Gas-filled microstructured hollow-core fibers (MS-HCFs) have demonstrated many advantages as a minimized gas cell and shown the promising potential of reforming the bulky gas laser^[1] and sensors into a more compact and robust all-fiber configuration^[2]. The gas–light interaction along MS-HCF could be possibly enhanced by ${10}^6$ times stronger than that in the Rayleigh zone of the free-space laser beam^[3]. In MS-HCF, both the gas and light field could be tightly confined in a core region of tens of ${\mathrm{\mu m}}^2$ area. Meanwhile, its effective interaction length depending on the fiber loss is much extended to tens of meters. The latest development of anti-resonant hollow-core fiber (AR-HCF) demonstrated 0.22 dB/km minimum fiber loss that implies more than a thousand effective length to use^[4].

The stimulated Raman scattering (SRS) of gas was first observed, to the best of our knowledge, in 1963^[5], which is proven as an effective method to generate new wavelengths especially in the ultraviolet and mid-infrared spectral ranges^[6,7]. Both hollow capillaries^[8] and high-finesse cavities^[9] were often used to reduce the pump threshold, while the Stokes conversion efficiency was still weak around 5%^[9].

In 2002^[3], SRS in ${\mathrm{H}}_2$-filled Kagome hollow-core fiber (Kagome-HCF) was successfully demonstrated, and a record of 30% conversion efficiency of the Stokes wave at the 683 nm wavelength was measured using a pulsed laser source (532 nm, pulse duration 6 ns) as the pump. Such a leap of conversion efficiency gave birth to the concept of the fiber gas Raman laser (FGRL)^[10–24].

Given the nature of SRS, the $Q$-switched pulsed laser of nanoseconds of pulse duration is often preferred as the pump of the FGRL, where the high peak power and narrow linewidth are favorable for low threshold and high slope efficiency. Pumped by a $Q$-switched microchip laser at 1 µm, FGRL based on vibrational SRS (Vib-SRS) of ${\mathrm{H}}_2$ at 1.9 µm was demonstrated in 2014^[20], with a maximum quantum conversion efficiency of 48%. In 2016^[21], Vib-SRS output at 1553 nm was generated in the ${\mathrm{C}}_2{\mathrm{H}}_6$-filled AR-HCF with a 38% conversion efficiency reported, pumped by a microchip laser at 1064 nm with a linewidth of 6 pm and a maximum peak power of about 400 kW. Recently, Edelstein and colleagues explored the use of ${\mathrm{SF}}_6$ and ${\mathrm{CF}}_4$ in the FGRL by pumping with a high peak power $Q$-switched fiber laser at the 1030 nm wavelength^[22].

In spite of continuous interest in the pulsed FGRL, continuous-operation FGRL was seldom reported. Lack of proper pump sources with spectral brightness comparable with the pulsed laser makes the CW SRS threshold a challenge to reach. In 2007^[17], Benabid et al. firstly reported, to the best of our knowledge, the continuous running of rotational-SRS (Rot-SRS) of H₂ gas filled in a photonic-bandgap hollow-core fiber (PBG-HCF) by pumping with a single-frequency fiber laser at 1064 nm. The brightness of the pump source reached 250 W/MHz with a linewidth less than 100 kHz. Later, they successfully raised the continuous output up to 55 W^[18]. It is noted that in both works, the PBG-HCFs with a small core of 6 µm in diameter played an important role, where the spatially tight confinement of light contributed to the effective reduction of the lasing threshold.

In this Letter, we report the continuous Rot-SRS lasing in H₂-filled AR-HCF in a single-pass configuration with a large-core diameter exceeding 35 µm for the first time, to the best of our knowledge. Up to 26 W of Rot-SRS output at 1135 nm was demonstrated, pumped by a circularly polarized single-frequency fiber laser at 1064 nm with 10 kHz linewidth. A maximum slope efficiency of 40.3% was obtained in a 50 m fiber length filled with 2.5 bar (1 bar = 10⁵ Pa) H₂ gas. The threshold of FRGL was measured as 31.5 W. We found that the emergence of backward Rot-SRS and the cascaded Rot-SRS would ultimately limit the FRGL output.

The homemade AR-HCF in this paper was fabricated by the stack-and-draw technique method, using Heraeus F300 fused silica tubes, as shown in the inset of Fig. 1. The cladding of AR-HCF consists of seven capillaries, with an average capillary diameter of about 17 µm and core wall thickness of around 335 nm. The fiber attenuation was measured by a cut back and around 0.11 dB/m and 0.13 dB/m at 1064 nm (pump wavelength) and 1135 nm (first Stokes of Rot-SRS) wavelengths, respectively. The core diameter of AR-HCF is about 35 µm, and the fundamental-like mode diameter at 1064 nm wavelength is simulated around 26 µm.

Figure 2 shows the schematic of the experiment setup. The pump source was a homemade single-frequency continuous fiber laser operating at the 1064 nm wavelength, with a linewidth of 10 kHz and a maximum output power above 70 W^[25]. The laser output is linearly polarized and near diffraction limited with an ${\mathit{M}}^2$ of 1.2. A $\lambda /4$ waveplate (QWP) is used before the incidence of AR-HCF to convert the linearly polarized laser beam into circular polarization to enhance the Rot-SRS effect. A $4\text{-}f$ lens system is applied to scale the pump beam size, and, finally, around 80% coupling efficiency at pump wavelength at the incident end of AR-HCF was realized.

The single-pass FGRL consists of a 50 m AR-HCF, as shown in Fig. 2. The AR-HCF was loosely rewound on a steel plate in a circle with a radius of about 50 cm to avoid any notable bend loss^[26]. The two fiber ends were mounted in homemade gas cells and air-tight sealed. The whole length of AR-HCF could be vacuumed and pressurized via the gas cells. Cooling water was circulated inside the shell of the gas cell for heat management.

Before the incident end of the AR-HCF, a sampler (M1) (9:1) was used to monitor the pump power and collect the possible backward laser beam with a long-pass filter (LPF) together. Coated sapphire glass plates were used as the windows of gas cells with transmission efficiencies of 99% and 98.5% at 1064 nm and 1135 nm, respectively.

At the output end, the laser beam was collimated by the lens (L4) and then characterized by the thermal power meter, optical fiber spectrometer (Ideaoptics NIR2500), and pyroelectric array camera (Pyrocam IIIHR), respectively. Another LPF was used to separate the Stokes laser from the residual pump, with an average 97% transmission efficiency at wavelengths longer than 1100 nm and $<{10}^{-4}$ at 1064 nm (FELH1100).

In our experiment, the dependence of SRS on hydrogen pressure is explored at 2.5, 5, 7.5, and 10 bar, respectively. Typical spectra of forward emission at 5 bar for different pump powers are shown in Fig. 3. Figure 4 are the corresponding far-field beam patterns at the pump and Rot-SRS wavelengths. We attribute the slightly degraded output beam profile to the stress applied on AR-HCF when mounted by V-grooves in the gas cell.

As shown in Fig. 3, the cascaded Rot-SRS (RS2) at 1216 nm could be measured by the spectrometer at high pump power over 60 W while it was too weak for the thermal meter to respond. Vib-SRS at 1907 nm failed to be found for any ${\mathrm{H}}_2$ pressure, even when the pump power rose to 70 W. We attribute the absence of Vib-SRS to (1) the circular polarization of the pump that is expected to effectively suppress the Vib-SRS^[27] and (2) 1 dB/m high loss at 1907 nm that is almost ten times of that at 1135 nm.

Figure 5 summarizes the power of the Stokes and residual pumps as a function of coupled pump power measured by a power meter. In the experiment, an LPF was used to record the Stokes power, and the residual pump power was then calculated by subtracting the total power of the Stokes from the total output.

In Region I of Fig. 5, only the forward and backward Rot-SRS Stokes were measured at all pressures. At higher pump power, the cascaded second Rot-SRS started to appear in Region II but remained weak in the forward direction at 2.5 and 5 bar pressures. In Region III, the pump laser failed before the cascaded Rot-SRS because the quick rise of Rot-SRS in the backward direction strongly interfered with the pump laser operation. In Fig. 6, conversion efficiencies of bidirectional Rot-SRS are replotted from the data in Fig. 5.

The maximum output of forward Rot-SRS appeared at 2.5 bar pressure, with a conversion efficiency of 40.4%. At higher pressures, despite a reduced threshold from 50.6 W down to 31.5 W, the rising of backward Rot-SRS started to compete and exhausted the gain of the single pass. Moreover, the backward Rot-SRS resulting in the pump failure also stopped the potentially higher FRGL output.

According to the threshold formula^[18], $$P_{\mathrm{th}}=\frac{A_{\mathrm{eff}}{G}_{\mathrm{th}}}{g_s{L}_{\mathrm{eff}}}=\frac{A_{\mathrm{eff}}{G}_{\mathrm{th}}}{g_s}\frac{{\alpha }_p}{1-\exp (-{\alpha }_{p}L)},$$(1)where $A_{\mathrm{eff}}$ is the effective area of the modal field, $g_s$ is the peak steady-state Raman gain coefficient, and ${\alpha }_p$ is the fiber attenuation at the pump wavelength. At bending, extra loss of AR-HCF gives rise to a shorter effective length, resulting in a higher threshold. Based on the characterized pump thresholds for different pressures, when we assume that the pump power threshold rises to 70 W, the losses should increase to 0.15 dB/m, 0.18 dB/m, 0.21 dB/m, and 0.25 dB/m for 2.5 bar, 5 bar, 7.5 bar, and 10 bar pressures, respectively. According to our simulation of AR-HCF, the corresponding minimal bend radii are estimated as 7.6 cm, 6.3 cm, 5.8 cm, and 5.4 cm, respectively.

The maximum power of backward Rot-SRS reached 1.16 W at 10 bar pressure, with quantum efficiency close to 3%. Our continuous backward SRS generated in the experiment is much more efficient than reported previously^[17,18]. In Fig. 6, the measured thresholds of backward Rot-SRS were very close to the forward, implying a strong backward gain comparable with the forward. The tendencies of bi-directional Rot-SRS conversion are found very similar with the noise-seeded bi-directional Vib-SRS in Ref. [28] pumped by a $Q$-switched laser at 532 nm.

In Table 1, we summarize and compare the pump thresholds of H₂-filled FRGLs in the continuous and typical pulsed operations^{[13,18,19,29]}. By normalizing the threshold with the core area and pump laser linewidth, we find that the continuous pumping could be more efficient in terms of pump power utilization. It is noted that the pressure broadening effect and effective fiber length are not included in the normalization process.

In conclusion, we demonstrated a low threshold and efficient Rot-SRS in ${\mathrm{H}}_2$-filled large-core AR-HCF in a single-pass configuration at 1135 nm. A maximum power conversion efficiency of 40.3% was achieved by using a 50 m length of fiber filled with 2.5 bar H₂, and the maximum power reached 26.2 W. The noise-seeded backward Rot-SRS was observed with a maximum power of 1 W. A lower pressure is favorable to suppress the backward Rot-SRS, which competes with the forward counterpart and limits further increase of FRGL output.