failure as discussed in Sect. 4.3.1. Since the emission spectrum from the Er³⁺ ion overlaps strongly with a number of water vapour absorption lines, this intracavity absorption could cause issues with the reliability and stability of the output from the ultrafast pulse fiber system. Alternatively, it has been shown [167] that the ⁵I₆ → ⁵I₇ transition of the holmium ion [see Fig. 6(b)] which provides emission at a wavelength which is approximately 100 nm longer (see Fig. 7) can provide pulses as short as 180 fs and record peak powers of 37 kW (i.e., 21 GWcm⁻² at the core-air interface). The emission spectrum from the Ho³⁺-based NPR fiber oscillator is in a region nearly void of strong atmospheric water vapor absorption. Table 5 lists the important parameters relating to ultrashort pulse generation from fibre-based laser oscillators emitting in the MIR. Of particular note is the recent demonstration of 126 fs pulses directly from an Er³⁺-doped ZBLAN NPR oscillator in which the cavity dispersion was controlled using a Martinez stretcher [168].

Table 5. Reported output characteristics of short pulsed MIR fiber sources emitting with pulse durations <1 ps

View Table | View all tables in this article

Despite their excellent performance, NPR-based mode-locked MIR fiber lasers require careful alignment and usually frequent adjustment. To provide a more convenient arrangement, many research groups are developing two-dimensional nanomaterials as saturable absorbers [175] for ultrashort pulse generation that removes the need, at least in principle, for constant cavity re-alignment. These materials have broad absorption bands, low saturation intensities, and ultrafast recovery times. After depositing a multilayer graphene saturable absorber on a gold mirror, which acts as one of the cavity mirrors, Zhu et. al. [176] demonstrated pulse durations of 42 ps and an average output power of 18 mW at 2.8 µm using Er³⁺-doped ZBLAN fiber. Black phosphorus can also be deposited onto a gold mirror and act as a saturable absorber mirror for the creation of short pulses. Using this approach, pulses with pulse duration of 42 ps have been produced from an Er³⁺-doped ZBLAN fiber laser at a higher average power of 613 mW [177]. Black phosphorus can also be deposited using a liquid exfoliation process and, in combination with Ho³⁺, Pr³⁺-co-doped ZBLAN fiber, 8.6 ps pulses have been demonstrated [178]. Many more demonstrations of short MIR pulse emission have been reported using two-dimensional nanomaterials as saturable absorbers [see for example 179,180], but a number of these materials are unstable in an atmospheric environment. Morever, the methods in which the saturable absorber are applied to the substrate do not always result in controlled layering of the material. As a result, the pulse stability is typically quite low meaning that direct pulse measurement in the time domain is difficult. These problems have led to the laser pulse durations being typically inferred from the measured optical spectrum under the assumption that the pulses are Fourier transform limited. Despite these problems, steady progress with regards to the processing of these materials is being made.

4.4. MIR coherent sources based on nonlinear optical interactions

4.4.1. MIR Raman fiber lasers

4.4.1.1. High power and high efficiency broadband tunable MIR Raman fiber lasers

Efficient RFLs and rare-earth-doped tunable fiber lasers and amplifiers have been demonstrated frequently at numerous wavelengths and over relatively broad spectral ranges both at NIR and MIR wavelengths [68–85]. Multiple Stokes or cascaded Raman oscillators were first demonstrated at visible and NIR wavelengths [81,82] to enable generation of very efficient wavelength extension to longer wavelengths and ultrabroadband tunable coherent laser-like sources, albeit with relatively broad bandwidths. Subsequently, because of the absence of high-power pump sources for efficient in-band pumping of 1.55 μm high-power EDFAs (erbium-doped fiber amplifiers) needed for telecom applications, a high-power multi-Watt silica fiber-based cascaded Raman laser source—based on a series of successively longer wavelength nested Raman fiber oscillators facilitated by in-line FBGs—was demonstrated at 1.48 μm for EDFA pump applications [83].

After frequent demonstrations of the use of small core As₂Se₃ and other chalcogenide fibers for efficient Raman amplification by several researchers [84,92,94], a very efficient chalcogenide (As₂Se₃) fiber based fourth-order cascaded Raman laser, pumped by a high power 2 μm Tm³⁺-doped silica fiber laser was demonstrated by Duhant et. al. in 2011 [87], followed by a proposal for the use of such cascaded Raman fiber laser sources to cover the entire spectral range from 2 to 4 µm [95]. Yan et. al. studied transient Raman shifting and soliton self-frequency shifting in microstructured tellurite [93] and ZBLAN [85] fibers, and efficient Raman amplifiers and lasers were demonstrated by O’Donnell et. al. [86] in tellurite and fluorotellurite glass fibers, followed more recently by predictions of very efficient power scaling at MIR wavelengths as long as 5 µm in tellurite glass-based fiber RFLs [181].

In summary, relatively broadly tunable (∼100 nm tuning range) MIR Raman fiber lasers—in the CW and quasi-CW mode—at multi-Watt average power levels at wavelengths up to 5 µm have either been demonstrated or anticipated to be developed relatively easily in fluoride (ZBLAN), tellurite, and chalcogenide fibers, but the output spectra are generally limited to relatively broad linewidths (of the order of a few nm) because of the use of fiber lengths of several tens of meters and the onset of SBS and FWM in such long fibers, even with the use of NLW seed sources and NBW (narrow bandwidth) FBGs [182–184].

4.4.1.2. Narrow linewidth MIR Raman fiber lasers

The achievement of NLWs in most RFL sources are limited in part by the broad Raman gain bandwidths and in part by the extremely high gain and low threshold of SBS which occur readily when NLW radiation is propagating in very long nearly-single mode fibers [185,186], and finally by the onset of readily-phase-matched FWM processes [187–190] in long-fiber based laser cavities.

Narrow-linewidth RFLs, with sub-GHz linewidths, have been proposed and demonstrated at NIR wavelengths in silica and germanosilicate fiber [191–194] devices via the use of appropriate distributed feedback (DFB) structures—notably  π-phase shifted (PPS) FBGs—which enable strong feedback over a very narrow band of wavelengths (enabling effective discrimination against lasing at wavelengths outside this narrow bandwidth)—in relatively short fiber (∼1 m long) lasers, circumventing the onset of SBS and FWM processes.

In principle, one anticipates that it should be possible to extend this PPS-FBG-RFL platform to MIR Raman fiber lasers, but extension of the concept and optimal designs of such NLW RFLs is not straightforward. Behzadi et. al. [90] have quantified the critical parameters that will enable extension of the PPS-DFB RFL source technology platform to MIR wavelengths well beyond 3 µm by using optimized PPS-DFB RFL designs based on advanced MIR glass (tellurite, chalcogenide) fibers, including establishing both the requirements on the mode areas and losses in such fibers as well as the optimal requirements for low-pump threshold single frequency PPS-DFB designs. In particular, these authors showed that with carefully optimized PPS-DFB gratings, single mode sub-MHz linewidths, with >20 dB discrimination of side-order modes should be attainable at wavelengths up to 6.5 µm in such NLW RFLs fabricated with currently available low-loss single mode chalcogenide fiber-based devices; the threshold pump powers—using appropriate wavelength shifted broadband cascaded RFLs as pumps—were estimated to be as low as 50 mW.

Figure 15 shows a specific schematic diagram for a practical three-nested cascaded broadband Raman MIR fiber laser to generate broadband MIR pump wavelengths to generate NLW MIR laser radiation at any wavelength up to 6.5 µm, and Fig. 16 depicts a schematic arrangement or “roadmap” of possible pump and material systems that will enable extension of this scheme—with higher order cascaded RFL pumps—for the generation of wavelengths as long as 9.5 µm.

 

Fig. 15. A proposed schematic arrangement for a mid-IR RFL system based on three (n = 3) nested cascaded broadband RFLs to generate pump broadband pump wavelengths as long as 5 µm, and narrow linewidth laser radiation at any arbitrary wavelength up to 6.5 µm.

Download Full Size | PDF

 

Fig. 16. Ranges of cascaded Raman wavelengths for various Stokes orders in specific mid-IR glass-based RFLs pumped by Er³⁺-doped ZBLAN and Tm³⁺-doped silica fiber lasers. The solid rectangles (or bars) indicate wavelength ranges of efficient Raman conversion, and the hollow rectangles correspond to wavelength ranges that may be limited by transparency limits of specifically chosen glasses (approx. 4.5 µm for most tellurites and 6.5 µm for As₂S₃). The last (6^th) row corresponds to the use of both tellurite and As₂Se₃ based RFLs, with 2 orders of nested tellurite fiber-based RFLs followed by 3 orders of nested cascaded As₂Se₃ RFLs (inside the large green rectangle) to yield pump wavelengths as long as 7 µm to pump NLW As₂Se₃ DFB-RFLs at target NLW wavelengths between 7 and 9.5 µm.

Download Full Size | PDF

4.4.2. Ultrashort pulsed MIR fiber sources based on nonlinear optical interactions

4.4.2.1. Self-frequency shifting MIR fiber sources

Soliton self-frequency shifting (SSFS), which involves a Raman shifting process so that the high frequency components of the pulse Raman shift to low frequencies resulting in a pulse shifting to longer wavelengths as it propagates along the fiber [195], has been utilized to extend the wavelengths of ultrashort generation to longer wavelengths [85,93]. A key parameter is the input pulse peak power as this determines, for a given nonlinear fiber length, the emitted centre wavelength of the pulse. This technique allows more conventional NIR fiber lasers and amplifiers to be used [169] although starting with a MIR oscillator or amplifier has the advantage of a longer initial pump wavelength [170]. One major achievement using the SSFS concept is the recent demonstration of an all oxide all spliced fiber system that emitted 3.08 µm [172].

4.4.2.2. Fiber-based MIR continuum sources

MIR supercontinuum sources in which the nonlinear medium is an optical fiber, offer high-brightness, broadband light in a compact fiber package which is vital to applications in optical coherence tomography, characterisation of optical components, broadband spectroscopy and hyper-spectral imaging. Whilst supercontinuum sources in the MIR have historically lagged behind the progress made in the NIR, in recent years, a dramatic shift in focus to the MIR has occurred. Major advances in the manufacturing of optical fibers made from the fluoride and chalcogenide glass families have played a major role driving down both optical loss and financial cost. Designing the fiber for slightly anomalous dispersion and high nonlinearity is required for each pump wavelength [98]. Usually, the longer the MIR wavelength required from the supercontinuum the longer the pump wavelength needed.

Table 6. Parameters relating to fiber-based supercontinuum generation in the MIR

View Table | View all tables in this article

The typical approach to MIR supercontinuum generation is to employ an OPO or OPA with pulse durations of a few hundred femtoseconds and peak powers ranging from 10 kW to 1 MW as the excitation source; chirped pulse OPA’s [196] can deliver peak powers up to 4 GW. These excitation sources are however complex systems requiring constant re-alignment but the wavelength tunability combined with the high available peak power from these excitation sources has driven the performance of MIR supercontinuum sources involving step-index [197–200], suspended-core [201] and photonic crystal fiber [202]. Table 6 lists the important experimental parameters relating to a range of fiber-based MIR supercontinua.

The introduction of ultrafast NIR fiber laser excitation for the generation of a MIR supercontinuum, in particular Yb³⁺-based fiber lasers that emit in the 1 µm region, Er³⁺-based fiber lasers that emit in the 1.5 µm region and Tm³⁺-based fiber lasers that emit in the 2 µm region, represents a move towards the creation of “all fibre” arrangements. With further amplification of the pulses using fiber amplifiers and the extension of the emission to longer MIR wavelengths using SSFS in fibres, all fiber arrangements can potentially create pulse parameters approaching those produced from OPO’s and OPA’s but with the added bonus that the overall system is less complex, more robust and more efficient. One clear advantage of an all fiber laser excitation source is the potential for higher average power levels. This creates a significant increase in the power spectral density of the supercontinuum that is produced.

In the quest for less complex optical arrangements, it is better to generate the ultrafast MIR pulse directly and use that to excite the supercontinuum. The demonstration of an ultrafast 3-µm class fiber laser in 2014 [207] and the subsequent creation of much shorter pulses [162,163,167] opens up the opportunity to use these sources, with their higher average power levels, for the creation of a MIR supercontinuum. Combining these light sources with SSFS [169] will enable extension of the emitted MIR supercontinuum wavelength deeper into the MIR. The first demonstration involving a MIR ultrafast fiber oscillator for the creation of a supercontinuum was in 2017 when Hudson et. al. [206] combined an ultrafast Ho³⁺, Pr³⁺-doped ZBLAN fiber laser with a tapered chalcogenide microwire that was made robust by coating the taper with a polymer. Nonlinear broadening through SPM, optical wave-breaking, SRS, and FWM resulted in a supercontinuum spanning from 2 µm to 12 µm with the advantage of being close to an all-fiber package.

5. Key future needs and possible fiber-source based solutions

5.1. Introduction

From an applications perspective, there is always a need for higher power and higher pulse energy and future MIR fiber source research will endeavour to increase both but, the need to extend the emission wavelength further into the MIR region is critical. This latter objective is particularly reliant on the development of new materials because, as we have discussed in Sect. 2.3.1, it is the lowering of the maximum phonon energy of the host glass (or crystal) that is essential to the creation and propagation of light in the MIR. Future materials will be required to perform many tasks simultaneously including supporting lower phonon energies and low optical loss whilst also being able to be drawn into optical fiber. The surface damage thresholds will also be required to be sufficiently high so that light can pass through the glass-air interface without catastrophic failure of the fiber tip. In this section we explore potential directions for fiber-based research that aims for the production of high-power light emission much deeper in the MIR.

5.2. Fiber laser arrangements for higher power and higher efficiency

The first 100-W-class fiber laser [208] reported over two decades ago marked the beginning of high-power fiber laser research. This landmark result demonstrated that the ideal arrangement for the emission of high power NIR light involved directly diode pumping Yb³⁺ ion-doped silicate fiber. In this context, the future demonstration of the first 100-W-class MIR fiber laser will set another benchmark in fiber laser research—it will open up a plethora of new applications and broaden MIR research to many more opportunities. In this review, we have showcased the research that has been dedicated to increasing the output power from an Er³⁺-doped ZBLAN oscillator, but as we have seen in Sect. 4.3.1, the emitted output power has saturated and new laser arrangements are needed. We could use for example, an optical configuration well known to NIR optics by combining multiple MIR fiber oscillators together using a MIR transparent beam combiner. Alternatively, multiple Er³⁺ (or Ho³⁺) oscillators could pump a single Dy³⁺-doped fiber oscillator for a more refined emitted wavelength in a region of the MIR less dense with water vapor absorption features. This arrangement would be useful for applications requiring longer free space propagation distances for example. The 91% slope efficiency recently obtained from an inband pumped Dy³⁺-doped ZBLAN fiber laser [133] suggests that using four 30-W Er³⁺-doped ZBLAN pump lasers to pump a Dy³⁺-doped ZBLAN fiber oscillator could produce 100-W-level output at a wavelength just beyond 3 µm. Of course, this demonstration would be heavily reliant on advanced end-capping techniques [26] and the advanced splicing methods discussed in Sect. 4.2.1. Coherently combining the phase-controlled output from a number of MIR fiber oscillators should also be developed.

5.3 Longer and new wavelength MIR sources: needs and possible solutions

5.3.1. Optical fibers based on next-generation solid state host materials

5.3.1.1. Crystalline fibers

Whilst practically all optical fiber is composed of glass, exploiting the low phonon energies of crystals by creating crystalline fiber is attractive; it has the benefit of providing a host that can accept higher RE³⁺ ion concentrations compared to most glasses. One of the first demonstrations of growing crystalline fiber [209] involved single crystal sapphire fiber that was grown using a computer-controlled laser-heated pedestal-growth technique. The fibre displayed a low loss of 0.3 dB/m at 2.94 µm and was able to propagate approximately 5 W of average power emitted from a solid state Er:YAG laser. The recent [210] fabrication of double clad fibre using YAG crystal opens up opportunities for diode pumping directly. The double clad fiber was composed of an Yb³⁺-doped YAG core and undoped YAG was used for the cladding. The fibre presented a minimum optical loss of only 0.24 dB/m. The fiber was fabricated using a combination of laser-heated pedestal-growth and liquid phase epitaxy; a total optical conversion efficiency to a wavelength of 1 µm of 68.7% was demonstrated using this fiber in a laser configuration. The real challenge, however, will be attempts to scale the output power from crystalline fiber-based oscillators. Crystalline fibers are fragile and can only be supplied in limited lengths. To date, there has been no reported demonstration of MIR fiber laser emission from a crystalline fiber.

5.3.1.2. New host glasses

Opportunities to extend the power and wavelength of MIR fiber sources have primarily involved glass optical fiber. As we have seen in Sect 4.3.1, laser emission from experiments involving fluoride (ZBLAN) all glass optical fibre is currently limited to approximately 4 µm (and potentially longer with InF₃-based glass) and chalcogenide-based fibers are yet to create laser emission in the MIR. As a result, there is an urgent need to develop glass-based optical materials that can support appropriate levels of RE³⁺ or TM²⁺ ion doping and be drawn into optical fiber and that can emit and propagate with low loss, light at wavelengths longer than 4 µm.

5.3.1.3. MIR transparent glass ceramics (TGCs)

This Section is excerpted from (a condensed version of) a recent review article [67] on MIR TGCs, co-authored by one of the authors of this publication, and has been added here for the convenience of the readers; please see Reference [67] for additional details on this highly relevant topic.

5.3.1.3.1. Background and relevance

Glass ceramics were first developed in the middle ages as “colored glasses” or “stained glasses” for decorative purposes, notably for the use as stained glass windows, although it is clear that the initial “developers” of such “stained glasses” (such as gold ruby glasses) and their unusual optical properties were aware of the glass ceramic nature of these materials, first proposed and expounded by Stookey [211]. More recent optical applications include their use as “long pass filters” (Corning, Schott) and nonlinear optical materials, when the nature of their ceramic properties were accidentally discovered by experiments on transient gratings related to FWM [212].

As now understood, glass ceramics (GCs) are essentially “multicomposite materials” with semiconductor, metal, oxide, or other crystallites uniformly dispersed or “embedded” in a super-cooled liquid [213] or an “essentially amorphous” solid glass matrix. The crystallization phase is formed during the heat treatment or cooling of the constituents of a specially prepared melt (containing all the desired constituents) at a temperature T_x that is higher than the glass transition temperature that forms the “base” glass matrix. As such, these materials can be tailored in such a way that they are endowed with optical properties that are determined essentially by the nano/microcrystallites (NMCs) while their macroscopic or “bulk” properties, including physical, chemical, and mechanical properties, are determined by the glass matrix. In particular, TGCs are created by using high purity constituents and controlling the fabrication processes and parameters in a way that minimizes spurious absorption and optical scattering losses [214]. Scattering losses are minimized in TGCs by reducing pores and grain boundaries and making sure that the nanocrystals (NCs) are very small compared to the wavelengths of interest, i.e., typically less than 50 nm for MIR TGCs.

Many functional ceramics, mainly those based on silicates were developed at the turn of the 21^st century [215,216], both to improve the mechanical properties—such as hardness and fracture toughness—of these “glasslike” materials, as well as to develop new optical, electrically, and magnetically active materials based on the NMC constituents of these glass ceramics. In contrast with the silicate glass ceramics, the glassy matrix of MIR glass ceramics is based on “soft glass” constituents such as those used to make low phonon energy tellurites and chalcogenides and other low phonon energy constituents such as alkali halides [217]. Functionalization of such glasses for efficient MIR luminescent materials was then achieved by developing methods for incorporating appropriate RE³⁺ and TM³⁺ MIR emitting ion dopants (which are subsequently optically excited by appropriate pump wavelengths) into optimal crystal sites in the MIR-transparent nanocrystallites of the desired composition (usually wide bandgap semiconductors) [218,219]. The challenge in the last step lies in the extreme difficulty of developing the best materials processing techniques that enable formation of stable high quality TGCs while doping ions of the desired valence at appropriate sites of the nanocrystallites (such as +2 valence transition metal ions such as Cr²⁺ in tetrahedral sites of II-VI zinc-blende structure nanocrystallites) with the desired dopant densities.

As stated above, fabrication of the desired luminescent low phonon energy-based glass ceramics is a difficult task with no a priori method of fabricating the desired TGCs, and processes are currently developed empirically and laboriously by multiple cycles of trial and error. As such, there is a strong need for controlling the design and fabrication process via theoretical methods based on understanding the crystal nucleation and growth method [213] via studies of molecular dynamics and detailed understanding of nanoscale heterogeneous structure models of glasses [220]. The theoretical basis for predictive development of processing recipes based on theoretical methods is still very much in its infancy. As such, significant advances in the development of the best MIR luminescent glasses—including those suitable for MIR lasers based on materials such as Fe²⁺:ZnSe nanocrystallites in appropriate glass matrices, which appear to be very promising for the next-generation MIR fiber lasers and, will most likely be developed using novel glass compositions (such as those reported recently by [221]) by empirical methods.

Table 7 shows some of the most promising MIR-relevant TGCs with low phonon energy crystallites, along with the degree of crystallinity and method of fabrication of these TGCs [67,222]. Note that the phonon energies of the NCs (column 2 in Table 7) in these TGCs provide the low phonon energy crystalline environment and appropriate lattice sites for potential ionic emitters.

Table 7. Summary of the reported crystallinities of TGCs. The crystallinity is defined as the volume fraction of the embedded crystals in TGCs. CC: conventional in-situ crystallization

View Table | View all tables in this article

5.3.1.3.2. Rare-earth doped chalcogenide GCs

Due to the heavy masses of the constituent atoms, the vibration frequencies of metal-chalcogen (S, Se) bonds are very low, i.e., chalcogenide glasses possess extremely low phonon energies (typically in the range of 150-450 cm⁻¹). As stated in Section 1, low phonon energy hosts are critical for MIR optical transitions from dopant ions (RE³⁺, TM²⁺, or other), particularly at transitions corresponding to wavelengths longer than 4 µm, where the corresponding luminescence is strongly quenched in oxide and fluoride hosts due to fast multiphonon relaxation [229]. Thus, most observations of MIR luminescence at wavelengths longer than 3.5 µm have been reported primarily in chalcogenide crystals or glasses [20]. As listed in Table 8, the MIR optical transitions of the following RE³⁺ ions: Dy³⁺, Er³⁺, Pr³⁺, Tm³⁺, Tb³⁺, and Ho³⁺ can cover the entire spectral range between 3 and 5 µm (although laser inversion is much more difficult at the longer wavelengths, such as the ⁴I_9/2 to ⁴I_11/2 transition of Er³⁺ [20].

Table 8. Representative MIR luminescence of rare-earth ions in chalcogenide glasses. σ_em= emission cross-section (×10⁻²⁰ cm²)

View Table | View all tables in this article

Transparent chalcogenide GCs have been achieved via precipitation of appropriate crystalline phases, and widely studied because of their promising properties of increased mechanical strength, permanent IR frequency-doubling, and enhanced luminescence—with appropriate rare-earth dopants—in the MIR region [235]. Enhanced luminescence of RE³⁺ in chalcogenide GCs was firstly observed in Nd³⁺ doped 70GeS₂·8Ga₂S₃·12Sb₂S₃·10CsCl nanocrystallized samples [236] In addition, with the precipitation of Ga₂S₃ NCs, a large enhancement (∼20x) of the upconversion luminescence was achieved in Er³⁺ doped 70GeS₂·20Ga₂S₃·10CsCl GCs [237], resulting in several studies of the enhanced luminescence of RE³⁺ in chalcogenide GCs during the past decade [238–241].

In 2011, Dai, et. al., reported the first observation of enhanced MIR luminescence in 65GeS₂·25Ga₂S₃·10CsI chalcogenide GCs doped with 0.6 wt% Tm³⁺ ions [242]. The presence of ∼50 nm Ga₂S₃ nanocrystals promoted a two-fold increase in 3.8 µm MIR emission, which originates from the optical transition of ³H₅ → ³F₄ of Tm³⁺ ions. The enhancement of this optical transition was further improved to more than 5-fold in well-crystallized 80GeS₂·20Ga₂S₃ GCs [243,244]. In the glass-ceramic sample with a similar composition of Ge_28.125Ga_6.25S_62.625, emissions at 2.9 µm and 3.5 µm originating from appropriate optical transitions of Dy³⁺ ions (see Fig. 4), were also enhanced by ∼12 times because of the precipitation of GeGaS nanocrystals [219].

In recent years, Ge-Ga-S GCs have been investigated quite thoroughly to study and clarify the fundamental mechanisms leading to enhanced MIR emissions [219,243,244]. It is worth mentioning that the Ga-based chalcogenide glasses were selected specifically because Ga not only favors the solubility of RE³⁺ [245], but also provides essential structural units related to control of the crystallization of Ga₂S₃ [216]. In particular, in Ge-Ga-S glasses, the Ga-related units (i.e., [GaS₄] tetrahedral and ethane-like [S₃Ga-GaS₃]) play a vital role—due to their structural similarity—for the formation of isochemical Ga₂S₃ nuclei, which grow subsequently under further thermal treatments.

5.3.1.3.3. Transition metal ion - doped chalcogenide GCs

Divalent transition metal ions (TM²⁺, such as Cr²⁺, Co²⁺, Fe²⁺) doped II-VI chalcogenide (such as ZnS, ZnSe etc.) semiconductors have emerged as excellent MIR gain media, due to their broad MIR emission bands (∼0.5λ₀, λ₀ is the emission peak wavelength), large stimulated emission cross-sections (∼10⁻¹⁸ cm²), high quantum efficiencies (>70%) and the ready “pumpability” of the upper laser levels by commercially available cost-effective laser sources (including diode lasers, and diode-pumped Er³⁺ and Tm³⁺ doped high power fiber lasers [246]). As such, TM²⁺: II-VI crystals now appear to rival the “gold standard” Ti³⁺-doped sapphire crystals (Ti³⁺: Al₂O₃) used for solid state lasers at NIR wavelengths.

In 2010, R. Mironov et. al. produced As₂S₃ glass ceramic fibers embedded with Cr²⁺: ZnSe or Cr²⁺: ZnS crystals for the first time [247]. The fiber preforms were based on chalcogenides fabricated by the glass powder (GPD) method. The fibers contained crystals of about 1 µm in size [248], and emitted broadband luminescence in the 1.8-3 µm (centered at 1.9 µm) with a 2 to 4 dB/m optical loss (between 1.5 µm and 2.7 µm), but were still far from satisfactory due to significant scattering losses (∼4 dB/m) [248]. On the flip side, strong interparticle scattering caused by microcrystals has been used favorably to demonstrate room temperature (RT) random lasing at 2.4 µm in Cr²⁺: ZnSe/As₂S₃: As₂Se₃ micro-composites [249]. The use of hot-pressing methods has also been suggested [250] to prepare II-VI crystal/chalcogenide glass composites, and emission in the 1.8 - 2.8 µm spectral range has been observed in Cr²⁺:ZnSe/As₄₀S₅₇Se₃ glass composites. However, due to the presence of strong scattering, these materials are very lossy, and an optical transmission of less than 40% was observed in the 2-10 µm wavelength region in a 2 mm thick sample. Emission in the 1.9 - 3.1 µm emission range has also been reported in Cr²⁺: ZnSe/95%As₂S₃:5%As₂Se₃ composites prepared by a uniaxial hot-pressing method [251]. These results demonstrate the feasibility of making chalcogenide-based composites with reduced scattering loss by matching the refractive indices between the embedded II-VI crystals and the chalcogenide glasses.

As discussed in Sect. 2.3.1, in contrast to RE³⁺ ions, the electronic transitions of TM²⁺ ions are extremely sensitive to local crystal-field environments [252]. Recently, tunable broadband MIR (2.5 to 4.5 µm) emissions of Co²⁺ in chalcogenide GCs containing a variety of II-VI NCs (such as ZnS, ZnSe, ZnSSe and ZnCdS) was reported [253]. Cost-effective and commercially available Er³⁺-doped fiber lasers can be used as the excitation source for Co²⁺. By crystal-field engineering of the embedded NCs through cation- (Zn²⁺ ↔ Cd²⁺) or anion-substitution (S^2- ↔ Se^2-) (Fig. 17(a)), the emission properties of Co²⁺ including its emission peak wavelength and bandwidth can be tailored in a broad spectral range (Fig. 17(b)) [253].

 

Fig. 17. (a) Scheme of crystal field engineering of TM²⁺ (e.g., Co²⁺) emission via cation or anion substitution. CB: conduction band; VB: valence band. The indicated energy (in eV) correlates with the emission peak wavelength of Co²⁺ in the corresponding crystals. (b) Emission spectra of Co²⁺-doped chalcogenide GCs containing different II-VI NCs [253].

Download Full Size | PDF

For Fe²⁺-doped crystals, the desired pump wavelengths are around 3.0 µm and necessitate the use of Er³⁺ or Cr²⁺ doped crystalline (e.g., YSGG: Er³⁺, Cr²⁺: ZnS) solid state lasers, or Er³⁺-doped ZBLAN fiber lasers [254]. The large spectral overlap of 2.5 - 4.5 µm emission band of Co²⁺ and 2.0 - 5.0 µm absorption band of Fe²⁺, suggests the use of Co²⁺ as a sensitizer for Fe²⁺, whose MIR emission, peaked at 3.5 µm, and temperature dependence, is shown in Fig. 18(c). Based on the efficient Co²⁺ → Fe² ET [Fig. 18(a)], Lu et. al. [255] demonstrated broadband MIR (2.5 - 5.5 µm) emission in Co²⁺-Fe²⁺ codoped chalcogenide GCs pumped by a commercially available erbium doped fiber amplifier emitting at 1.57 µm [Fig. 18(b)]. These authors also described the first observation of a unique “anomalous” increase in the MIR luminescence intensity as a function of temperature [Fig. 18(d)]. In addition, gas sensing was also demonstrated for multiple target analytes such as butane and carbon dioxide using the developed chalcogenide GCs.

 

Fig. 18. (a) Schematic of ET between Co²⁺ and Fe²⁺. (b) RT MIR emission spectra of Co²⁺-singly doped (multiply by 1/5), Fe²⁺-singly doped, and Co²⁺/Fe²⁺ codoped chalcogenide GCs under the 1570 nm excitation. Temperature dependence of MIR emission spectra of the Co²⁺-singly doped (c), and (d) the Co²⁺/Fe²⁺ codoped samples [255].

Download Full Size | PDF

In summary, chalcogenide GCs and fibers embedded with TM²⁺ doped II-VI crystals can be obtained by in-situ crystallization and GPD methods. Hot pressing has been suggested to make chalcogenide GCs, which has the potential of mitigating glass dissolution of NCs (also referred to as “melt corrosion” by some researchers). Thanks to the large range of compositions accessible to practical chalcogenide glasses, it is possible to match the refractive indices between the II-VI crystals and the chalcogenide glasses. As such, GCs and fibers with low scattering losses can be expected even for particle sizes that are of micrometer scales [248]. Because of their sensitivity to the crystal field environment, MIR emissions of TM²⁺ can be tuned over a very broad spectral range by cation-substitution (Zn²⁺ ↔ Cd²⁺) or anion-substitution (S^2- ↔ Se^2-) [255]. Broad MIR (1.9 - 5.5 µm) emissions of TM²⁺ have been observed in TM²⁺: II-VI/ chalcogenide GCs and fibers at room temperature, offering a platform of gas sensing for multiple target analytes. However, only lasing in random mode has been realized so far [249]. Future studies should focus on exploring new chalcogenide GCs with a much larger crystallinity—and ultrahigh dopant densities—for enhanced MIR emissions of TM²⁺, as needed for efficient lasing in these materials.

As such, ion-doped functional NMCs in TGCs can meet requirements that are impossible to fulfill by their glass and crystal counterparts. This is particularly true for MIR luminescence, which is significantly quenched in high phonon energy glasses. The crystalline nature of the sites is critical for TM²⁺ ions, because of the sensitivity of these ions to local field effects and the need for sites of tetrahedral coordination. For photonic applications, GCs with low optical attenuations—i.e., high optical transparency—are critical. This need for high optical transparency requires that: 1) the particle size of the embedded crystallites be kept as small as possible; and 2) the refractive index of the embedded crystal be very well matched (ideally with a difference of less than 0.001) to that of the surrounding glass matrix. However, the difficulty in molecular dynamics computational methods and the lack of fundamental knowledge to predict which types of crystals can be grown in GCs by in-situ crystallization precludes the possibility of GCs materials by design. On the other hand, because the refractive index is generally known for the crystal that is to be synthesized before its incorporation into the glass, it is possible to match the refractive indices by tuning the glass composition using the GPD method. However, to avoid the corrosion or dissolution of the incorporated crystals in the glass host matrices, and for ease of fabrication—and high MIR transparency—soft glasses with low melting temperatures are preferred. Enhanced MIR emissions of RE³⁺ and TM²⁺ have been observed in various TGCs, and theoretical simulations also point to their feasibility for future MIR lasers.

5.3.2. Alternative ionic transitions

5.3.2.1. Ho³⁺ in InF₃-based glass fiber

The pursuit for longer emission wavelengths from MIR fibre sources is driven by the need to develop robust applications deeper in the MIR, e.g., towards the molecular fingerprint region where a significant number of sensing and security applications would benefit. To date, the longest emission wavelength from a MIR fiber source using the electronic transition of a RE³⁺ ion is based on the ⁵I₅ → ⁵I₆ transition of the Ho³⁺ ion, see Fig. 4(a). The first experiment involving this transition used a ZBLAN host [56] however it was necessary in this experiment to cryogenically cool the fiber because a small number of phonons could bridge the energy gap between these energy levels and this resulted in high rates of nonradiative decay. Cooling the fiber reduced the occupancy of the higher energy phonon states consequently reducing the rate of nonradiative decay (thus more phonons would be required to bridge the gap). Irrespective of the cooling, the lifetime of the lower laser level is still longer than the upper level lifetime for this transition and evidence of self-saturation was observed.

Recently [125] a breakthrough demonstration was reported in which the output power and slope efficiency from this transition were significantly increased, see Table 4. This advance was due to a number of important factors. Firstly, an InF₃-based fluoride glass fibre was used which has a (slightly) lower maximum phonon energy compared to ZBLAN meaning that the rate of nonradiative decay from the ⁵I₅ upper level was lower compared to ZBLAN. Secondly, the incorporation of a high Ho³⁺ concentration (in this case 10 mol.%) meant that the ETU process ⁵I₆, ⁵I₆ → ⁵I₈, ⁵F₅ [256] could be exploited and recycle the energy away from the lower laser level back up to the upper laser level. Thirdly, a pump ESA process [257], see Fig. 19 alleviated potential population bottlenecking at the lower states which have much longer lifetimes. This latter process helps mitigate the problem of possible bleaching of the ground state absorption as a result of large population densities residing in these lower energy levels above the ground state. These results are promising but recent spectroscopic measurements on Ho³⁺-doped ZBYA (ZrF₄-BaF₂-YF₃-AlF₃) glass [258] suggest that ZrF₄-based glasses may still be useful for extending the output power emitted from this transition; the measured lifetime of 330 µs is very encouraging but drawing this glass into low loss optical fibre may still prove challenging. Future power scaling experiments of this transition will be assisted by the wide availability of high-power diode pumps that can pump the upper state directly and optical amplifier development (for quantum cascade lasers for example). The demonstration of ultrafast fibre lasers of the “4-µm-class” using this transition are also on the horizon.

 

Fig. 19. (a) Schematic diagram of the energy level diagram of the Ho³⁺ ion showing the pump, excited state absorption (ESA), energy transfer upconversion (ETU) and laser processes [125].

Download Full Size | PDF

5.3.2.2. New rare earth transitions

Figure 20 shows the ground state terminating transition of the Tb³⁺ ion which, like many of the other rare earth ions, is ideally suited to a gain medium in the form of an optical fiber. Like Dy³⁺, there exists a strong overlap between the ground state emission and absorption cross section spectra for this transition. This forces the need for a large population inversion which for bulk solid-state systems is more difficult. The recent demonstration [259] of laser emission from this transition using bulk chalcogenide (in this case a selenide) glass required pulsed pumping at 2.94 µm from an Er:YAG laser as a result of the strong overlap between the emission and absorption spectra and the consequent need for a large population inversion. The ⁷F₄ level has a peak absorption at a wavelength that can be addressed using the highly mature Er³⁺-doped ZBLAN fiber laser. If fluoride glass (in which all levels are non-radiatively quenched) is replaced with chalcogenide glass or TGC fiber, the ⁷F₄ level would have a relatively short lifetime of <10 µs whilst the ⁷F₅ level a much longer lifetime [260] resulting in the potential for fiber laser emission at 4.7 µm. With cavity mirrors and a fiber length of Tb³⁺-doped chalcogenide fiber optimised from spectroscopic measurements, a linear cavity could be constructed to resonate the ⁷F₅ → ⁷F₆ transition, for the demonstration of the first fiber laser of the “5 µm class”. However, there are still concerns regarding the actual nonradiative relaxation rates for this RE³⁺-glass system based on the theory of Layne, Lowdermilk and Weber [58], as elaborated in Section 2.3.1 above.

 

Fig. 20. Terbium energy levels, showing the pump and laser transition for a potential 5-µm-class fiber laser.

Download Full Size | PDF

There are number of additional transitions that are worth discussing. The ⁶H_11/2 → ⁶H_13/2 transition of Dy³⁺ with an emission wavelength of approximately 4.3 µm, see Fig. 6(c), has attracted a lot of interest from the bulk solid-state laser community [53], but from a fibre perspective, theoretical modelling studies and basic luminescence measurements have dominated the literature. Extracting efficient laser emission from this transition will be difficult because of its reliance on very low phonon energy glasses but, with much progress already obtained using chalcogenide glass fibres [261] particularly those based on the selenides, we anticipate that MIR fibre laser emission from this transition is certainly on the horizon too. Note that using a fluoride glass fibre may also result in fibre laser emission from this transition however, despite fluorescence having already been measured [262], it is becoming increasingly clear that high intensity pulsed pumping will be required when a fluoride host is used. The Pr³⁺ ion is also worth mentioning. A detailed spectroscopic analysis [263] has suggested that emission on the ¹G₄  → ⁴F₃ transition centred at 3.6 µm may also lase using a fluoride host, however, as yet, the highest possibility for MIR fibre laser emission from the transitions of Pr³⁺ relates to the use of a chalcogenide glass host where mW-level superluminescence has already been observed [264].

5.3.3. Ionic transitions and Raman shifters in gas-filled hollow core fibers

5.3.3.1. Fiber lasers based on inversion in ionic transitions in gases

As discussed above, the current limit of fiber laser emission from RE³⁺ doped glass is currently 3.9 µm, a “limiting value” that was achieved decades ago [59]. While there is significant promise that appropriate chalcogenide based TGC options (as discussed in Section 5.3.1.3. above) may result in ionic transition-based fiber laser emission wavelengths longer than 4 µm, other options, namely gas-filled hollow-core fibers also appear promising for fiber-based laser systems [265]. Gases have high damage thresholds, relatively large emission cross-sections and a wide range of emission wavelengths from the NIR to LIR. Loss measurements [4] of hollow-core fibers made from silicate glass reveal attenuation coefficients at a pump wavelength of 1.53 µm and MIR lasing wavelengths >3 µm as low as 0.11 dB/m and 0.10 dB/m, respectively. These loss values are more than 10,000 times lower than that of bulk silicate glass and are key enabling parameters opening many opportunities. Table 9 lists a number of important recent demonstrations of fiber lasers that employ low loss hollow-core silicate fiber.

Table 9. Representative list of MIR lasers based on gas-filled hollow core silica fiber lasers

View Table | View all tables in this article

Hollow-core silicate fibers are versatile in that any gas can be filled into the core whilst maintaining the strength and durability of silicate glass. The confinement of the light over a long distance is therefore possible (typically 10 m) allowing low gas pressures to be used. This is an important consideration because it has been shown [267] that at high gas pressures, the gain per mbar gas pressure is reduced. Additional requirements include the use of narrow linewidth pump sources so that enough spectral overlap with the narrow rotational-vibrational absorption lines of the gas is made and the fiber must be enclosed at each end using a gas cell that also provides optical access through a window. The gas cells and hollow-core fiber are evacuated and then filled with a low pressure (typically <1 mbar) of gas. With highly resonant feedback, cw emission with a relatively low pump power requirement at threshold is possible. Note that longer wavelength emission is expected in the future when low loss hollow core fibers based on negative curvature anti-resonant waveguide designs using soft glasses are developed.

5.3.3.2. Fiber lasers and amplifiers based on SRS in gas-filled hollow-core fibers

Long-wavelength lasers —particularly MIR ultrashort lasers at wavelengths near 10 µm — are a critical need, for many applications. Starting with amplified high-power femtosecond NIR fiber lasers, or amplified MIR short pulse lasers, SRS is an efficient modality for achieving such longer MIR sources via wavelength or frequency conversion in gas-filled hollow-core fibers. There are several candidates for the NIR or MIR “starter” short pulse sources that may be used. The choice of the gas medium to be used and the exact pump frequency or wavelength of the initial ultrashort pulse pump may be determined in part by the simple relationship:  ν_s= ν_p - nΔν_R, where ν_p is the NIR or MIR pump frequency, ν_s is the MIR signal frequency, Δν_R is the Raman shift (usually expressed in wavenumbers), and n is a simple integer (preferably of low order).

In order to get ultrahigh final peak powers at the MIR wavelength, it is highly desirable that the final Stokes-shifted wavelength ν_s matches an appropriate CO₂ laser gain wavelength, that in turn is optimized for “final stage” high peak power short pulse amplification; the latter condition—along with the desirability of using an atmospheric transmission window—limits the choice of the final MIR wavelengths to a wavelength in the 9.2 and 10.6 µm band of the CO₂ gain spectrum. As such, the actual choice of initial SRS pump wavelengths is determined in part by the Raman gain medium that can be used for efficient wavelength conversion, and in part by the availability of an efficient short pulse pump source at the appropriate wavelength or frequency. For this application, the choice for the Raman pump laser is based on a nearly conventional “baseline” mode-locked fiber laser (as an initial “seed” fs or ps laser source) followed by a suitably matched high gain multistage fiber amplifier system.

Figure 21 below depicts a 2.9-µm based short pulse seed source, although any mode-locked fiber laser sources at various wavelengths ranging from 1 to 3.8 µm may alternatively be used as the starting source, provided there is an efficient gas-based Raman shifter that helps achieve the precise MIR wavelength needed for high-gain MIR amplification in appropriate final-stage transversely-excited CO₂ waveguide gas amplifiers. The starting short pulse source depicted in the illustrative example of Fig. 21, namely a mode-locked Ho³⁺-doped ZBLAN fiber laser, has been recently demonstrated, see Sect. 4.3.3. Some promising choices of gases to be used for gain media for such an application are shown in Table 10 below. The choice of Raman gain media – namely high-pressure hydrogen and nitrogen gases —are dictated partly by the large Stokes shifts, and high Raman gains and saturation intensities achievable [270]. Note that additional flexibility can be achieved with the use of multiple Stokes cascaded SRS methods or multiple Raman gain stages. As seen in this table (Table 10), a high power pump source at 3.8 µm is an extremely desirable option not only because it is a highly desirable MIR wavelength for directed energy weapons (DEW) applications, but also because this wavelength is itself located in a near ideal atmospheric transmission window [271], while also enabling efficient generation of a short pulse “signal source” at a highly desirable MIR wavelength (9.3 µm) for amplification in a CO₂ amplifier and for atmospheric propagation studies in a highly transmissive spectral MIR region [271].

 

Fig. 21. Schematic diagram of an ultrashort pulse MIR fiber laser arrangement based on nonlinear polarisation rotation. OC refers to the output coupler. HT and HR refer to high transmission and high reflectivity, respectively [167].

Download Full Size | PDF

Table 10. Some promising choices of pump and signal wavelengths, and Raman gain media for 10 µm fiber-based sources

View Table | View all tables in this article

Very high efficiency multistage or cascaded Raman shifting, limited only by the quantum defect between input and final wavelengths can be achieved with the use of: (i) high-pressure gas (nitrogen/hydrogen/ deuterium/carbon dioxide)-filled hollow-core fibers [272]. Low loss waveguides at MIR wavelengths can be achieved with the use of low-loss MIR glasses and anti-resonant/negative curvature guiding structures [273], (ii) chalcogenide glass photonic crystal fibers [274], or (iii) hollow core silver coated silica fibers [275]. Figure 22 below shows a confocal microscope image —and a schematic diagram—of a negative curvature chalcogenide fiber [276], which exhibited a loss of 2 dB/m at 10 µm, and which should yield conversion efficiencies of over 20% at GW/cm² peak input powers for many of the Raman conversion options indicated in Table 10 above. Note that the conversion efficiencies of such fiber Raman shifters are limited only by the quantum defect and the insertion loss of the hollow core fibers to be used, and conversion efficiencies of over 20% are anticipated with the use of MIR hollow core fibers with insertion losses of 0.5 dB/m [277] at the target wavelengths.

 

Fig. 22. Schematic of a negative curvature fiber designed and optimized for MIR gas laser and MIR Raman shifters with ultralow losses at wavelengths between 2 and 12 µm [276].

Download Full Size | PDF

5.3.4. Some specific MIR short pulse needs and MIR short pulse system design options

In recent years applications involving high harmonic generation (HHG) have driven a strong need for short pulses in the MIR. The ponderomotive force is the force experienced by an electron in an anharmonic potential when subject to a sinusoidal electric field, e.g., light. This force is central to energising the electrons that constitute molecular bonds [278] and driving, with single cycle pulses, the electrons of noble gas atoms for HHG [279]. The amount of energy transferred to an electron in an anharmonic potential varies inversely with the frequency of light. Thus, for a given electric field amplitude (i.e., intensity), electrons experience larger forces and receive much more energy when the wavelength of the electric field is longer. For example, it was recently shown by [280] that the energy deposited from 110 fs pulses at 2.2 µm is equivalent to the energy from single cycle pulses at 800 nm.

In addition, there are several basic science applications—and defense mission needs—for high peak power ultracompact short pulse MIR and LIR wavelengths, such as for complex infrared counter-measure scenarios and annulment of incoming hostile weaponry and surveillance and communication devices in above-ground and underwater scenarios. In particular, ultrashort pulse MIR and LIR laser sources are particularly promising for high peak power and high peak intensity DEW systems because of their potential for propagation over relatively long distances relative to their NIR counterparts because of the well-known wavelength-squared (λ²) dependence of the critical power limit for atmospheric transmission.

Basic science applications include a strong need to understand numerous laser-matter interactions in the LIR, including novel propagation phenomena and controlled filamentation in several gaseous, liquid, and solid media—including the atmosphere—particularly because of the strong reduction in the ionization effect and the strong increase (λ²) in the ponderomotive forces at the longer wavelengths, and unusual effects such as shock wave generation, enhanced electron tunneling, and novel x-ray generation options with ultrashort pulses with long wavelength short pulse sources. Short pulse MIR sources are also of strong interest in the study of strong field physics and for the generation of compact sources of energetic electron and ion beams that might complement conventional particle accelerators and accelerator-based radiation sources, particularly since short pulse MIR sources enable a favorable scaling of pondermotive energy and critical density. MIR filaments with propagation lengths of approximately 200 meters to several kilometers have been demonstrated [281] and the creation of filament lengths of several hundred meters with reduced ionization losses have also recently been proposed [282]. Peñano and co-workers have recently suggested the possibility of long-range nonlinear atmospheric propagation through strong turbulence with the use of the phenomenon of nonlinear self-channeling [283], in which highly collimated propagation over many Rayleigh lengths can be achieved when the laser power is close to the self-focusing power of air and the transverse dimensions of the pulse are smaller than the coherence diameter of turbulence. Because the coherence diameter and laser spot size are centimeters in scale, and because there are no ionization losses, the “channeling effect” can persist for several kilometers through strong turbulence, allowing for an unusual new prospect for the use of such short pulse MIR sources for DEW applications.

Despite the fact that petawatt peak power levels have been achieved by many researchers in different laboratories around the world for several years now with NIR laser sources based on mode locked solid state lasers, and have been used in applications ranging from nuclear energy sources to compact particle accelerators and radiation sources, the development of multi-TW ultrafast MIR sources is still in a relative infancy, with peak power levels ranging from a few TW to 30 TW or so, depending on the pulsewidth and chosen wavelengths. The highest powers to date have been achieved via the amplification of femtosecond to picosecond pulses, with some pulsewidth degradation, in high pressure CO₂ amplifiers excited by UV pre-ionized transverse electric discharges [284].

Nevertheless, despite the above-cited achievements in output powers, the overall sizes of the lasers demonstrated by previous researchers are extremely large, often occupying complete labs of relatively large size, and as such there is still a strong need for a relatively compact “table-top” solid state laser based LIR source at power levels approaching 1 TW. There is a critical need for fiber laser type solutions, based on short pulse MIR fiber lasers, optical fiber amplifiers Raman shifters, mainly in gas media similar to those discussed above, to demonstrate a compact TW-level short pulse MIR source at wavelengths in the vicinity of 10 µm, at specific wavelengths corresponding to the best transmission windows in the atmosphere. One such schematic design of a potential compact 10 µm TW laser source, based on MIR fiber laser and fiber amplifier components is illustrated in Fig. 23 below.

 

Fig. 23. Schematic “block diagram” illustrating the “basic” design of an ultracompact high peak power fiber-based near-10 µm MIR source.

Download Full Size | PDF

As seen in Table 10, the Stokes shift in N₂ is 2331 cm⁻¹ [270], and a single Stokes shift will convert the 3 µm radiation to 10.6 µm radiation, within the gain bandwidth of previously demonstrated electrically-driven CO₂ TE discharge amplifiers that may be used for post-amplification in the final stages of such a short pulse MIR system [285]. As an alternative example, if one starts with readily achievable MW-level peak powers from a short 1 µm pulses from an amplified Yb³⁺ fiber laser-amplifier system, the use of H₂ (Stokes shift ∼ 4156 cm⁻¹) as the SRS gain medium will yield MIR radiation at 10.6 µm via just 2 Stokes shifts (which can be obtained via a single stage or 2-stage Raman amplifier/gain medium).

6. Closing thoughts

The MIR region of the electromagnetic spectrum is of growing interest to many sectors in science, medicine, defense and industry. The development of advanced coherent MIR sources that are compact, powerful, wavelength agile, and cost effective is therefore required in response to this interest. There are also a growing number of exciting applications in nonlinear optics and materials modification that require ultrashort MIR pulses delivered in a beam of high spatial quality and from a system of high reliability. The optical fiber geometry is a favored arrangement for the NIR that meets these needs but translation of NIR fiber technology to the MIR is not a straightforward process. As we have showcased in this review, a whole range of new innovations are being tested and an enormous number of new materials are being fabricated in order to fully release the opportunities contained in the MIR. Perhaps the most important parameter for the development of coherent sources for the MIR—relative to the visible and the particularly the NIR regions—is the material that makes up the glass fiber. The development of new low phonon energy materials is vital. In this regard, we have explored exciting options based on MIR luminescent TGCs, or more explicitly MIR luminescent ion-doped NC-embedded glasses.

Laser physics is experiencing a revolution due to the enormous opportunities that are being provided by fiber-based sources of NIR light; the creation of the anticipated fiber sources of MIR light will inevitably impact many more applications than can be easily anticipated at the present. Launching light into optical waveguides designed for “on chip” sources of broadband MIR emission for all-optical information processing and frequency comb generation is highly dependent on the temporal and spatial qualities of the optical pulses used, and MIR fiber sources appear ideal for many of these applications. New experiments in the modification of small bandgap materials based on nonlinear (i.e., multiphoton) or linear absorption are also heavily reliant on high beam quality ultrashort MIR pulses of moderate pulse energy, and are expected to benefit from fiber-based sources of MIR light in the foreseeable future. In fact, many applications involving the modification of biological tissues, polymers and other industrial materials remain largely unexplored because of the lack of efficient, conveniently compact and powerful sources of ultrafast-pulsed MIR light. Single mode optical fiber-based MIR sources are a natural choice to meet this requirement because the beam quality can be pre-defined and made extremely high by appropriate design and fabrication of optimized fiber cores in future low-loss fibers of the types either used currently (Section 4), or those based on the advanced materials expected in the future, as discussed in Section 5 of this review.

Another critical problem is still the lack of broad commercial availability of fiber components designed specifically for the MIR e.g., couplers and beam combiners, although recent work [286,107] is generating encouraging results. This problem has historically deterred many researchers and applications engineers away from choosing MIR fiber componentry as their first choice, but this scenario is now changing. Whilst the vast majority of fiber processing and handling equipment is still designed for silicate glass fibers, many companies are testing soft glass fibers and modifying their equipment and processing recipes accordingly. Many companies are delivering spliced MIR fibre solutions and systems direct to the market causing a major rethink of the commercial opportunities that are available in the MIR. The number of commercial suppliers of off-the-shelf mirrors, lenses and detectors created for the MIR is rising, suggesting that the market for these products is strong. We are now entering a new phase in the exploitation of the MIR with many exciting opportunities ahead and fiber-based sources of coherent MIR radiation are likely to play an underpinning role.