Fiber lasers offer numerous advantages over traditional bulk lasers in terms of efficiency, compactness, robustness, beam quality, and other characteristics. Systems based on silica fibers doped with rare-earth ions have developed rapidly over a range of wavelengths in the spectral bands around 1 μm, 1.5 μm, and 2 μm. At longer wavelengths, attention has been on fibers formed from different host glasses that have higher transparency and lower phonon energies. Progress in this area is currently very rapid, with multiwatt performance being demonstrated over the last few years [1,2]. However, the performance of systems in the mid-IR based on these other glasses is well behind that at shorter wavelengths using silica. Further development of mid-IR fiber lasers faces challenges including a typically high quantum defect, thermal management, and fiber failure [3].

Gas-filled hollow-core fiber (HCF) lasers are emerging as an alternative technology [4], thanks to the development of silica-based HCFs [5] with low attenuation at mid-IR wavelengths. This type of laser source retains the advantages of fiber lasers, including high power conversion efficiency [6], near-diffraction-limit beam quality [7], compactness, and flexibility. It offers potential advantages over rare-earth-doped soft-glass fiber systems in the spectral band 3–4 μm, including very low fiber attenuation [5], a high damage threshold, and excellent power handling characteristics [8]. Instead of using rare-earth ions doped into a glass matrix, a gas in the HCF provides the gain for lasing.

The first CW gas lasing in HCF was achieved using molecular iodine in the 1280–1340 nm region pumped at 532 nm, but with a slope efficiency of only a few percent [9]. A 55 W output power was demonstrated by rotational stimulated Raman scattering (SRS) of ${\mathrm{H}}_2$ around 1 μm in HCF [10]. The large Stokes shift and rich content of ro-vibrational components make SRS in HCF a promising alternative technology for laser generation above 4 μm.

An optically pumped acetylene-filled HCF gas laser at 3 μm wavelength was first demonstrated in the pulsed region using an optical parametric oscillator pump at 1.52 μm [11]. Later, up to 30% conversion efficiency was demonstrated using a wavelength-stabilized diode laser pump source [6]. In 2016, we observed CW lasing of acetylene at a 3 μm wavelength in a low-loss antiresonant (AR)-HCF with feedback [12] and reported preliminary results using a single-pass configuration in early 2017 [13]. In this Letter, we use a systematic study of the performance of our laser system, scaled to over 1 W output power, to determine the underlying performance limits of this form of laser. A 10 W customized high-power erbium-doped fiber amplifier (EDFA) seeded by a tunable diode laser is used as the pump source. We measure 1.12 W output power at 3 μm, with 33% slope efficiency relative to the absorbed pump power. We systematically characterize the pump power absorption and output power scaling of the laser system, and suggest that the dynamics of the gain molecules ultimately determine the laser performance, with low fiber attenuation becoming the key to efficient laser operation for high-power output.

Acetylene molecules emit 3 μm photons via a vibrational mode transition from $v_1+v_3$ to $v_1$. The $P(9)$ absorption line from the $v_0$ ground vibrational state to the $v_1+v_3$ vibrational state is the preferred pump line due to its high thermal population and large absorption cross section [14]. As shown in Fig. 1, the $P(9)$ absorption line, the $J=8$ rotational state of the $v_1+v_3$ vibrational state is populated, and according to the selection rules, the excited molecules are allowed to transit to the $J=7$ and $J=9$ rotational states of the $v_1$ vibrational state, giving the $R(7)$ and $P(9)$ emissions, respectively [16]. These are the lasing lines around 3.1 μm.

The relaxation transition from $v_1$ to the $v_0$ ground level is dipole forbidden [17]. Efficient CW laser operation relies on nonradiative mechanisms to depopulate the lower laser level. Collision-induced energy transfer processes are therefore key to realizing a CW laser and producing a high-efficiency and high-power laser output.

Our pump source comprises a tunable distributed Bragg reflector diode laser (ID Photonics GMBH, CoBrite DX1) as the seed for a customized CW EDFA (Bktel Photonics, HPOA-S370ac). The seed laser has a specified maximum output power of 40 mW and a linewidth of less than 100 kHz. The maximum output power of the EDFA is 9.6 W with a strong amplified spontaneous emission (ASE) background [Fig. 2(b)]. When the seed laser wavelength is tuned to the acetylene absorption $P(9)$ line at 1530.385 nm and the amplified pump is passed through a 31 m HCF filled to around a pressure of 0.5 mbar, approximately 4.9 W is absorbed. We therefore consider that the maximum effective pump power for $P(9)$ absorption is around 4.9 W.

The coupling efficiency of pump light at the AR-HCF input end is generally above 90% using a coated ${\mathrm{CaF}}_2$ lens with a 50 mm focus length. At the output end, a dichroic mirror (94% transmission at 1.53 μm, 99% reflection at 3.16 μm) is used to separate the pump light from the mid-IR laser beam for measurement. A long-pass filter is used to remove residual pump light before the 3 μm laser power is measured by the thermal detector.

A 10-cell-cladding HCF made of fused silica was fabricated for the experiments, as shown in Fig. 2(c) (inset). Figure 2(c) shows measured attenuation curves around the pump and lasing wavelengths. Two cutbacks from 148 m to 30 m and from 118 m to 30 m give 37, 63, and 69 dB/km attenuations at 1.53, 3.12, and 3.16 μm wavelengths, respectively. To our knowledge these are the lowest combined attenuations (i.e., at pump and laser wavelengths) yet reported for HCFs for this application [4,6,7,11–13,16,18]. In fact, measurements of the pump transmission (tuned off-resonance) combined with this measured attenuation value can only be explained by assuming 100% pump coupling efficiency, indicating that the actual attenuation might be somewhat lower than the value measured using a white light source.

To characterize the pump absorption, we adopt a side-scattering measurement [19], as shown in Fig. 3(a). Pump light lost from the fiber leaves the fiber through the side walls. We collect some of this light locally using a standard multimode fiber, and analyze it using an optical spectrum analyzer (OSA).

Figure 3(b) displays typical spectra of pump light collected from the side of the fiber when the seed laser is tuned on and off the $P(9)$ absorption line. The wavelength shift is not resolved by the OSA, which is operated with low resolution for this measurement. $P_{\text{on}}/P_{\text{off}}$, the ratio of local pump light peak intensities on- and off-resonance, is used to measure the pump absorption as a function of the fiber length. A detuning of 0.05 nm from the absorption line is used to ensure that the seed laser is away from the absorption line for the off-resonance measurement, but that the amplification performance of the EDFA is unchanged.

Figure 3(c) plots the pump attenuation as a function of fiber length at 0.6 mbar pressure for different incident pump powers. The absorption mainly reflects the population of the lower state. For low incident pump powers (e.g., 15 dBm), the majority of molecules in the fiber are in thermal equilibrium and the state is thermally populated. Absorption is roughly consistent with the $-4.66\mathrm{dB}/\mathrm{m}$ at 0.6 mbar predicted in Fig. 1 (only Doppler broadening assumed). At increased pump power levels, populations are determined by the laser dynamics with a reduction in the ground-state population, as an increasing fraction of the population is shelved in post-lasing intermediate states and the ground state is not fully repopulated.

 

Fig. 1. Simulated absorption of acetylene gas in a 1.5 μm spectral region at 20°C using data from the HITRAN database [15]. For 0.6 mbar pressure, the linewidth is assumed to be decided by the Doppler broadening only. Inset: energy diagram of acetylene energy levels for 3 μm emissions when pumped at the 1.5 μm absorption band.

Download Full Size | PDF

 

Fig. 2. (a) Experiment layout of the single-pass configuration. The AR-HCF in the experiment is laid on an optical bench in loops about 1 m in diameter to remove any possible bend loss. (b) Measured spectrum of 9.6 W pump light at the output of the EDFA. Around half of the total pump power falls within a 3 nm spectral band centered at the seed wavelength. The pump line is not resolved in the measurement. (c) Measured attenuation of the AR-HCF. Inset: scanning electron microscope picture of the AR-HCF in the experiment. The fiber’s outer diameter is 199 μm, and the core diameter is 75 μm.

Download Full Size | PDF

 

Fig. 3. (a) Schematic of the side-scattering measurement. An optical spectral analyzer (OSA; Yokogawa AQ6370D) is used to measure the collected side-scattering spectra. (b) Typical measured side-scattered spectra of pump light near $P(9)$ in on-/off-resonance conditions. The ratio of peak intensities is used to represent the relative absorbed pump power. (c) Pump attenuation along the fiber length at 0.6 mbar pressure for different incident pump powers. Points showing 0 dB pump attenuation at zero distance are added rather than measured directly.

Download Full Size | PDF

Figure 4(a) shows the measured 3 μm laser power for different gas pressures as a function of the total incident pump power (including the broadband ASE). Figure 4(b) replots the slope efficiency curves with the $x$ axis changed to the measured absorbed pump power at $P(9)$. The maximum slope efficiency is 33% when the pressure is 0.6 mbar, with 1.12 W maximum output power at 3 μm. When calculated versus the assumed total usable pump power, the efficiency is 23%. At the lowest pressures [0.23, 0.42 mbar; see Fig. 4(a)], the reduced pump absorption at high pump powers means that the pump is only partially absorbed and the gain saturates. At 0.6 mbar, there is no longer evidence of gain saturation (although there would be for a shorter fiber length) and a maximum output power of 1.12 W is recorded. At higher gas pressures (above 0.6 mbar, but particularly above 1 mbar) the increased intermolecular collision rate increases the internal losses and decreases the gain, leading to a reduced output power and an increased laser threshold.

 

Fig. 4. For 15 m fiber length and various pressures: (a) incident pump power—mid-IR laser output curves, (b) absorbed pump power—mid-IR output power curves.

Download Full Size | PDF

The measured output powers are the sum of both $R(7)$ and $P(9)$ emissions. The relative strengths of $R(7)$ and $P(9)$ emissions follow the same trends as reported in Ref. [13].

Figure 5(a) summarizes the measured slope efficiencies and maximum absorbed power as functions of the pressure. As pressure is increased, collisions increasingly shorten the lifetime of the upper laser level. The $J=8$ rotational state of the $v_1+v_3$ vibrational state has been reported to be depopulated at a total removal rate of $12.3\times {10}^{10}{\mathrm{cm}}^3{\mathrm{s}}^{-1}$ (measured when pressure is lower than 0.4 mbar) in Ref. [20], by the vibrational and rotational energy transfer processes via intermolecular collisions. However, when the gas is confined in the hollow core, the mean free path between intermolecular collisions is greater than the core radius for pressures lower than about 2.2 mbars. In this case, the inelastic relaxation via collisions of acetylene molecules with the silica core wall dominates the collision rate and may have a major impact on the laser operation. Inelastic relaxation affects both the upper and lower laser levels as well as the ground state. On the one hand, it assists repopulation of the ground level and plays a valuable role in maintaining CW laser operation. On the other hand, it simultaneously reduces the population of the upper laser level.

 

Fig. 5. (a) Summary of measured slope efficiencies and maximum absorbed pump power for two different fiber lengths as a function of acetylene pressure. Slope efficiencies in the saturation condition are not included. (b) Pump threshold for two different fiber lengths as a function of acetylene pressure.

Download Full Size | PDF

Compared to 15 m, a 6 m gain fiber length slope efficiency has a similar dependence on pressure [Fig. 5(a)], but the absolute slope efficiencies at low pressures are different. The reduction of conversion efficiency due to a shorter gain fiber length has also been confirmed in the pulse-pumped single-pass configuration [18].

In the experiments, a threshold occurs when ASE into the low-loss fiber mode becomes the dominant route for de-excitation of the upper lasing level, reducing fluorescence and nonradiative transition processes and increasing the efficiency of converting pump photons into coherent fiber output. Figure 5(b) summarizes the measured pump threshold for two fiber lengths and different pressures. In Fig. 5(b), the measured pump threshold is found to be almost independent of gain fiber length but dependent on the gas pressure, as expected. The minimum pump threshold is measured to be 48 mW when the pressure is 0.23 mbar for a 15 m fiber length, larger than the 34 mW reported in a ring cavity [12] (where, of course, the threshold represents a slightly different effect).

In the quasi-CW region, the important role of acetylene pressure on the laser’s power-scaling performance has also been confirmed for a pulsed pump with pulse durations of tens of nanoseconds [6,18]. For pulse widths of approximately 1 ns, a constant 20% slope efficiency was reported for pressures in ranging from 1.6 to 18.7 mbars [7]. As the pulse length becomes shorter and enters the transient region, the pump duration becomes shorter than the collisional kinetic process. We can then assume that no collision-induced changes take place (over a certain pressure range) during the pulse duration. In that case, only a small fraction of molecules can participate in the lasing process, determined by the thermal equilibrium. In this case, a higher pressure may be preferable for more efficiently utilizing the pump power.

In this Letter, a systematic study of the power-scaling dependence on gas pressure, fiber length, pump intensity, and pump absorption reveals a reduction in the pump absorption (and therefore fiber gain) with increasing pump flux and a degradation of gain performance at high gas pressure. Taken together, these necessitate the use of increasingly long lengths of fiber for efficient laser operation at increased powers. We have demonstrated 1.12 W output power at 3 μm wavelength with 33% slope efficiency (relative to absorbed pump) in a 15 m low-loss AR-HCF filled with acetylene gas at 0.6 mbar. Demonstrating significant further power scaling of this system would require the use of longer fiber lengths and ultimately lower-loss fiber. One possible future direction might be to investigate the use of a buffer gas to affect the molecular dynamics and to improve the de-excitation rates back into the ground state.

Acetylene-filled HCF presents an alternative route to high-power mid-IR laser generation with the potential to overcome the barriers of phonon absorption, optical nonlinearity, and fiber damage in a rare-earth-doped solid fiber laser. The data in this Letter provide information that will enable a deeper understanding of the molecular dynamics in this system [21] and offer guidance for further power scaling.

All data underlying the results presented in this Letter can be found in Ref. [22].