Mid-infrared (mid-IR) lasers (3–5 μm) have become an indispensable tool in science and technology, for example, in medical treatments, free-space communication, and trace gas detection [1–3]. Recently, the hollow-core-fiber gas laser (HCF-GL) based on population inversion has attracted extensive attention, owing to its advantages of wide wavelength range, high output power, and good beam quality. In that regard, a vast amount of research oriented toward the 3–5 μm HCF-GL has been successfully achieved, including those employing acetylene (${\mathrm{C}}_2{\mathrm{H}}_2$) [4–7], carbon dioxide (${\mathrm{CO}}_2$) [8–10], hydrogen bromide (HBr) [11,12], and nitrogen dioxide (${\mathrm{NO}}_2$) [13] as the filling gases. However, most of these studies have been conducted using a free-space coupling arrangement. Generally, a pump laser is coupled into HCF through a telescope system constructed using a pair of coated planoconvex lenses and two plane mirrors. Although this approach can flexibly regulate the mode field and numerical aperture of the pump laser to achieve a high coupling efficiency, such a system is bulky and highly sensitive to the external environment. In addition, even tiny external perturbations can lead to misalignments, decreasing the pump coupling efficiency and increasing the overlap between the incident pump beam and microstructure of the HCF, thereby damaging it. Hence, the pursuit of an all-fiber-structure is a crucial development direction for mid-IR HCF-GLs.

Recently, a new technique called nanospike, which theoretically achieves high-efficiency direct coupling between the single-mode fiber (SMF)-28 and HCF, has been reported [14]. Such a technique has been utilized to realize mid-IR all-fiber HCF gas light sources at 3 μm [15,16]. Nevertheless, due to the tapering fiber being inserted into the core of the HCF, the coupling efficiency is significantly affected by the relative spatial position between the HCF and the tapering fiber. Thus, the actual coupling efficiency of the pump laser is only approximately 40%. Additionally, because the tapering fiber cannot withstand a pump laser with a high power, the mid-IR HCF gas laser can achieve amplified spontaneous emission (ASE) light output power only on the order of hundreds of milliwatts at 3.1 μm [15].

Fusion splicing, a robust method to connect the solid-core fiber (SCF) with the HCF, enhances the stability and compactness of the pump laser coupling and maintains high coupling efficiency. Up to now, the graded index (GRIN) fiber bridging [17] and fiber reverse-tapering techniques [18] have shown excellent results for conventional SMF and HCF fusion splicing. However, because SMF has a small core diameter, the mode-field scaling ability that can be achieved by the above two technologies is limited, and they are not suitable for splicing with a large-core HCF commonly used in mid-IR HCF gas lasers. Fusion splicing of a large-mode-area (LMA) SCF with a hollow-core anti-resonant fiber (HC-ARF) using the reverse tapering technique opens potential avenues for realizing the coupling of high-power pump lasers with low losses; this is helpful in promoting the realization of high-power all-fiber-structure mid-IR HCF gas lasers.

However, the Fresnel reflection of a splice from the air–silica interface between the HCF and the LMA SCF needs to be addressed, as it limits the forward coupling power of the pump laser, leading to limiting the power enhancement of the all-fiber-structure HCF-GL. Currently, the angle-cleaved HCF fusion splicing technique has been developed to reduce back-reflection losses. In 2022, Zhang et al. spliced a 2°-angle-cleaved SMF to a nested HC-ARF, achieving a forward coupled loss of 1 dB and back-reflection loss of $-41\mathrm{dB}$ [19]. In 2023, Wang et al. spliced an HC-ARF to an antireflective (AR)-coated SMF without damaging the coating, achieving an average fusion splicing loss of 0.30 dB at each joint and back-reflection loss of $-28\mathrm{dB}$ at two fusion joints [20]. It can be seen that the AR coating deposited on the end face of the SMF effectively reduced the Fresnel reflection, but the damage threshold of the AR coating limited the increase in the laser transmission power.

This study adopted the reverse tapering and angle-cleaved fusion splicing techniques to reduce the coupled losses and back-reflection between the LMA SCF and the HC-ARF to 0.45 dB and $-31.91\mathrm{dB}$ at 1535 nm, respectively. In addition, a high-power (52.8 W) single-frequency laser transmission was obtained at 1535 nm, realizing a transmission efficiency of 83.8% in a 5-m-long nested HC-ARF. Based on these findings, a high-power all-fiber-structure ${\mathrm{C}}_2{\mathrm{H}}_2$-filled HC-ARF ASE source at 3.1 μm was experimentally fabricated. By optimizing the ${\mathrm{C}}_2{\mathrm{H}}_2$ pressure, an ASE output power of 6.59 W was observed; the slope efficiency was 19.74%, and the beam quality was $M^2=1.05$. By using the fusion-splicing scheme, we further extend the generation of 4.3 μm ASE light, facilitating an output power of 1.43 W with a slope efficiency of 8.76%. All-fiber fusion splicing between the LMA SCF and the nested HC-ARF is highly likely to pave the way for the possibility of mid-IR HCF-GL miniaturization and high-power laser all-fiber delivery in HCFs.

A layout of the reverse tapering process is illustrated in Fig. 1(a). The two ends of the LMA SCF were fixed symmetrically to the fusion machine (Fujikura, Fiber Fusion Splicer, FSM100P+) and pulled in the same direction at different speeds while the discharge center discharged at the appropriate charge. With the motor pulling the fiber, the waist diameter of the LMA SCF at the center of the discharge area gradually thickened and the mode-field diameter (MFD) increased. The un-tapered fiber was not heated and was connected to the tapered waist by taper transitions at both ends. Subsequently, by cutting the waist in the middle, two reverse-tapered fibers were obtained. The adiabatic tapering of the LMA SCF can be achieved by setting appropriate fiber tapering parameters [21]. Figure 1(b) presents the dependence of the theoretical and measured MFDs on the reverse-tapered cladding diameter of the LMA SCF (Changjin, CJGDF-LMA-25/300, core diameter of 25 μm, and cladding diameter of 300 μm). When the cladding diameter of the LMA SCF increased from 300 μm to 380 μm, the theoretical MFD obtained by RSoft simulation software increased from 23.2 μm to 30 μm. However, the MFD measured using the beam quality analyzer (Thorlabs, BP109-IR) was slightly larger than the theoretical simulation value. The linear fitting results show that the slope efficiency of the variation in the MFD with the cladding diameter is 9.3%, which indicates that an increase of 100 μm in the cladding diameter increases the MFD by 9.3 μm.

Figure 1(c) shows a scanning electron microscope (SEM) image of the homemade HC-ARF. It comprises six silica tubular cladding elements with average outer and inner diameters of 29 μm and 13 μm and tube thicknesses averaging 1.17 μm and 1.22 μm, respectively. The laser is guided through the hollow core with a diameter of 39 μm. Figure 1(d) details the measured transmission loss characteristics of the six-tube nested HC-ARF, characterized using the back-cut method in the near-IR (550–10 m) region. The minimum transmission loss in the 1500–1600 nm range is 0.85 dB/km, indicating a transmission loss of 0.85 dB/km at 1535 nm (blue curve). In the mid-IR region (red curve), the theoretical transmission loss of the fundamental mode is calculated using COMSOL simulation software, indicating that the theoretical minimum transmission loss in the 2400–4000 nm region is 37.5 dB/km, and the theoretical transmission loss at 3100 nm is 47.8 dB/km.

The corresponding theoretical MFD of the nested HC-ARF at 1535 nm is 29.2 μm. An MFD mismatch loss of 0.75 dB occurs between the HC-ARF and LMA SCF (23.2 μm). The theoretical and measured coupled losses according to the tapered cladding diameter of the LMA SCF are shown in Fig. 1(e). Because of the reverse tapering of the LMA SCF, its MFD gradually approaches that of the HC-ARF, leading to a significant reduction in the MFD mismatch loss. When the cladding diameter of the LMA SCF was enlarged to 368.06 μm, the theoretical minimum coupling loss reached 0.083 dB. However, owing to the discrepancy between the theoretical and measured MFDs of the LMA SCF and HC-ARF, the measured minimum coupled loss was 0.29 dB at the LMA SCF diameter of 360 μm. As evident, the measured and theoretical minimum coupled losses have a significant difference since the HCF structure undergoes slight deformation during the fiber fusion process. Additionally, the cleave angles of the LMA SCF and HCF are not completely consistent; thus, a certain loss is introduced after fiber fusion. The inset in Fig. 1(d) exhibits the cross sections of the HC-ARF after breaking the splice point, as viewed under a DVM6 digital microscope (Leica Microsystems). The cladding structure of the HC-ARF has no notable collapse or deformation after fusion. Meanwhile, a tensile test (tensile strength of 2.9 N) was conducted on the joint using the FSM100P+ fusion machine; the joint did not exhibit fracture, indicating that the joint has good strength.

Fresnel reflection is a crucial factor affecting the coupling power between the LMA SCF and HCF. The back-reflected light is amplified backward along the fiber amplifier and damages the pre-amplified stage. Therefore, utilizing the angle-cleaved fusion splicing technique reduces the back-reflected light power in the fiber core and improves the transmission power in the all-fiber-structure HCF. The measurement setup for the coupled loss and back-reflection is shown in Fig. 2(a). We used a homemade 1535 nm single-frequency fiber laser (SFL) as the pump source, whose maximum output power is approximately 63 W, and equipped by an LMA SCF with a core diameter of 25 μm and cladding diameter of 300 μm. To evaluate the back-reflection losses of the spliced point of the LMA SCF and HC-ARF, a circulator is used. The output of the 1535 nm SFL was passed through the circulator to the input of the HC-ARF; the tail fiber of the circulator was reverse tapered and angle cleaved, that is, spliced to a 5-m-long angle-cleaved HCF. The corresponding coupled loss was calculated by measuring the power at the output end of HCF. The power of the reflected signal was measured at the third port of the circulator using a power meter, enabling the calculation of the back-reflection losses.

Figure 2(b) shows the coupled loss and back-reflection dependence on cleave angles. As the cleave angle of the LMA SCF increases from 0° to 2.6°, the coupled loss increases from 0.29 dB to 1.05 dB, with a net increase of 0.76 dB. And the back-reflection decreases from $-18.34\mathrm{dB}$ to $-38.91\mathrm{dB}$, with a net decrease of 20.57 dB. When the cleave angle was greater than 2°, the coupled loss increased rapidly with the cleave angle, and the proportion of back-reflected light decreased significantly. These features appear because the cleave angles of the LMA SCF and HC-ARF do not match perfectly, causing a significant increase in the intrinsic loss during the coupling process. Considering the balance between the coupled loss and back-reflection, the cleave angles for the splice were 2° for the LMA SCF and HC-ARF; then high-power laser transmission was achieved. Figure 2(c) presents the output power of the HC-ARF and transmission efficiency as functions of the 1535 nm incident power. Under an incident power of 3 W at 1535 nm, the transmission efficiency of the HCF was 91%, whereas with increase in the 1535 nm laser power to 63 W, the transmission efficiency gradually decreased to 83.8%. This was primarily because the beam quality factor of the pump laser source deteriorated from 1.05 to 1.3 during power boosting, as shown in Fig. 2(d). Since the HC-ARF can support larger cores with effective single-mode operation, the higher-order modes of the pump laser are lost during the transmission process. At an incident power of 63 W, the output power of the HC-ARF was 52.8 W, with a corresponding transmission efficiency of 83.8% and transmission slope efficiency of 83.55%.

Indeed, this reverse-tapering and angle-cleaved fusion splicing approach can be extended to other wavelength bands. We have achieved high-power and low-loss laser transmission at 2 μm by utilizing a homemade high-power single-frequency thulium-doped fiber laser (SF-TFL) and 5-m-long nested HC-ARF. The SF-TFL possessed a central wavelength of 2000.6 nm and a maximum output power of 56.8 W, and the corresponding beam quality was 1.38. Similar to the setup of the 1535 nm SFL, the SF-TFL using an LMA SCF (Nufern, LMA-GDF-25/250-09M) acts as the output port. The LMA SCF has a core diameter of 25 μm and a cladding diameter of 250 μm, whose MFD at 2000.6 nm is approximately 24.8 μm. The five-tube nested HC-ARF for a mid-IR ASE light source at 4.3 μm is depicted in Fig. 3(a). It can be seen that the HC-ARF is composed of two rings of five untouched cladding silica tubes; the average diameters of the outer and inner nested tubes are 53.6 μm and 30 μm, with average tube thicknesses of 1.5 μm. Additionally, the core diameter is 50 μm and the cladding diameter is 290 μm. Figure 3(b) displays the measured transmission loss characteristics in the region from 1950 nm to 2200 nm, which has the minimum transmission loss of 0.85 dB/km at 1985 nm and 2040 nm, with a 200 nm bandwidth below 2 dB/km [22]. Moreover, a simulation of the mid-IR transmission loss of the HC-ARF is presented in Fig. 3(b), which indicates that the theoretical transmission loss is 1.06 dB/m at 4.3 μm. The MFD of the nested HC-ARF at 2000.6 nm is 35.2 μm, which indicates a large MFD mismatch loss between the nested HC-ARF and LMA SCF (24.8 μm). By employing the reverse tapering and fusion splicing techniques, the theoretical coupled loss can be reduced to 0.069 dB at 2 μm. However, limited by the beam quality of the 2 μm SF-TFL test source ($M^2=1.15$ at 1.05 W), the lowest coupled loss we could achieve was 0.67 dB [as shown in Fig. 3(c)]. Moreover, the microstructure of the HCF after splicing did not change significantly, and the splicing point exhibited strong mechanical properties.

Figure 3(d) presents the 2 μm laser power transmitted through a 5-m-long HC-ARF versus the variation in the incident laser power. It is interesting to note that the output power increases linearly with respect to the incident power; inversely, the transmission efficiency is gradually reduced, which is attributed to the deterioration in the beam quality of the pump laser from 1.15 to 1.32. Under the maximum pump power of 56.8 W, the output power of the nested HC-ARF is 44.8 W; the corresponding transmission efficiency is 78.87%, and the transmission slope efficiency is 79.32%. The transmission efficiency of the 2 μm laser is slightly lower than that of the 1.5 μm laser because the higher-order mode filtering capability of the five-tube nested HC-ARF is better than that of the six-tube structure. In comparison with the results of prior studies [15], the laser transmission power was improved by nearly an order of magnitude. The temperatures at and near the splice point at the highest pump power were measured using a thermal imager (FOTRIC, 220S), as shown in Fig. 3(e). Because the beam quality of the pump laser source decreases at high powers, part of the pump laser fails to effectively couple with the HCF and accumulates at the splice point, which significantly increases the splice point temperature. The splice point is covered with a high-refractive-index UV adhesive to strip the pump laser that fails to couple with the HCF. Moreover, the splice point is placed on water-cooled plate with grooves to ensure effective heat dissipation. When the pump power increases to 56.8 W, the splice point temperature remains at 20.8°C, indicating that this type of an all-fiber HCF structure has the potential to withstand high-power-laser transmission.

Figure 4 illustrates the schematic of the proposed mid-IR gas-filled HCF ASE source with an all-fiber structure. We utilized ${\mathrm{C}}_2{\mathrm{H}}_2$ gas as the gain medium to generate mid-IR emission at the 3.1 μm wavelength band. An in-house erbium-doped fiber single-frequency laser with a central wavelength of 1535 nm and the maximum output power of 63 W was employed as the pump source. An in-house 5-m-long six-tube nested HC-ARF with a core diameter of 39 μm was used as the gas cell, exhibiting a transmission loss of 0.85 dB/km at 1535 nm. And the nested HC-ARF was coiled onto an optical platform with a bending diameter of 60 cm. The output end of the fiber was sealed in a gas chamber with an uncoated calcium fluoride window ($\mathrm{HT}>92\%$ at 1–5 μm). To prevent the return of the laser, the gas chamber was designed with an 8° angle structure. A mid-IR bandpass filter (Thorlabs, FB3000-500) was placed at the output end to selectively transmit the mid-IR laser signal.

By changing the pump source, ${\mathrm{CO}}_2$ gas, and nested HC-ARF, we can also achieve an ASE light emission at the 4.3 μm wavelength band with the all-fiber structure. The pump source is a self-developed 2 μm single-frequency fiber laser with a maximum output power of 56.8 W. The 2 μm SFL can be tuned between 1999.5 nm and 2001.6 nm, corresponding to the R(30) absorption lines of ${\mathrm{CO}}_2$ gas. A 5-m-long, 50-μm-core-diameter, five-tube nested HC-ARF was employed as the gas cell, characterized by a cladding diameter of 290 μm, a transmission loss of 3.2 dB/km at 2000 nm, and a simulated transmission loss of 1.06 dB/m at 4.3 μm.

The ${\mathrm{C}}_2{\mathrm{H}}_2$ molecule has five types of vibrational modes, each of which has abundant rotation modes, leading to discrete and quantized energy levels. Figure 5(a) illustrates the absorption spectral line corresponding to the transition from the vibrational ground state (${\nu }_0$) to the ${\nu }_1+{\nu }_3$ vibrational state [23]. Among these spectral lines, the P(17) spectral line was chosen as the pump line in our experiment because other stronger absorption lines located at the edge of the gain band were difficult to amplify using the erbium-doped fiber amplifier. Figure 5(b) depicts a simplified energy-level diagram of ${\mathrm{C}}_2{\mathrm{H}}_2$ and its corresponding transition processes. Upon excitation by the P(17) absorption line, the ${\mathrm{C}}_2{\mathrm{H}}_2$ molecules transfer from the $J=17$ rotational state of the vibrational ground state ${\nu }_0$ to the $J=16$ rotational state of the ${\nu }_1+{\nu }_3$ vibrational state. After that, they leave from the $J=16$ rotational state of the ${\nu }_1+{\nu }_3$ vibrational state to the $J=15$ and $J=17$ rotational states of the ${\nu }_1$ vibrational state, emitting photons at $\sim 3.105\mathrm{\mu m}$ [R(15)] and $\sim 3.181\mathrm{\mu m}$ [P(17)], according to the selection rule ($\mathrm{\Delta }J=\pm 1$) [24].

Figure 6 shows the output characteristics of the 3.1 μm ASE source. Owing to the difficulty in obtaining a high-power passive fiber isolator, the 1.535 μm SFL was not equipped with an output isolator. To protect the pump source and keep it operating in a stable state, we utilized only 42 W to pump the ${\mathrm{C}}_2{\mathrm{H}}_2$-filled HCF. Figures 6(a) and 6(b) show the output power of the 3.1 μm ASE light and 1.535 μm residual pump laser versus the coupled pump power at different acetylene pressures. The 3.1 μm ASE light power is the sum of the R(15) and P(17) line emission powers. Under the coupled pump power of 42 W, the maximum output power of the 3.1 μm ASE light exhibits a trend of initial increase followed by decrease, when the pressure increases from 2 mbar to 6 mbar. The highest output power of 6.59 W is obtained at 4 mbar. Notably, the value of 6.59 W is taking into account the loss from the mid-IR bandpass filter. When the gas pressure is lower than 4 mbar, the pump laser cannot be fully absorbed owing to the limited number of ${\mathrm{C}}_2{\mathrm{H}}_2$ molecules within the nested HC-ARF, resulting in a relatively low output power of the 3.1 μm ASE light. As the gas pressure increases to 4 mbar, the residual pump power decreases significantly, and the mid-IR ASE source achieves a maximum output power of 6.59 W. However, increasing the gas pressure in the HCF enhances the collision losses. Consequently, the output power of the mid-IR ASE source appears to decline when the gas pressure exceeds 4 mbar. Figure 6(c) plots the signal power as a function of the absorbed pump power under varying gas pressures. When the gas pressure increases from 2 mbar to 4 mbar, the slope efficiency of the mid-IR ASE source decreases by only 0.56% (from 20.3% to 19.74%) because of the low effect of collision losses caused by the low pressure. As the gas pressure reaches 6 mbar, the collisions between the gas molecules intensify, resulting in an increase in the loss of the mid-IR light source, and the slope efficiency decreases to 14.43%.

Furthermore, Fig. 6(d) compares the slope efficiencies of the coupled and absorbed pump powers as functions of the ${\mathrm{C}}_2{\mathrm{H}}_2$ gas pressure. At lower gas pressures, the slope efficiency of the absorbed pump power is greater than that of the coupled pump power because a large proportion of the pump laser is not absorbed. With increasing gas pressure, the increased molecular density in the HCF results in higher pump power absorption, demonstrating similar tendencies for the two slope efficiencies. However, the mid-IR ASE source cannot obtain the maximum output power owing to enhanced intermolecular collisions. In fact, the gas molecular collision loss can be effectively reduced by increasing the length of the HCF instead of the gas pressure, thus allowing for the slope efficiency of the mid-IR ASE source to be improved. Nevertheless, the large transmission loss of the mid-IR HCF limits its useful length; thus, increasing the core diameter appears to be a potential approach to ensure the gain of the gas-filled HCF with a small HCF length. Figure 6(e) shows the long-term stability test results of the mid-IR ASE source at 6.59 W output power. The calculated root-mean-square (RMS) is 1.35% over a time span of 120 min, and the peak-to-peak deviation ($\mathrm{\Delta }\mathrm{pp}$) is 8.8%, indicating that the laser has good power stability.

Figure 7 depicts the typical spectrum characteristics of the 3.1 μm ASE source. The optical spectra of the 6.59 W output power were measured using a mid-IR optical spectral analyzer (Thorlabs, OSA 205C), and the results are illustrated in Fig. 7(a). Two separate spectral lines with center wavelengths of 3105.48 nm and 3181.72 nm can be observed, which are consistent with the theoretically calculated center wavelengths. Figure 7(b) shows the output spectra measured at different gas pressures. It can be clearly seen that the intensity ratio of the two signal lines is clearly correlated with the gas pressure. When the gas pressure is 2 mbar, both the R(15) and P(17) emission lines occur simultaneously. The intensity of the P(17) emission line increases with respect to the gas pressure, but decreases in the case of R(15). This is because the R(15) and P(17) emission lines have the same upper energy level, and the Einstein coefficient of the P(17) emission line is larger than that of R(15), leading to the generation and amplification of the P(17) emission line, thus suppressing the intensity of the R(15) emission line. When the gas pressure exceeds 15 mbar, intense molecular collisions lead to a significant increase in the lasing threshold. The R(15) emission line with a smaller Einstein coefficient is completely suppressed, and only the P(17) emission line exists.

The radio frequency (RF) spectrum characteristics of 3.1 μm ASE, measured by an extended HgCdTe photodetector (Thorlabs, PDAVJ10; 100 MHz bandwidth) and a frequency analyzer (Agilent Technologies, N8030A, 3 Hz–44 GHz), are recorded in Fig. 8. The 3.1 μm signal power was attenuated to yield a PD output of 10 mV DC voltage. Figure 8(a) illustrates the RF spectrum of the 3.1 μm ASE light at 6.59 W of output power. The spectrum has a downward trend in the scanning range of 0 to 50 MHz with a bandwidth (BW) of 1 kHz, decreasing from $-120\mathrm{dBm/}\mathrm{Hz}$ to $-135\mathrm{dBm/}\mathrm{Hz}$. The relative intensity noise (RIN) characteristics of the 3.1 μm ASE at different output powers are characterized in the scanning range of 0–1 MHz, as shown in Fig. 8(b). One important remark is that the RIN of the 3.1 μm ASE at 1 MHz increased approximately by 10 dB at 6.59 W output power when compared with the output power at 1.5 W, which indicates that the RIN of the HCF gas laser is related to the output power. As mentioned above, a higher output power causes the gas molecules to absorb more of the pump laser and introduces a strong thermal effect, which intensifies the collision between the gas molecules, resulting in an increased ASE RIN.

To evaluate the beam quality of the 3.1 μm ASE source under the output power of 6.59 W, the beam was monitored using a scanning slit beam profiler system (DataRay Inc., BeamScope-P8) and the results are presented in Fig. 9. The $M^2$ factors in the $x$- and $y$-directions were obtained through polynomial fitting of the measured data, giving $M_x^2=1.03$ and $M_y^2=1.06$; the values indicate that the mid-IR ASE source operates in the fundamental mode state. The inset of Fig. 9 shows the near-field spot of the ASE source, as measured using a beam profiler (Spiricon, PY-III-C-A). The near-Gaussian shape of the intensity distribution and good ellipticity indicate that the laser has excellent spot characteristics.

The ${\mathrm{CO}}_2$ molecule is a linear triatomic molecule with four vibrational and multiple rotational modes. In addition, the transitions between different vibrational states generate multiple absorption spectral lines. Figure 10(a) depicts the absorption spectrum between the $2{\nu }_1+{\nu }_3$ state and the ground vibrational state of ${\nu }_0$. The R(30) absorption line was selected as the pump wavelength after accounting for the intensity and wavelength of the absorption line. Figure 10(b) illustrates the simplified energy levels of the ${\mathrm{CO}}_2$ gas, outlining the transition process under the R(30) pump line between the ground vibrational state (${\nu }_0$), the $2{\nu }_1$ vibrational state, and the $2{\nu }_1+{\nu }_3$ vibrational state. When the R(30) absorption line is excited, the ${\mathrm{CO}}_2$ molecules leave from the $J=30$ rotational state of the ground vibrational state (${\nu }_0$) to the $J=31$ rotational state of the vibrational state ($2{\nu }_1+{\nu }_3$). Since the population in the $2{\nu }_1$ vibrational state that conforms to the Boltzmann distribution is small at room temperature, the population inversion is formed between the $2{\nu }_1+{\nu }_3$ vibrational state and $2{\nu }_1$ vibrational state. According to the selection rule $\mathrm{\Delta }J=\pm 1$ [24], the emission lines with wavelengths of 4295.85 nm [R(30) emission line] and 4388.3 nm [P(32) emission line] are simultaneously generated.

The corresponding output characteristics of the ${\mathrm{CO}}_2$-filled nested HC-ARF ASE are shown in Fig. 11. The output power of the 4.3 μm ASE light and 2 μm residual pump power as functions of the coupled pump power under varying ${\mathrm{CO}}_2$ pressures are displayed in Figs. 11(a) and 11(b). The 4.3 μm signal power is the sum of the R(30) and P(32) emission powers. As shown in Fig. 11(a), the optimal gas pressure inside the nested HC-ARF was 4.1 mbar. When the gas pressure operates at lower than 4.1 mbar, the limited number of gas molecules cannot sufficiently absorb the pump laser to provide sufficient gain to generate the high-power 4.3 μm ASE light. Therefore, as the gas pressure in the HC-ARF is increased from 1.1 mbar to 4.1 mbar, the conversion efficiency and output power of the ASE source are notably improved. When the gas pressure exceeds 4.1 mbar, the increase in the gas pressure improves the absorption rate of the pump laser; however, the collision loss between the gas molecules increases simultaneously. Since the gain originating from the increase in the gas pressure is less than the collision loss, the 4.3 μm CW power does not increase with the gas pressure. At the highest pressure of 8.3 mbar, the maximum output power of the 4.3 μm ASE light is only 0.91 W with a residual pump power of 12 W. The small-core-diameter HC-ARF used in this experiment can easily match the MFD to obtain higher laser coupling efficiency than that observed in a prior study [10]. Nevertheless, a small core diameter also increases the overlap between the laser field and quartz, resulting in an increase in the transmission loss of the mid-IR laser; thus, the conversion efficiency and maximum output power of the mid-IR ASE source are degraded. In the future, it is expected that low-loss fusion splicing can be achieved with an LMA HC-ARF ($\mathrm{core}\mathrm{diameter}>100\mathrm{\mu m}$) by adopting an LMA SCF as the bridging fiber, thus obtaining high-efficiency mid-IR light emission.

Figure 11(c) plots the output power of the 4.3 μm mid-IR ASE source with respect to the absorbed pump power under varying gas pressures. One can see that the limited number of gas molecules can absorb only a small proportion of the pump laser to achieve the conversion of mid-IR ASE under lower pressure. However, the slope efficiency of the mid-IR ASE source is higher owing to the lesser collision loss of the gas molecules at a lower gas pressure. With the increase in gas pressure, the absorption pump power increases significantly but the slope efficiency of the mid-IR ASE source decreases. In Fig. 11(d), we further summarize the characteristics of the variation in the slope efficiency with gas pressure (blue curve). At the ${\mathrm{CO}}_2$ gas pressure of 1.1 mbar, the slope efficiency of the mid-IR ASE source was 17.18%, whereas at the gas pressure of up to 8.3 mbar, the slope efficiency decreased to 3.68%. This was due to the higher gas pressure shortening the lifetime of the upper energy level and reducing the emission cross section of the gas molecules; the gain capacity decreases, leading to the decrease in the slope efficiency. Furthermore, Fig. 11(d) shows the splicing point temperature as a function of the ${\mathrm{CO}}_2$ gas pressure (red curve). It can be seen that the ${\mathrm{CO}}_2$ gas absorbs a large proportion of the pump laser within a short time at the front end of the HCF, resulting in a strong thermal effect and rapid increase in the splice point temperature. To protect the splice point, it is placed on a water-cooled plate for heat dissipation. When the gas pressure increases from 1.1 mbar to 8.3 mbar, the splice point temperature rises from 26°C to 34°C, further indicating that the splice point can withstand the higher pump laser power and support mid-IR ASE light generation of higher power.

Figure 12 illustrates the output spectral characteristics measured using a mid-IR spectrum analyzer (Thorlabs, OSA205). The mid-IR ASE spectrum at the output power of 1.43 W is shown in Fig. 12(a). Two discrete narrow spectral lines at 4296.91 nm and 4383.65 nm, belonging to the R(30) and P(32) transitions, are measured, which is a little different from the theoretical predictions because the pump wavelength does not fully overlap with the gas absorption peaks. The inset of Fig. 6(a) shows the near-field spot of the mid-IR ASE source, which was characterized using a laser beam profiler (Spiricon, PY-III-C-A). The ASE light exhibited good spot characteristics at a high-power output. The mid-IR ASE light source spectrum varied with the ${\mathrm{CO}}_2$ pressure at varying pump powers, as shown in Fig. 6(b). Similar to the ${\mathrm{C}}_2{\mathrm{H}}_2$-filled HC-ARF ASE source, both the R(30) and P(32) emission lines were generated at a lower pressure. With increasing gas pressure, the larger collision loss gradually inhibited the generation of the R(30) emission line. Thus, the P(32) emission line became dominant.

In summary, we have designed and implemented two all-fiber-structure mid-IR gas-filled HCF ASE source systems. By combining the reverse tapering technique and angle-cleaved fusion splicing method, high-power single-frequency laser transmissions of 52.8 W and 44.3 W were obtained at 1.535 μm and 2 μm, respectively, realizing the corresponding transmission efficiencies of 83.8% and 78.8% in 5-m-long nested HC-ARFs with core diameters of 39 μm and 50 μm, respectively. Furthermore, output powers of 6.59 W at 3.1 μm and 1.43 W at 4.3 μm were achieved by using ${\mathrm{C}}_2{\mathrm{H}}_2$ and ${\mathrm{CO}}_2$ gases, respectively, at appropriate pressures; the corresponding slope efficiencies were 19.74% and 8.76%, respectively. Thanks to the used HC-ARFs having small core diameters, high transmission losses occurred in the high-order modes; this ensured good beam quality for the mid-IR ASE light. In particular, the beam quality of the 3.1 μm ASE source was indicated by $M_x^2=1.03$ and $M_y^2=1.06$. This study marks the first instance of a high-power all-fiber-structure gas-filled HCF ASE source. Such an all-fiber structure system can also be harnessed to design light sources in the range of 3–5 μm. In the future, we will further optimize the reverse tapering and fusion splicing techniques to achieve hundreds of watts of laser transmission, thereby establishing a new power record for mid-IR HCF-GLs. This technology will rapidly advance miniaturized HCF lasers, which are promising and powerful new tools for laser medicine, space communication, and other scientific research applications.