Review of mid-infrared mode-locked laser sources in the 2.0 μm–3.5 μm spectral region Open Access

Ultrafast mid-infrared (mid-IR) laser sources have recently attracted tremendous interest due to their promising applications in molecular spectroscopy, mid-IR supercontinuum generation, mid-IR frequency combs, material processing, laser surgery, biodiagnostics, etc.^1–3 Passive mode locking is a commonly used, straightforward method to generate ultrashort laser pulses in various spectral regions. However, limited by the availability of gain materials and saturable absorbers (SAs) in the mid-IR region, passively mode-locked ultrafast mid-IR laser sources were seldom available before the early 2000s. In order to obtain mid-IR ultrashort pulses, the most common approaches were nonlinear conversion processes such as optical parametric oscillation,⁴ optical parametric amplification,^5,6 and four-wave mixing.^7–9 However, compared with the mode-locked lasers, especially the diode-pumped mode-locked lasers, these nonlinear approaches are usually very expensive and complex, and require precise optical synchronization, which prevents their widespread implementation. Thus, the development of high-efficiency, low-cost, compact mode-locked lasers with picosecond or femtosecond pulse durations, diffraction-limited beam quality, and high average power in the mid-IR region has been the goal pursued by laser researchers.

Over the past decade, benefiting from the rapid advances in material research, various mid-IR laser gain materials with excellent optical and mechanical properties have been developed with the aim to generate mid-IR ultrashort pulses.^10–14 And significant developments have also been achieved in the mid-IR mode-locked laser sources, driven by new materials, technologies, and continued innovations in laser research field. Nowadays, less than 10-optical-cycle and even ∼3-optical-cycle ultrashort pulses have been directly generated in mode-locked laser oscillators around 2.0 μm and 2.4 μm regions, respectively,^15,16 while the wavelength of the mode-locked lasers has been extended to ∼3.5 μm.¹⁷ These newly developed mid-IR ultrafast laser sources, together with ongoing progress in related technologies, can open new research avenues and offer novel opportunities for applications in scientific research, industry, and other fields. In this work, we review the progresses of mid-IR mode-locked lasers in the last decade, including all-solid-state and fiber mode-locked laser sources that operate around the 2.0 μm, 2.4 μm, and 2.8 μm–3.5 μm spectral regions.

The structure of this review is as follows. In Sec. II, the current main mid-IR mode-locking techniques are introduced. Section III presents and discusses the recent progress in mid-IR passively mode-locked lasers, including Tm³⁺-, Ho³⁺-, and Tm³⁺/Ho³⁺-doped mode-locked all-solid-state and fiber lasers around 2.0 μm, Cr²⁺:ZnSe and Cr²⁺:ZnS lasers around 2.4 μm, and recently developed Er³⁺-, Ho³⁺/Pr³⁺-, and Dy³⁺-doped fluoride fiber lasers in the 2.8 μm–3.5 μm region. In Sec. IV, some selected applications of ultrafast mid-IR lasers are presented to show the statuses and prospects in diverse application areas. Section V provides a summary and outlook regarding the development of mid-IR mode-locked laser sources.

Unlike active mode locking that typically produces longer (∼100 ps) pulses, passive mode locking has the ability to generate pulses as short as a few femtoseconds. In passive mode locking, a saturable absorption effect is typically required to modulate losses inside the laser cavity. Therefore, continuous-wave (CW) and low intensity laser radiation suffer non-negligible absorption losses, while short pulses with high peak power saturate the absorber and achieve low losses. Meanwhile, individual short pulses are shaped because the high-intensity portion at the pulse peak saturates the absorber more than its low-intensity wings, further shortening the pulse width in the time domain. With current technology, saturable absorption can be induced either by real SAs such as semiconductor saturable absorber mirrors (SESAMs) and various low-dimensional materials or via nonlinear optical effects such as the Kerr-lens effect, nonlinear polarization rotation (NPR), and nonlinear optical/amplifying loop mirrors (NOLM/NALM). These real SAs and nonlinear optical effects in passive mode-locked lasers tend to exhibit very fast (femtosecond to picosecond) response times, which aid in ultrashort pulse generation. Although they are similar to those used in near-infrared (near-IR) mode-locked lasers, passive mode-locking techniques for mid-IR lasers have several different characteristics and requirements, which are introduced in this section.

SAs are key to achieving picosecond and femtosecond pulse operation in both all-solid-state lasers and fiber lasers at various wavelengths. The SESAM was first demonstrated as a mode locker in an all-solid-state Nd³⁺:YLF laser in 1992.¹⁸ Since then, it has been widely used in various passively mode-locked lasers and is considered a revolutionary technology for ultrashort pulse generation. In particular, the broad design freedom associated with SESAMs provides compelling advantages such as flexible, precise parameter customization and robust functionality that help to satisfy specific laser requirements. SESAMs enable improvements in the output characteristics of various ultrafast lasers in different wavelength regions. However, due to lattice mismatch and growth-related issues,¹⁹ commercial GaAs-based SESAMs in the mid-IR region have not yet reached the same maturity level as those used in the near-IR region. Their relatively poor saturable absorption properties, especially relatively long absorption recovery times, inevitably degrade the laser mode-locking performance in the mid-IR region. In order to obtain SESAMs with better capabilities for mid-IR mode-locking, new GaSb-based SESAMs with appropriate postprocessing techniques, such as ion irradiation/implantation and incorporation of quantum wells or quantum dots, have been explored.^20–22 Compared with standard GaAs-based devices, GaSb-based SESAMs can achieve better lattice-matching, exhibit lower nonsaturable losses, and offer significantly shorter recovery times. They have been used successfully in several mid-IR mode-locked all-solid-state and semiconductor disk lasers for picosecond and even sub-10 optical-cycle femtosecond pulse generation.^15,19–30 Nevertheless, the limited operation wavelengths of GaSb-SESAMs hinder their applications in mid-IR mode-locked lasers beyond 3 μm.^31,32

In addition to continued SESAM improvement and optimization, significant efforts have been made to exploit SAs based on novel materials. This has been done in order to develop reliable SAs with broader operating wavelengths (e.g., beyond 3.0 μm) and saturable absorption characteristics that are comparable to or better than those of current SESAMs. Newly emerged low-dimensional and bulk materials such as carbon nanotubes (CNTs),^33–41 graphene,^42–53 topological insulators (TIs),^54–58 transition metal dichalcogenides (TMDCs),⁵⁹ black phosphorus (BP),^60–62 and topological Dirac semimetals⁶³ have recently attracted significant attention and been extensively explored for mid-IR mode-locked lasers. Their advantages of broadband absorption region, ultrafast recovery time, low saturation fluence, and easy fabrication and integration make them potential candidates as SAs for various mid-IR lasers. However, some nonnegligible problems such as poor production repeatability and performance degradation in long-term applications remain to be solved.

Carbon-based low-dimensional materials, particularly CNTs and graphene, have shown significant promise as SAs that can generate ultrashort laser pulses at various wavelengths, including in the mid-IR region. Their unique electronic structures endow them with interesting linear and nonlinear optical properties,^64–68 including broadband operating regions, large optical nonlinearities, ultrafast carrier relaxation times, and low saturation intensities. These properties make them attractive SESAM substitutes for ultrashort mid-IR pulse generation. CNTs are direct bandgap materials with excellent chemical stabilities and high optical damage thresholds. Their bandgaps depend on their diameters and chiralities.⁶⁴ Natural mixtures of CNTs with various diameters and chiralities can provide a broadband operation region that covers approximately 0.8–2.4 μm and can be used as SAs for Tm³⁺-, Ho³⁺-, and Cr²⁺-doped mid-IR mode-locked lasers. Graphene, a two-dimensional allotrope of carbon atoms with a honeycomb lattice, has a zero bandgap and a linear energy dispersion in momentum space.⁶⁹ These properties make it applicable as a SA for mode locking over an ultrawide spectral range from the visible to the mid-IR region. Compared with CNTs, graphene exhibits additional advantages of broader operating bandwidth, uniform linear and nonlinear optical response, and easy integration into photonic components and systems. Ultrafast pulse generation around the mid-IR regions of 2.0 μm,⁴² 2.4 μm,⁴⁷ and 2.8 μm⁵³ has been demonstrated with graphene SAs. In addition, the fascinating electronic and optical properties of these carbon-based low-dimensional materials have created new opportunities for optoelectronic applications and triggered an explosion of research to explore other novel low-dimensional materials for mid-IR ultrashort pulse generation.

Over the past few years, TIs such as Bi₂Te₃, Bi₂Se₃, and Sb₂Te₃ have also emerged as promising broadband SA candidates for the near-IR to mid-IR regions.^54–58 TIs represent a new class of electronic material in which the bulk state has a bandgap like that of an ordinary insulator but the surface or edge state is topologically protected conducting states and exhibits a Dirac cone-like electronic band structure.^70,71 Due to a combination of spin–orbit interaction and time-reversal symmetry, TIs exhibit unique electronic and optical properties such as the quantum spin Hall effect,⁷² topological magnetoelectric effect,⁷³ high nonlinear refractive indices,⁷⁴ and ultrafast relaxation times.⁷⁵ This new type of Dirac material was first successfully demonstrated as a SA in a passively mode-locked fiber laser around 1550 nm in 2012.⁷⁶ Since then, several trials have been conducted by various groups to investigate their nonlinear absorption features in pulsed laser systems in the near-IR region. Due to Dirac-like linear dispersions which are similar to that of graphene, TIs can provide ultrabroadband operating bandwidth up to the mid-IR region. In addition, investigations of Bi₂Se₃, Bi₂Te₃, and Bi₂SeTe₂ indicate high modulation depths and relatively small nonsaturable absorption losses.^74,77,78 With these unique properties, TIs are attractive SAs for mode-locked pulse generation in the mid-IR region.

TMDCs are semiconductors with MX₂ stoichiometry, where M is a transition metal atom (e.g., Mo or W) and X is a chalcogen atom (e.g., S, Se, or Te). They have recently been demonstrated as SAs for mid-IR mode-locked lasers.^79–85 TMDC bulk crystals are bonded together by weak van der Waals force and can be separated into monolayer and few-layer forms via mechanical exfoliation.⁸⁶ Unlike graphene, TMDCs exhibit layer-dependent bandgaps that generally range from direct in monolayer to indirect in the bulk form.⁸⁷ This creates new opportunities for photonic and optoelectronic applications in various wavelength regions.^88–90 In addition, the optical properties of ultrafast carrier dynamics and high nonlinear optical response discovered in monolayer and few-layer TMDCs further support their development as SAs for mode-locked lasers. However, the optical bandgaps of most traditional 2H-phase TMDCs typically appear in the visible to near-IR spectral range due to interband optical transitions.⁸⁸ They are generally not suitable for mid-IR photonic applications. Recently, Wang et al. demonstrated that wideband saturable absorption extending to the mid-IR region can be achieved in few-layer TMDCs by introducing suitable defects.⁹¹ This finding provides a feasible approach for the use of TMDCs in mid-IR mode-locked lasers. Meanwhile, some novel TMDC materials and new absorption mechanisms (e.g., sub-bandgap optical absorption) have been proposed and developed to scale the operating band into longer (3.0 μm to 10.0 μm) wavelength ranges.^92–94 This significantly extends the application of these materials to mid-IR ultrafast laser technology.

Like TMDCs, BP is a layered material in which individual atomic layers are stacked via van der Waals interactions. It also demonstrates strong nonlinear absorption in various spectral regions. Monolayer BP is a semiconductor with a puckered honeycomb structure in which one phosphorus atom is covalently bonded to three adjacent phosphorus atoms.⁹⁵ Few-layer black phosphorus has a direct bandgap,⁹⁶ high carrier mobility (up to ∼1000 cm² V⁻¹ s⁻¹),^96–98 and a large on/off ratio (>10⁵) at room temperature.^96,99 These characteristics make it a promising two-dimensional semiconductor material for electronic and optoelectronic applications. In addition, theoretical and experimental investigations have shown that the direct bandgaps of the BP materials vary from ∼2.0 eV for a monolayer to ∼0.3 eV for bulk material due to layer-layer coupling,^100,101 and this bandgap can be controlled by tuning the number of stacked layers. Such a layer-dependent direct bandgap can lead to a broad range of applications in photonics in the ∼0.6 μm to ∼4 μm region. Recently, broadband nonlinear optical absorption from 0.8 μm to 2.1 μm has been examined by Wang et al.,¹⁰² showing a clear wavelength-dependent nonlinear optical response. The combination of strong saturable absorption and a broadband operation bandwidth makes BP a promising candidate for ultrashort pulse generation across the visible, near-IR, and mid-IR spectral regions.

In addition to using real SAs, one can also exploit artificial SAs to generate saturable absorption effects. The most prominent artificial saturable absorption effect applicable to mode locking is the Kerr-lens effect, which is induced by Kerr nonlinearity.¹⁰³ The physical origin of Kerr nonlinearity is nonlinear polarization generated in the medium under high electric field intensity. This phenomenon changes the refractive index inside the medium in response to the intensity of the applied electric field. Considering that the laser beam power densities of mode-locked lasers typically conform to Gaussian distributions, the refractive index change induced by the beam is greater in the center than around the edge, and thus produces a self-focusing effect referred to as the Kerr-lens effect. Since Kerr-lens mode locking was first demonstrated in a Ti:sapphire laser in 1991,¹⁰⁴ it has become a common ultrashort-pulse generation technique in the near-IR spectral region. However, KLM is more challenging to achieve in the mid-IR wavelength range than in the near-IR because the magnitude of the Kerr-lens effect in the cavity is inversely proportional to the cavity mode area. Since the cavity mode radius is proportional to the square root of the laser wavelength, the Kerr-lens effect is not usually sufficient for KLM in mid-IR lasers, except for lasers whose gain materials have particularly high nonlinear refractive indices, such as Cr²⁺:ZnSe and Cr²⁺:ZnS. KLM remains an attractive mode locking technology for ultrashort few-cycle pulse generation in the mid-IR region because it offers extremely fast response time in the few-femtosecond range and ultrabroadband wavelength operation even though critical cavity alignment and initiation via external perturbations are usually required.

NPR is a commonly used artificial SA for mode-locked fiber lasers. Its saturable absorption effect arises from nonlinear phase delay-induced light polarization rotation. Linearly polarized light is converted to elliptically polarized light by a quarter waveplate before entering the fiber. This elliptical polarization can be treated as a combination of two orthogonal polarizations with unequal intensities. Self-phase modulation in the fiber introduces a certain amount of nonlinear phase delay between the two orthogonal polarizations, resulting in intensity-dependent polarization rotation within the fiber. Nonlinear amplitude modulation is achieved by rotating a half waveplate to select the high-intensity pulse peak through a polarizer while blocking low-intensity pulse edges and CW waves. NPR can not only initiate the mode-locking operation but also narrow the pulse for achieving shorter pulse generation. Since its first demonstration, NPR has been extensively applied to near-IR mode-locked fiber lasers.^105,106 With the development of mid-IR fiber lasers, particularly fluoride fiber lasers that operate beyond 3 μm where commercial InGaAs-based SESAMs are unavailable, NPR has become increasingly essential due to its wavelength independence and simple structure. Currently, NPR is a common way of generating femtosecond mode-locked pulses in 3.0 μm fluoride fiber lasers where traditional SAs and pulse generation approaches are limited.

Another simple way to realize mode locking in fiber lasers is to implement a nonlinear optical loop mirror (NOLM), which provides intensity-dependent transmission via interference between two counter-propagating waves in a fiber loop. A laser beam is fed into one input port of an X fiber coupler and split into two waves that counterpropagate in the fiber loop, which return back to the coupler and interfere with each other. Transmission to the output port is determined by the interference condition, e.g., the phase delay between the two counterpropagating waves. If a coupler with unequal coupling is employed, the nonlinear phases induced by self-phase modulation are different for the two counterpropagating waves, resulting in a nonlinear phase delay. Since the amount of the nonlinear phase delay depends on the laser intensity, intensity-dependent transmission is formed and the system can act as an artificial fast SA for mode-locked lasers. By introducing an amplification fiber section at one end of the fiber loop in NOLM, a nonlinear amplifying loop mirror (NALM) is formed. It increases the difference between the accumulated nonlinear phase shifts of the two counterpropagating waves and, thus, significantly reduces the power required to initiate mode locking. Thus far, NOLM and NALM have been widely used in mode-locked fiber lasers that cover the 1.0 μm to 2.0 μm spectral region.^107–111 With the development of fiber components and related technology, particularly the fluoride fiber-based X coupler and related splicing, there is potential to extend NOLM and NALM operating wavelengths to the longer mid-IR region.

The trivalent rare earth ions Tm³⁺ and Ho³⁺ are the most well-known elements that provide emissions for high-power CW and pulsed laser operation in the ∼2.0 μm wavelength range. Various types of high-efficiency and high-power laser emissions have been successfully demonstrated in a variety of host crystals, ceramics, and fibers around this “eye-safe” spectral region since the first Tm³⁺- and Ho³⁺-doped lasers were reported in the 1960s.¹¹² In the pulsed operation regime, generation of mode-locked pulses at room temperature around 2.0 μm was first demonstrated in the Cr³⁺/Tm³⁺:YAG and Cr³⁺/Tm³⁺/Ho³⁺:YAG bulk lasers via an acousto-optic modulator in 1991, and 45 ps and 800 ps pulses were obtained, respectively.¹¹³ At almost the same time, 35 ps pulses were also achieved in an acousto-optic modulator mode-locked Tm³⁺:YAG laser with an average output power of 70 mW at 2.01 μm.¹¹⁴ A few years later, a milestone in the generation of ultrashort pulses around 2.0 μm was accomplished with a Tm³⁺-doped fiber ring laser that was passively mode-locked via additive-pulse mode locking,¹¹⁵ which produced 350–500 fs pulses at ∼1.8–1.9 μm. Since then, many efforts have been made to produce ultrashort pulses in the 2.0 μm spectral region using various lasers. In this section, recent progress in mode-locked all-solid-state and fiber lasers around 2.0 μm is reviewed.

Although mode-locked pulses with 190 fs pulse width were produced from a Tm³⁺-doped silica fiber laser as early as 1996,¹¹⁶ only active mode locking with tens to hundreds of picosecond pulses output was achieved in all-solid-state lasers in the following decade.^117–121 For passive mode locking, which can generate ultrashort picosecond and femtosecond pulses, Z-folded or X-folded resonators were commonly used to achieve suitable laser mode sizes in the laser gain material and on the SA. However, due to the relatively low stimulated emission cross sections of the Tm³⁺- and Ho³⁺-doped bulk laser gain media, it is not a trivial task to suppress Q-switching instabilities and achieve stable CW mode-locking operation in 2.0 μm all-solid-state lasers. Another challenge to achieve mode locking is providing reliable SA, which is the key element and plays a crucial role in generating ultrashort mode-locked pulses.

Among various SAs mentioned in Sec. II, SESAM is the most commonly used one for passively mode-locked lasers. However, it was not mature enough to support stable mode locking in solid-state lasers in the 2.0 μm spectral region until the late 2000s due to the relatively poor performances in absorption recovery time, nonsaturable loss, and damage threshold. In 2009, an InGaAsSb-based quantum well SESAM was developed, in which GaSb/AlAsSb distributed Bragg reflectors and Te-doped GaSb substrates were adopted to avoid lattice mismatch, significantly improving the SESAM quality around the operating wavelength of 2.0 μm. With this optimized SESAM as the mode locker, Lagatsky et al. achieved passive mode locking in a Tm³⁺/Ho³⁺-doped KY(WO₄)₂ crystal laser.²³ Short pulses with a duration of 3.3 ps and an average power of 315 mW were produced at 2057 nm. Subsequently, transform-limited 570 fs pulses were achieved with a lower output power of 123 mW in the same SESAM mode-locked laser by introducing a pair of fused silica prisms to compensate for intracavity dispersion.²⁰ Since this first femtosecond operation of a mode-locked bulk laser was demonstrated, significant efforts have been made to generate ultrashort pulses with a shorter pulse duration toward less than 100 fs directly from mode-locked all-solid-state lasers around 2.0 μm. By using a Tm³⁺/Ho³⁺:NaY(WO₄)₂ disordered crystal which has broader and flatter emission spectra, 191 fs transform-limited mode-locked pulses around 2060 nm were produced with an 82 mW average output power.²⁴ Later on, utilizing the smooth and broad gain of Tm³⁺-doped sesquioxide materials, Lagatsky et al. further reported soliton pulses as short as 105 fs centered at 2010 nm from a Tm³⁺:LuScO₃ crystal laser in 2013. This represented considerable progress in the development of mode-locked all-solid-state lasers in the 2.0 μm spectral region.¹²² After that, passive mode locking was achieved in various Tm³⁺-, Ho³⁺-, and Tm³⁺/Ho³⁺-doped bulk laser materials with SESAM and semiconductor saturable absorbers as mode lockers. Although various mode-locked pulse durations from femtoseconds to dozens of picoseconds were achieved (Table I), no shorter mode-locked pulses were produced for nearly four years. Until 2018, the aim to break through sub-100 fs SESAM mode-locked pulses was successfully achieved by Wang et al. from a Tm³⁺:(Lu_2/3Sc_1/3)₂O₃ mixed ceramic laser.¹²³ This milestone was achieved using a Ti:sapphire pump laser source and an InGaAsSb quantum well-based SESAM. Five chirped mirrors were implemented for precise intracavity dispersion management, as shown in Fig. 1. Sub-10-optical-cycle (63 fs) pulses were generated at 2.06 μm for the first time, which are the shortest pulses obtained from the all-solid-state lasers around 2.0 μm up until now. The obvious Kelly sidebands on the mode-locked spectrum shown in Fig. 1(c) demonstrate the soliton mode locking operation of the laser. In the soliton mode-locking regime, the obtained shortest pulses are mainly attributed to the balance between strong nonlinear self-phase modulation and net negative cavity dispersion, as well as the laser gain, losses, and gain bandwidth,^124–126 rather than the SESAM which typically acts as a slow SA and is only used to start and stabilize the soliton pulses.

It is worthwhile to note that most of the 2.0 μm passively mode-locked all-solid-state lasers, particularly those with femtosecond operation, were pumped using Ti:sapphire lasers due to their good beam quality and high brightness. These high-brightness pump sources can well match a tightly focused laser mode. This helps to suppress a tendency toward Q-switched mode locking instability and aids in achieving stronger self-phase modulation inside the gain medium, benefitting for obtaining desired stable CW mode-locked operation and very short mode-locked pulses.¹²⁷ However, the high costs and bulky volumes of these pump sources represent major obstacles to development of practical 2.0 μm mode-locked ultrafast lasers and significantly limit their extensive applications. With the rapid development of AlGaAs-based laser diodes (LDs) in recent years, direct pumping of Tm³⁺- and Tm³⁺/Ho³⁺-doped materials with compact, low-cost, efficient commercial LDs near 790 nm has become very attractive, even though it will be more challenging to suppress Q-switched mode locking and obtain shorter pulses than previous works pumped by the Ti:sapphire laser system. In 2010, LD-pumped passively mode-locked operation around 2.0 μm was achieved in a Tm³⁺:GdLiF₄ laser with a transmission-type InGaAs quantum well SA. It generated 17 ps mode-locked pulses with an average power of only 38 mW.¹²⁸ Two years later, Ma et al. presented the first diode-pumped femtosecond operation of a 2 μm mode-locked solid-state laser using a Tm³⁺:CLNGG disordered crystal as the gain medium.¹²⁹ The laser generated stable 479 fs ultrashort mode-locked pulses with average output power as high as 288 mW at 1994 nm. The pump source was a 790 nm single-emitter AlGaAs LD, and mode locking was performed using a commercial SESAM. Since then, several LD-pumped mode-locking results with various laser gain materials and SAs in the 2.0 μm spectral region have been reported.^{42,43,46,59,130–133} Kong et al. reported the first dual-wavelength synchronously mode-locked laser around 2.0 μm in 2015, as shown in Fig. 2(a). Benefitting from the dual-wavelength emission of a Tm³⁺:CaYAlO₄ disordered crystal, the mode-locked laser delivered temporally synchronous dual-color pulses at 1958.9 nm and 1960.6 nm [Fig. 2(b)].¹³⁰ Optical beating between the dual-color pulses was clearly observable in the form of beating pulses with 3.5-ps duration and 0.13-THz repetition rate [Fig. 2(c)]. The dual-wavelength synchronously mode-locked laser provides a potential laser source for efficient generation of coherent terahertz radiation via optical difference frequency. Given that a Tm³⁺:CaYAlO₄ disordered crystal has a broadband wavelength tuning range of 185 nm (from 1861 nm to 2046 nm),¹³⁴ it is also considered a good candidate for femtosecond pulse generation around 2.0 μm. Using SESAM mode-locking, 496 fs pulses were achieved in a diode-pumped Tm³⁺:CaYAlO₄ laser with an average output power of 531 mW at 1975 nm.¹³³ Recently, Qiao et al. found that the diode-pumped femtosecond mode-locked Tm³⁺:CaYAlO₄ laser could operate in the high-order transverse mode [Figs. 3(a)–3(c)].¹³² This is an interesting discovery because it is traditionally thought that femtosecond mode-locked lasers work only in the fundamental transverse mode. The high-order transverse-mode femtosecond laser provides a platform for generation of ultraclean, topological charge-tunable mid-IR femtosecond vortex beams [Figs. 3(d)–3(f)]. Figure 3(a) shows a schematic of a high-order transverse-mode femtosecond laser. Using a noncollinear pumping configuration, a pure and mode-order-tunable femtosecond Hermite-Gaussian beam was generated from the mode-locked laser. It was then converted into a femtosecond Laguerre-Gaussian vortex beam with a cylindrical lens mode converter, as shown in Fig. 3(d). The resulting femtosecond vortex beam had a donut profile and a clean intensity node with a ring-to-center intensity contrast of 36 dB [Fig. 3(f)], which approached the theoretical limit of a vortex beam. The ultraclean femtosecond vortex beam may create new possibilities for high-resolution nonlinear imaging, femtosecond trapping, and vortex chirped pulse amplification.

In recent years, low-dimensional materials have attracted tremendous research interest for use as next-generation optoelectronic devices such as photoelectric detectors, field-effect transistors, and optical modulators. As SAs for mid-IR mode-locked lasers, they generally have the advantages of broadband absorption, fast relaxation time, low saturation fluence, and simple fabrication methods. These characteristics make them potential replacements for SESAMs in ultrashort pulse generation. In fact, the first passively mode-locked operation of solid-state laser near 2.0 μm was achieved using a transmission-type single-walled carbon nanotube (SWCNT)-based SA in 2009.³³ Stable mode-locked ∼10 ps pulses with a maximum average power of up to 240 mW were generated from a Tm³⁺:KLu(WO₄)₂ crystal pumped by a Ti:sapphire laser. After that, ultrashort mid-IR pulses with pulse durations from several picoseconds to less than 100 fs were achieved in various SWCNT-based mode-locked lasers,^34–40 demonstrating the excellent saturable absorption performances of SWCNTs around the 2.0 μm region. At the same time, other low-dimensional material SAs such as graphene have also been successfully employed in all-solid-state mode-locked lasers to generate ultrashort mid-IR pulses. In 2012, Ma et al. presented the first graphene mode-locked femtosecond laser around 2.0 μm with a Tm³⁺:CLNGG disordered crystal,⁴² delivering stable 729 fs pulses at 2018 nm with an average output power of 60.2 mW. Later, graphene mode locking was also achieved in a Tm³⁺:Lu₂O₃ crystal laser,⁴⁴ a Tm³⁺:LiYF₄ crystal laser,⁴⁵ and a Tm³⁺:YAG ceramic laser⁴⁶ near 2.0 μm. Remarkably, with graphene SA, the first generation of sub-100 fs pulses from solid-state oscillators around 2.0 μm was achieved in a Tm³⁺:MgWO₄ laser by a research group from Max-Born-Institute in 2017.¹³⁵ Near transform-limited pulses as short as 86 fs operated at 2017 nm were produced by using chirped mirrors for dispersion compensation, as shown in Fig. 4. Since then, these researchers have demonstrated a series of ultrashort sub-100 fs mode-locked lasers in various Tm³⁺- and Tm³⁺/Ho³⁺-doped materials with SWCNTs^37–40 and SESAMs.¹²³ They have even propelled the pulse duration down to sub-10 optical cycles.¹²³ The laser materials used in these works generally had flat, smooth, broadband gain spectra profiles that could support ultrashort femtosecond pulse generation. In addition, the tightly focused laser mode matched with the high-brightness Ti:sapphire pumping beam in the laser media, allowing for suppression of Q-switched mode locking and achieving strong self-phase modulation during laser operation. Moreover, all of these ultrashort sub-100 fs pulses were obtained with very low output coupling, resulting in high pulse energy in the cavities. Given the high energy and ultrashort duration of the pulse in the cavity, strong self-phase modulation could be induced to further broaden the mode-locked pulse spectrum. With the balance between self-phase modulation and anomalous group delay dispersion, ultrashort soliton mode-locked pulses were finally achieved. It should be noted that the pulse shaping was dominated by the interplay between group delay dispersion and self-phase modulation under these conditions, and the SA was no longer the crucial factor that determined the pulse duration.

Compared to mode-locking techniques based on real SAs, KLM shows a significant advantage in generating shorter pulses due to its intensity response rather than energy fluence response. To obtain ultrashort pulses around 2.0 μm, KLM technique has also been tried in various Tm³⁺-, Ho³⁺-, and Tm³⁺/Ho³⁺-doped laser materials. However, due to the larger divergence of the mid-IR beam, Kerr-lens effect is usually not strong enough to achieve KLM in mid-IR lasers, except for the lasers whose gain materials have remarkably high nonlinear refractive indices, such as ZnSe and ZnS. Thus, it is generally a challenging work to achieve KLM in the 2.0 μm spectral region. Until 2017, remarkable progresses were obtained. Zhang et al. introduced a 1 mm-thick sapphire plate as the Kerr medium to provide the necessary Kerr self-focusing effect in a Ho³⁺:YAG thin-disk oscillator, and successfully realized KLM, delivering 220 fs pulses at 2090 nm with a record average output power of 20 W.¹³⁶ Later, benefitting from the high nonlinear refractive index of Re₂O₃ material, KLM operation was also demonstrated in a Tm³⁺:Sc₂O₃ crystal laser, which was in-band pumped by an Er³⁺/Yb³⁺-doped fiber master oscillator power amplifier.¹³⁷ Pulses as short as 166 fs with an average output power of 440 mW were obtained at 2124 nm. At almost the same time, Canbaz et al. used KLM to produce a new source of 2.3 μm femtosecond pulses from a Tm³⁺:YLF laser.¹³⁸ An undoped 2 mm-thick polycrystalline ZnSe plate was used in the cavity to enhance the Kerr effect and help to achieve KLM. The mode-locked laser generated 514 fs pulses with an average power of 14.4 mW at 2.3 μm, which is believed to be the longest mode-locking wavelength from Tm³⁺-doped solid-state lasers to date.

Table I summarizes the main results of passively mode-locked Tm³⁺-, Ho³⁺-, and Tm³⁺/Ho³⁺-doped all-solid-state lasers published over the past decade. For each reference, we quote the shortest pulse width result.

Compared to solid-state laser materials, optical fiber is an excellent platform that is capable of generating laser emission with many exclusive advantages such as compactness, high power, excellent beam quality, and outstanding heat dissipation capability. Moreover, the guiding property of fiber systems allows one to easily integrate fiber components (e.g., fiber gratings and fiber couplers) to build all-fiber laser systems. These systems are compact, robust, economical, and require minimal maintenance. With these advantages, fiber-based laser systems have attracted significant attention and been well developed in the near-IR wavelength region in recent decades. Their output power has been scaled up to kilowatt-class and their pulse widths span the range from nanoseconds to femtoseconds. In the mid-IR spectral region, ultrashort pulse generation in fiber lasers began in the 1990s when Nelson et al. demonstrated a Tm³⁺-doped mode-locked fiber laser that produced sub-500 fs pulses around 2.0 μm.¹¹⁵ Since then, ultrafast fiber lasers have evolved rapidly toward ever-higher performances in this spectral region, and a number of mode-locked Tm³⁺-, Ho³⁺-, and Tm³⁺/Ho³⁺-doped fiber laser systems have been reported. Since a few reviews have been published on fiber laser sources around 2.0 μm,^156–159 our discussion focuses primarily on critical results and recent progress in mode-locked fiber lasers in this wavelength region.

In 1995, the first 2.0 μm femtosecond fiber laser was achieved using the NPR mode-locking technique. It produced 350 fs–500 fs pulses over a tuning range of 1.8 μm–1.9 μm.¹¹⁵ Subsequently, Sharp et al. presented an InGaAs SESAM mode-locked Tm³⁺-doped silica fiber laser capable of generating 190 fs pulses with an emission wavelength extending to 1.997 μm.¹¹⁶ A specific feature of optical fiber operating around 2.0 μm is that silica fibers exhibit anomalous dispersion, which causes typical soliton operation without dispersion management. The generated ultrashort soliton pulses were the results of balance between group delay dispersion and self-phase modulation, and no critical SA parameters were required under these conditions. However, due to the lack of optical fiber components in the 2.0 μm range, there were few mode-locked results reported in the following decade. Until 2007, transform-limited 650 fs–850 fs soliton pulses were achieved in a Tm³⁺/Ho³⁺-doped fiber oscillator with a Sb-based SESAM.¹⁶⁰ Since then, SESAM mode locking has been successfully demonstrated in a series of fiber lasers near 2.0 μm,^161–168 including the first Tm³⁺/Ho³⁺-doped fiber oscillator operating beyond 2.0 μm,¹⁶³ the first diode-pumped mode-locked Ho³⁺-doped fiber laser,¹⁶⁷ and a mode-locked Tm³⁺-doped multimode fiber oscillator with a record average power that exceeded 10 W.¹⁶⁸ In the meantime, the booming development of low-dimensional nanomaterials over the past decade has triggered an explosion of investigation into their saturable absorption properties as SAs for mid-IR mode-locked fiber lasers. CNTs,^169–174 graphene,^175–184 TIs,^185–188 TMDCs,^79–83,189 BP,^190–192 and other novel materials^193–195 have been used to achieve mode-locking operation in various fiber lasers around 2.0 μm. They have generated various pulse patterns such as conservative solitons, dispersion-managed solitons, dissipative solitons, bound-state solitons, vector solitons, and harmonic soliton pulses. Among them, with a normally dispersive germanium-silicate fiber to control intracavity dispersion, 152 fs ultrashort pulses with a 52.8 nm bandwidth were generated in a Tm³⁺-doped all-fiber oscillator mode locked by CNTs.¹⁷³ This was the shortest pulse duration with the broadest bandwidth directly obtained from CNT mode-locked 2.0 μm all-fiber lasers. Implementing dispersion management in this laser yielded obvious improvement in pulse duration. Stretched-pulse operation, which could overcome the pulse energy limitation of the fundamental soliton regime, was also achieved in a graphene mode-locked Ho³⁺-doped all-fiber laser.¹⁸⁴ It delivered a high pulse energy reaching to 2.55 nJ with 190 fs pulse duration at 2060 nm. In addition, ultralow noise 58 fs soliton pulses were produced¹⁹⁶ using a graphene electro-optic modulator and a real SA. This is the shortest pulse obtained from Tm³⁺-doped silica fiber oscillators thus far.

Another important mode-locking technology for ultrafast fiber lasers is use of an artificial SA based on nonlinear phase shift, such as NPR,^{115,197–202} NOLM,¹¹¹ and NALM.^110,164 Compared to real SAs, these mode-locking approaches have the advantages of ultrafast response time and wavelength-insensitive characteristic. In particular, NPR has been demonstrated to be a powerful technique that is able to generate versatile ultrashort pulses in mode-locked fiber lasers around 2.0 μm although bulk components are usually required in the laser setup. To overcome the soliton energy limitation in the NPR mode-locked fiber laser demonstrated by Nelson et al.,¹¹⁵ Engelbrecht et al. introduced a pair of gratings within the cavity to compensate for anomalous dispersion of the fiber, achieving a 4.3 nJ pulse energy.²⁰³ The generated 1.2 ps chirped pulses were centered at 1976 nm and could be externally compressed to a minimum pulse duration of 294 fs. By further shifting the cavity dispersion to the normal regime, dissipative soliton operation and self-similar pulse evolution could be obtained. These conditions allow the best combinations of high energy and short pulse duration. By employing ultrahigh numerical aperture fibers with normal dispersions, Tang et al. presented a Tm³⁺-doped fiber laser operating at a large normal dispersion with hybrid self-similar evolution and achieved 7.6 nJ, 130 fs pulses.²⁰⁰ Pulses as short as 160 fs with pulse energy above 1 nJ at 2060 nm were also produced in a Ho³⁺-doped fiber oscillator operating in the dispersion-managed soliton regime. Ultrashort sub-100 fs pulses were further obtained after passing through an additional nonlinear compressor stage.²⁰⁴ In 2014, an ultrafast mode-locked Tm³⁺-doped ZBLAN fiber laser oscillator was produced via NPR (Fig. 5), producing 45 fs pulses via external compression with an average power of 13 mW.¹⁹⁹ This represents the shortest pulses generated from fiber laser oscillators operating around the 2 μm wavelength region up until now.

Table II summarizes the important results of passively mode-locked Tm³⁺-, Ho³⁺-, and Tm³⁺/Ho³⁺-doped fiber lasers. For each reference, we quote the shortest pulse width result.

In the above section, we have discussed the ultrafast mode-locked laser sources around the 2.0 μm spectral region. Simultaneously, mid-IR mode-locked lasers operating around 2.4 μm have also attracted significant interest over the past decade due to their wide potential applications such as spectroscopy, gas sensing, high harmonic generation, etc. In this longer mid-IR wavelength region, transition-metal Cr²⁺-doped ZnSe and ZnS, which were introduced in the 1990s by Lawrence Livermore National Laboratory,²⁰⁸ represent a simple and viable route to achieve high-efficiency, high-power, cost-effective ultrafast mid-IR laser sources.

Cr²⁺:ZnSe and Cr²⁺:ZnS have highly similar spectroscopic properties. They feature four-level energy structures, broad vibronic emission spectra, the absence of excited state absorption, close to 100% fluorescence quantum efficiency at room temperature, good chemical and mechanical stability, and high thermal conductivity. These advantages, especially the ultrabroad emission spectrum, make them dominant media for femtosecond pulse generation around the 2 μm–3 μm spectral region and be referred to as “Ti:sapphire of the mid-IR.” Thus far, ultrabroad tuning ranges of 1973 nm–3349 nm in Cr²⁺:ZnSe and 1962 nm–3195 nm in Cr²⁺:ZnS have been obtained,^11,209–212 which means they are very good candidates for ultrashort femtosecond mode-locked pulse generation. In addition, due to the broad absorption band around the 1.5 μm–2.0 μm wavelength region, they can be conveniently pumped by cost-effective and reliable Er³⁺- and Tm³⁺-doped fiber lasers.

Since the first Cr²⁺:ZnSe laser was demonstrated in 1996,²⁰⁸ significant progress has been made with regard to various Cr²⁺:ZnSe and Cr²⁺:ZnS laser systems, covering a broad range of operation modes and output parameters. The ultrabroad tuning operating regime is particularly attractive because the broad tuning range is not only useful for mid-IR spectroscopic and sensor applications but also very important for the gain medium to generate few-cycle ultrashort femtosecond pulses with passive mode locking techniques. In 2005, a 1100 nm tuning bandwidth from 2000 nm to 3100 nm was demonstrated in a CW Cr²⁺:ZnSe laser.²¹³ To achieve broader wavelength tuning range, an intracavity-pumping configuration was used in a pulsed Cr²⁺:ZnSe laser, and 1220 nm smooth and continuous tuning from 1880 nm to 3100 nm was reported.²¹⁰ After that, Sorokin et al. demonstrated the broadest tunable operation of Cr²⁺:ZnSe and Cr²⁺:ZnS materials with over 1400 nm wavelength tuning range in 2010.²¹¹ These extremely broad tuning ranges indicate the potential of Cr²⁺:ZnSe and Cr²⁺:ZnS to generate few- to single-cycle pulses in the mid-IR region around 2.4 μm.

Pumped by a CW NaCl:OH⁻ color center laser at 1.58 μm, the first mode-locked Cr²⁺:ZnSe laser was demonstrated with an acousto-optic modulator in 2000. It generated transform-limited Gaussian-shaped 4.4 ps mode-locked pulses at 2470 nm.^214,215 Next, active mode locking was achieved in Cr²⁺:ZnSe single crystal²¹⁶ and ceramic²¹⁷ using an acousto-optic modulator. The mode-locked pulses obtained around 2.5 μm were approximately 4 ps and 21 ps in duration, respectively. In comparison to active mode locking, passive mode locking is the mainstream of ultrashort-pulse generation technique and can produce shorter pulses in various spectral regions. In 2005, the first passively mode-locked Cr²⁺:ZnSe laser was produced by Pollock et al. using a SESAM as the mode locker. It generated 10.8 ps mode-locked pulses near 2.5 μm.²¹⁸ Since then, SESAM mode-locked Cr²⁺:ZnSe and Cr²⁺:ZnS ceramic and single crystal lasers have been continuously reported, delivering mode-locked pulses with various pulse widths in the 2.4 μm region.^219–224 Shortly after the first SESAM mode-locked Cr²⁺:ZnSe laser was demonstrated, Sorokina et al. presented a passively mode-locked Cr²⁺:ZnS laser with an InAs/GaSb-based multiple quantum well SESAM. It produced ∼1.1 ps pulses in the 2.45 μm region.²¹⁹ Subsequently, ultrashort 106 fs pulses were obtained by introducing a 5 mm sapphire plate for intracavity dispersion compensation under similar cavity conditions.²²⁰ In 2007, the same group used two chirped mirrors and a 3.1 mm YAG plate to further compensate for intracavity dispersion and achieved shorter pulses in a SESAM mode-locked Cr²⁺:ZnSe laser.²²¹ Ultrashort ten-optical-cycle 80 fs pulses around 2.4 μm were obtained, which are the shortest pulses achieved from the SESAM mode-locked Cr²⁺:ZnSe and Cr²⁺:ZnS lasers thus far.

Compared to the SESAM mode-locking technique, KLM is more suitable to generation of ultrashort few-cycle mode-locked pulses due to its ultrafast response time and ultrabroad operation bandwidth. It is especially applicable to Cr²⁺:ZnSe and Cr²⁺:ZnS materials since they have high nonlinear refractive indices and particularly broadband emission spectra. KLM operation around 2.4 μm was first achieved in Cr²⁺:ZnSe lasers by two independent groups in 2009.^225,226 They achieved 95 fs and 100 fs mode-locked soliton pulses in the experiments, respectively. Later, by replacing the MaF₂ prism pair with a CaF₂ prism pair for dispersion compensation, the mode-locked Cr²⁺:ZnSe laser delivered shorter 92 fs pulses.²²⁷ In 2013, with Cr²⁺:ZnS gain medium and KLM technique, transform-limited pulses with 69 fs pulse width and 550 mW average power were produced at 2.39 μm.²²⁸ Subsequently, the average power was increased to 1 W in a shorter cavity, generating mode-locked pulses with 75 fs pulse duration.^229,230 In recent years, polycrystalline Cr²⁺:ZnSe and Cr²⁺:ZnS have attracted significant interest due to their advantages of high dopant concentrations, uniform dopant distributions, low losses, and mass-production with preassigned parameters. These properties support high-efficiency, high-power laser operation. In addition, because of their disordered polycrystalline structures that consist of a multitude of microscopic single-crystal grains with a broad distribution of grain sizes and orientations, a random quasi-phase-matching process can be achieved. This random quasi-phase-matching process has been demonstrated in several polycrystalline Cr²⁺:ZnSe and Cr²⁺:ZnS KLM lasers and enables direct nonlinear frequency conversion within the laser gain medium.^16,231,232 Since the first KLM polycrystalline Cr²⁺:ZnSe and Cr²⁺:ZnS lasers were reported in 2014,²³³ the output parameters of ultrafast KLM polycrystalline Cr²⁺:ZnS and Cr²⁺:ZnSe lasers have been significantly improved.^16,231–235 In 2015, ultrashort 29 fs mode-locked pulses were obtained from a KLM polycrystalline Cr²⁺:ZnS laser (Fig. 6),¹⁶ which are believed to be the shortest pulses obtained directly from mode-locked lasers in the 2.4 μm region. Under 4.6 W of pump power, the mode-locked laser had a spectrum as broad as 950 nm (1900 nm–2850 nm) at −30 dB level, which is a good source for mid-IR spectroscopy.

Low-dimensional materials, such as graphene and CNTs, have also been demonstrated as promising SAs for Cr²⁺:ZnSe and Cr²⁺:ZnS mode-locked lasers around 2.4 μm in recent years. Using monolayer graphene deposited on a CaF₂ plate as a SA, graphene mode-locking was first achieved in a Cr²⁺:ZnSe laser around 2500 nm in 2013.⁴⁷ Mode-locked 226 fs pulses with an average output power of 80 mW were produced. Given that graphene material had ultrabroadband saturable absorption, Ma et al. fabricated a graphene-gold film saturable absorber mirror (GG-SAM) with a demonstrated operating bandwidth that exceeded 1300 nm and achieved mode locking in various all-solid-state lasers.⁴⁸ Figures 7(a)–7(c) show the structure, image, and characteristics of the GG-SAM. Based on the GG-SAM, a tunable femtosecond Cr²⁺:ZnSe laser was realized. It produced ultrashort pulses of 116 fs and supported a mode-locking tuning range of 116 nm (2310 nm–2426 nm), as shown in Figs. 7(d)–7(f). In 2014, mode-locking operation in the positive dispersion regime around 2.4 μm was achieved in a graphene mode-locked Cr²⁺:ZnS laser, which produced chirped pulses with a high pulse energy of 15.5 nJ.⁴⁹ After compression by a pair of low OH-content infrasil prisms, 189 fs pulses were obtained. Later, the combination of a YAG wedge pair and a chirped mirror was used for dispersion compensation, and ultrashort 41 fs pulses centered at 2.4 μm with 190 nm spectral bandwidth were obtained in a similar graphene mode-locked Cr²⁺:ZnS laser.⁵⁰ To improve the average output power, a ceramic Cr²⁺:ZnS material was employed as the gain medium. The resulting 140 fs mode-locked pulses around 2.35 μm had an average power as high as 1.0 W.⁵¹ Besides graphene, CNTs which generally operate around 1.0 μm–2.0 μm have also been shown to support mode-locked operation in the 2.4 μm spectral region. In 2014, a Cr²⁺:ZnS laser mode-locked with a CNT SA was reported for the first time. The resulting laser pulses had a high average power of 950 mW and a pulse duration as short as 61 fs at 2.35 μm,⁴¹ indicating the excellent performance of a CNT SA around 2.4 μm. The obtained result is the shortest pulse produced from CNTs mode-locked lasers so far.

Table III summarizes the results of the passively mode-locked Cr²⁺:ZnSe and Cr²⁺:ZnS lasers. For each reference, we quote the shortest pulse width result.

With the rapid development of mid-IR laser materials, particularly rare-earth-doped bulk gain media and fluoride fibers, laser emission in the 2.8 μm–3.5 μm spectral region has become a reality. Laser emission in this region is based primarily on optical transitions offered by Er³⁺-, Ho³⁺/Pr³⁺-, and Dy³⁺-doped gain materials, as shown in Fig. 8. Because the upper laser level has a shorter lifetime than the lower laser level in Er³⁺- and Ho³⁺-doped gain media, lasing processes are usually self-terminated. To overcome this limitation, several methods for achieving efficient laser operation have been explored, including energy transfer up-conversion, cascaded lasing, and co-doping with Pr³⁺ ions. Thus far, various all-solid-state and fiber laser systems have been demonstrated in the 2.8 μm–3.5 μm spectral region using Er³⁺, Ho³⁺, and Dy³⁺-doped gain materials.^238–245 

Owing to the water vapor absorption in the atmosphere, all-solid-state lasers that operate around 2.8 μm suffer considerable losses that make CW mode-locking formation particularly difficult. Thus far, no CW mode locking has been reported in all-solid-state lasers around 2.8 μm. Alternatively, fiber laser systems represent an ideal platform for ultrashort pulse generation in this region because the waveguide structure of fiber can eliminate the deleterious effect of intracavity water vapor absorption. Besides, fiber lasers are featured with high power, good beam quality, inherent simplicity and flexibility, and outstanding heat-dissipating capability. In 1996, Frerichs and Unrau reported the first Q-switched mode-locked fluoride fiber laser in the 3.0 μm spectral region by using flying-mirror technique and an InAs SA.²⁴⁶ However, low-quality fluoride fiber and an immature SA resulted in failure of realizing CW mode-locking operation. With the advances in fluoride fiber fabrication in recent years, commercial fluoride gain fibers are available with low loss (<50 dB/km), single-transverse-mode operation, and double-cladding structure, and different mid-IR SAs are developed, creating the conditions for ultrashort mid-IR pulse generation. Using a newly developed GaAs-SESAM, mode-locking operation was demonstrated in a Ho³⁺/Pr³⁺-doped ZBLAN fiber laser in 2012. The system delivered mode-locked pulses with a duration of 24 ps and a peak power of 0.2 kW at 2.87 μm.²⁴⁷ In the following years, mode-locked fluoride fiber lasers had been optimized continuously via adoption of various SAs, cavity structures, output couplers, and gain fiber types. However, the resulting pulses were still a few to tens of picoseconds in duration with low average power.^248–250 In 2015, Tang et al. demonstrated a mode-locked Er³⁺-doped ZBLAN fiber laser at 2.8 μm (Fig. 9), with a record average output power of 1.05 W in single-pulse operation.²⁵¹ This represented a significant milestone for CW mode-locked ZBLAN fiber lasers in the mid-IR region. The mode-locked pulses showed a pulse duration of 25 ps at 2780 nm. Since the Er³⁺-doped ZBLAN fiber possesses a broad emission bandwidth, wavelength-tunable mode-locking operation was achieved by incorporating a diffraction grating with a tunable range from 2710 nm to 2820 nm.²⁵² 

Although various SAs such as Fe²⁺:ZnSe, InAs, and SESAM have been tested in Er³⁺- and Ho³⁺/Pr³⁺-doped fluoride fiber lasers, these mode-locked lasers typically operated in the picosecond regime. Femtosecond mode-locked fluoride fiber lasers were not achieved around 2.8 μm until NPR was employed. NPR mode-locked femtosecond fluoride fiber laser around 2.8 μm was first demonstrated In 2015.^253,254 Figure 10(a) shows the schematic of an NPR mode-locked Er:ZBLAN fluoride fiber laser, where 207 fs pulses were generated with a peak power of 3.4 kW at 2.8 μm [Fig. 10(b)]. Since fluoride fiber has an anomalous dispersion, the femtosecond mode-locked lasers generally operated in a soliton regime, as confirmed by the Kelly sidebands in the mode-locked spectrum [Fig. 10(c)]. Since the emission spectrum of the Er³⁺-doped fluoride fiber overlaps the absorption lines of water vapor, the mode-locked spectrum includes some absorption-related artifacts. This issue was mitigated by red shifting the laser wavelength toward 2.9 μm using a Ho³⁺/Pr³⁺-doped fluoride fiber. Antipov et al. reported an NPR mode-locked Ho³⁺/Pr³⁺-doped fluoride fiber laser in 2016. The laser produced 180 fs pulses with a peak power of 37 kW at 2876 nm.²⁵⁵ Recently, we also demonstrated an NPR mode-locked Er³⁺-doped fluoride fiber laser at 2.8 μm. The laser generated pulses with a duration of 131 fs by optimizing the gain fiber length, which are believed to be the shortest mode-locked pulses in the 2.8 μm wavelength region thus far.²⁵⁶ 

As previously mentioned, low-dimensional materials have been successfully used as SAs in mode-locked lasers that operate around 2.0 μm and 2.4 μm. Due to their broadband saturable absorption, ultrafast recovery time, and easy fabrication and integration, most of these materials have also been demonstrated as mode lockers for generation of ultrashort pulses at longer wavelength around 2.8 μm. Unlike most mode-locked all-solid-state lasers, fluoride fiber lasers operating around 2.8 μm generally require relatively large modulation depths for the SAs to initiate and sustain mode-locking operation. One effective way to increase the modulation depths of low-dimensional material SAs is to increase the sample thicknesses, such as by stacking single-layer graphene and directly synthesizing thicker TI samples. With multilayer graphene and Bi₂Te₃ nanosheets as SAs, CW mode-locking operation was achieved in fluoride fiber lasers around 2.8 μm, delivering mode-locked pulses with pulse durations of ∼42 ps and ∼6 ps, respectively.^53,257 BP, which exhibits a layer-dependent bandgap that varies from ∼2.0 eV for monolayer to ∼0.3 eV for bulk, has also been demonstrated as a promising mid-IR saturable absorption material. In 2016, Qin et al. demonstrated the first BP mode-locked Er:ZBLAN fiber laser around 2.8 μm. It delivered 42 ps pulses with an average output power of 613 mW, as shown in Fig. 11.⁶⁰ Subsequently, shorter 8.6 ps mode-locked pulses were achieved with BP as the mode locker in a Ho³⁺/Pr³⁺-doped fluoride fiber laser at 2866.7 nm.⁶¹ Cd₃As₂ film, a topological Dirac semimetal, has recently shown potential as a robust mid-IR SA for the development of ultrafast mid-IR laser sources. A mode-locked Ho³⁺/Pr³⁺-doped fluoride fiber laser was demonstrated with Cd₃As₂ film as SA, delivering 6.3 ps mode-locked pulses around 2.86 μm.⁶³ Due to the immaturity of fluoride fiber-based optical devices and challenges in fluoride fiber splicing, most previous studies of mid-IR fluoride fiber lasers were performed using free-space optical elements that do not rival the advantages of all-fiber lasers with regard to flexible delivery, stability, and compactness. Recently, an all-fiber-designed fluoride fiber laser was constructed by taking advantage of BP flake integration and CW mode-locking operation was achieved at 2.8 μm, generating mode-locked pulses with estimated tens of picoseconds duration.²⁵⁸ 

The emission wavelengths of Er³⁺- and Ho³⁺/Pr³⁺-doped fluoride fibers are generally below 2.9 μm. This covers only a small part of the so-called molecular fingerprint region. There is a strong demand for compact, robust ultrafast laser sources at longer wavelengths where some important functional groups locate. One promising candidate is the Dy³⁺-doped fluoride fiber laser, which has a particularly broad tunable range from 2.8 μm to 3.4 μm.²⁵⁹ Mode locking of Dy³⁺-doped fluoride fiber laser was first demonstrated using frequency shifted feedback.²⁶⁰ This actively mode-locked laser delivered pulses with an average output power of 120 mW and a pulse duration of 33 ps at 3.1 μm. Based on the NPR mode locking technique, a femtosecond Dy³⁺-doped fluoride fiber laser was also demonstrated recently. The laser generated 724 fs transform-limited soliton pulses with a pulse energy of 2.3 nJ at 3.05 μm.²⁶¹ Laser emission beyond 3.4 μm was also achieved using a dual-wavelength pumped Er³⁺-doped fluoride fiber laser, showing a wide tunable range from 3.4 μm to 3.8 μm.²⁶² However, due to the lack of practical SAs in this region, the development of ultrafast laser was blocked for a long time. Until 2018, Qin et al. developed a bulk BP SAM and demonstrated CW mode locking in a dual-wavelength pumped Er³⁺-doped fluoride fiber laser at 3.5 μm (Fig. 12).¹⁷ The pulse duration was estimated to be tens of picoseconds according to the narrow spectral bandwidth of 4.7 nm. Femtosecond mode-locked fluoride fiber laser at 3.5 μm has not been achieved yet so far, and additional efforts need to be made due to shortage of mature optical devices around this wavelength region.

Table IV summarizes the results of CW mode-locked Er³⁺-, Ho³⁺/Pr³⁺-, and Dy³⁺-doped fluoride fiber lasers in the 2.8 μm–3.5 μm spectral region. For each reference, we quote the shortest pulse duration result.

Since many important molecules have their characteristic vibration and rotation spectra in the midinfrared region, ultrafast midinfrared lasers provide new opportunities for a host of application areas such as molecular spectroscopy, chemical and biomolecular sensing, etc. In addition, ultrafast midinfrared sources have also been proved as a powerful tool to generate midinfrared supercontinuum and laser frequency combs, and to investigate ultrafast dynamic processes in solid-state physics. In this section, we will highlight some representative applications of ultrashort mid-IR mode-locked laser sources.

Supercontinuum source, featured with an extremely broad emission bandwidth, is considered a very important tool for many applications and attracts a growing attention in the mid-IR region. Although supercontinuum generation technology is well developed and has even reached commercialization in the visible and near-IR spectral regions, it remains challenging to provide high-efficiency, high-power mid-IR supercontinuum sources. The limitations concern both highly transparent nonlinear media in the longer mid-IR region and reliable ultrafast pump sources that can be used to pump nonlinear media efficiently around zero-dispersion wavelengths. Benefiting from the development of ultrafast mid-IR mode-locked lasers, remarkable progress has been made in supercontinuum generation in recent years. Pumping by a CNT mode-locked Tm³⁺-doped fiber laser around 2.0 μm, Thapa et al. reported ultrawide mid-IR supercontinuum generation spanning from 0.9 μm to 4.2 μm in a low-loss W-type tellurite fiber.²⁶⁵ They successfully developed fiber fusion splicing technology capable of splicing tellurite fibers directly to silica fibers, paving the way for all-fiber mid-IR supercontinuum sources. Compared to tellurite glass with a transmission window limited to ∼5 μm, chalcogenide materials such as As₂Se₃ and GaSe not only exhibit much broader transmission windows up to 15 μm and even into the low-frequency terahertz range^266–268 but also have much higher nonlinear refractive indices. This makes them the most promising candidates for broadband mid-IR supercontinuum generation. By combining a newly developed mode-locked Ho³⁺/Pr³⁺-doped ZBLAN fiber laser with a dispersion-engineered As₂Se₃/As₂S₃ tapered fiber device, Hudson et al. achieved a supercontinuum with a bandwidth of over 10 000 nm (125 THz) in the mid-IR region,²⁶⁹ as shown in Fig. 13. The pump laser emitted 230 fs pulses around 2.9 μm at a repetition rate of 42 MHz with an average power of 140 mW. It generated a 2.4 octave-spanning supercontinuum covering from ∼2 μm to 12 μm range. More recently, Zhang et al. demonstrated broadband, coherent, and carrier-envelope-phase-stable supercontinuum radiation from a 1 mm-thick GaSe crystal driven by ultrashort self-compressed pulses from a KLM Ho³⁺:YAG thin-disk oscillator.²⁷⁰ The generated supercontinuum had a spectrum of 500 cm⁻¹–2250 cm⁻¹ (−30 dB), which corresponded to a wavelength range from 4.5 μm to 20 μm. This suggests excellent potential for future spectroscopy applications.

Laser frequency combs have dramatically improved precise measurements of frequency and time. They have been used widely in metrology and physics research since their introduction in the late 1990s. In the mid-IR range, optical frequency combs are of significant interest for molecular spectroscopy. They are expected to enable ultrasensitive, rapid, precise molecular sensing across the broadband mid-IR molecular fingerprint region. Although several approaches to mid-IR frequency comb generation have previously been reported,^271–274 mid-IR mode-locked laser sources are probably the most general, reliable, naturally low-noise, and compact choices for frequency comb generation. Using SESAM as a mode locker, Bernhardt et al. presented frequency-comb Fourier-transform spectroscopy based on two ceramic Cr²⁺:ZnSe femtosecond oscillators emitting in the 2.4 μm mid-IR region.²⁷⁵ Ultrashort pulses as short as ∼3 optical cycles have been recently reported from a Cr²⁺:ZnS mode-locked laser, emitting a very broadband spectrum around 2.5 μm.¹⁶ This unique mid-IR mode-locked laser generated intrinsically coherent frequency combs, further extending potential applications to high-resolution, high-sensitivity spectroscopy and optical clocks. In the 2.0 μm–3.5 μm wavelength region, mode-locked lasers have also been achieved using a variety of techniques based on Tm³⁺, Ho³⁺, Er³⁺, and Dy³⁺-doped materials, which may directly extend the spectrum coverage region and offer potential sources for laser frequency combs in the mid-IR region.^271,276

The development of ultrafast mid-IR laser sources significantly advances progress in mid-IR spectroscopy, especially ultrafast mid-IR spectroscopy, which are very important sample characterization and analytical techniques in the fields of molecular science and solid-state physics.^277–282 They have been used extensively in applications that involve qualitative and quantitative analyses, investigating the structural information of molecules and its governing physical laws, and providing the time-dependent physical parameters for dynamic processes investigations. For example, structural dynamics of molecules such as vibrational mode coupling and charge transfer were investigated via time-resolved mid-IR spectroscopy at the femtosecond time scale.^279,283 Free-carrier dynamics were clearly observed in semiconductors via ultrafast mid-IR pump-probe spectroscopy.^284,285 Furthermore, two-dimensional infrared spectroscopy, a novel analytical technique based on time-resolved detection of mid-IR signals, has also been developed benefitting from the ability to generate ultrashort mid-IR pulses.^{277,280,286,287} Two-dimensional infrared spectroscopy can provide direct insight into structural dynamics such as vibrational mode coupling, anharmonicities, bond formation and dissociation, and hydrogen network alternation at the femtosecond time scale, offering much more precise information than linear mid-IR spectroscopy. As mentioned in Sec. III, compared to ultrafast sources generated by nonlinear processes, mid-IR mode-locked lasers have the advantages of compactness, reliability, low cost, and the ability to directly generate ultrashort femtosecond pulses with good stability and beam quality. Using a femtosecond Cr²⁺:ZnSe laser, Sorokin et al. demonstrated mid-IR spectroscopy for gas detection and concentration analysis, as shown in Fig. 14.²⁸² They achieved high sensitivity and resolution over a broad spectral domain. In addition, with mid-IR mode-locked laser sources, one can extend the spectrum to longer mid-IR regions and generate pulses as short as a few cycles via nonlinear techniques.²⁶⁹ This has the potential to further aid ultrafast mid-IR spectroscopy.

Material processing using ultrashort mid-IR laser pulses has also attracted significant interest in recent years due to its potential applications in science, technology, and industry.^288–292 Ultrafast mid-IR laser sources extend existing ultrafast laser processing technology into the mid-IR region. This opens the possibility of expanding high-precision processing techniques to materials (e.g., silicon, germanium) that are opaque in the visible and near-infrared spectral ranges but optically transparent in the mid-IR region. Employing a femtosecond laser source at 2.4 μm, Nejadmalayeri et al. fabricated a single-mode optical waveguide for the 1.55 μm and 1.32 μm telecommunications bands in bulk crystalline silicon via ultrafast mid-IR laser inscription.²⁸⁸ The waveguide exhibited a low insertion loss of ∼1.2 dB/cm or less, which made it applicable to silicon-based optical circuits.²⁹³ With further optimization, laser inscription with ultrashort mid-IR laser pulses could be applied to in-chip microstructure fabrication, where it could be used to create complex three-dimensional optical circuits directly inside silicon, overcoming problems that current lithography methods cannot solve. Recently, a numerical study of three-photon absorption vs two-photon absorption for in-bulk modification of silicon indicated that mid-IR femtosecond lasers have the potential to be used for efficient kerfless silicon processing.²⁹¹ This technique can be employed to replace the currently used diamond saws that separate wafers from the silicon bulk and minimize wasted material. Polymer processing, including cutting and welding, was also investigated using ultrashort mid-IR laser pulses. Using an all-fiber ultrafast mid-IR laser around 2 μm, two 5 mm-thick polymethyl methacrylate blocks were butt-welded with a stable joint, and Kapton film was cut with a cut width of only 11 μm.²⁹² 

In the medical field, ultrafast mid-IR laser sources have been demonstrated to be powerful tools and shown great advantages for the applications of laser surgery and biodiagnostics.^294–304 The mid-IR wavelength around 3.0 μm corresponds to the vibrational absorption band of water molecules within tissue. Thus, laser radiation can be substantially absorbed within a few microns of incidence. Furthermore, the picosecond-level pulse duration is short enough to drive tissue ablation faster than the thermal and acoustic energy transfer processes that cause damage to the surrounding tissue, but long enough to avoid tissue ionization and the risks of creating toxic free radicals and fragmenting constituent proteins.^294,295,305 Therefore, high-energy mid-IR picosecond laser pulses around 3 μm are of particular interest for minimally invasive surgery and may be better than conventional medical lasers with pulse durations longer than nanoseconds. Figure 15 shows schematics of three cutting modalities by using a mid-IR picosecond laser system, a conventional long-pulsed surgical laser, and a mechanical surgical scalpel.²⁹⁵ The mechanical surgical scalpel causes more than 400 μm damage zone from the borders of the incision and the conventional laser damages skin border up to 800 μm away from the ablated edge. In contrast, cuts using the mid-IR picosecond laser system have sharp edges and a damage zone of less than 10 μm, which is even smaller than the diameter of a single skin fibroblast. In addition, an investigation of wound healing has demonstrated that wounds produced by a mid-IR picosecond laser healed faster than those produced via either a conventional surgical laser or a mechanical surgical scalpel. The scars formed were half the size of those from the aforementioned conventional methods. These advantages can decrease the risk of infection after surgery and increase patient comfort. Aside from promising applications as surgery tools, ultrafast mid-IR laser sources have significant potential for early real-time biodiagnostics.^302,304 Mouras et al. recently used a high-repetition-rate (80 MHz), tunable ultrafast mid-IR laser to develop a high-speed, chemical selective, label-free bench-top mid-IR microscope for chemical imaging.³⁰⁴ This microscope had a subwavelength resolution and could be applied to living cells and tissue, where it offered real-time tracking of drugs and drug carriers within cells and tissues.

The development of robust, compact, cost-effective ultrafast mid-IR laser sources with advanced performances has a rich history. Passively mode-locked lasers in the mid-IR region act as ideal sources as they can generally satisfy these requirements and provide high power, ultrashort pulses, and good beam quality. However, there was little progress in the development of passively mode-locked mid-IR laser sources until the recent decade. Driven by the rapid development of mid-IR laser gain materials and mode locking techniques, significant advances were made in ultrafast mode-locked lasers for the 2.0 μm–3.5 μm mid-IR region by using various rare-earth-ion (Tm³⁺, Ho³⁺, Er³⁺, and Dy³⁺)-doped solid-state and fiber materials and transition-metal Cr²⁺-doped ZnS and ZnSe gain media. Remarkably, pulse duration from mid-IR mode-locked laser has reached down to ∼3 optical cycles in the Cr²⁺:ZnS laser, and the average output power from a mid-IR mode-locked laser has been as high as 20 W from a Ho³⁺:YAG laser. The wavelength range, which is particularly important for molecular spectroscopy, has also been extended to the 3.5 μm region. The mid-IR mode-locked laser achievements summarized and highlighted in this review represent crucial and recent advances in this research field. Furthermore, these significant progresses on mid-IR ultrashort laser sources open the door for a series of applications including mid-IR supercontinuum generation, mid-IR frequency combs, mid-IR spectroscopy, material processing, laser surgery, and biodiagnostics.

In the meantime, although significant progress has been achieved in recent years, developing compact and robust ultrafast mid-IR laser sources with better performances such as broader wavelength range covering 5–10 μm, shorter pulse duration down to one optical cycle, higher pulse energy, and peak power levels that can reach the strong-field regime, is still on the way. Several challenges remain to be addressed. First, the development of mid-IR mode-locked laser sources has historically been hindered by poor laser material availability. Thus, generation of ultrashort mid-IR pulses via nonlinear frequency conversion has become a general method, especially in the long wavelength region that current gain materials cannot support. To realize direct ultrashort laser generation with high efficiency and high power, novel gain media that can operate in the longer mid-IR wavelengths need to be developed. Meanwhile, few-cycle ultrashort mid-IR pulse generation and amplification have become increasingly attractive in recent years, which depend on the development of novel broadband gain materials with smooth, flat gain profiles and excellent mechanical, thermal, and optical properties. Second, significant advances have been made in the mid-IR mode-locking devices and techniques in recent years which are the critical factors influencing the mode-locked laser performance. However, compared to the mature SAs available for the near-IR region, there is still significant room for further improvement and optimization of current mid-IR SAs in terms of recovery time, nonsaturable loss, and damage threshold. Considerable efforts are still required for development of novel saturable absorption materials and mode-locked techniques for ultrashort mid-IR pulse generation. In addition, mid-IR SAs based on low-dimensional materials are in their infancy. Exploring reliable approaches to reproducible SA fabrication with precisely customized parameters is still an important task in the future work. Third, ultrashort mid-IR laser sources have become of interest to strong-field physics, as strong laser-matter interactions are dependent on the laser wavelength (high-harmonic generation cutoff frequency ∝ λ², attosecond pulse duration ∝ λ^−1/2, electron kinetic energies ∝ λ²).^306–308 Thus, developing novel ultrashort mid-IR pulse generation and amplification technologies that can achieve high energy and peak power levels is increasingly important for strong field science. Although some preliminary work such as development of Cr²⁺:ZnSe regenerative amplifiers³⁰⁹ and Cr²⁺:ZnSe thin disk lasers³¹⁰ has been performed, a lot of work remains to be done for achieving robust, compact, and practical tabletop ultrashort mid-IR laser sources with high energy and peak power levels, such as development of mid-IR dispersion devices, exploration of mid-IR pulse and spectrum shaping techniques, realization of carrier envelope phase control, etc. With continued development of novel materials and technologies in the mid-IR region, we believe that ultrashort mid-IR pulses and even few-cycle pulses with high average power, high pulse energy, and longer operation wavelength can be expected in the future. Breakthroughs in ultrafast mid-IR laser sources will open new research avenues and lead to broad applications in science, industry, medicine, and other fields.

This research was partially supported by funds from the National Natural Science Foundation of China (Grant Nos. 61575089, 61675130, and 11721091), the Jiangsu Specially-Appointed Professor Program, the National Postdoctoral Program for Innovative Talents (Grant No. BX20170149), the China Postdoctoral Science Foundation (Grant No. 2017M620150), and A^*Star Singapore (Grant No. A1883c0003).