As a valuable spectral region, mid-infrared 2–20 μm contains strong characteristic vibrational transitions of a series of important molecules and two atmospheric transmission windows (i.e., 3–5 μm and 8–13 μm) ^[1]. Motivated by growing demands in the fields of medicine, industry, security, etc., lasers in this spectral region have attracted increased interest in the last several decades. Nowadays, many laser technologies are serving this range, e.g., optical parametric oscillator ^[2], quantum cascade laser ^[3], and doped insulator laser ^[4]. The fiber laser, as one of the doped insulator lasers ^[5], has been strongly developed because of its great power scalability, near-diffraction-limited beam quality, compact and robust structure, and recently, towards the mid-infrared region ^[6]. Among them, the ∼2.8μm transition (I11/24→I13/24) in Er3+-doped ZrF4 fiber laser was paid most attention to owing to the mature 976 nm diode pumping technique and efficient energy transfer upconversion (ETU), enabling lasing beyond the Stokes limit ^[7]. The continuous wave (CW) laser at 2.82 μm was scaled to 40.6 W power recently by Aydin et al. ^[8], and the power was further boosted to a record 65 W under quasi-continuous wave (QCW) by Newburgh et al. very recently ^[9]. Based on this platform, the comparative efforts on pulse operation by Q switching, gain switching, and mode locking have been widely reported ^[10]. For example, Tokita et al. presented an actively Q-switched Er3+-doped ZrF4 fiber laser at 2.8 μm, giving an average power of up to 12 W and a pulse energy of 100 μJ at the repetition rate of 120 kHz ^[11]. In the same way, Lamrini et al. scaled the pulse energy to 560 μJ at the repetition rate of 1 kHz ^[12]. Wei et al. utilized a Fe2+:ZnSe crystal to realize high-power passive Q switching, tunable in the range of 2762.5–2852.5 nm, where the maximum average power of 5.16 W and pulse energy of 27.7 μJ were achieved at 2.81 μm ^[13]. Paradis et al. demonstrated a gain-switched Er3+-doped ZrF4 fiber laser at 2.826 μm with an average power of 11.2 W and a pulse energy of 80 μJ ^[14]. Duval et al. reported the first femtosecond-level mode-locked Er3+-doped ZrF4 fiber laser in a ring cavity using nonlinear polarization rotation ^[15], yielding a peak power of 3.5 kW with a pulse width of 207 fs. Because of the longer emission band around 3 μm, I65→I75 and H13/26→H15/26 transitions in Ho3+- and Dy3+-doped ZrF4 fibers have also received much attention ^[16–25]. So far, their CW and pulsed versions have covered the broad range of 2.7–3.4 μm, with some shining results, such as a record peak power of 37 kW from a mid-infrared fiber oscillator ^[18], and a record tuning range of ∼600nm (2.8–3.4 μm) from a rare-earth-doped fiber oscillator ^[21].

At longer wavelengths of ∼3.5μm, Többen et al. demonstrated lasing from an Er3+-doped ZrF4 fiber laser based on F9/24→I9/24 transition pumped at 653 nm at a temperature of 77 K as early as 1991 ^[26]. Poor performance due mainly to the ions bottlenecking issue caused by the long-lived excited states, however, limited its development in the following several decades. As a result, its applications in polymer processing (e.g., methyl methacrylate, polypropylene, polyethylene) ^[27] and gas monitoring (e.g., methane, propane, formaldehyde) ^[28] are also blocked. Until 2014, efficient emission potential of this transition was unlocked by Henderson-Sapir et al. for the first time using the proposed dual-wavelength pumping (DWP) (985nm+1983nm) concept, where the ions bottlenecking issue was solved ^[29]. And beyond 200 mW, CW lasers around 3.5 μm with an optical-to-optical efficiency of 16% were reached at room temperature ^[29]. Subsequently, Henderson-Sapir and colleagues realized ∼450nm (3.33–3.78 μm) CW wavelength tuning ^[30], and Maes et al. scaled the average power to 5.6 W at 3.55 μm ^[31]. Meanwhile, the theoretical analyses were also made to deeply understand the dynamics of this system ^[32–34]. Along with the success in CW, its pulsed versions were constructed owing to higher peak power and pulse energy. Jobin et al. presented the first gain-switched DWP Er3+-doped ZrF4 fiber laser at 3.55 μm ^[35]. A high peak power of up to 204 W and a pulse energy of 6.83 μJ were achieved at the repetition rate of 15 kHz ^[35]. Then we scaled the average power to 1.04 W by a factor of 10 by using microsecond-scale pulse pumping to relieve power quenching during one pump pulse ^[36]. In comparison to gain switching, Q switching just needs a cost-effective CW pump source while avoiding the possible relaxation oscillation spikes. And the availability of all-fiber passive modulators using the emerging low-dimensional materials also pushed the progress of Q switching technique. In fact, self-Q-switching behavior was observed by Henderson-Sapir et al. in their first demonstration of DWP ∼3.5μm emission ^[37]. Then they numerically predicted the feasibility of active Q switching in such a system ^[38] and experimentally realized actively Q-switched pulses using an acousto-optic modulator, yielding a peak power of 14.5 W and a pulse energy of 7.8 μJ ^[39]. Qin et al. used black phosphorus (BP) to present passive Q switching and mode locking from a DWP Er3+-doped ZrF4 fiber laser at ∼3.5μm ^[40]. However, the low damage threshold of BP led to the limited average power of 120 mW and pulse energy of 1.8 μJ ^[40]. Low-dimensional materials represented by BP have been serving pulsed fiber lasers as saturable absorbers (SAs) from the visible to the mid-infrared region for almost one decade owing to their broadband feature ^[41,42]. Although overcoming the narrow operation band of the traditional semiconductor SA mirror, they most contribute to quite moderate output due to their low damage thresholds.

Fe2+:ZnSe is another alternative SA applicable in the mid-infrared region as a result of its strong absorption in the range of 2.5–4.5 μm. More importantly, it has a damage threshold of up to ∼2J/cm2 with a small saturation, suitable for powerful and energetic Q switching. To date, ∼2.8μmEr3+- and ∼2.9μmHo3+-doped ZrF4 fiber lasers Q-switched by Fe2+:ZnSe have been widely reported ^[13,43–46], of which output scaling was limited by pump power or damaged fiber ends instead of Fe2+:ZnSe itself. At wavelengths beyond 3 μm, although the absorption decreases with the wavelength ^[47], which implies an increased initial transmission, and hence a decreased modulation depth for a certain Fe2+:ZnSe crystal, it should still be effective as an SA at <4.5μm in principle. However, no relative efforts have been reported so far, to the best of our knowledge.

In this paper, we extend the Fe2+:ZnSe-based Q-switching technique to operate in a tunable region of 3.4–3.7 μm, based on a DWP (976nm+1981nm) Er3+-doped ZrF4 fiber laser. This is the first tunable pulsed rare-earth-doped fiber laser in this band. Attributable to the high damage threshold of the Fe2+:ZnSe SA, the output average power and pulse energy of passive Q switching are significantly scaled to 583.7 mW and 9.07 μJ (not simultaneously), only limited by the available pump power at 1981 nm. The output characteristics, including pulse width, repetition rate, average/peak power, and pulse energy with the varied pump power and operation wavelength, were also studied.

Figure 1 shows the experimental setup of our constructed tunable passively Q-switched Er3+-doped ZrF4 fiber laser based on a Fe2+:ZnSe crystal around 3.5 μm. The pump unit includes two laser sources, one of which is a commercial 976 nm laser diode (LD) (BWT, China) with a fiber pigtail [core/numerical aperture (NA): 105/0.22] yielding a maximum output power of 30 W; the other is an in-house CW 1981 nm Tm3+-doped fiber oscillator whose cavity is formed by a pair of fiber Bragg gratings (FBGs) with an SMF-28e fiber pigtail yielding a maximum output power of 8 W. The 976 nm laser is collimated using an objective lens (L1, Melles Griot) with a focal length of 10 mm, and the 1981 nm laser is collimated using an aspheric mirror (L2, C240TME-C, Thorlabs) with a focal length of 8 mm. They are combined using a dichroic mirror (DM1, DMSP1500, Thorlabs) placed at an angle of 45° relative to the light axis and then launched into the cladding and core of the gain fiber using an objective lens (L3, LFO-5-6, Innovation Photonics) (transmission 90% at 1981 nm, 70% at 976 nm, and 95% at ∼3.5μm) with an efficiency of 76% and 86%, respectively. A dichroic mirror (DM2, transmission 97% at 1981 nm, 85% at 976 nm, and reflection >98% at ∼3.5μm) is placed between L3 and DM1 to steer the laser from the F9/24→I9/24 transition (see the inset of Fig. 1). After that, a commercial bandpass filter (Thorlabs, FB3500-500) is used to purify the ∼3.5μm laser. The gain medium is a piece of 3 m double-cladding Er3+-doped ZrF4 fiber (Le Verre Fluoré, France) with a dopant concentration of 1.5 mol.%. The core has a diameter of 16.5 μm with an NA of 0.15, and the inner cladding has a diameter of 260 μm with two parallel flats separated by 240 μm. Its outer cladding is a low-index fluoroacrylate polymer with a diameter of 290 μm. The fiber end close to the pump is perpendicularly cleaved and butted against a dichroic mirror (DM3, transmission 95% at 1981 nm, 96% at 976 nm, and reflection ∼60% at ∼3.5μm) acting as the cavity feedback and output coupler. The other fiber end is cleaved at an angle of 8° to suppress ∼2.8μm oscillation and ∼3.5μm parasitic laser. The ∼3.5μm laser from the angle-cleaved fiber end is collimated using an off-axis parabolic reflector (L4, Thorlabs, MPD00M9-P01) with a focal length of 15 mm to avoid focal position shift over the entire tuning range. A dichroic mirror (DM4), the same as DM2, is placed at an angle of 45° to reflect the ∼3.5μm laser while the transmitted residual 976 and 1981 nm laser is collected using a beam trap. The reflected laser is resonated by a plane-ruled grating (Thorlabs, 450 lines per mm, blaze wavelength λB=3.1μm, blaze angle θB=32°) in a Littrow configuration. The wavelength can be tuned by rotating the grating. A confocal scheme including two commercial uncoated CaF2 plano–convex lenses (L5, L6, Thorlabs, LA5763) with a focal length of 50 mm is placed between the grating and DM3. The Fe2+:ZnSe crystal with a size of 6.7mm×6.7mm×1.6mm and a Fe2+ doping concentration of 0.9×1018cm−3 (provided by the manufacturer) is mounted and connected with a one-dimensional platform, and then inserted into the confocal scheme and carefully adjusted near the focal point. Its initial transmission at ∼3.5μm was measured to be ∼80% using a CW laser source with a low average power. Due to lack of a high energy pulsed source at ∼3.5μm, its saturable absorption parameters were not measured. Nevertheless, its modulation depth could be roughly estimated to be close to ∼20% considering the small unsaturation loss ^[48]. Notice that no heat management was imposed on it. Although the relatively long upper state lifetime of 370±25ns for Fe2+:ZnSe at room temperature ^[47] makes it difficult to realize megahertz-level high repetition rate Q-switching, this is in turn beneficial for higher pulse energy owing to lower photon loss caused by the SA. In our experiment, the position of the Fe2+:ZnSe crystal needs to be optimized to maximize the output power while producing stable Q-switched pulses. The laser power was recorded using a high-resolution thermal power sensor (Thorlabs, S405C), the temporal pulses were captured using an InAs detector (Judson J12D) with a response time of 2 ns connected with a 500 MHz bandwidth digital oscilloscope, and the optical spectrum was measured using a monochromator (Princeton Instrument Acton SP2300). Note that the laser powers mentioned in this paper are all corrected.

First, we adjusted the launched 1981 nm pump power to the available maximum 5.25 W and then increased the launched 976 nm pump power to 2.47 W while rotating the grating to maximize the ∼3.5μm output power. After that, the laser output characteristics as the varied launched 1981 nm pump power were recorded, keeping the launched 976 nm pump power at 2.47 W and the position of the grating unchanged.

Increasing the launched 1981 nm pump power to 2.10 W, a Q switching sign with quite low amplitudes and serious fluctuations first appeared. Further increasing the pump to 2.24 W, stable Q-switched pulses were achieved as displayed in Fig. 2(a), yielding a repetition rate of 30.12 kHz and a pulse width of 2.67 μs. This stable Q-switching regime could be maintained until the maximum launched 1981 nm pump power of 5.25 W, as shown in Fig. 2(d), giving a low amplitude fluctuation of ±1.4%. The temporal behaviors at the middle 1981 nm pump powers of 2.78 and 4.19 W were also recorded, as shown in Figs. 2(b) and 2(c), respectively. At the maximum 1981 nm pump power, the optical and RF spectra were measured, as shown in Figs. 3(a) and 3(b), respectively. It is seen that only one strong spectral peak centered at 3478.0 nm with a narrow FWHM of 0.4 nm is observed, different from the several spectral lines in free-running cases ^[36,39,40]. This suggests that the plane-ruled grating in our case takes effect. The signal-to-noise ratio (SNR) of >35dB for the fundamental frequency is also located within the typical range of stable passive Q switching ^[43–46].

Figure 4(a) plots the variations of repetition rate and pulse width as the 1981 nm pump power was launched. It is observed that both undergo the same variation trend as standard passive Q switching. At the maximum launched 1981 nm pump power, the largest repetition rate of 71.43 kHz and shortest pulse width of 1.18 μs were achieved, while the average output power reached the maximum 583.7 mW, with a slope efficiency of 15.2% relative to the launched 1981 nm pump power, as shown in Fig. 4(b). The corresponding pulse energy was calculated to be 7.54 μJ. Further power and energy scaling was limited by the available 1981 nm pump power.

Then wavelength tuning of the Q-switched fiber laser was performed. Figure 5 displays the recorded optical spectra and corresponding average output powers when rotating the grating at the launched 1981 nm pump power of 5.25 W. The wavelength of stable Q-switched pulses can be continuously tuned within a ∼300nm range of 3407.2–3701.1 nm. The average power increases first and decreases with tuning towards the long wavelength, in agreement with the gain spectrum of this transition ^[29]. To identify the factors limiting further extension of the tuning range, considering that it is much narrower than the ∼450nm CW tuning range of 3330–3780 nm ^[30], we removed the confocal scheme and Fe2+:ZnSe crystal to reduce the intracavity insertion loss. As a result, only a moderately extended ∼345nm CW tuning range of 3385.2–3730.6 nm was obtained. It suggests that the tuning range extension in our system is mainly limited by the available 1981 nm pump power, for which it is quite difficult to make gains at the wavelengths outside the range reaching laser thresholds.

Figure 6(a) shows the variations of repetition rate and pulse width with the tuned wavelength. With tuning from 3407.2 to 3701.1 nm, the repetition rate changes between 16.67 and 71.43 kHz, with a similar variation trend to that of the average power shown in Fig. 5. This is because higher intracavity laser power leads to faster bleaching of the Fe2+:ZnSe crystal, thus to a larger repetition rate, while the pulse width varies between 2.98 and 1.16 μs. Over this wavelength range, the varied pulse peak powers and energies were also recorded, as shown in Fig. 6(b). Almost the same variation trends as that of the average power are observed. At the tuned wavelength of 3598.3 nm, the maximum peak power of 7.82 W and pulse energy of 9.07 μJ were achieved.

In summary, we have presented a tunable passively Q-switched DWP Er3+-doped ZrF4 fiber laser around 3.5 μm using a Fe2+:ZnSe crystal as the SA and a plane-ruled grating in a Littrow configuration as the tuning element. To the best of our knowledge, this is also the first tunable pulsed laser based on rare-earth-doped fibers at ∼3.5μm. At the tuned wavelength of 3478.0 nm, stable Q-switched pulses could yield a maximum average power of 583.7 mW with the corresponding pulse width, repetition rate, and pulse energy of 1.18 μs, 71.43 kHz, and 7.54 μJ, respectively. Further power scaling is only limited by the available 1981 nm pump power. By rotating the grating, the wavelength of Q-switched pulses could be tuned continuously from 3407.2 to 3701.1 nm; further extension is mainly limited by the available 1981 nm pump power. The work in this paper has extended the Fe2+:ZnSe-based Q-switching technique to operate in the longest 3.5 μm band for the first time, to the best of our knowledge. In the next step, power scaling will be continued, aided by the advantageous high damage threshold of the Fe2+:ZnSe crystal. Moreover, we will also attempt to utilize the Fe2+:ZnSe crystal to realize passive mode locking at ∼3.5μm, considering its success in a 2.8 μm mode-locked Er3+-doped ZBLAN fiber laser ^[49]. Specifically, we plan to further increase the intracavity intensity based on a ring cavity while improving the SA initial transmission by reducing its doping concentration and thickness to make the CW mode-locking condition easier.

Acknowledgment. H. Y. Luo and J. Yang contributed equally to this work. H. Y. Luo and J. F. Li designed the experiment. H. Y. Luo prepared the paper and discussed it with J. F. Li. J. Yang built up the system and finished the measurements with H. Y. Luo. Y. Liu supervised the project.