The mid-infrared spectral range is known as the “molecular fingerprint” region. Particularly, the 2–3 μm section overlaps the fundamental vibrational absorption lines of water vapor, carbon dioxide, and the other gases^{[13" target="_self" style="display: inline;">–3]}. Broad-emission-bandwidth lasers operating in this region have great significance in numerous applications, such as attosecond transient absorption spectroscopy, atmospheric remote sensing, frequency metrology, laser surgery, and industrial processing^{[4–9" target="_self" style="display: inline;">–9]}. Nonlinear optical technologies, including optical parametric oscillators and difference frequency generators, are normally employed to generate mid-infrared lasers due to their wide wavelength tunability, but they suffer from a lack of robustness and compactness. Sustained effort has been expended to find solid-state laser materials for direct mid-infrared light generation. In 1996, DeLoach et al., from Lawrence Livermore Laboratory, demonstrated the spectroscopic characteristics of divalent transition metal ions doped II-VI semiconductors and indicated the advantages of Cr2+-doped chalcogenide crystals as a novel class of tunable mid-infrared gain media for their broad and smooth emission band located in the 2–3 μm region^[10]. Then interest was inspired in the research of this class of materials, and laser oscillations of Cr2+-doped ZnSe, ZnS, CdSe, CdMnTe, etc., were gradually realized^{[1113" target="_self" style="display: inline;">–13]}. Among the materials, Cr2+:CdSe attracted plenty of attention for its large emission cross section and ultra-broad emission band (2.2–3.6 μm)^[14,15].

After McKay et al.^[16,17] reported the broadly tunable laser oscillation of Cr2+:CdSe, a series of subsequent research on the gain-switched Cr2+:CdSe laser was demonstrated, while pulsed Tm3+ or Ho3+ lasers were usually employed as a pump source^[18,19]. In 2018, an x-fold Cr2+:CdSe resonator pumped by a nanosecond Tm:YAP laser was reported by Martynova et al.^[20], with the average output power of 420 mW and a pulse duration of 200 ns. In the same year, Yumoto et al.^[21] reported the output energy scaling of a Cr2+:CdSe laser pumped by a high-energy 420 ns Tm:YAG laser, producing a maximum output energy of 4.4 mJ at 2.64 μm. They also presented the multi-peak temporal profile of the output pulse, and the duration of the primary spike pulse was 32 ns. In 2019, Antipov et al. reported the highest output power of a 6.3 W gain-switched Cr2+:CdSe laser with a pulse duration of about 50 ns^[22]. These works have shown a good performance of Cr2+:CdSe for pulsed operation, but the laser pulse durations in most of the above works are more than dozens of nanoseconds. A narrow laser pulse that is shorter than 10 ns has been achieved in the recent research^[22]; however, there has not been any detailed discussion about the generation method and its parameter optimization. Since the mid-infrared narrow pulses are crucial for the applications with time resolution requirements, in 2015, Mirov et al. reported the narrow-pulse laser in the gain-switched regime with Cr2+:ZnSe and Cr2+:ZnS as the gain medium, and the pulse duration was in the 3.3–5.3 ns range depending on the repetition rate^[23]. It is also worth achieving the narrow-pulse operation of a Cr2+:CdSe laser, as Cr2+:CdSe has a longer peak-emission-wavelength and a broader wavelength tuning range than Cr2+:ZnSe/ZnS.

In our previous work, a quality Cr2+:CdSe crystal was analyzed and applied as a saturable absorber in a Ho3+:YAG Q-switched laser system^[24]. In this Letter we attempted to study the pulse narrowing phenomenon in the Cr2+:CdSe laser pumped by 50 ns holmium laser pulses and verify its feasibility for generating a several-nanosecond-level mid-infrared laser. After that, we discussed the parameter optimization for obtaining single-peak output pulses.

In the schematic diagram of the gain-switched Cr2+:CdSe laser shown in Fig. 1, a linear cavity consisting of a plane dichroic mirror [DM1, antireflection (AR)-coated from 1.9 to 2.1 μm and high reflection (HR)-coated from 2.5 to 2.9 μm] and a plane output coupler (OC, AR-coated from 1.9 to 2.1 μm, reflectivity R=80% from 2.5 to 2.9 μm) was employed, and the cavity geometrical length was 45 mm. The 16 mm long Cr2+:CdSe crystal was grown by the vertical Bridgman method at a high speed, which sacrifices the crystalline structure. The doping concentration of Cr2+ is 1.0×1018cm−3. Both the end faces were AR-coated at 1.9–2.1 μm and 2.5–2.9 μm, and the crystal was placed 15 mm away from DM1. The intrinsic loss of the crystal should be relatively large, due to its imperfect crystalline structure and polishing quality. The crystal was wrapped with indium foil and was mounted on a copper heat sink, then kept at a constant temperature of 14°C by using a thermoelectric cooler (TEC). The laser cavity was designed with the consideration of stability conditions, while the thermal lens effect of the crystal was estimated by the expression^[25]fT=2πKcdn/dT+αTn×ωeff2ηQηaPin,(1)where fT represents the thermal focal length, Kc=9W·K−1·m−1 is the thermal conductivity of the crystal, dn/dT=1×10−4K−1 is the thermo-optical coefficient, αT=6.3×10−6K−1 is the coefficient of thermal expansion^[26], n is the refractive index, ωeff is the effective pump radius, ηQ is the quantum efficiency, ηa is the absorption efficiency, and Pin is the incident pump power.

Fig. 1. Schematic diagram of the gain-switched Cr2+:CdSe laser. L: plano-convex lens; DM: dichroic mirror; OC: output coupler; BS: beam splitter; PD: photodetector.

下载图片 查看所有图片

A Tm:fiber-laser-pumped Q-switched Ho3+:YAG laser operating at a pulse repetition frequency (PRF) of 1 kHz with a maximum output power of 2.7 W worked as the pump source. The pump pulse profile and spectrum are shown in Fig. 2 with a pulse duration (FWHM) of about 50 ns and a central wavelength of 2090.1 nm. The pump laser was focused into the crystal by an AR-coated plano-convex lens (L) with a focal length of 100 mm. The estimated focal spot radius was about 250 μm, resulting in a maximum pump density of nearly 1.4J/cm2, which was lower than the damage threshold of the film (about 2.5J/cm2). The unabsorbed pump laser was removed by the dichroic mirror (DM2) after the OC. Then the beam splitter (BS), with a reflectivity of 1%, sampled the output laser, and the HgCdTe photodetector (VIGO System S.A. PVM-10.6) measured the pulse profile. The output power was measured by a power meter (Ophir, 30A-BB-18), while the output spectrum was measured by a monochromator (BOIF, WDM1-3). By substituting the related parameters into Eq. (1), we found that the thermal focal length of the crystal at 2.7 W pump power was about 68.9 mm. It was verified that the cavity remained stable when the pump power varied from zero to 2.7 W.

Fig. 2. Pump laser pulse profile and spectrum.

下载图片 查看所有图片

The output power of the gain-switched Cr2+:CdSe laser is shown in Fig. 3, and optical damage was not observed on the surfaces of the crystal. The output power increased linearly with the pump power, then reached a maximum value of 100 mW with a pump power of 2.7 W. The optical-to-optical conversion efficiency was only 3.7%, which resulted from the imperfect crystalline structure and the low polished quality. Without the cavity, we measured that the transmissivity of the crystal to the pump laser was about 18%. It was given that the absorptivity of the crystal was about 77.8%, and we estimated that the intrinsic loss coefficient was about 0.13cm−1. We assumed that the loss coefficient at the pump wavelength was the same as that at the signal wavelength; therefore, the signal laser would suffer an intrinsic loss of 34% in a round trip. The inset (a) in Fig. 3 exhibits the normalized temporal profile of the output pulse versus the pump pulse under the maximum pump condition. The profile of the output pulse was featured by multiple peaks, which was mentioned by many of the literatures about gain-switched lasers^[23,27,28]. Owing to the high pump rate of the nanosecond pump pulse, the medium could achieve a high optical gain that was far above the threshold. Therefore, the signal laser intensity could increase rapidly and form a spike pulse that is narrower than the pump pulse. Then, with the injection of the subsequent part of the pump pulse, the relaxation oscillation process modulated the profile of the signal pulse, resulting in the formation of a multi-peak tail. For gain-switched laser operation, the output pulse is normally characterized by the pulse duration and the peak power of the primary spike pulse^[23,27]. In our experiment, the duration of the spike pulse was only 4.15 ns, which was below one-tenth of the pump duration. This result indicated that over 10 times pulse narrowing could be realized in the Cr2+:CdSe laser; therefore, we can obtain several-nanosecond-level pulses with much broader pump pulses. The inset (b) in Fig. 3 depicts the pulse sequence under the maximum pump power. The standard deviation of the pulse peak intensity was used to describe the pulse stability, and it was calculated as 1.85% here. The spectrum of the Cr2+:CdSe laser is depicted in Fig. 4; the peak wavelength was 2.65 μm and the bandwidth (FWHM) was about 50 nm without a wavelength-selection element. The beam quality of the laser was measured with the knife-edge method, and the result is shown in Fig. 5, where the M2 factor was 3.54. This relatively high M2 factor was due to the employment of the short cavity. The imperfect crystalline structure and the low polished quality would also hinder the improvement of the beam quality.

Fig. 3. Power performance of the gain-switched Cr2+:CdSe laser. Inset (a): the temporal profiles of the Cr2+:CdSe laser and the Ho3+:YAG laser under the maximum pump power. Inset (b): the pulse train under the maximum pump power.

下载图片 查看所有图片

Fig. 4. Output spectrum of the Cr2+:CdSe laser.

下载图片 查看所有图片

Fig. 5. Beam quality of the Cr2+:CdSe laser.

下载图片 查看所有图片

The temporal profiles of output pulses under different pump powers, with a step of 0.3 W, are shown in Fig. 6(a). Meanwhile, based on the laser properties of the Cr2+:CdSe crystal, we built a theoretical model and conducted a numerical investigation of the pulses profiles. The validity of the model was verified by comparing the theoretical and experimental results, and then the model helped to perform the parameter optimization of the laser.

Fig. 6. Temporal profiles of the output pulses under different pump powers: (a) experimental results; (b) simulation results.

下载图片 查看所有图片

Free Cr2+ ions with 3d4 configuration have a ground state of D5. When Cr2+ ions are doped into the chalcogenide crystals, the tetrahedral crystal field will split the D5 state into triplet T52 and duplet E5. Transitions between the two states are the main kinetic processes relating to the mid-infrared laser operation of Cr2+-doped chalcogenide crystals, since the other transitions from the ground state T52 or the excited state E5 to upper lying states are spin forbidden^[14,29]. The Jahn–Teller effect causes further splitting of the states^[30,31], making the states no longer a line energy level but a broad energy band. Therefore, Cr2+-doped chalcogenide crystals have an ultra-broad absorption and emission band.

The studies of laser dynamics showed that the two-energy-band structure in Cr2+-doped chalcogenide crystals has similar transition properties to four-level systems because the in-band nonradiative decay exists in both bands. The four-level rate equations have been applied in the theoretical investigation of the Cr2+:ZnSe laser^[32].Although the single frequency does not conform to the situation of the wide-bandwidth laser, it provides an approximation for the laser oscillation since all the frequency components have nearly the same gain. Here we also utilize a four-level system to model the energy level structure of Cr2+:CdSe. By neglecting the spatial distribution, the rate equations of the population inversion density and photon density can be expressed as^[28]dΔN(t)dt=Rp(t)-cσeΦ(t)ΔN(t)-cσesaΦ(t)ΔN(t)-ΔN(t)τ,(2)dΦ(t)dt=cσeΦ(t)ΔN(t)L0L-cσesaΦ(t)ΔN(t)L0L-Φ(t)τc.(3)ΔN(t) and Φ(t) represent the population inversion density and photon density, respectively. σe is the emission cross section of Cr2+:CdSe, and σesa is the absorption cross section of the excited level at the lasing wavelength. The excited-state absorption is negligible because the transition is spin-forbidden and σesa is much smaller than σe. This has been confirmed by the study of Cr2+:ZnSe, which works as the gain medium and saturable absorber^[32,33]. c is the velocity of light in vacuum, Rp(t) is the pump rate, τ is the fluorescence lifetime of the excited state, and τc is the photon lifetime of the cavity. L0 and L represent the length of the Cr2+:CdSe crystal and the cavity, respectively. For the gain-switched operation pumped by Gaussian nanosecond pulses, the pump rate can be expressed as Rp(t)=PpeakηaVhνpexp(-4ln2×ttp)2,(4)where Ppeak is the peak pump power, tp is the pump pulse duration, V is the gain volume, h is the Planck constant, νp is the pump laser frequency, and ηa is the pump absorption efficiency, which is defined as ηa=1-exp{-σa[N0-ΔN(t)]L0},(5)where σa is the absorption cross section of Cr2+:CdSe and N0 is the Cr2+ doping concentration.

The parameters listed in Table 1 were used to conduct the numerical simulation. The values of the parameters come from Refs. [15,26,34]. The pump duration and pump power were the same as that in the experiment. The simulation results are shown in Fig. 6(b). It can be found that compatible results are obtained from the experiment and simulation. The narrow spike pulse could be always observed at the rising edge of the pulse profile, whereas the shape and intensity of the pulse tail varied with the pump power. At a low pump power of 1.2 W, the pulse had a single-peak profile. As the pump power increased, the pulse build-up time of the signal pulse decreased, and the location of the spike pulse moved forward. More pump power was converted into the pulse tail, which formed after the primary spike pulse, leading to a gradual increase in the intensity.

The pulse duration of the spike pulse was measured in the experiment and is depicted in Fig. 7(a) by triangle dots. The minimum pulse duration of 4.15 ns was achieved at the maximum pump power. The simulation results of the spike pulse duration are illustrated by the solid line, which have the same tendencies as the experimental results. Nevertheless, the calculated values were lower than the measured values, which we determined for the following two reasons. One was the calculation error induced by neglecting the spatial distribution of the laser beam, and the other one was that we simplified the broadband signal laser as monochromatic light. Based on the pulse energy and the temporal profile measured in the experiment, the peak power of the spike pulse was calculated by using the expression Ppeak=Esingle∫−∞∞Wnorm(t)dt,(6)where Esingle is the single-pulse energy and Wnorm(t) is the normalized profile of the output pulse. The results are shown in Fig. 7(b), where the maximum peak power of 5.7 kW was obtained at the pump power of 2.7 W.

Fig. 7. (a) Pulse duration and (b) peak power of the output pulse under different pump powers.

下载图片 查看所有图片

Even though we have obtained several-nanosecond-level pulses in the experiment, a practical problem is that the pulse tail always exists, and the intensity of the tail increases with the increase of the pump power. However, in the applications of the mid-infrared laser, the pulse tail would act as the noise, and hence the cavity should be optimized to obtain the single-peak output pulses. The high intrinsic loss of the crystal hinders the experimental optimization, and here we used a better-quality crystal to conduct a numerical simulation to discuss the appropriate parameter for producing the single-peak pulses.

The core idea of suppressing the pulse tail in the gain-switched laser is to control the build-up time to let the signal pulse form at the falling edge of the pump pulse^[28]; therefore, the signal pulse tail is not able to extract enough gain and is suppressed. The main factors that determine the pulse build-up time include the gain and loss of the cavity and the cavity length. We fixed the pump power at 2.7 W and assumed that the intrinsic loss coefficient was 0.01cm−1; as a result, the output loss became the dominant loss term. We set up four different combinations of the cavity length and the output transmissivity. By substituting the parameters into the theoretical model, we calculated the temporal profiles of the output pulses.

For combination 1, the output transmissivity was 20%, and the cavity length was 45 mm, which was the same as that of the above experiment. The normalized temporal profiles of the signal and pump pulse are shown in Fig. 8(a). It could be found that the signal pulse formed at the rising edge of the pump pulse. The primary spike pulses had a duration of 3.14 ns, and were followed by an intense tail with a nearly-Gaussian shape. In addition, the average output power was 509.7 mW. For combination 2, the output transmissivity was 80%, and the cavity length was 45 mm. The simulation result is shown in Fig. 8(b), where the signal pulse had a multi-peak profile, the duration of the primary spike pulse was 1.24 ns, and the average output power was 397.8 mW. With these two conditions, although the output transmissivity has an obvious influence on the pulse profile, the pulse tail is still not suppressed.

Fig. 8. Temporal profiles of the output pulse and the pump pulse with (a) 20% output transmissivity, 45 mm cavity length; (b) 80% output transmissivity, 45 mm cavity length; (c) 20% output transmissivity, 500 mm cavity length; (d) 80% output transmissivity, 500 mm cavity length.

下载图片 查看所有图片

For combination 3, the cavity length was enlarged to 500 mm with the output transmissivity of 20%, and the result is shown in Fig. 8(c). The signal pulse with a duration of 40.52 ns formed at the falling edge of the pump pulse. The profile had a single peak, but the pulse duration was relatively large, while the average output power was 323.1 mW. Figure 8(d) shows the result of combination 4; the output transmissivity was 80% and the cavity length was 500 mm. A single-peak signal pulse with a duration of only 8.17 ns formed at the falling edge of the pump pulse, and the average output power was 117.2 mW. From these results we know that the signal pulse tail could be suppressed by employing a long cavity, as large cavity length corresponds to long pulse build-up time. Then high output transmissivity could help to reduce the pulse duration; therefore, the several-nanosecond-level single-peak pulses could be obtained at the cost of output power.

In conclusion, for a fiber-bulk Ho3+:YAG laser pumped gain-switched Cr2+:CdSe laser, the generation of high-peak-power nanosecond mid-infrared pulses was demonstrated, and the experimental results highly agreed with the simulation. Under 2.7 W pump power, a pulse duration of 4.15 ns, which was shorter than one-tenth of the pump duration was obtained. To the best of our knowledge, this is the shortest pulse duration ever reported for a Cr2+:CdSe laser, which shows that the gain-switched Cr2+:CdSe laser could achieve over 10 times pulse narrowing, while the corresponding peak power was 5.7 kW. The optical-to-optical conversion efficiency was 3.7%, which should be further improved with better crystal surface polishing quality. The laser central wavelength was 2.65 μm with a bandwidth (FWHM) of 50 nm. Meanwhile, the parameter optimization for generating single-peak pulses was discussed by numerical simulations. When a 500 mm long cavity and 80% output transmissivity were employed, the tail of the output pulses could be suppressed and 8.77 ns single-peak pulses could be obtained with a better-quality crystal. Our demonstration has great significance for the time-resolution applications of mid-infrared lasers, and future efforts will be spent on the further optimization of the crystal quality, the Cr2+ doping concentration, the output transmissivity, and the cavity length to suppress the tail of the output pulse and to improve the conversion efficiency.