A coherent laser source emitting mid-infrared (MIR) radiation in the 2.7–3 µm wavelength band has several unique features, including (i) locating at the well-known atmospheric transparency window^[1]; (ii) corresponding to the strong rovibrational absorption lines of various gas molecules^[2,3]; (iii) overlapping the strong absorption peak of water^[4]. Based on the above merits, it has recently attracted tremendous attention owing to the potential applications in remote sensing, laser surgery, gas monitoring and detection, precision spectroscopy measurement, material processing and countermeasures, etc. Up to now, the most advanced MIR solid-state laser technologies that are used for 2.7–3 µm laser generation include but are not limited to the rare-earth-doped crystalline and fiber lasers^[5–21], semiconductor laser diodes (LDs)^[22–26], quantum cascade lasers (QCLs)^[27,28], nonlinear optical frequency conversion {optical difference frequency generation (DFG)^[29,30], optical parametric sources [OPSs, including optical parametric oscillators (OPOs)^[31–33], optical parametric generators (OPGs)^[34,35]], etc.}, and so on.

The GaIn(As)Sb/AlGaAnSb system-based strained multi-quantum-well LD is regarded as the most established semiconductor laser technology for 2–3 µm MIR laser generation^[23,26], which relies on the inter-band laser transition and depends on the composition of the selected alloys. However, the Auger effect can strongly reduce the efficiency when the emitting wavelength is longer than 2 µm^[26,32]. In contrast to semiconductor LDs, the laser transition in QCLs is in the conduction band, which is commonly named the inter-sub-band transition. One type of carrier and a multistage cascade scheme are the two fundamental features where QCLs differentiate from semiconductor LDs^[32]. The shortest wavelength of QCLs is limited by the conduction band offset height between different heterostructure materials, while there is no fundamental limitation for the long-wavelength side. For example, the shortest wavelengths for InP- and GaAs-based QCLs are 3.4 and 8 µm, respectively. As a consequence, QCLs are more suitable for longer MIR laser generation. However, the biggest critical issue of QCLs is the large amount of dissipated heat, which makes it difficult to achieve continuous-wave (CW) laser operation at room temperature. Compared with semiconductor LDs and QCLs, converting the most mature 1 µm laser to the MIR region through nonlinear optical frequency conversion is the commonly used technique for 2.7–3 µm laser generation, especially for high-power and high-energy MIR lasers. However, either DFG or OPSs need high power and a high-beam-quality fundamental laser and a nonlinear optical convertor, which make the system very complex and with high cost. Therefore, researchers have endeavored to pursue 2.7–3 µm MIR laser sources with the advantages of being robust and compact, high efficiency, high beam quality, low cost, being easy to use, etc.

Over the past two decades, benefiting from the great progress in material science and technology, a lot of MIR laser gain materials, especially rare-earth-doped crystalline and fiber materials with excellent optical, thermal, and mechanical properties have been developed^{[5–21,36–39]}. Thus far, direct laser emissions in the 2.7–3 µm spectral region have become reality^{[9,16,21,32,40,41]} and thus have the significant advantages of low cost, high efficiency, simple structure, and low loss. Direct laser emission around 2.7–3 µm is primarily based on the optical transitions offered by Er3+-, Ho3+-, and Dy3+-doped gain materials. The typical emission spectrum and wavelength coverages of Er3+-, Ho3+-, and Dy3+-doped lasers around 3 µm are shown in Fig. 1^[19,42]. Nowadays, driven by the new materials and sustainable innovations in laser technology, a lot of crystalline and fiber lasers operating around 2.7–3 µm have been realized. Compared with fiber lasers, all solid-state lasers with bulk crystals have the main advantages of low undesirable nonlinear effects and a large mode area, which make them more suitable for generating high-energy and high-peak-power ultrashort pulses. Here, in this work, we mainly summarize and review the recent progress of MIR bulk laser sources based on the rare-earth ions-doped crystals in the 2.7–3 µm spectral region, including Er3+-, Ho3+-, and Dy3+-doped all solid-state crystalline lasers.

At present, the rare-earth ions that can achieve room temperature MIR laser operation in the 2.7–3 µm spectral region are mainly Er3+, Ho3+, and Dy3+, among which the Er3+ ion is mostly studied. Nevertheless, the MIR emission spectrum of the Er3+ ion is a line, and the corresponding output wavelength is relatively short. In contrast, the number of electrons in the 4f shell of Ho3+ and Dy3+ ions is even, resulting in the Stark-level splitting being greatly influenced by the crystal fields. Thus, the fluorescence spectra of the Ho3+ and Dy3+ ions-doped crystals are usually smooth and broadband, which enables the tunable laser output and also expands the wavelength towards the infrared direction.

Besides the well-known laser transition of I413/2→I415/2 emitting wavelength around 1.55 µm, the Er3+ ion can also provide 2.7–3 µm MIR emission with the I411/2→I413/2 transition. The simplified energy-level diagram of the Er3+-doped gain medium is shown in Fig. 2(a). Easy growing and the pumping wavelength being around ∼970nm (commercial LD operation wavelength) are the two fundamental merits that make Er3+-doped crystals much more extensively studied for the 2.7–3 µm laser generation compared to that of Ho3+- and Dy3+- doped crystals. As early as 1967, Robinson and Devor realized the first, to the best of our knowledge, laser oscillation of an Er3+-doped crystalline MIR laser at 2.69 µm with a CaF2:ErF3 mixed crystal^[43]. However, for the Er3+:I411/2→I413/2 transition, there is a so-called self-terminated effect leading to a population bottleneck issue caused by the lifetime of the upper laser level (I411/2) being much shorter than that of the lower level (I413/2)^[44,45], which is the main obstacle preventing the development of Er3+-doped 2.7–3 µm crystalline lasers. Up to now, several approaches have been developed to solve this detrimental feature. One is increasing the doping concentration of the Er3+ ion to generate the “quenching effect”, which could decrease the lifetime of Er3+:I413/2 so as to relieve the self-terminating behavior^[11,46–49]. The energy-transfer up-conversion (ETU) process between Er3+ and Er3+ ions in highly doped crystals can also effectively depopulate Er3+:I413/2 to solve the population bottleneck issue. But, it should be noticed that it is a double-edged sword because of the severe thermal effects and the lifetime reduction of the upper laser level Er3+:I411/2. Another approach is co-doping with sensitized ions (typically Yb3+) with the efficient population of Er3+:I411/2 or deactivated ions (typically Pr3+) with efficient depopulation of Er3+:I413/2 for establishing and sustaining population inversion, as shown in Fig. 2(a)^[50–60]. Besides, the host material is another essential factor for the Er3+:I411/2→I413/2 transition, which should have low photon energy, high radiative emission rate, and low absorption in the 2.7–3 µm band. To date, the oxide [primarily the garnet structure: Y3Al5O12 (YAG), Gd3Ga5O12 (GGG), Gd3Sc2Ga3O12 (GSGG), Y3Sc2Ga3O12 (YSGG), etc. and YAlO3 (YAP)]^{[11,46,50,60–67]}, fluoride [CaF2, SrF2, LiYF4, LiLuF4 (LLF), etc.]^{[15,43,48,55,58,68–74]}, and sesquioxide crystals (Sc2O3, Lu2O3, Y2O3, etc.)^[49,75–79] or ceramics have been proved to be the promising and most spread host materials for Er3+-doped 2.7–3 µm crystalline lasers by considering the thermal conductivity, opto-mechanical properties, photon energy, etc. Moreover, the cascading laser operation of I413/2→I415/2 and I411/2→I413/2 transitions is another way to suppress the saturation of the I411/2→I413/2 transition. Such cascade oscillation was demonstrated for instance in Er:YLiF4 (YLF) Q-switched lasers and Er-doped fluoride fiber lasers^[80,81].

After the first, to the best of our knowledge, realization of an Er3+:CaF2:ErF3 mixed crystal operating at 2.69 µm in 1969, efficient operations both in the CW and pulsed regimes with Er3+-doped crystals have been demonstrated with the development of the crystal design and growth and the innovations of the laser technology. Figure 2(b) shows the room temperature CW output power and the corresponding slope efficiency obtained with Er3+ ions-doped crystalline lasers in the 2.7–3 µm region. In 1992, Dinerman et al. reported the first, to the best of our knowledge, CW operation of monolithic Er:YAG, Er:GGG, and Er:YSGG lasers near 3 µm with output powers of 143, 155, and 190 mW^[61], which were further promoted to 171, 293, and 511 mW in 1994^[46]. In 2010, Sousa et al. realized a maximum CW output power of 1.5 W with an Er:YAG crystal at 2.94 µm^[82]. In 2014, You reported a diode-end-pumped MIR multi-wavelength Er,Pr:GGG laser with CW output power of 324 mW^[57]. In 2015, Shen et al. studied the CW laser performance of an LD side-pumped Er:YSGG slab at 2.79 µm, in which the maximum output power of 1.84 W was obtained with a slope efficiency of 10.2%^[83]. The corresponding experiment setup is shown in Fig. 2(c), in which an Er:YSGG slab with dimensions of 1mm×2mm×12mm was dual-side-pumped by 970 nm LDs. In 2018, Yu et al. realized a high-efficiency Er:YGG laser at 2.82–2.92 µm with output power of 1.38 W and slope efficiency of 35.4%, approaching the theoretical quantum limits^[84]. For fluoride crystals, in 2006, Basiev presented a continuously tunable CW laser operation near 2.75 µm of diode-pumped Er:SrF2 and Er:CaF2 crystals with output powers of 0.4 and 2 W^[70]. In 2018, Švejkar et al. promoted the Er:SrF2 laser output power up to 1.3 W with a slope efficiency of 9.2% and tuning range of 123 nm^[72]. Liu et al. realized an efficient CW laser performance of a diode-end-pumped Er:CaF2-SrF2 crystal with an output power of 712 mW and a slope efficiency of 41.4%^[71]. However, as shown in Fig. 2(b), the output powers of the garnet structure and fluoride-crystals-based Er3+-doped crystal lasers in the 2.7–3 µm region are limited to ∼2W, mainly because of the severe thermal effect. Benefiting from the high thermal conductivity and low phonon energy, sesquioxide crystals have been proved to be more suitable for high-power MIR laser operation. In 2012, Li et al. presented a 5.9 W CW output power with Er:Lu2O3 as the gain medium and a slope efficiency of 27%^[49]. Besides sesquioxide crystals, high-quality polycrystalline transparent sesquioxide ceramics also show huge opportunities for high-power 3 µm lasers because of the advantages compared to single crystals, such as excellent mechanical strength and easy fabrication process^{[76,77,79,85–87]}. In 2011, a 2.8 µm Er:Y2O3 ceramic laser with an output power of 14 W was reported with a cooling temperature of 77 K^[87], which was promoted to be 24 W in 2016^[86]. Very recently, Yao et al. demonstrated an Er:Lu2O3 ceramic laser at 2845 nm with 6.7 W output power and >30% slope efficiency, which is the highest output power ever achieved from Er-doped sesquioxide ceramics at room temperature^[88]. Besides, YAP is another attractive host candidate for high-power Er3+-doped crystal lasers owing to low phonon energy and excellent thermal properties. In 2019, Yasuhara et al. presented a 1.17 W CW Er:YAP crystal laser operating at 2.9 µm with a slope efficiency of 29%^[89]. Subsequently, due to the anisotropic thermal properties, a b-cut Er:YAP crystal was chosen, and the output power was promoted to 6.9 W [as shown in Fig. 2(d)], which is the highest CW output power generated from Er-doped solid-state lasers at room temperature^[15]. Table 1 summarizes the important results of CW Er-doped solid-state lasers.

In the pulsed regime, flash-side pumping is an effective and commonly used architecture to produce high-energy 2.7–3 µm laser pulses at low repetition rate. As early as 1990, pulse energy as high as 400 mJ was obtained with an Er:YAG crystal under the pump energy of 92 J^[92]. However, the pulse width was always in the scale of hundreds of microseconds or even milliseconds without any cavity Q-factor modulation. Combining with the active or passive Q-switching technique, nanosecond 2.7–3 µm pulsed lasers could be achieved. In 2004, a pulse energy of 137 mJ with a pulse width of ∼90ns and repetition rate of 3 Hz was obtained from a single xenon flashlamp-pumped, actively Q-switched Er:YAG laser at 2.94 µm^[93]. In 2005, Koranda et al. reported a 60 ns laser pulse with energy of 60 mJ generated to form a LiNbO3 (LN) electro-optically (EO) Q-switched 2.94 µm Er:YAG laser^[94], as shown in Fig. 3(a). An Er:YAG crystal with dimensions of Φ4mm(diameter)×89mm(length) was placed along the Xe flashlamp in a Linear Matrix Inequality diffuse ceramic cavity, which was the key part of the laser oscillator. The LN crystal with both faces cut under a Brewster angle acted as a Pockels cell and optical polarizers. The specially designed delay circuit was another key element to provide precise switching of the EO shutter at the time when the population inversion inside the Er:YAG crystal reached the maximum value. In 2007, a giant pulse width of 35 ns and an output energy up to 30 mJ were obtained from a Fe2+:ZnSe passively Q-switched Er:YAG laser at 2.94 µm^[95]. In 2013, Wang et al. realized a 2.79 µm high-peak-power langasite (LGS) EO Q-switched Cr,Er:YSGG laser with pulse energy of 216 mJ and pulse duration of 14.36 ns^[96]. For the flash-pumping laser system, it is difficult to achieve high pulse energy at a high repetition rate due to the low operation repetition rate, low conversion efficiency, and strong thermal effect.

Compared to flash pumping, pulsed LD pumping has the merits of high efficiency, better beam quality, and high repetition rate. Hence, an efficient and compact diode-laser-pumped 2.94 µm Er:YAG laser with energy up to 9 mJ was realized in 2010, consequently making the hermetically sealed windowed package^[82]. In 2015, a pulse energy of 562 mJ at 16 Hz was obtained from an LD side-pumped 2.79 µm Er:YSGG laser^[97]. In 2017, an output peak power of 1.2 W was obtained by a quasi-CW LD end-pumped Er:Lu2O3 laser^[77]. In 2018, a maximum output power of 8.86 W was achieved at 125 Hz with a slope efficiency of 14.8% from an LD-side-pumped Er,Pr:Gd_1.17YSc_1.284Ga₃O₁₂ laser^[60]. As shown in Fig. 3(b), the main part was the diode-side-pumped Er,Pr:GYSGG module, in which three LD arrays were symmetrically placed with intervals of 120° to alleviate the thermal effect. For CW LD pumping, pulsed Er-doped laser generation at 2.7–3 µm is mainly focused on the passive Q-switching technique. The typical traditional saturable absorber (SA), semiconductor SA mirror (SESAM), and Fe:ZnSe have been applied in Er-doped crystal laser resonators to generate nanosecond pulsed lasers at 2.7–3 µm. In 2018, Qin et al. realized a passively Q-switched Er:Y2O3 ceramic laser by using SESAM as the SA, generating a pulse energy of 1.7 µJ and a pulse duration of 350 ns at 2709.3 nm^[98]. In 2019, Zhang et al. reported a sub-15-ns passively Q-switched Er:YSGG laser at 2.8 µm with Fe:ZnSe as the SA, in which a pulse energy of 5.05 µJ and a pulse width of 14.6 ns were obtained^[99]. The schematic of the experimental setup and the passive Q-switching output characters are shown in Figs. 3(c) and 3(d). The Fe:ZnSe crystals with a doping concentration of 0.18% and thickness of 0.6 and 0.8 mm (corresponding to the initial transmissions of 55.4% and 50.2%) were used as the SAs. In order to get the short pulse width, the output coupler was directly attached to the rear surface of the Fe:ZnSe crystal to form the “microchip” structure and compress the cavity length to be approximately equal to the length of the laser crystal and SA (∼5.6 and 5.8 mm). Nevertheless, SESAM and Fe:ZnSe suffer several drawbacks, including complex and costly fabrication process, narrow saturable absorption band, and slow recovery time, which significantly limit their applications. Since the first, to the best of our knowledge, demonstration of a graphene SA in 2009^[100], studies on 2D materials-based SAs have experienced a boom in development due to the advantages of fast relaxation time, proper modulation depth and saturation intensity, broad operation wavelength band, and easy fabrication. In 2017, the passively Q-switched Er:YSGG lasers at ∼2.8µm were demonstrated by utilizing the composite Bi2Te3/graphene and ReS2 as SAs, generating the pulse width of 243 and 324 ns, respectively^[101,102]. By using bismuth nanosheets (Bi-NSs), MXene Ti3C2Tx, and black phosphorus (BP), Liu et al. realized Q-switched Er3+-doped lasers with pulse widths of 980, 814, and 702 ns, respectively^[103–105]. Besides, as the typical materials of transition metal dichalcogenides (TMDs), MoS2, WS2, ReSe2, and TiSe2 were also applied in MIR passively Q-switched lasers to generate the shortest pulse durations of 335, 679, 202, and 160 ns, respectively^[106–109]. The performances of diode-end-pumped passively Q-switched Er-doped crystalline lasers are summarized in Table 2.

Besides the Q-switched pulsed lasers, the mode-locked Er-doped ultrafast lasers are of great interest for some practical applications, owing to the ultrashort pulse width and high peak power. Picosecond or even femtosecond CW mode-locked Er-doped fiber lasers have been extensively studied and realized^[111–116], while only Q-switched mode-locked operation was obtained for Er-doped crystalline lasers because the intracavity pulse energy could not be high enough to reach the stable CW mode-locking regime caused by the relatively low gain and absorption of H2O. In 2018, Xue et al. realized a stable Q-switched mode-locked 2.7 µm Er:Y2O3 ceramic laser with SESAM, generating an average output power of ∼92mW with 100% modulation depth and 130 MHz repetition rate embedded in the Q-switched envelope of ∼1.2µs width^[85]. Liu et al. reported passively Q-switched mode-locked Er:CaF2-SrF2 lasers with the repetition rate of 136.3 MHz inside the Q-switched envelope and a pulse width estimated to be 1.78 ns^[71].

The Ho3+ ion is another promising candidate for generating 2.7–3 µm lasers related to the I65→I57 transition, as shown in Fig. 4(a). However, Ho-doped crystalline lasers emitting in the 2.7–3 µm spectral region are much less studied compared to those emitting around ∼2.1µm and Er3+-doped lasers. The main limiting issues are the pumping wavelength around ∼1150nm (not the commercial emitting wavelength of a LD) and the same self-terminated effect occurring with the Ho3+:I56→I57 transition (Ho3+:I57 has a longer lifetime than Ho3+:I56, resulting in the lower laser level during oscillation). The same as the Er3+:I411/2→I413/2 transition, the saturation of the Ho3+:I56→I57 transition can also be suppressed by cascade lasing (Ho3+:I56→I57 and Ho3+:I57→I58 transitions) or co-doping with sensitized (typically Yb3+ ions) or deactivated ions (typically Nd3+ and Pr3+ ions)^{[5,9,39,117–125]}. The host materials are also focused on the garnet structure and fluoride.

In the beginning, Ho-doped crystalline lasers operating in the 2.7–3 µm spectral region were mainly pumped by a flashlamp or pulsed laser due to lack of pumping source and the population bottleneck effect. In 1987, Machan et al. realized the simultaneous lasing of Nd3+ and Ho3+ ions at 1.064, 1.339, 2.94, and 3.011 µm with a flashlamp-pumped Ho:Nd:YAG crystal, indicating that the strong ion-ion interaction could produce efficient 3 µm lasing^[117].

In 1990, Anthon reported the first laser (Q-switched Nd:YAG laser operating at 1123 nm) pumped 3 µm Ho:YAG and Ho:GGG laser^[126]. In 1996, Umyskov et al. demonstrated a flashlamp-pumped Cr3+:Yb3+:Ho3+:YSGG laser with the emission wavelength continuously tuned from 2.84 to 3.05 µm, in which co-doping with Cr3+ and Yb3+ ions could efficiently increase the absorption of the pump light by the energy transfer of Cr3+→Yb3+→Ho3+^[118,127]. In 2002, Lukashev demonstrated a flashlamp-pumped Cr,Yb,Ho:YSGG laser at 3 µm with an output energy of 62 mJ^[128].

In 1998, Diening et al. realized 11 and 2.5 mW CW laser output at 2.84 µm with an Yb3+:Ho3+:KYF4 crystal pumped by a Ti:Al2O3 laser and LD, respectively^[129]. In 2000, they reported another lasing of an Yb,Ho:YAG crystal around 1.2 and 3 µm, in which the quasi-CW laser emission at 2844 nm with pulse energy up to 10.5 mJ was obtained^[119]. In 2017, our group realized a 1150 nm LD end-pumped Ho,Pr:LLF laser with CW output power of 172 mW and a slope efficiency of 10.8%^[9]. Then, by using a high-power and high-beam-quality 1150 nm Raman laser as the pump source, the CW output power was promoted to 1.15 W with a slope efficiency of 15.5%^[130]. In 2019, by optimizing the doping concentration of Ho3+ and Pr3+ ions in the Ho,Pr:YLF crystal, the lifetime of Ho3+:I56 was designed to be larger than that of Ho3+:I57, as shown in Fig. 3(b). Thus, as high as 1.27 W CW laser output with a slope efficiency of 28.3% was obtained^[131], as shown in Fig. 4(c). By using the dual-end-pumping configuration, a maximum output power of 1.46 W was obtained with a slope efficiency of 7.7%. To the best of our knowledge, it is the largest CW output power ever obtained with Ho-doped crystalline lasers^[132]. Table 3 summarizes the flashlamp-pumped and CW laser performance of Ho-doped MIR lasers in the 2.7–3 µm region.

In the pulsed regime, besides the microsecond pulse generated by pumping with the flashlamp and pulsed LD, nanosecond pulses were obtained with the active and passive Q-switching techniques. For passive Q-switching operation, SAs are mainly focused on low-dimensional materials. In 2017, our group realized a 2.95 µm diode-end-pumped passively Q-switched Ho,Pr:LLF laser with graphene as an SA, generating a maximum average output power of 88 mW with pulse width of 937.5 ns and repetition rate of 55.7 kHz^[9]. Then, by using BP as the SA and a Raman fiber laser as the pump source, a pulse width of 194.3 ns with a repetition rate of 158.7 kHz and average output power of 385 mW was obtained^[135]. With other low-dimensional materials as SAs, such as 2D TMDs (MoSe2, TiSe2, etc.), graphitic carbon nitride (g-CN), and gold nanospheres, passively Q-switched laser operation with the pulse widths of 731.5, 160.5, 420, and 743 ns was obtained^[136–139]. With the SESAM as the SA, Liu realized a passively Q-switched Ho,Pr:LLF laser at 2.9 µm with a pulse duration of 395 ns and repetition rate of 7.29 kHz^[140]. For active Q-switching operation, a high-repetition-rate (compared to flashlamp pumping) kilohertz (kHz) actively Q-switched Ho,Pr:YLF laser at 2.9 µm was realized with LN as the electro-optical Q switch, in which the shortest pulse width of 25.2 ns was obtained with the repetition rate of 500 Hz and single pulse energy of 0.4 mJ^[132].

The corresponding schematic experimental setup and the relationship between the output power and incident pump power are shown in Figs. 4(d) and 4(e). Table 4 summarizes the actively and passively Q-switched laser performance of Ho-doped crystalline lasers in the 2.7–3 µm region.

The Dy3+ ion is also a most promising and efficient candidate for emitting MIR laser wavelengths around ∼3µm based on its energy-level structure with the Dy3+:H13/26→H15/22 transition. The study of Dy-doped MIR lasers is much less than that of Er3+ and Ho3+ ions, basically because of the lack of high-quality crystals and pump sources. Figure 5(a) shows the simplified energy-level diagram of the Dy3+ ion. The absorption bands of the Dy3+ ion are located in the near-infrared region (around 1.1, 1.3, and 1.7 µm). Same as the Er3+ and Ho3+ ions, the possibility of realizing laser emission from the Dy3+:H13/26→H15/22 transition around ∼3µm depends on the host crystal material choice, which should possess low photon energy and weak ion to crystal lattice orbital coupling and therefore can efficiently decrease corresponding non-radiative losses and increase the quantum efficiency. To date, the Dy3+:H13/26→H15/22 transition has been obtained in fluoride crystals, such as Dy:BaYb2F8 and Dy:BaY2F8. In 1973, Johnson et al. demonstrated a flashlamp-pumped Dy:BaY2F8 laser operating at 3.022 µm^[141]. In 1982, a Dy3+:BaYb2F8 MIR laser at 3.02 µm pumped by a 1.06 µm Nd:YAG laser was demonstrated^[142]. In 1997, Djeu et al. realized a room temperature Dy3+:BaYb2F8 laser at 3.4 µm pumped by a pulsed 1.3 µm Nd:YAG laser^[142]. Unfortunately, there are no reports about CW Dy3+-doped crystalline lasers. But, for the fiber laser, Jackson has realized room temperature 2.9 µm CW laser emission with output power of 0.275 W and slope efficiency of 4.5% from a 1100 nm fiber-laser-pumped Dy:ZBLAN fiber in 2003^[18]. In 2006, they realized a CW Dy:ZBLAN fiber laser with a maximum output power of 180 mW and a slope efficiency of 20%, in which the pump source was an ∼1.3µm Nd:YAG laser^[143]. In 2016, they reported a high-efficiency 3.04 µm Dy:ZBLAN fiber laser with a record slope efficiency of 51% pumped by a 2.8 µm Er:ZBLAN laser^[144]. Moreover, they also realized the acousto-optically and passively Q-switched Dy:ZBLAN fiber laser with the central wavelength tunable from 2.97 to 3.23 µm^[145]. The actively Q-switched Dy:ZBLAN fiber laser schematic is shown in Fig. 5(b), where the acousto-optic tunable filter (AOTF) comprises an anisotropic TeO2 crystal in “slow shear operation.” The corresponding actively Q-switched laser characterization with 450 mW pump power is shown in Fig. 5(c). In 2019, Luo et al. realized a gain-switched Dy3+-doped ZBLAN fiber laser around 3 µm by using an actively Q-switched Yb3+-doped fiber laser at 1.1 µm as the pump source, yielding a repetition rate of 80 kHz and a pulse width of 300 ns^[146].

However, it is a real drawback for establishing a compact MIR laser because of the pumping wavelength not corresponding to any commercially available high-power LDs. Therefore, researchers try to study the sensitized ions that can transfer pumping energy to the Dy3+ ion to allow optical pumping with commercially available LDs. Yb3+ ions have been proved to be the most efficient sensitized ions for Dy3+ ions-doped 3 µm MIR laser emission with energy transformation from Yb3+:F25/2 to Dy3+:H65/2, which enables it to be pumped by the commercial 970 nm LD^[147,148]. The simplified energy-level diagram of Yb3+, Dy3+ co-doped crystals is shown in Fig. 5(b). In 2015, Zhang et.al presented the successful growth of a Yb3+, Dy3+:PbF2 crystal, in which the energy-transfer efficiency from Yb3+ to Dy3+ was as high as (97.7±0.3)%. In addition, this crystal possesses long fluorescence lifetime of 15.4 ms and high quantum efficiency of 95%^[148].

Expanding the laser wavelength to the MIR region is one of the most important developing trends of laser technology. To date, laser sources with directly emitting wavelengths at 2.7–3 µm are mainly based on Er3+, Ho3+, and Dy3+ rare-earth ions-doped gain media, in which Er3+ ions are mostly studied, Ho3+ ions take second place, and Dy3+ ions are the least studied. Compared to the rare-earth-doped fiber lasers, the rare-earth-doped crystalline lasers experience a relatively slow development mainly because of the lack of high-quality laser crystals. But, the solid-state crystalline lasers are compact and efficient all solid-state coherent laser sources with the merits of low undesirable nonlinear effects and large mode area and therefore have great advantages in producing high-energy and high-peak-power ultrafast lasers. In this review, we mainly summarize the state-of-the-art developments of all solid-state MIR crystalline lasers in the 2.7–3 µm spectral region based on Er3+, Ho3+, and Dy3+-doped crystals. However, there are still several challenges, and a series of potential studies need to be further pursued in the future.

First, the host material selection and the preparation of the high-quality crystals are the basis for high-power and high-efficiency solid-state MIR crystalline lasers in the 2.7–3 µm region. The longer the emitting wavelength, the narrower the bandgap between the upper and lower laser level, which, thus, results in the larger non-radiative transition loss. Therefore, for MIR laser emission, the host material should have low phonon energy to reduce the probability of non-radiative transitions. In addition, the host materials should have large thermal conductivity to mitigate the relatively heavy thermal effect of the MIR crystalline lasers. The damage threshold is another important issue for high-power and high-energy laser operation.

Second, the selection of sensitized and deactivated ions and the doping concentration are also important for rare-earth-doped crystalline lasers at 2.7–3 µm. For Er3+ and Ho3+ ions-doped crystals, the deactivated ions are important for solving the “self-terminated” bottleneck to realize the high-efficiency laser operation. For Ho3+ and Dy3+ ions-doped crystals, the sensitized ions are important for selecting the commercial LD as the pump source. Besides, the doping concentrations of both excited and sensitized or deactivated ions should be further optimized.

Third, cascade laser operation is very attractive for multi-wavelength MIR laser generation. Based on the energy-level diagram of Er3+, Ho3+, and Dy3+ ions, the cascade laser operation not only provides a multi-wavelength MIR laser source, but also enhances the 2.7–3 µm laser generation. The cascade laser operation of Er3+ and Ho3+ ions-doped fiber lasers has been realized, while it still remains a big challenge for crystalline lasers.

Fourth, mode-locked laser operation is another challenge for rare-earth-doped crystalline lasers at 2.7–3 µm. The mode-locked laser operation in the near-infrared (1.0, 1.3, 1.5 µm) and MIR (2.0, 2.4 µm) regions has been widely studied, and picosecond or even femtosecond pulses have been generated. Due to the lack of suitable SAs and the absorption of H2O, it is very difficult to achieve the mode-locking operation of rare-earth-doped crystalline lasers at 2.7–3 µm. However, with the innovations of ultrafast laser technology and material science, it is something to look forward to and can be widely applied in the fields of strong field physics, optical frequency comb, ultrafast spectroscopy and microimaging, etc.