Fluoride and chalcogenide glass fiber components for mid-infrared lasers and amplifiers: Breakthroughs, challenges, and future perspective

Driven by an ever-increasing demand for photonic solutions for applications in medicine, micromachining, telecommunications, and science, research into developing efficient mid-infrared (mid-IR) laser systems has grown dramatically in recent years. Particularly, laser sources that operate at the short-wavelength end of the mid-IR range, spanning from 2.5 to 5 μm, are required in infrared spectroscopy, including environmental gas monitoring^1,2 and breath analysis³ as well as in microscopy, minimally invasive laser surgery,⁴ and non-metal laser processing (cutting, drilling, surface treatments, etc.). Complementary to UV absorption, selective ablation in the mid-IR can play a noteworthy role in the laser processing of multilayer materials in the electronics, photonics, or medical industries, where it is critical to process a particular layer without affecting adjacent ones.^5,6 Mid-IR lasers also serve as ideal laboratory tools for scientific applications, including nonlinear optics, silicon photonics, quantum optics, high-field physics, and metrology, e.g., using frequency combs.^7–9 These advanced applications are driving the development of new laser architectures that will eventually be translated into real-world applications.

To date, only a few mid-IR laser technologies capable of generating light beyond 2.5 μm have reached a high level of maturity, primarily quantum cascade lasers (QCLs)¹⁰ or interband cascade lasers (ICLs),¹¹ as well as optical parametric oscillators, crystal-based solid-state lasers,^12,13 and sources employing nonlinear effects, such as supercontinuum generation¹⁴ or soliton-self-frequency shift.¹⁵ In these regards, fiber lasers present an emerging technology entering the race as a promising mid-IR source, with their inherent benefits of compactness, robustness, user-friendly operation, diffraction-limited beam quality, and high efficiency. The operational wavelength range of the current state-of-the-art systems is summarized in Fig. 1.

Among crystal-based lasers, the most effective mid-IR emitters are transition metal doped II–VI compounds,¹² such as Fe²⁺:ZnSe, Cr²⁺:ZnSe, or Cr²⁺:CdSe, as well as Er-doped yttrium based crystals, particularly Yttrium–Aluminum–Garnet (Er³⁺:YAG) or Yttrium–Scandium–Gallium–Garnet crystals (Er³⁺:YSGG). Transition metal doped II–VI compounds can enable generation within the wavelength range spanning from 2.8 to 4.4 μm (potentially expandable up to 5.5 μm in the case of Fe²⁺:ZnSe).¹³ Presented in 1975, Er³⁺:YAG lasers¹⁷ have become a golden standard for applications in dentistry, dermatology, and cosmetology as their operation wavelength of ∼2.94 μm overlaps with the absorption maximum of liquid water.

The operation principle of QCLs relies on inter-subband transitions, i.e., transitions occurring within the conduction band.¹⁰ As a result, the emission wavelength of the systems is determined not by the bandgap of materials, as in the case of other semiconductor lasers, but by the thickness of the quantum wells. The most widely used materials for QCLs are III–V compound semiconductors such as InP and GaAs, which have a minimal height of the conduction band offset of 3.4 and 8 μm.¹⁸ Therefore, the shortest possible wavelength for QCLs is limited, and the above-mentioned minimal values can only be achieved under cryogenic cooling. ICLs are similar in many ways yet rely on interband transitions instead of inter-subband transitions.

Optical parametric sources, such as optical parametric oscillators (OPOs) and generators (OPGs) and difference frequency generators (DFGs), exploit χ⁽²⁾ nonlinear interaction to achieve widely tunable coherent generation at different time domains, from CW to ultrashort pulsed.¹⁹ The research in birefringent nonlinear media and quasi-phase-matched materials has enabled modern parametric sources to operate at high optical powers with great conversion efficiency and excellent beam quality. Nevertheless, the nonlinear material itself presents the key limitations of parametric systems, as it has to satisfy the requirements of mechanical hardness, chemical and thermal stability, optical damage threshold, and phase-matching.²⁰ Further limitations come from frequency adjustments and environmental (temperature and vibration) sensitivity.

Fiber-based laser systems present a more promising platform compared to solid-state lasers due to their robustness, high stability, cost-effectiveness, and tuneability. Silica-based fiber lasers and amplifiers, operating in the near-IR wavelength range, have reached a high level of maturity. Delivering high-power ultrashort pulse output, these photonic systems can seed nonlinear effects in optical fibers, crystals, or chips, yielding efficient frequency conversion mechanisms into the mid-IR wavelength range, such as supercontinuum generation^14,21 or soliton-self-frequency shift.^15,22 However, the long wave transmission edge of silica-based optical fibers and, therefore, direct laser generation in silica fiber-based laser systems is limited to 2.5 μm due to the high phonon energy of ∼1100 cm⁻¹.

As the above-mentioned approaches to generate in the mid-IR wavelength range are summarized in Fig. 1, the apparent gap can be identified in the spectral range from 3 to 5 μm. These wavelengths are covered by direct laser generation only discretely or by broadband nonlinear sources, each of which features certain limitations for potential application. Alternatively, this wavelength gap in the mid-IR range can be directly accessed by potential high-brightness fluoride and chalcogenide-based fiber systems, which were the subject of extensive research over the past three decades.^23–26 While careful analysis of the current state-of-the-art demonstrates a significantly deepened understanding of the properties of fluoride and chalcogenide fibers (which are referred to as soft glass fibers throughout this Perspective) and laser dynamics in the mid-IR range, the actual laser designs have not significantly improved since the first demonstrations of all-fiber linear laser configurations with butt-coupled pump source²⁷ or ring mode-locked lasers employing nonlinear polarization evolution and free-space components,²⁸ schematically shown in Fig. 2. The free-space elements in the laser cavity limit the robustness and user-friendliness of a laser. More importantly, the absorption bands of the atmosphere introduce additional spectral filtering, limiting the minimal achievable pulse duration²⁹ and pulse quality. The demonstration of fully flexible, reconfigurable, and maintenance-free all-fiber lasers still presents an ongoing challenge. Over the past decade, several review papers have covered the progress in achieving mid-IR generation in fiber lasers at different spectral and temporal domains.^25,30–33 In this Perspective, we aim to present an overview of the existing challenges and possibilities of adapting existing silica-based fiber processing and functionalization infrastructure to fluoride and chalcogenide fibers to fabricate the essential building blocks for all-fiber lasers. By reviewing the current progress in the field, the prospect of future research and development in mid-IR all-fiber lasers will be evaluated.

The outline of the current Perspective is the following: Sec. II will summarize the properties of the soft-glass fibers, focusing on fluoride and chalcogenide materials relevant for designing and fabricating the laser components. Section III will discuss the functionalization of mid-IR glass fibers, mainly focusing on the inscription methodologies of fiber grating structures. Here, different modalities of fiber gratings will be reviewed with an outlook toward enhancing their functionalities. Further in Sec. IV, we will discuss the challenges and prospects of conventional heat-pull methodologies for fiber-based laser components, as well as project the pathway for developing alternative fuse-free approaches. Section V will provide the readers with a broader landscape of components for fiber lasers, going beyond the pure fiber designs toward waveguide systems. Here, chip-based components with nonlinear response will be covered, followed by Sec. VI portraying saturable absorbers operating in the mid-IR spectral range for ultrashort pulse initiation in more detail. Finally, Sec. VII will discuss further potential developments in the field.

Up until now, a manifold of different glass compositions with potentially exceptional transmission in mid-IR has been introduced in the scientific literature. This includes already mature heavy metal fluoride and chalcogenide glasses and fibers, which will be discussed in the current section. The characteristic transmission ranges for these fibers are summarized in Fig. 3. Despite the significant progress, the development of new glass compositions with improved stability and transmittance remains under continuous research. It is worth mentioning that many more glass compositions were investigated, including heavy metal oxide, fluoride phosphate, fluorotellurite, oxyfluoride, and other glasses. This chapter will provide concise information about the most widely used fiber glass matrices compositions and their properties, which are relevant particularly for fiber laser and amplifier component fabrication.

The fluorozirconate, or so-called ZBLAN, glass has become the most mature composition of the mid-IR glasses. The original ZBLAN glass attributes to the specific composition of 53% ZrF₄, 20% BaF₂, 4% LaF₃, 3% AlF₃, and 20% NaF.³⁷ At present, this term is attributed to a variety of glass compositions based on ZrF₄, which are used interchangeably in modern literature. In this context, it is essential to highlight that changes in the concentration of main fluorides, as well as the addition of several dopants, may significantly vary the properties of the glass. Therefore, for all glasses in Sec. II, we discuss the impact of specific glass compositions, primarily affecting the glass stability and thermal and optical properties, which are important for component fabrication.

The transmission edge in the mid-IR wavelength range is determined by the multiphonon absorption, shown in Fig. 3. Modern fluorozirconate optical fibers demonstrate good transparency up to ∼4.5 μm, where the loss reaches 3 dB/m before increasing sharply. Reducing the concentration of the most absorbing units in the glass makes it possible to shift the multiphonon absorption edge to longer wavelengths. Following this idea, aluminum-free fluorozirconates should be more transparent than ZBLAN.³⁸ However, elaborate research is required to make such glass compositions stable.

One of the frequently used substitutions for ZrF₄ is HF₄. Hafnium is chemically similar to zirconium, but its atomic mass is almost twice that. Adding Hafnium decreases the glass refractive index due to its lower atomic polarizability³⁹ and increased glass transition temperature (Tg). Therefore, it can be utilized to tailor the numerical aperture of the fibers by ranging refractive index within 1.49–1.515 span.^34,40 Along with hafnium, the refractive index difference between fiber core and cladding can be managed by doping the core with PbF₂ or by increasing the concentration of NaF or AlF₃ in the cladding.

It is well known that glasses with high concentrations of glass formers, due to their high bond strength, would possess the lowest thermal expansion and highest values for their transition temperatures. The average values of the thermal expansion coefficient for ZBLAN glass vary from 1.5 to 2 × 10⁻⁵ K⁻¹ but can be reduced down to 10⁻⁵ K⁻¹ for glasses free of barium and sodium.⁴¹ Furthermore, NaF and PbF₂ lower the glass transition temperature with a consequential decrease in viscosity.⁴² On the contrary, a high alkali content in the glass reduces the glass transition temperature and increases the thermal expansion coefficient. LiF presents an exception to this rule, and while replacing NaF, it lowers the expansion coefficient and raises Young’s modulus, but at the same time leads to a decrease in Tg.⁴³ 

Rare earth elements can be doped into fluoride glass as a direct substitute for the existing lanthanide of the glass matrix, i.e., LaF₃, without a considerable decrease in glass’s stability. Notably, the rare-earth doping concentration can reach 10 mol. %, which is much higher than in silica glass.

In recent years, substantial interest has been shown in glasses with a high content of indium fluoride, InF₃. Due to the lower phonon energy of fluoroindate glasses (510 cm⁻¹) in comparison to fluorozirconate (580 cm⁻¹), they are more transparent toward the longer mid-IR wavelength range (extending from UV to 7–8 μm for bulk glass and up to 5.5 μm for fibers, as depicted in Fig. 3) and have lower values of nonradiative decay rates. Furthermore, the calculated values of minimum optical losses are significantly lower in fluoroindate glasses than in other fluoride-based glasses.⁴⁴ They are also characterized by better optical quality, stability against atmospheric moisture, and mechanical stability than fluorozirconate glasses. However, despite the slightly higher thermal stability of fluoroindate glass (T_g ≈ 300 °C for InF₃ and T_g ≈ 280 °C for ZBLAN glass⁴⁰), it requires faster cooling rates of glass-forming melts than fluorozirconate glass.

Contrary to ZBLAN in the fluorozirconate glass family, there is no golden standard for InF₃-based glass composition. Therefore, the literature demonstrates quite a broad range of properties related to different glass compositions. Here, we briefly review the roles of possible glass doping materials.

The first works have identified the essential presence of GaF₃ with at least 2 mol. % concentration for the glass stability,⁴⁵ although this negatively affects the mid-IR transmission edge. Typically, the total concentration of InF₃ and GaF₃ ranges from 35 to 45 mol. %. Furthermore, NaF, ZnF₂, and GaF₃ have been used for efficient glass stabilization in various fluoroindate glass matrices.^46,47 The addition of Sr, Ca, and Ba fluorides allows for a decrease in the glass refractive index to obtain the required NA of the fiber.⁴⁵ The typical refractive index of fluoroindate fiber core is close to one of ZBLAN glass and is about 1.48–1.52 at 633 nm wavelength.^34,40

The incorporation of rare-earth ions into fluoroindate glasses has also been thoroughly investigated, particularly of Er³⁺ ions. In earlier studies, while the 2 mol. % doping concentration of Er³⁺ ions showed a promising enhancement in InF₃-based glass stability, further increase in dopant concentration resulted in a significantly less stable glass.^48,49 Critically, the incorporation of rare-earth ions at high concentrations, above 8 mol. %, into fluoroindate glass fibers proved to be challenging due to detrimental crystallization processes in the glass during fabrication, as confirmed by x-ray diffraction. Therefore, doping concentrations just above 2 mol. % for a range of ions, such as Er³⁺, Yb³⁺, Nd³⁺, and Pr³⁺, had been chosen as a trade-off between high gain and good stability of the glass matrices.^48,49 The most recent improvements in glass composition and preparation techniques, as well as advances in the purity of the materials, led to a substantial increase in possible doping concentrations and outstanding minimum losses below 20 dB/km, justifying good glass stability. Particularly, the concentration of Er³⁺ ions has recently been increased up to 9 mol. % and of Ho³⁺—to 10 mol. % in multicomposite fluoroindate fibers, comprising also BaF₂, ZnF₂, SrF₂, and other metal fluorides in the core.^50–52 

Fabrication of fluoroaluminate glass was first reported by Sun et al. in the 1950s.⁵³ Similarly to the above-discussed members of the fluoride glass family, they exhibit extended transmission from UV to mid-IR (up to 6 μm for a 3 mm-thick glass)⁵⁴ with a slightly lower refractive index of 1.4–1.45. The key differences that attracted the research attention are higher thermal stability (T_g ≈ 450 °C), better hydrolysis resistance, and mechanical strength of fluoroaluminate glasses⁵⁵ than other fluoride matrices. Due to the above-mentioned advantages, fluoroaluminate fibers are often used as endcaps for InF₃ and ZBLAN fibers to prevent tip degradation.⁵⁶ 

To date, fluoroaluminate glasses remain less developed than fluoroindate or fluorozirconate matrices due to their lower stability against devitrification. The glass-forming region of fluoroaluminate glasses is small and limited. As typical in glass development work, to improve the glass stability, multicomponent compositions were adapted.⁵⁷ In Ref. 58, the effect of the addition of YF₃, LaF₃, and YbF₃ on the glass crystallization was studied. It was shown that the incorporation of LaF₃ even in small concentrations leads to crystallization. On the other hand, doping with YF₃ or YbF₃ increased the glass stability at concentrations as high as 15 mol. %. Furthermore, the glass forming region was enlarged with the addition of MgF₂, which allowed obtaining bulk glasses with different compositions without devitrification.⁵⁴ The incorporation of barium and strontium fluorides into the glass significantly decreased the critical cooling rates close to the values of fluorozirconate glass.⁵⁹ 

Among the other potentially beneficial properties of fluoroaluminate glass are the highest solubility and stability upon doping with rare-earth fluorides. Glasses with concentrations of ErF₃ as high as 20 mol. % were successfully drawn into fibers, demonstrating the highest emission efficiencies at 2.7 and 3.5 μm with 18 mol. % doping concentration.⁶⁰ 

Chalcogenide glasses are a large group of materials containing one or more of the chalcogens (S, Se, or Te) combined with elements from IV, V, or VII groups of the Periodic Table (Ge, As, Sb, Ga, In, P, F, Cl, etc.). Chalcogenide glasses possess the widest transmission window among other mid-IR glasses, typically spanning from 1 to 15 μm, low phonon energy (200–300 cm⁻¹), and high nonlinearity, with Kerr constant n₂ two to three orders of magnitude higher than of silica fibers. Refractive indices range from 2 to over 3. Figure 3 represents an example of the loss spectra of two characteristic chalcogenide fibers within the spectral range of interest of the current Perspectives. The resulting high Fresnel reflection from the fiber end face sets a requirement for antireflective coatings to avoid it.

In contrast to less polarisable fluorine atoms in the fluoride glasses, chalcogen atoms possess a lone pair of electrons, which can participate in light-induced reactions, producing structural modifications.⁶¹ These effects lead to changes in the optical properties of the material by altering the optical bandgap. Some of the reactions can be reversed by annealing.⁶² The incorporation of heavier chalcogens in the glass network leads to weaker average bond strength and lower thermo-mechanical stability. For instance, in the Ge–As–Se–Te system, the glass-transition temperature decreases both with an increase of Se level when it substitutes Ge or As or as Te replaces Se.⁶³ 

It was shown that weaker bonding promotes longer wave transmission. Therefore, Te-based glass matrices exhibit extended transmission as far as 20 μm in bulk glass,⁶⁴ yet at the cost of drastically lower glass-transition temperature (typically on the order of 10–100 °C). Among other chalcogenides, binary As₂S₃ or ternary As₂Se_1.5Te_1.5 sets of glasses are ones of the most promising due to wide transmission range (1–18 μm), better stability against crystallization, and moderate glass transition temperatures (145–185 °C).⁶⁵ 

With the incorporation of rare-earth elements, chalcogenide glass enables laser generation further to the longer edge of the mid-IR wavelength range. It is worth noting that the solubility of rare-earth ions in chalcogenide fibers is limited to the order of 1000 ppm due to a decrease in the glass stability. A small concentration of gallium was reported to help integrate more rear-earth ions, yet to a still restrained extent.⁶⁶ 

Similar to well-established fiber lasers in the near-IR, the incorporation of gain fibers doped with rare-earth ions presents an efficient and straightforward technique to achieve narrow or broadband direct laser emission in the mid-IR spectral range. Here, we briefly discuss emission and absorption cross sections and energy level lifetimes of the most widely used ions, as well as how to create and, most importantly, maintain a population inversion to achieve high efficiency and high output from a laser cavity. Note that the transitions with emission wavelengths shorter than 4 μm are generally considered in the context of the fluoride matrix. Emission at longer wavelengths is discussed in the scope of chalcogenide glass.

Laser emission from Er³⁺ transition in fluoride fibers at 2.75 μm is one of the most mature in mid-IR fiber lasers (Fig. 4). The upper laser level ⁴I_11/2 is easily accessible from the ground state via convenient and cost-effective laser diodes operating at 975 nm wavelength. The shortcoming of this laser transition is that the intrinsic lifetime of the upper level ⁴I_11/2 is lower than of the lower level ⁴I_13/2 (6.9 and 9 ms, respectively), limiting the population inversion. One of the approaches to reduce the population of the lower laser level ⁴I_13/2 relies on utilizing heavily Er³⁺-doped fibers with concentrations above 4 mol. % and, therefore, taking advantage of the energy transfer upconversion (ETU) processes.^67,68 An alternative approach introduces cascaded lasing at 2.8 and 1.6 μm with a lower level of Er-doping (around 1 mol. %) to avoid ETU.⁶⁹ Further emission from the Er³⁺ transition at 3.5 μm wavelength is located between ⁴F_9/2 and ⁴I_9/2 levels. The most efficient way to access this transition was found in dual-wavelength pumping, exploiting 980 and 1980 nm simultaneously.⁷⁰ 

Being in the 1990s among candidates of 1.3 μm-active ions, Dy³⁺ reemerged later as a promising ion enabling generation at a longer mid-IR range, demonstrating emission at 4.4 and even 5.5 μm in both fluoride and chalcogenide fibers.^71–73 Dy³⁺ ions exhibit many absorption lines convenient for laser pumping, spanning from 0.8 to 2.9 μm. Importantly, dysprosium features efficient ground-terminated transition ⁶H_13/2 → ⁶H_15/2 at 3 μm, which can be excited by in-band pumping with Erbium fluoride fiber lasers with minimal quantum defect.⁷⁴ Furthermore, the population inversion of the ⁶H_13/2 level can be further increased by co-doping of Dy³⁺ fluoride fibers with other rare-earth ions. Particularly, the sensitization with Er³⁺ is of great interest,⁷⁵ yet co-doping with Yb³⁺ and Tm³⁺ has also been demonstrated in different hosts.⁷⁶ It is worth noting that the application of such co-doped active fibers enables the additional benefit of using cost-effective pump sources operating at 980 nm. Lasing transitions of Dy³⁺ ions corresponding to longer wavelengths, such as those mentioned earlier at 4.4 and even 5.5 μm, are prone to self-termination in both fluoride and chalcogenide glasses, as the lifetimes of the upper laser levels are significantly shorter than those of the lower ones. Therefore, it is challenging to achieve laser generation from these transitions. For example, the cascade lasing on the 3 μm transition, when pumped at 1.7 μm, has been suggested as one of the routes to achieving 4.1–4.4 μm laser generation.⁷⁷ 

Holmium-doped fluoride fibers are another promising platform for mid-IR laser development. Specifically, Ho³⁺ exhibits three transitions in the mid-IR, located at 2.9, 3.2, and 3.9 μm wavelength. The 3.2 μm transition of Ho³⁺ is located between the thermally coupled (⁵F₄,⁵S₂) levels and the ⁵F₅ level. Since the direct pumping to the upper level with 532 nm is inefficient, Zhou et al. recently suggested a dual-wavelength pumping scheme at 1150 and 980 nm. However, its efficiency is yet to be experimentally proven.⁷⁸ Both Ho³⁺ transitions at 2.9 (⁵I₆ → ⁵I₇) and 3.9 μm (⁵I₅ → ⁵I₆) wavelengths are self-terminating due to the longer lifetimes of the lower laser levels. After the first demonstration of emission at 3.9 μm in liquid nitrogen-cooled ZBLAN fiber,⁷⁹ further research led to a change of the matrix toward InF₃, to decrease the nonradiative decay rates.⁵⁰ Furthermore, high doping levels promote ETU processes, enabling more efficient depopulation of the lower levels. Recently, Luo et al. proposed a novel approach, utilizing a dual-wavelength pumping scheme at 1650 and 1950 nm and cascaded lasing at 3.9 and 2.9 μm, in lightly doped (1 mol. %) fluoroindate fiber.⁸⁰ This technique is expected to increase the efficiency of the 3.9 μm emission up to 32%. The incorporation of Pr³⁺ was found to be another effective solution to depopulate the ⁵I₇ level of holmium through the resonance with the Pr³⁺³F₂ level.⁸¹ 

Despite mainly being researched for applications in visible and near-IR wavelength regions, Pr³⁺ ions incorporated into chalcogenide glass matrices possess a few favorable emission lines in the mid-IR. Pr³⁺ can be pumped advantageously using well-developed silica-based fiber lasers at 1.06, 1.5, and 2 μm wavelengths. One of the obstacles for laser generation is a small luminescent branching ratio of transitions ending on the upper laser level ³H₅, which results in the bypassing of this level by the pump energy.⁸² An increase in the doping concentration can help to overcome this bottleneck by taking advantage of the cross-relaxation processes, resulting in up to 3 photons on the upper laser level ³H_15/2 when pumped in the near-infrared. Furthermore, to increase the emission bandwidth in the 2–6 μm band, Pr³⁺ can be co-doped with Dy³⁺.⁸³ 

Nd³⁺ in chalcogenide fibers demonstrates a similar energy level structure to that of Pr³⁺. Since the energy gaps between the four levels of Nd³⁺ (⁴I_15/2 → ⁴I_9/2) are similar, the implementation of a cross-relaxation inducing pumping scheme at 1.7 or 2.5 μm makes it possible to achieve 5.5 μm emission.⁸² 

Tb³⁺-doped chalcogenide fibers are another promising platform for mid-IR laser generation. Strong absorption bands located at 2 and 3 μm wavelengths made it possible to achieve 5 μm emission in a 3-level scheme.⁸⁴ Further transitions at 7.5 μm and a shorter one at 3.2 μm were experimentally observed but have not been shown in a fiber laser configuration yet.⁸⁵ 

A further potential candidate for emission in the 4.5–5.5 μm band is the Ce³⁺ ion. However, a long luminescence lifetime of the upper laser ²F_7/2 level of 4.7 ms is overshadowed by a challenging pumping scheme operating in the 3.5-5-5 μm band, limiting the resulting output power. Sensitization of Ce³⁺ ions when co-doped with Dy³⁺, as demonstrated in selenide glass, can increase the pump efficiency in the future.⁸⁶ 

Finally, the energy level structure of Sm³⁺ exhibits a variety of mid-IR transitions located at 2.5, 3.5, and 7.7 μm bands with convenient pumping schemes in the near-IR (1.3, 1.45, and 2 μm). The highest emission intensity in the 7–8 μm band makes Sm³⁺ a promising solution for laser generation at longer wavelengths.⁸⁷ 

The initial attempts to transfer the established inscription techniques to ZBLAN fibers concluded that the fiber is not photosensitive to UV inscription.^91,92 One of the proposed solutions to artificially boost the photosensitivity of fluoride glass fibers was doping with certain rare-earth ions, e.g., cerium⁹³ or cerium–thulium.⁹⁴ Sufficient reflectivity of the resulting grating could be achieved only at very high concentrations of rare-earth ions (more than 4 wt. %). However, it is worth noting that such a heavily doped core glass was prone to crystallization and required high quenching rates during the preform fabrication to mitigate it.⁹⁵ 

A substantial step forward was made by Davis et al. by applying tightly focused femtosecond (fs) pulses to induce a stable refractive index change in transparent glasses.⁹⁶ Eventually, permanent refractive index structures could be induced by employing ultrashort pulses in glasses and optical fibers with different chemical compositions, regardless of their photosensitivity. The most widespread state-of-the-art approaches utilizing fs-laser for fiber grating inscription are described in detail below.

Direct inscription of FBGs implies modification of a fiber with a tightly focused fs laser beam. Depending on the exact spatial parameters of the induced modification, different techniques can be found in the literature. For instance, the point-by-point (PbP) method utilizes a single fs laser pulse, typically focused via an objective, to create a highly localized single micro-void in a fiber core. A grating is formed by creating a periodic sequence of such individual modifications along the fiber, e.g., while moving it on a translation stage with respect to the laser beam focal point. A schematic representation of the inscription setup is shown in Fig. 5(a).

The key advantage of the inscription methodology is its flexibility, particularly in applied pulse energy and translation speed. It yields any diffraction pattern with high efficiency in fiber, e.g., chirped and apodized gratings. Nevertheless, since the inscription relies on the processes occurring above the damage threshold of the material, the resulting structures have a high level of scattering losses due to Mie scattering from the individual micro-voids.⁹⁷ To mitigate the scattering losses, the technique of “continuous core-scanning” has been introduced,⁹⁸ where the inscription laser intensity was kept below the damage threshold of the material. In addition to the axial direction, the fiber was scanned transversely to its axis using an oscillating piezo actuator, allowing for enhanced overlap between the modified cross section and the guided mode. Overall, the scattering loss could be significantly decreased, but at the cost of increased fabrication time. Further improvements in fiber translation techniques led to demonstrations of “line-by-line”⁹⁹ and “plane-by-plane”¹⁰⁰ methods.

The first inscription in fluoride-based fibers using the PbP technique was performed by Hudson et al.¹⁰¹ This work reported on the inscription of a 20 mm long, first-order FBG with a Bragg wavelength of 2.9 μm in Ho/Pr co-doped ZBLAN fiber. The inscription setup was based on a Ti:Sapphire fs laser, emitting 110-fs pulses at 800 nm wavelength with a repetition rate of 1 kHz. It is worth mentioning that the inscribed structure consisted of type I modifications, contrary to micro-voids observed in silica fibers.

The flexibility of the direct inscription has been once again confirmed by inscribing a 16-mm long, 45°-tilted FBG in Ho/Pr co-doped ZBLAN fiber.¹⁰² Since each inscribed line is at the Brewster angle for the incident light, the tilted structure acts as an in-fiber polarizer, allowing it to achieve a 21.6-dB polarization extinction ratio in the laser cavity.

Another implementation of the direct inscription approach was the fabrication of a long-period grating in Er-doped double-clad ZBLAN fiber.¹⁰³ Here, the focus of the inscription light was adapted to cover the entire cross section of the fiber core, and by translating the fiber along the fiber axis, a smooth modification with a period of 630 μm over 75 mm length was obtained. Inscription was performed at a high repetition rate of the fs laser of 250 kHz, which led to heat accumulation effects in the focal volume. However, these effects were taken advantage of by increasing the overlap of the modification with the fiber core. As a result, strong attenuation peaks with a magnitude of 24 dB were shown at 2.8 μm wavelength. After thermal annealing at 150 °C for 30 min, the grating strength was reduced by about 20%.

In addition to ZBLAN, FBGs were also inscribed into dysprosium-doped fluoroindate fibers via the line-by-line technique. Narrow-linewidth (∼100 pm) wavelength stabilized lasing at 3.2 μm was achieved using a high-reflectivity fiber Bragg grating (R = 98%).¹⁰⁴ 

A typical setup of the phase mask inscription method comprises a diffractive element—a phase mask—that creates an interference pattern between its ±1 diffraction orders, as schematically represented in Fig. 5(b). The depth of the phase mask grooves is designed in such a way as to maximize the diffraction efficiency in both first orders and suppress the intensity in the 0th order. The cylindrical lens focuses the incident light along the transverse axis of the fiber. Therefore, the interference pattern is transferred to the fiber core. The simple setup makes the inscription process robust and easy to produce high-quality gratings. A tailored design of the phase mask allows the fabrication of more complex grating structures, such as chirped, tilted, pi-shifted, or apodized. The combination of the phase mask with the translation stage produces the so-called phase mask scanning method¹⁰⁵ to achieve several centimeters long gratings with extremely narrow reflection bands or, on the opposite, acquire a large amount of chirp for a broad reflection spectrum. On the downside, the period of the generated modification pattern is permanently fixed at the value of half of the phase mask period, so the individual structure requires a unique phase mask for its inscription.

The first inscription of a FBG in fluoride-based fiber using the phase mask technique was reported by Bernier et al. in 2007.¹⁰⁶ The inscription setup was based on a Ti:Sapphire regenerative amplifier system, emitting 35 fs pulses at 800 nm with a repetition rate of 1 kHz. The polymer coating of the fiber was removed prior to the inscription. A series of gratings were inscribed in passive and Tm-doped ZBLAN fibers, demonstrating the impressive capabilities of the phase mask technique. The reflectivity of these gratings reached as high as 98%, and a negative refractive index change of 10⁻³ was achieved. After annealing at 125 °C for 30 min, the initial modification of the refractive index was reduced by about 50%. This successful inscription procedure was later optimized for the FBGs writing through the polymer coating, further expanding the potential applications of the phase mask technique by preserving the mechanical strength of the fiber.^107,108 Overall, the phase-mask inscription method enabled the development of the record high-power fiber lasers to date.^108–111 

Inscription of FBGs using a Talbot interferometer, shown in Fig. 5(c), has been first introduced for germanosilicate fibers using UV excimer lasers.^89,112 Similarly to phase mask inscription, the system employs a phase mask as a beam splitter. While the 0th order of the phase mask is blocked, reducing the heat load onto the fiber, the ±1 diffraction orders are launched onto adjustable mirrors of the Talbot interferometer. Combined at the fiber position, their interference pattern modifies the refractive index of the fiber core. Overall, this technique combines the benefits of great flexibility in possible Bragg wavelengths, typically within 200 nm, via adjusting the position of the interferometer mirrors and gentle modification of the fiber through an interference pattern. As the inscription lasers moved from nanosecond to femtosecond pulse duration, the requirements on fiber photosensitivity were relaxed, which allowed the implementation of the inscription technique to more exotic fibers, such as sapphire¹¹³ and recently fluoride fibers.¹¹⁴ 

For the first time, Chiamenti et al. demonstrated the FBG inscription utilizing the interferometric technique for passive InF₃-based fibers.¹¹⁴ The inscription was performed using the second and third harmonics of the Ti:Sapphire regenerative amplifier at 266 and 400 nm, with a pulse duration of 350 fs and a repetition rate of 1 kHz. Due to the high absorption of fluoride glass in the UV wavelength range, the heat-induced bending of the fiber during the inscription produced multiple peaks in the reflection spectrum. An additional shutter was used to decrease the heat load, reducing the illumination time to only 100 ms per second. Overall, the maximum achieved reflectivity could not reach above 55% even during hours of laser exposure. It is worth noticing that UV light has not been shown to be suitable for inscription in undoped ZBLAN fibers so far.

On the contrary, visible femtosecond irradiation at 400 nm wavelength allows inscription of 7 mm long, first-order gratings in InF₃ fiber with reflectivity >99%. The inscription-induced refractive index modification was measured to be positive with a peak value of 12 × 10⁻⁴. Thermal annealing at 120 °C for 10 h revealed different dynamics of the grating reflectivity, depending on the fs-laser exposure time. Therefore, after cooling down to room temperature, the authors recorded an increase in reflectivity of about 16% and a blue shift of the Bragg wavelength of ∼795 pm for FBG inscribed during 1 h. The Bragg grating experienced 2-h laser exposure expanded by 100 pm in bandwidth, blueshifted only by 120 pm, and received almost 1.5 dB enhancement in reflectivity.¹¹⁴ These observations suggested that the inscribed FBGs belong to different types of grating structures.

Moreover, an interferometer associated with visible fs-laser enabled FBG inscription in ZBLAN fibers with a reflectivity of 95%.¹¹⁵ Thermal treatment of the FBG in ZBLAN fiber caused similar changes in its spectral characteristics to the grating in InF₃ with comparable reflectivity. After annealing, FBG in ZBLAN fiber exhibited an increase in reflectivity by 2 dB, a blue shift of the Bragg wavelength of 450 pm, and a decrease of the spectral width by 30 pm. Finally, the same setup has been applied for the inscription in active Dy-doped InF₃ fiber, with a maximum induced refractive index change and reflectivity of 8 × 10⁻⁴ and 85%, respectively.¹¹⁶ 

The properties of chalcogenide fibers pose certain limitations toward the translation of the techniques mentioned earlier for grating inscription. For instance, due to the high absorption of chalcogenide glass in the UV and partially in the visible spectral region, the inscription is only possible using near-IR lasers. Furthermore, the high optical nonlinearity of the chalcogenide matrix should be taken into consideration, as it leads to a significantly lower threshold for self-focusing, which is almost inevitable when using fs laser irradiation. Owing to the high values of the fiber refractive index, the resulting FBG pitch for a fundamental Bragg resonance in the near-IR is shorter than the inscription wavelength. This restricts the inscription of gratings operating in near-IR wavelengths to higher-order FBGs. Similarly to fluoride-based fibers, the inscription should be performed through the coating of the fiber to maintain the mechanical properties.

The pioneer FBG inscription in chalcogenide fiber utilized the holographic approach using a CW He–Ne laser with 50-mW power.¹¹⁷ Gratings with reflectivity higher than 90% were achieved in As₂S₃ fiber, with an estimated induced refractive index change of the order of 10⁻⁴.

The phase mask technique has also been successfully implemented for chalcogenide fibers using He–Ne^118,119 and Ti:Sapphire laser irradiation.¹²⁰ The work¹¹⁸ implemented a specially designed phase mask for the inscription of a first-order FBG at 1.55 μm, which produced the interference pattern between its 0th and −1st diffraction orders. However, grating reflectivity in single mode As₂S₃ fiber was limited to only 50%. Inscription of second-order FBGs in a similar fiber resulted in stronger gratings with reflectivity up to 97%.¹¹⁹ Grobnic et al. were the first to combine the phase mask approach together with fs-laser irradiation for FBG inscription in chalcogenide fibers,¹²¹ but a strong grating with a reflectivity higher than 99% in As₂S₃ fiber has been thus achieved only later.¹²⁰ The estimation of the refractive index change was found to be 1.04 × 10⁻³. Thermal annealing performed at 100 °C for 64 min resulted in a reflectivity decrease down to 90%, maintaining about 60% of initial refractive index modulation.

Recently, Liu et al. demonstrated the direct inscription of FBGs in double-D shaped As₂S₃ fiber.¹²² The inscription setup was based on a Ti:Sapphire laser, emitting 50-fs pulses at 800 nm with a 200-Hz repetition rate. The authors reported a very high sensitivity of the grating morphology to the average inscription power in the order of microwatt, which significantly affects the grating strength. After several optimization steps, 99% reflective gratings were achieved.

The first demonstration of FBG inscription in fluoride glass fibers has become an enabling factor for monolithic all-fiber laser system development.^108,123 To date, many grating designs have been implemented in fluoride-based fibers, including chirped, 45° tilted, π-shifted, and long-period gratings. Remarkably, in Ref. 124, Bernier et al. have demonstrated the first all-fiber single-frequency distributed feedback laser at 2.8 μm, based on 30-mm long π-phase-shifted FBG, inscribed with the dithering of the phase mask against the fiber. However, the state of the art in fluoride-based fiber Bragg gratings is far from the technology developed for silica fibers. To achieve better quality of the inscribed gratings and enhance their reproducibility, it is crucial to understand the mechanisms of laser-induced modification of the glass and be able to tailor it for specific needs.

Characteristics of the inscription laser, such as central wavelength, pulse energy, and repetition rate, as well as the chosen fabrication technique itself, can play a significant role in the laser-induced modification of the fiber.

Reference 125 studied the impact of the fluence of the femtosecond laser irradiation on ZBLAN glasses during direct inscription. The authors observed a negative refractive index change with Δn in the order of 10⁻³ in a wide range of pulse energies and repetition rates. However, when the fluence increased above 375 kJ/cm², the sign of the refractive index modification changed to positive [Fig. 6(a)]. The authors attributed positive refractive index change to heat accumulation processes at high repetition rate exposure, while the local expansion of the irradiated region at low repetition rates leads to negative change.

These results were opposed by the results of Gross et al., which showed smooth and negative refractive index change with amplitude of 10⁻³ in ZBLAN glass as a result of point-by-point laser irradiation at 800-nm wavelength with pulse energy ranging from 1 to 2.5 μJ.¹²⁶ Raman microscopy confirmed that fs laser irradiation causes local melting of the glass followed by rapid quenching, breaking one of the two bridging bonds between adjacent zirconium fluoride polyhedra during resolidification. Since the displacement of zirconium cations is more substantial for single bridging links, it leads to the expansion of the glass network and a simultaneous decrease in optical density. The observations of the refractive index modification in InF₃-based fiber caused by direct inscription also showed a similar negative change.¹²⁷ In Ref. 114, Chiamenti et al. observed a positive refractive index change after the holographic inscription of an FBG using 400-nm fs irradiation [Fig. 6(c)].

The nature of the modification of the fluoride fiber refractive index is yet to be clarified and requires further extensive research. A particular challenge presents the complexity of the fiber drawing processes, as during it, fluoride glass might undergo a slight structural transformation, which changes the photosensitivity of the fiber and, therefore, influences the obtainable refractive index change.¹²⁶ Furthermore, as we discussed earlier, the composition of fluoride fibers is very diverse. The addition of rare-earth ions or other elements, such as hafnium, into fluorozirconate glass (more details see in Sec. V) can alter the amplitude and even the sign of the laser-induced refractive index modification.

Mid-IR glass fibers, apart from extended transmission windows, provide a range of promising properties for laser wavelength tuneability or sensing applications compared to conventional silica-based fibers, such as higher thermal expansion or thermo-optic coefficient. Furthermore, soft glasses usually have a lower elasto-optic coefficient, making the inscribed FBGs more strain-sensitive. Such environmental sensitivity has to be accounted for in laser systems comprising fiber gratings to ensure their stability or tuneability.

An experimental investigation of Ho:Pr-co-doped ZBLAN fiber laser tuneability demonstrated ∼1.39 pm/μϵ strain sensitivity of ZBLAN fibers.¹²⁸ Applying the compression to a 15-mm long FBG in ZBLAN fiber so that it produced a maximum strain of 21.57 mϵ corresponding to a displacement ΔZ, normalized by the length of the fiber L over 0.03, resulted in ∼30-nm blueshift of the laser spectrum, depicted in Fig. 7(a). Stretching of the fiber with the FBG was performed up to a normalized displacement of ∼0.005, which enabled 7 nm tuneability. Notably, the line-by-line inscription technique was performed through fiber polymer coating, preserving its rigidness during strain application.

The thermal sensitivity of FBG inscribed in standard silica fiber is 12 pm/°C. The first numerical simulations of the FBGs spectral response in chalcogenide fibers showed the highest thermal sensitivity of 175.1 pm/°C (in the case of As₂S₃ fiber).¹²⁹ FBGs inscribed in fluoride-based fibers showed thermal sensitivity of the same order as in silica at room temperature [Fig. 7(b)]. However, both ZBLAN and InF₃ matrices demonstrated a tendency to increase thermal sensitivity at longer wavelengths, which is promising for further enhancement of the laser wavelength tuneability or the performance of thermal sensors, given their wide transmission window.

The further benefit of FBGs in mid-IR fibers has been recently revealed at cryogenic temperatures.¹¹⁵ While FBGs in bare silica fiber can operate only up to liquid nitrogen temperatures (∼50–70 K), sensing the lower temperatures requires the application of additional coatings to induce the stress. The experimental measurements have confirmed that the Bragg wavelength of the gratings inscribed in ZBLAN and InF₃ fibers demonstrates continuous blueshift with the temperature decreasing to 4 K. The recorded FBG sensitivity was 1.4 and 1.7 pm/K for ZBLAN and InF₃-based fibers, respectively, within a temperature range spanning 25–50 K [Fig. 7(b)]. At lower temperatures (4–25 K), the sensitivity decreased to 0.5 pm/K and 0.8 pm/K for fluorozirconate and fluoroindate fibers, respectively.

Pump combiners are essential components in all-fiber lasers, facilitating the injection of pump radiation into the laser cavity. The primary challenge in fabricating pump combiners operating at mid-IR wavelengths is creating the interface between the mid-IR glass fibers of the laser cavity and conventional silica fibers commonly used with commercial pump diodes. The substantial difference in thermal properties makes it almost impossible to post-process these types of fibers together using standard heat-pull techniques. Therefore, alternative approaches to side-couple the pump radiation have been considered, generally based on double-clad (DC) fibers,¹³⁰ which allow the application of high-power multimode diodes as pump sources. The recent progress is summarized in Fig. 8 and will be discussed in the current section. It is worth noting that, initially, a range of similar methodologies have been suggested for conventional silica fiber-based laser component configurations.^131–133 However, such fiber arrangements have not received wide application due to the prevalence of the fiber heat-pull method.

One of the pioneers to demonstrate a fluoride-based pump combiner was Schäfer et al., who introduced a method for side-pumping through an angle-polished fiber.¹³⁴ For this, a tip of a multimode step-index passive ZBLAN fiber was polished at a 10-degree angle and spliced onto the second cladding of an Er-doped DC fiber [Fig. 8(a)]. The coupling efficiency reached 83% at a wavelength of 980 nm, notably demonstrating effective operation even at a pump power as high as 83 W. In this work, the radiation from the pump diode was launched through auxiliary lenses into pump-delivering ZBLAN fiber. The recent advances in the splicing of double-clad fluoride and multimode silica fibers led to minimal losses of only 0.2 dB.^138,139 Low loss and long-term stability of the spliced fiber interface eliminate the need for bulk optics within the demonstrated pumping arrangement, underscoring the promising potential of such a side-pump technique for all-fiber laser development.

Another technique for introducing pump radiation also involved preparing a special pump diode fiber tip.¹³⁵ This process entailed drawing a taper from a coreless silica fiber from an initial diameter of 125 μm down to 12 μm waist diameter. The tapered fiber was cleaved to remove the up-taper portion. The resulting length of the pump fiber tip comprised 40 mm of down taper region and a 30-mm long waist. In the next step, the prepared tapered fiber, spliced to the pump diode multimode port, was wrapped around a DC fiber approximately three times, as shown schematically in Fig. 8(b). The resulting combiner was immersed in a low-refractive-index polymer, which was then cured to enhance connection efficiency and maintain the relative positioning between the fibers. Within the pump power range up to 40 W at 981 nm, the coupling efficiency increased from 87.5 to a maximum of 92.8%, making it essentially the most efficient fluoride-based side-pumping combiner ever reported.

The previous technique has been further adapted and modified for pumping tellurite fibers.¹⁴⁰ It is worth noting that signal and pump fibers in the pump combiner were based on tellurite glass, and in the original arrangement, fibers were not twisted. Instead, after tapering the pump fiber, it was spliced straight onto signal fiber using graphite filament with the power of 15–22 W over 3–10 s. By optimizing the tapering ratio and lengths of the down-taper region and the waist, a coupling efficiency of 67.52% could be achieved experimentally with the signal insertion loss of 0.52 dB when pumped by a 100-mW source at 1310 nm. Furthermore, the authors investigated the impact of the length of the waist and twisting angle on coupling efficiency.¹⁴¹ Therefore, by twisting the pump fiber with a 15-mm long waist around the signal fiber at rotation angles of 30°, the maximum coupling efficiency could reach 76.58%.

Later, Luo et al. suggested an alternative pump combiner design based on a tapered pump fiber. Here, one end of the Er-doped DC ZBLAN fiber with a diameter of the first cladding of 250 μm was simultaneously spliced to passive step-index InF₃-based passive fiber with a 125 μm cladding diameter and a multimode silica pump fiber taper. The splice of rare-earth doped DC ZBLAN and passive fluoroindate fibers was performed with the core alignment, which left ∼60-μm wide ring of the ZBLAN fiber cladding spare. Therefore, a tapered down to a waist diameter of 40 μm multimode pump fiber could be attached to the ZBLAN end face, as shown in Fig. 8(c).¹³⁶ The maximum efficiency reached 75%. However, the authors observed significant variability in coupling efficiency due to the bending of the tapered section of the silica fiber. The same methodology was used by N. Karampur et al., who successfully implemented such a pump combiner into an all-fiber laser with a ring resonator, generating radiation at a wavelength of 2.7 μm.¹⁴² 

Finally, a hybrid silica-fluoride side-polished fiber pump combiner has been demonstrated,¹³⁷ shown schematically in Fig. 8(d). Multimode silica and DC ZBLAN fibers were polished along the length for ∼10 mm and physically connected using index matching oil. In this study, coupling efficiency reached above 80% when pumped with 12 W at a 980 nm pump wavelength and demonstrated excellent stability over 8 h of continuous operation under active thermal control. An excess loss was 0.65 dB. However, the proposed pump combiner design cannot incorporate multiple pump ports with several pump sources in high-power laser systems.

Beam combiners and couplers combine the generated mid-IR radiation from several mid-IR lasers into a single fiber port or split it between several fibers with a controllable power ratio. Since these components are wavelength selective and, thus, have to be designed to operate at mid-IR signal wavelength, they comprise only one type of fiber, i.e., either fluoride or chalcogenide. This makes it favorable to adapt the conventional fiber heat-pull fabrication techniques.¹⁴³ 

The method for producing fluoride glass optical couplers was first proposed and described in 1989.¹⁴⁴ Similarly to the heat-pull method applied for silica fibers, a fused coupler is formed from two twisted optical fibers of fluoride glass by heating them in the slot of a temperature-controlled heater within a shroud containing an oxygen-free atmosphere. However, it took nearly 25 years and the increasing demand for compact and stable lasers operating in the mid-IR spectral range to decrease their excess losses, as discussed in the current section. It is worth mentioning that apart from fused coupler designs, more exotic ways of producing mid-IR fiber couplers and beam combiners have been introduced over the last decades. These include Y-extrusion of an unclad silver halide (AgCl_0.8Br_0.2) optical fiber preform,¹⁴⁵ shown in Fig. 9(a), or side-polishing of fluoroindate fibers, which ensures a controllable coupling ratio by managing the relative positions of the fiber cores.¹⁴⁶ However, the exposure of fiber cores, particularly transmitting laser irradiation to atmospheric water or other gas absorption bands, makes such configurations prone to degradation.

Despite the pioneer demonstration and more mature technology of fluoride fibers, the first fused fiber couplers and fiber arrays operating at the mid-IR wavelength range have been implemented using multimode chalcogenide fibers,¹⁵⁰ shown in Fig. 9(b). Thermal treatment of chalcogenide fibers demonstrated several technological differences when compared to conventional silica fibers. First, the fragility of the chalcogenide glass challenged the in situ twisting of the fibers during heating and tapering. Instead, the fibers were twined before clamping into the experimental system. The purging of nitrogen (or other inert gas) was essential for eliminating water vapors from the furnace chamber. After the fibers were heated to the point of turning dark, the tapering tension was applied to achieve the required coupling ratio. Reaching softening temperature in this way allowed for achieving low diffusion between core and cladding and, therefore, preserving the waveguiding properties of the fibers. The heating filament was immediately switched off as soon as the fusion was complete, allowing the resulting coupler to cool down to room temperature under the inert gas flow. The resulting couplers demonstrated an ∼3:1 splitting ratio, with the excess loss as low as 0.27 dB at 2.65 μm.

Due to the low melting temperature and a high coefficient of thermal expansion, the thermal treatment during the tapering procedure has been continuously improved. One of the fruitful techniques is the heat-brush approach, or flame-brush, as introduced in 1988.¹⁵¹ Here, a heat source, such as a flame, a CO₂ laser, or a resistive heater, locally heats a very short section of fiber. While traveling along the fiber axis, the heat source gradually modifies the fiber, “brushing” off the diameter of the taper within each sweep. This approach ensured the successful fabrication of quasi-single-mode As₂Se₃ couplers with coupling ratios ranging from 1% to 99% with little or large wavelength-dependence in the short-wave infrared range.¹⁵² 

Further research has been pursued to achieve low-loss fused couplers based on single-mode fluoride glass fiber. Nevertheless, implementing low loss came at the cost of the low coupling ratio. Therefore, Stevens et al. achieved only a 1% coupling ratio with 0.72 dB excess losses, measured at a wavelength of 2 μm.¹⁵³ The increase of the coupling ratio to nearly 50% led to a significant rise in excess loss, reaching 2.8 dB. Furthermore, achieving a more substantial coupling ratio resulted in a narrower taper waist of the fibers and, therefore, critical crystallization and mechanical weakness, leading to frequent fiber breakage during production. Figure 9(c) compares the fluoride fibers tapered in the standard and inter-gas atmosphere, confirming that water vapors and gas impurities act as nucleation sites for the crystallization process during thermal treatment.¹⁴⁸ 

In the following work, Anelli et al. successfully addressed the challenge of taper waist mechanical weakness by placing two fluoride fibers into a capillary.¹⁴⁹ Drawn together under the controlled heating, the fibers and capillary formed a single taper, demonstrating a minimal excess loss of 0.88 dB at 3.34 μm and a coupling ratio of 27.8:72.2. Despite this work having been performed using multimode fibers, such a technique has the potential to reduce the excess loss further when utilized for single-mode optical fibers. An alternative approach, as demonstrated in Ref. 148, relied on pre-fusing the side surfaces of two fibers under high heat for a short time prior to tapering. The resulting fiber couplers with 40%⋯45%:60%⋯55% final coupling ratios and 0.33–0.35 dB loss were measured using a broadband (450–5500 nm) tungsten light source.

Despite the imperfections of the fiber coupler and beam combiner fabrication technology, the very first prototypes have become an essential element for demonstrated monolithic all-fiber laser systems.^134,135,137 The demonstration of the fused ZBLAN fiber coupler with reasonably low insertion loss led to the proof of concept ring mid-IR laser sources.¹⁴² 

The prevalence of hybrid configurations of pump combiners, i.e., when they comprise silica and fluoride glass fibers, is due to the lack of commercially available pump diodes pigtailed with fluoride fibers. The integration of fluoride fibers into pump diode modules will solve the main technological challenge of optical fiber incompatibility for joint thermal post-processing. This will provide flexibility for pumping schemes, enabling in-core pumping via wavelength-selective fused couplers, i.e., wavelength division multiplexers (WDM). In the context of high-power or ultrafast fiber lasers, the side-coupling approach is more advantageous than fused couplers, as it does not modify the fiber core diameter and, therefore, does not affect the mode field. Therefore, side-pumping combiners are free from parasitic local nonlinear effects, fusion-induced insert loss, and beam quality degradation. Finally, it brings the advantage of unlimited pump ports or the possibility for pump recycling.¹⁵⁴ 

As discussed, the basic technique of fabricating fused fiber components involves heating with simultaneous stretching, creating a biconical-tapered junction. The conditions of fusion and pulling determine the coupling parameters and excess loss. In Ref. 155, the author notes that the success rate for tapering the coupler without breakage is ∼80%. While this rate is considered high, it still indicates the immaturity of the technology and numerous remaining challenges for fluoride and chalcogenide glass fibers. As mentioned earlier, fiber crystallization during heating and tapering poses a trade-off between tapered fiber waist, i.e., resulting coupling efficiency, and insertion losses. Partially, the excess loss in the fused coupler can be minimized by thermal processing in an oxygen-free environment. Current development in the field also includes modification and enhancement of the splicing and tapering instruments, particularly by introducing controllable low-power filament operation with high-power resolution, which results in finer control over the heating temperature, as well as precise fiber holding blocks to enable repeatable processing results.

An alternative to the all-fiber pump combiners and couplers that have been described in detail in Sec. IV are fiber-coupled (pigtailed) optical waveguide chips. In that case, the desired functionality is provided by a bulk glass chip that is then integrated into the fiber laser cavity to form a monolithic hybrid waveguide/fiber laser architecture. A very simple example of such a system is schematically shown in Fig. 10 for the case of a monolithic erbium ring laser. Here, a wavelength-selective 4-port optical waveguide chip is used as a pump coupler (100% through coupling for 980 nm pump light) and, at the same time, also as an output coupler with optimized through/cross-coupling ratio at the signal wavelength of 2.9 μm. It is important to note that the availability of such a relatively simple device opens up the possibility to implement more complex laser cavities incorporating multiple couplers like figure-eight¹⁵⁶ or theta cavities.^157,158

While it comes with the added difficulty of having to ensure efficient fiber/chip coupling, the major advantage of this hybrid chip/fiber approach is that it allows for a more flexible coupler design and, as a consequence, that it enables the fabrication of devices that can provide additional functionalities, e.g., the combination of multiple pumps using a single chip. This is in particular true for chips fabricated via femtosecond laser direct inscription. Since its introduction by Davis et al. in 1996,⁹⁶ the technique has been widely adapted and expanded to enable the realization of full 3-D device designs.¹⁵⁹ The basic principle is the same in the case of femtosecond laser-inscribed FBGs: Nonlinear absorption facilitates an exchange of energy from the electromagnetic field of a tightly focused inscription laser pulse to the bulk material, which can subsequently lead to a permanent and highly localized (to the focal volume) modification of the refractive index of the glass.¹⁶⁰ Translating the sample transverse to the laser beam then eventually results in the formation of a line (or track) of the modified index, i.e., an optical waveguide in the case of a positive index change. In order to inscribe optical waveguides with circular cross sections, either very strong focusing of high-repetition-rate laser pulses can be used in the cumulative heating approach,¹⁶¹ or multiple elongated tracks are stacked to form a symmetric multiscan waveguide.¹⁶² 

Commercial fluoride optical fibers have numerical apertures ranging from 0.125 to 0.23.⁴⁰ Therefore, a prerequisite for the fabrication of fiber pigtailed chips in mid-IR transparent fluoride or silica glasses with n ≈ 1.5 is the ability to femtosecond laser-inscribe optical waveguides with a very high index contrast in the range Δn = 5 × 10⁻³⋯2 × 10⁻². For waveguides inscribed into glasses with a higher refractive index like highly nonlinear chalcogenides, e.g., gallium lanthanum sulfide (GLS) glass with refractive index of ∼2.4,¹⁶³ this requirement is somewhat relaxed to Δn = 3 × 10⁻³⋯1 × 10⁻². However, a mismatch in the (linear) refractive index between fiber and chip introduces additional Fresnel losses unless anti-reflection coatings, e.g., moth-eye nanostructures,¹⁶⁴ are utilized. In Sec. V A, we will outline the progress made in the fabrication of waveguides in fluoride as well as ultra-pure silicate glasses and their application as pump- and output couplers in mid-IR fiber laser systems. After that, we will provide an overview of work that has been performed in GLS glass with the aim of implementing fiber-integrated ultrafast nonlinear absorber chips for passively mode-locked mid-IR fiber lasers.

As fluoride glasses are the backbone of modern mid-IR fiber technology, fluoride glasses are also the obvious choice of material for fabricating fiber-pigtailed optical chips in that wavelength range. Indeed, the first demonstration of utilizing femtosecond laser pulses for the fabrication of optical waveguides into fluoride glasses was published as early as 1999 by Miura et al.,¹⁶⁵ where a positive index change of 3 × 10⁻³ was reported. However, subsequent reports were scarce, and attempts to fabricate waveguides mainly resulted in the creation of negative index change zones, as discussed in Sec. III. As a solution, a structurally engineered approach known as the depressed cladding structure was developed. This technique involves concentrically writing a set of negative index change structures to form a guiding zone in the center. While this technique enabled the fabrication of miniaturized waveguide lasers with large mode-field diameters, which is beneficial for pulsed applications,¹⁶⁶ the complex geometry and the limited NA (Δn ≈ −1 × 10⁻³)¹²⁶ make this approach unsuitable for a hybrid fiber/chip geometry. Other efforts in producing larger V-number waveguides focused on modifying the glass composition, as summarized in Table I. The first successful attempt to push the index change into the required range of Δn = 5 × 10⁻³⋯2 × 10⁻² was achieved by substituting a certain fraction of zirconium atoms in the ZBLAN glass matrix with hafnium, which facilitates a laser-induced distortion in the electron cloud distribution with an associated significant change in polarizability and, thus, refractive index.³⁹ Most recently, a further increase of the femtosecond laser-induced index change to over 10⁻² has been reported by further optimizing the fluoride glass composition.¹⁶⁷ 

An alternative material that has been considered in the past for the same application is Suprasil, a high-purity synthetic fused silica glass.¹⁶⁹ While the intrinsic absorption loss in the mid-IR in this glass is substantially higher as compared to typical fluoride glasses, its internal transmission of ∼90%/cm for wavelengths between 3 and 3.4 μm is still acceptable for the fabrication of waveguide chips that are typically only a few millimeters long. By scanning a wide range of process parameters, a fabrication window that allows for the inscription of multiscan waveguides with Δn of as high as 1 × 10⁻² has been identified.¹⁷⁰  Figure 11(a) shows that femtosecond laser-inscribed waveguides in Suprasil, therefore, feature mode field profiles that are virtually identical to those supported by commercial high-NA fluoride fibers. In addition, Figs. 11(b)–11(d) demonstrate the possibility of achieving arbitrary through/cross-coupling ratios at a signal wavelength of 3.2 μm in those chips. Based on these results, the first experimental realization of the concept introduced in Fig. 10 has been recently presented.¹⁶⁷  Figure 12 shows a photo of the corresponding setup involving the femtosecond-laser inscribed glass chip with a total of four optical fiber pigtails connected to both ends.

As will be discussed in more detail in Sec. VI, incorporating a saturable absorber inside a laser cavity typically results in the formation of a train of optical pulses by suppressing unwanted continuous-wave (CW) background against noise peaks in the initial build-up phase.¹⁷¹ In order to generate femtosecond pulses, saturable absorbers with ultrafast recovery times are needed.¹⁷² While various materials that exhibit saturable absorption due to their intrinsic properties will be discussed in Sec. VI, artificial devices that utilize the quasi-instantaneous optical Kerr effect to mimic the effect of a saturable absorber can also be created. An example that is widely used in bulk lasers is Kerr-lens mode-locking,¹⁷³ whereas nonlinear polarization rotation has been used in fiber lasers.^28,174 One particularly elegant solution beneficial due to its monolithic integration is nonlinear-coupled waveguide arrays.¹⁷⁵ As depicted in Fig. 13, when the light at low to moderate peak powers is launched into the central waveguide (WG 3), the input signal gets distributed to the neighboring waveguides due to evanescent wave coupling, resulting in negligible transmission through the central waveguide. However, as the input peak power increases, nonlinear self-focusing eventually overcomes linear diffraction, and the signal remains in the central waveguide. The result is a typical saturable characteristic as shown in Fig. 13(b) for waveguide 3. Gallium Lanthanum Sulfide (GLS) is a chalcogenide glass that features a wide transmission window (0.5–10 μm), a high nonlinear refractive index n₂, good thermal stability, and low multi-photon absorption, making it a promising material for the realization of nonlinear-coupled waveguide arrays for monolithic mode-locked mid-IR fiber lasers. While it has been demonstrated already in 2015 that femtosecond laser-inscription in GLS can result in the formation of low-loss waveguides with moderate NA in both the cumulative heating as well as in the multiscan regime,¹⁷⁶ the effect of femtosecond laser waveguide-inscription on the nonlinear index has remained largely unexplored. Very recently, it was shown that waveguides written using the multiscan method preserve the high optical nonlinearity of the bulk glass to a very high degree, whereas those fabricated via cumulative heating show a substantial decrease in n₂.¹⁷⁷ The numerical simulation shown in Fig. 13(b) is, therefore, based on the experimentally measured refractive index profile, nonlinearity, and loss values of a femtosecond laser-inscribed multiscan waveguide at 3.4 μm. Future steps will include the fabrication and characterization of the full waveguide array and, eventually, its integration into a monolithic mid-infrared fiber laser cavity.

High peak powers, pulse energy, and ultrashort pulse durations make mid-IR ultrafast fiber lasers even more attractive tools for material processing and spectroscopy. Typically, active and passive ultrafast modulators, which introduce intensity-selective losses in the laser cavity, are employed in the laser cavities to initiate pulse generation.

The first demonstration of pulsed generation in the mid-IR wavelength range is dated back to 1996, when an InAs saturable absorber was used to achieve passive Q-switching and potentially mode-locking in an Er-doped ZBLAN fiber laser.¹⁷⁸ The first mode-locked operation was demonstrated only in 2012 using Fe²⁺:ZnSe crystal as a saturable absorber.¹⁷⁹ Since then, various materials have been investigated to enable pulse generation in the mid-IR wavelength range, including the exploitation of properties of the gain fiber itself for gain-switching^180–182 or as a saturable absorber.¹⁸³ 

A SESAM represents a quantum well structure grown typically via molecular beam epitaxy on a Bragg reflector. In the context of the mid-IR operation wavelength range, InAs,¹⁸⁴ GaAs,^185–187 and GaSb^188,189 present potential alternatives to golden standard In_xGa_1−xAs-based SESAMs in the near-infrared range. While GaAs-based SESAMs are more mature and feature a modulation depth of 18%, 15% unsaturated loss, over 60% reflectivity, a saturation fluence of 70 J/cm² at 2.73–2.87 μm, and a damage threshold of 350 MW cm⁻²,¹⁸⁶ their relaxation time exceeds 1 ps and can reach as high as 10 ps. At the same time, GaSb-based SESAMs have the potential to provide much faster recovery time and reduced saturation fluence. While the measurements were performed only at a shorter wavelength of 2.4 μm, Sb-based SESAM can demonstrate relaxation time below 1 ps.¹⁹⁰ The linear reflectivity measurement allowed estimating its modulation depth as ∼14% at 2.76 μm.¹⁸⁹ 

One of graphene’s key advantages is constant absorption (of 2.3% of the vertically incident light, defined by the fine-structure constant) over a broad wavelength range. The ultra-broadband transmission of single-layered graphene is due to the semi-metallic band structure, where conduction and valence bands touch at the Dirac points.¹⁹¹ Still, the modulation depth of single-layered graphene is relatively low and further reduces with the extension to longer operation wavelengths. Increasing the sample thicknesses in multi-layered saturable absorbers offers an effective strategy for boosting the modulation depth and ensuring a better quality of the generated pulse. The exploration of more extended wavelength applications has started with the demonstration of mode-locking in Cr²⁺:ZnSe laser at 2.5 μm,¹⁹² shortly followed by the report on Q-switching in Er-doped ZBLAN fiber laser at 2.78 μm.¹⁹³ The nonlinear absorption of the graphene-saturable absorber with 4–6 layers was investigated in Ref. 194. In this work, the recorded modulation depth and saturation power intensity were around 10% and 2 MW/cm², respectively.

Finally, a new class of low-dimensional material saturable absorbers has been discovered—MXenes,¹⁹⁵ which are transition metal carbides, carbonitrides, and nitrides with a chemical formula of M_n+1X_nT_x (n = 1–3). Here, M stands for an early transition metal (Sc, Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, etc.), X—for carbon and/or nitrogen, and Tx—surface terminations (F, O, OH, etc.).¹⁹⁶ Among the benefits of MXenes are broadband operation range from visible to mid-IR, large modulation depth, tunable bandgap, outstanding photothermal effect, low optical absorption, and good thermal conductivity. Specifically, Ti₃C₂T_x demonstrated a comparable graphene negative nonlinear refractive index n₂ of approximately −10⁻²⁰ m²/W. Coupled with controllable saturation fluence via the thickness of the film and relatively high optical damage threshold compared to other 2D materials of 70 mJ/cm², it makes the performance of the saturable absorber very promising. Nonlinear absorption measurements revealed a modulation depth of 33.2%, nonsaturable loss of 25%, and saturation intensity of 43 MW/cm² at 2866 nm for the 600-nm thick sample.¹⁹⁷ 

(CNTs) are a direct bandgap material featuring a series of van-Hove singularities in the density of state. The bandgap and, therefore, the absorption band of the CNT saturable absorber are determined by its diameter.¹⁹⁸ To set an absorption peak position in the mid-IR wavelength range, careful growing procedures are required to produce tubes with large diameters (above 1.5 μm) and high aspect ratios. In this regard, the chemical vapor deposition method is the most promising technique.^199,200 The nonlinear response of CNTs investigated at 2.8 and 3.5 μm wavelengths demonstrated correspondingly 16.5%.

Similarly to graphene, recently emerged black phosphorus presents a single-component material with the atoms bonded covalently with three others, forming a puckered honeycomb structure.²⁰¹ The direct energy bandgap structure of black phosphorus depends on the number of layers, which allows for enhancing its absorption at longer wavelengths. Up until now, several works demonstrated passive Q-switching and mode-locking using black phosphorus in Er-^202,203 and Dy-doped mid-IR fiber lasers.²⁰⁴ 

Transition metal dichalcogenides, such as MoS₂, MoSe₂, WSe₂, WS₂, and others, have demonstrated quite limited performance at the mid-IR wavelength range due to their relatively large direct bandgaps corresponding to the visible to near-IR spectrum range. Only with the incorporation of some suitable defects can their operation range be extended to longer wavelengths but not beyond $∼2μ$m.

On the contrary, a narrow bandgap is characteristic of another class of Dirac materials—topological insulators (TIs), such as Bi₂Te₃, Bi₂Se₃, and Sb₂Te₃. Furthermore, these materials feature a gapless metallic state in their edges.²⁰⁵ These properties make topological insulators a promising saturable absorber for the mid-infrared spectral regions. The small bandgap, combined with its merits of large modulation depth and high damage threshold, makes it feasible as an effective SA. Recently, Q-switching with the aid of TI has been successfully performed in fiber lasers around 3 μm. However, it has the drawback of a complex preparation process due to the compound with two different elements.

The current limitation of ultrashort pulse generation in the mid-IR wavelength range is the relatively slow recovery time of common materials-based saturable absorbers. While the artificial modulators based on the optical Kerr effect feature fast response, they currently still suffer from the application of free-space components. The wavelength range spanning from 2.7 to 5 μm is a fingerprint region for the majority of greenhouse and other atmospheric gases,^206,207 which limits the width of the pulse spectrum and, as a consequence, pulse duration.²⁹ The same drawback refers to material saturable absorbers deposited onto laser mirrors. Furthermore, the realization of the majority of mature and newly emerging artificial modulators, such as nonlinear loop mirrors and multi-modal interference, has yet to be realized due to the lack of fiber-based laser components. We foresee that the progress described in this work toward the development of all-fiber laser configurations will bring flexibility to designing saturable absorbers.

From the perspective of basic laser power generation, it can be concluded from the preceding discussions that orders of magnitude power enhancements for single transverse mode operation in fiber lasers in the 2.7–4 μm wavelength range are within sight. The commercialization of the first mid-IR fiber lasers and their integration into innovative photonics systems confirm that such sources can impact a variety of areas. The commercialized application areas already cover polymer welding, surface texturing and patterning, patterning of thin films, semiconductor micro-processing, and medical treatment, particularly in dermatology, replacing Er:YAG lasers. Therefore, further extension of mid-IR laser development will boost not only the direct fields of their applications but also adjacent areas, such as electronics for designing fast and sensitive tools for mid-IR laser beam characterization. In this context, all-fiberized laser system configuration gradually becomes the key technical aspect missing on the way to further improving compact high-power mid-IR systems.

Despite the first successful laboratory demonstrations of integrated mid-IR components, their flexible design and mass production present an ongoing research and engineering challenge. Further enhancement of fiber components in terms of their loss reduction, parameter controllability, and repeatability of the fabrication process largely depend on the continuous optimization of optical post-processing platforms. In the current Perspective, we have discussed the current state-of-the-art and outlook for further development of couplers and beam combiners based on fiber or chip designs, fiber gratings, and optical fibers, in general, as essential components whose properties ensure distributed beam and pulse shaping. It is important to mention that the most significant breakthroughs in fiber grating inscription, coupler, and beam combiner fabrication discussed in this work have been achieved within the past decade, manifesting the vigorous development of the field. While most reports do not fully integrate demonstrated fiber-based components into systems, ongoing advances in manufacturing processes are expected to propel their further development.

Still, the full-scale flexibility of the laser beam control and ultrashort pulse dynamics management requires more elaborated designs and, therefore, sophisticated components, such as but not limited to tunable filters, optical circulators, mode adapters, cladding power strippers, and elements for polarization control. Further development must also include the enhancement of geometric structure and doping components engineering of fluoride and chalcogenide glass fibers for gain as well as nonlinearity and dispersion management, for example, designing chirally core, diameter oscillating, or large-mode-area fibers. It is important to mention the urgent need for the development of further essential components, such as

• Fiber-based isolators represent passive magneto-optical devices based on the Faraday effect. An isolator transmits light in only one direction, preventing the seed source or the generated beam from being affected by parasitic back-reflection or counter-propagating amplified spontaneous emission. As the typical fiber isolator employs bulk magneto-optical Faraday rotators and free space optics for light collimation, they have not yet been assembled for the mid-IR wavelength range. The recently emerged promising solutions for chip-based devices offer an alternative approach by breaking the reciprocity of waveguiding properties^208–210 or exploring magnetless Faraday isolators.²¹¹ 

• Fiber endcaps are required to prevent end-face damage at high powers. Particularly, fiber edge protection is essential for fluoride fiber systems operating at ∼2.8 μm as fibers are prone to water diffusion. Pure fluoride fiber endcaps (ZrF₄ and AlF₃-based) have shown mediocre enhancement of the tip durability, withstanding only 0.1 and 10-h operation of 20-W laser at ∼3 μm, respectively. At the same time, endcaps based on fluorogermanate (GeO₂–ZnO–PbO–K₂O–PbF₂), sapphire single crystal (Al₂O₃), and silica glass fibers could operate for 100 h, promising potential for power upscale up to 100 W in the case of GeO₂ endcaps.⁵⁶ Still, pure fiber endcaps suffer from water diffusion. Nanoscopic thin films, e.g., based on silicon nitride, can offer better protection against it, extending laser powers and operation time to a few months. Nevertheless, the high surface stresses have to be considered during deposition, counterbalancing with the film thickness or the introduction of oxygen in the matrix.

With the enhanced structure and combination of mid-IR glass fibers, their functionalization, and fiber-based component design, the next generation of mid-IR fiber lasers will be empowered with multi-domain flexibility of generation regimes while ensuring high-quality and high-power output. The largely unexplored field of mid-IR all-fiber laser technology offers a rich landscape of opportunities to harness multi-modal degrees of freedom of generated light, enabling advanced control and characterization of multi-dimensional spectral, spatial, and temporal interactions.

The authors have no conflicts to disclose.

Kirill Grebnev: Data curation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Boris Perminov: Data curation (equal); Methodology (equal); Writing – original draft (equal). Toney T. Fernandez: Data curation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Alex Fuerbach: Conceptualization (equal); Formal analysis (equal); Methodology (equal); Project administration (equal); Resources (equal); Writing – original draft (equal); Writing – review & editing (equal). Maria Chernysheva: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

Data sharing is not applicable to this article as no new data were created or analyzed in this study.