There are increasing demands for high-power solid-state mid-infrared lasers at 3–5 µm for their applications in laser radar, spectroscopy, remote sensing, infrared countermeasures, laser communication, etc.^[1–3]. Currently, the principal category of solid-state lasers at this band is based on the nonlinear effect, including optical parametric oscillators (OPOs, typically using PPLN and ZGP crystals as nonlinear media)^[4,5], difference frequency generation (DFG)^[6], and frequency doubling^[7]. Yet there are only a few direct solid-state mid-infrared 3–5 µm laser sources, merely including quantum cascade lasers^[8,9], fiber lasers^[10], and transition metal (TM) ion-doped crystalline lasers (typically Fe:ZnSe or Fe:ZnS lasers)^[1,11]. Compared with these lasers, Fe:ZnSe or Fe:ZnS lasers, emitting at the mid-infrared wavelengths of 4–5 µm, enjoy the impressive advantages of high efficiency, wide wavelength-tuning range, compactness of optical cavity, etc.^[12]. As such, much effort has been made in the development of Fe:ZnSe/Fe:ZnS lasers in the past decade. Since Fe:ZnSe crystal was demonstrated to be an effective gain medium for 4 µm lasers for the first time in 1999^[1], many Fe:ZnSe lasers in different regimes, including continuous-wave (CW)^[2], gain-switching^[3], Q-switching^[13,14], and mode-locking^[15], have been demonstrated by employing various pumping sources at ∼3μm. For some practical applications, high-power and highly stable CW 4 µm lasers are urgently required. The upper laser level lifetime of Fe2+ in ZnSe crystal is ∼60μs at 77 K and as short as ∼370ns at room temperature^[16]; as a consequence, it is necessary to cool Fe:ZnSe crystal down to ∼77K by liquid nitrogen in a cryostat for efficient CW operation. In 2008, Voronov et al. reported the first operation of a CW Fe:ZnSe laser pumped by a 2.97 µm Cr:CdSe laser^[17], of which the maximum output power and optical-to-optical (OTO) efficiency (i.e., the laser output power to the power of pump source output end) were 160 mW and 27%, respectively. Subsequently, Evans et al. demonstrated CW Fe:ZnSe lasers with output powers of 840 mW and 420 mW by using two Er:YAG lasers at 2.94 µm^[2] and an Er:Y₂O₃ laser at 2.74 µm^[18], and the corresponding OTO efficiencies were ∼28% and 13.3%, respectively. Martyshkin et al. developed a Cr:ZnSe laser at 2.94 µm with an output power of 23 W for power scaling. With the powerful pump source, they achieved a CW Fe:ZnSe laser with a maximum power of 9.2 W and an OTO efficiency of ∼40%^[19]. Recently, Li et al. presented a CW Fe:ZnSe laser pumped by an Er:YAP laser^[20], whose output power and OTO efficiency were ∼1W and 27.8%, respectively. Compared to solid-state pump lasers, fiber lasers possess the advantages of high beam quality and high stability, which are supposed to be ideal sources for pumping. Thanks to the dramatical development of fluoride fiber lasers at ∼3μm in the past decades^[21–23], of which the recorded CW output power was as high as 41.6 W^[23], it is natural to employ a fiber laser at ∼3μm to pump Fe:ZnSe crystals. In 2018, Pushkin et al. demonstrated a compact CW Fe:ZnSe laser with a maximum power of 2.1 W and an OTO efficiency of 32% pumped by an Er:ZBLAN fiber laser at 2.8 µm^[24]. Table 1 presents a comparison of state-of-the-art CW Fe:ZnSe lasers. The lack of a high-power CW pump source seems to be the main obstacle for 4 µm laser power scaling. However, promotion of OTO efficiency is more important for obtaining higher output power when the pump power is limited.

In this work, we report on a high-power, high-efficiency, and low-complexity Fe:ZnSe laser by adopting a series of innovation, including customization of a pair of special coatings into the Fe:ZnSe crystal working facets for lower cavity loss and a longer wavelength pump source for higher pump absorption and higher quantum efficiency. The maximum output power of 5.5 W was obtained under liquid nitrogen cooling, which is the highest in fiber-laser-pumped Fe:ZnSe lasers. The overall OTO efficiency with respect to the pump source was as high as ∼54.3%. The central wavelength was tuned from 3.96 to 4.15 µm by improving the pump power. The laser was considerably compact and nearly alignment-free, which led to fairly high long-term stability.

The output of our CW Fe:ZnSe laser was compared under two different pump schemes, i.e., forward pumping (FP) and backward pumping (BP). The experimental setup of the BP scheme is schematically shown in Fig. 1.

Compared with the pump source of a 2.8 µm fiber laser in the previous demonstration^[24], an all-fiber laser at 2.9 µm was developed to pump the Fe:ZnSe crystal in our experiment mainly based on the following considerations. One is that the longer wavelength of the pump results in a promotion of the quantum efficiency compared to pumping at the shorter wavelengths, and the other is that the absorption cross section at 2.9 µm of Fe:ZnSe at a low temperature (LT) of ∼77K is larger than at 2.8 µm^[18]. In 2015, a maximum output power of 30.5 W was obtained from an Er:ZBLAN all-fiber laser at 2.94 µm^[22], which was an ideal source for pumping Fe:ZnSe. As the techniques of writing fiber Bragg grating (FBG) in ZBLAN fiber and splicing between a ZBLAN fiber and a silica fiber continue to mature, it has been common to develop an all-fiber laser at 3 µm waveband. A single-mode all-fiber CW Er:ZBLAN fiber laser with a maximum output power of ∼10.1W at a central wavelength of 2907.3 nm was in-house built to be used as the pump source. The pump beam was collimated with a collimator (f=12.7mm) and then focused into the Fe:ZnSe crystal through an uncoated CaF2 lens (L1, f=50mm). The pump beam waist diameter was measured to be around 400 µm. The Fe:ZnSe crystal, grown from the vapor phase by using a concurrent-doping technology^[17], was ∼9mm in length with a labeled Fe2+ ion concentration of 5.0×1018cm−3. The one-pass absorption coefficient was measured to be 5.14cm−1 at an LT of ∼77K with a similar crystal with broadband anti-reflection (AR) coated for 2.7–4.8 µm. Consequently, the Fe2+ ion concentration was determined experimentally to be 5.5×1018cm−3, in which the absorption cross section at λ=2907nm (σgsa=0.94×10−18cm2) was taken from Ref. [25].

For more compact cavity design, both working facets (7mm×7mm) of the crystal were polished carefully to reduce the residual wedge as far as possible. Two series of special coatings were deposited onto the polished facets. The coating of S1 (see Fig. 1) had a high transmittance at the pump wavelength (T≥95% at 2.6–3 µm) and a high reflectivity at the output wavelength (R>99% at 3.7–4.8 µm), while the coating of S2 had a high transmittance at the pump wavelength (T≥95% at 2.6–3 µm) and a partial transmittance at the output wavelength (R∼65% at 3.7–4.8 µm), which served as the output coupler. The crystal was wrapped by a piece of indium foil and clamped to a U-shaped copper heat sink, and the temperature of the heat sink was controlled by a liquid-nitrogen-cooled cryostat from ∼77 to ∼300K to study the influence of crystal temperature on the laser conversion efficiency. The inset of Fig. 1 shows the partial energy levels of Fe2+ ions. The lifetime of upper laser level T52|2⟩ reduces significantly when the temperature of Fe:ZnSe crystal is higher than 120 K due to the increase of the rate of non-radiative relaxation^[26,27]. Theoretically, the longer lifetime of the upper laser level makes CW emission much easier. Therefore, almost all the CW Fe:ZnSe lasers are operated by cooling Fe:ZnSe crystals to LT, e.g.,  77 K. The vacuum windows (W1 and W2) of the cryostat were 3 mm CaF2 plates with broadband AR (T∼99%) coated for 2.7–4.8 µm. A dichroic mirror (DM, HR>99% at 2.6–3 µm, HR>95% at 3.7–4.8 µm) was placed with an incidence angle of 45° to separate the pump beam and output laser beam. FP was carried out by just turning around the cryostat in Fig. 1.

First, output characteristics of the Fe:ZnSe laser were compared under different pumping schemes at an LT of ∼77K. The laser cavity was self-aligned and had relatively low loss by customizing special coatings into the facets of the Fe:ZnSe crystal. It was easy to obtain the 4 µm laser output without any tedious alignment of the laser cavity in both schemes when the launched pump power was higher than the threshold of around 0.4 W. The output powers of FP and BP as a function of launched pump power are shown in Fig. 2. The maximum powers of FP and BP were 5.3 and 5.5 W at the full pump power of 10.1 W, respectively. The slope efficiencies were 61.9% in FP and as high as 66.3% in BP, corresponding to overall OTO efficiencies of 52.4% and 54.5%, respectively. Both the slope efficiency and the overall OTO efficiency are the highest ever reported in Fe:ZnSe lasers. The slope efficiency in BP was close to the limited laser efficiency, i.e., quantum efficiency (i.e., λpump/λlaser), which is 69.9% in our case. The results suggested that the performance of BP was slightly better than that of FP, which was similar to fiber lasers with different pump schemes^[28]. The laser cavity length, i.e., the length of the crystal, was ∼9mm, which was nearly 1 order of magnitude shorter than previously reported Fe:ZnSe lasers^[3]. Such a short cavity brought several advantages, including better cavity stability and lower cavity loss. In laser cavity design, an important parameter denoted as the Fresnel number can be expressed by N=d2/Lλ, where d is the diameter of the excitation region in crystal and L is the cavity length. The shorter the cavity, the higher the Fresnel number, which in turn leads to lower diffraction loss of the cavity. The geometric loss was very small as well due to the high parallelism of the cavity mirror. Thus, much lower cavity loss, through integrating the cavity mirrors and the crystal instead of the cavity mirrors, was mainly responsible for such high efficiency. Further in this work, if not specially stated, our considerations are confined to the laser in the BP regime. No thermal roll-off at the maximum pump power was observed, which indicated that further power scaling would be available by just increasing the pump power. The output beam profile of the BP at the maximum output power captured with a camera is shown in the inset of Fig. 2, from which we could find that the far-field beam spot had a good symmetrical Gaussian distribution. We measured the divergence of the output beam in the FP scheme with a CaF2 lens (f=100mm). The spot diameter at the focal plane was around 1.5 mm; therefore, the divergence was calculated to be 15 mrad (full angle).

The output spectra at various output powers were captured by a mid-infrared optical spectrum analyzer with a resolution as high as 0.2 nm, as typically shown in Fig. 3. The gain spectra of Fe:ZnSe at different temperatures are very broad, as shown in Fig. 4. The evolution trend of spectra was that the central wavelengths were red-shifted to longer positions with bandwidths broadening as the pump power increased.

The peak wavelength was 3.98 µm at the output power of ∼1W and then shifted to 4.15 µm at ∼5W. The red-shift of the central wavelength at different pump powers is very common in free-running solid-state lasers, which is interpreted as an increase of temperature of the pumping region by local heating of the gain crystal. A model was established to estimate the temperature distribution in the pumping area, and the highest temperatures generated in the center of pump beam spot were ∼82.5K at the output power of ∼1W and ∼96.1K at ∼5W, respectively. The increase of temperature was supposed to be responsible for such a wavelength shift. There was a dip in the spectrum at the output power of 5 W, which was attributed to the absorption by atmospheric CO2 in the path from the output chamber window to the optical spectrum analyzer when measuring the spectrum. Compared to the spectra of a spiky structure in previous reports, the spectra in our case had a relatively smooth curve. It is interesting that the curves had only one peak with an FWHM bandwidth (3 dB) of ∼20nm at a low output power, and the bandwidth was slightly narrower (∼12.3nm) at a high power. The signal-to-noise ratio (SNR) was measured to be roughly 25 dB for low power, and it was nearly 40 dB at the output power of 5 W.

The laser output was fairly stable by integrating the cavity coatings into the crystal to avoid any possible factors causing instability. The long-term stability of the laser at an output power of over 5 W over a 1 h period was measured with a power meter, as shown in Fig. 5. The peak-to-peak fluctuation was calculated to be approximately 0.7% at nearly the maximum output power.

The conversion efficiency of our experiment was even higher than the previous simulation of CW operation^[30], in which the slope efficiency at an LT of 77 K was only 42%. To gain insight into the laser dynamics and get an explanation for such high efficiency, we developed a numerical model of steady-state CW operation with the rate-equation theory. Only the ground state (E5|g⟩) and upper laser level (T52|2⟩) of Fe2+ (see the inset of Fig. 1) need to be considered in the calculation. The ground state population and the upper laser level population were denoted as N0 and N2, respectively.

The power evolution of pump Pp±(z) and laser signal Ps±(z) (+ forward, − backward) along the cavity axis (with longitudinal coordinate z) could be expressed as follows:±dPp±(z)dz=−σgsaN0(z)Pp±(z)−σesaN2(z)Pp±(z)−αpPp±(z),(1)±dPs±(z)dz=Ps±(z)(g(z)−l)+PASE,(2)where the gain g(z) isg(z)=ΓσseN2(z),(3)σgsa, σesa, and σse are the ground absorption, excited-state absorption, and emission cross sections of the Fe:ZnSe crystal, respectively; αp is the intrinsic absorption coefficient of the crystal; Γ is the overlap coefficient of pump spot and laser spot; and l is the total loss of the cavity including the intrinsic loss of crystal, geometry loss, and diffraction loss. PASE is the power of amplified spontaneous emission (ASE), which is given by^[31]PASE=Γ2hc2λs3σseN2(z)ΔASE,(4)where ΔASE is the ASE bandwidth centered at the signal wavelength, h is the Planck constant, c is the speed of light in vacuum, and λs is the laser wavelength. $$\begin{gathered} dN2(z,t)dt=λp(Pp+(z,t)+Pp−(z,t))πrp2hc(σgsaN0(z,t)−σesaN2(z,t)) \\ −λs(Ps+(z,t)+Ps−(z,t))πrs2hcσseN2(z,t)−N2(z,t)τf, \end{gathered}$$(5)where λp is the pump wavelength, τf is the fluorescence lifetime of the upper laser level, and rp and rs are the beam radii of the pump and laser, respectively.

The output power of the Fe:ZnSe laser could be expressed byPout=(1−R2)Ps+(L),(6)where R2 is the reflectivity of the output coupler and L is the position of the output end of the cavity.

We carried out the model with parameters in line with our experiment, as shown in Table 2.

The spectroscopic parameters of Fe2+ were taken from Ref. [25]. According to Ref. [32], the particles in the upper laser level (T52|2⟩) jump to the higher level via the excited-state absorption process by absorbing the pump photons not the laser photons, which is different from the process of the simulation in Ref. [30]. The ratio of σesa to σgsa was deduced to be 0.12 in terms of the experiment results in Ref. [32]. The simulation of output power with these input conditions under various cavity losses is shown in Fig. 6. The slope efficiency decreased as the total loss increased. The maximum slope efficiency was 63.1%, which was very close to our experimental result. That is, the cavity loss in our case was slightly lower than 0.05cm−1 due to such a compact cavity with a length of ∼1cm. The geometry loss and diffraction loss increased when the cavity length was lengthened by at least 1 order of magnitude in previous demonstrations^[18–20]. Therefore, reducing the loss as much as possible is beneficial for high conversion efficiency.

In conclusion, a high-power, high-efficiency fiber-laser-pumped all-solid-state CW Fe:ZnSe laser with a record overall efficiency as high as 54.5% was demonstrated. Output power of over 5 W with a central wavelength of around 4.1 µm was obtained from a Fe:ZnSe crystal under liquid nitrogen cooling. A numerical model was established to analyze factors influencing high-efficiency CW operation. The simulation was in line with our experiment, indicating that output power scaling to tens of watts or even higher could be possible by employing this configuration just by using higher pump power.