Mid-Infrared Fiber Amplification of a DFB Interband Cascade Laser

Abstract

The limited availability of powerful, tunable, and reliable mid-infrared sources has historically prevented their widespread adoption in spectroscopy applications, even if most greenhouse gases’ fundamental absorption lines are found in this region of the electromagnetic spectrum. However, both mid-infrared fiber lasers and ICLs have enjoyed substantial growth in available output powers in recent years. Since the two technologies have complementary benefits, combining them could prove to be an interesting avenue to explore toward the development of a powerful, easily tunable, and narrow linewidth mid-infrared source. We report what we believe to be the first demonstration of fiber amplification of a distributed feedback interband cascade laser (DFB-ICL) operating in the mid-infrared. The system, based on an in-band pumped dysprosium-doped fluoride fiber amplifier, yields 10 dB of gain and up to 30 mW of output power at 3240 nm. We believe this is an important milestone toward power scaling of single-mode, single-frequency, and rapidly tunable mid-infrared laser sources suitable for advanced gas spectroscopy.

1. Introduction

The mid-infrared (Mid-IR) region of the electromagnetic spectrum is of great interest because it encompasses most gases’ fundamental absorption lines [1]. These vibrations can be hundreds of times stronger than their harmonics found in the near-infrared, but the limited availability of mid-IR laser sources has somehow prevented the widespread adoption of this spectral region for gas spectroscopy.

However, in recent years, the mid-IR coherent sources have experienced a spectacular growth in terms of their output power, fueled by significant progress made in fiber and photonic components development. This has led to their use in numerous applications such as laser surgery [2,3], polymer processing [4,5,6], and remote sensing [7,8]. The fundamental vibrations of greenhouse gases (GHG) found in the mid-IR are thus relevant for sensing because they have the potential to yield better signal strength and accuracy. Interband cascade lasers (ICL) and fiber lasers have both proven to be interesting options for field applications, due to their robustness and portability [9,10].

Mid-IR ICLs also prove to be advantageous when it comes to remote gas sensing since they can emit a near diffraction-limited single-mode beam, with a single-frequency linewidth that is much narrower than that of a typical fiber laser. They are also easily and rapidly tunable by direct current modulation, which is a great advantage for advanced spectroscopy. However, powerful mid-IR distributed feedback (DFB) ICL sources are not yet available in the 3–4 µm range [11,12]. This region is of interest for sensing gases such as methane, ethylene, and water vapor, but achieving high power from an ICL while maintaining near diffraction-limited, single-frequency continuous wave (CW) emission has proven to be difficult at room temperature, with modest available output powers in the mW range [13,14,15,16].

Mid-infrared fiber lasers, on the other hand, have progressed significantly in terms of output power in recent years [17,18,19]. While their output spectrum is typically not as narrow as ICLs’ single-frequency spectrum, nor easily tunable, they do offer exceptional near diffraction-limited single-mode beam (M² ≈ 1.1–1.2) with a great potential for power scaling, along with the convenience of fiber-based delivery.

In this paper, we report on what is to our knowledge the first demonstration of an ICL laser operating in the mid-infrared amplified by a fiber device. The system, based on an in-band pumped dysprosium fluoride fiber, emits near 3240 nm, a wavelength range highly relevant for methane detection as it overlaps a strong methane absorption line that is isolated from other atmospheric interferents. Despite a low pump-to-signal optical conversion efficiency of only 0.4%, the amplifier provides a diffraction-limited output beam (M² < 1.2) and a gain of around 10 dB vs injected signal, representing the highest gain reported so far for a mid-infrared dysprosium-doped amplifier to our knowledge [20]. While the maximum output power is rather modest, i.e., 30 mW, we believe this shows a very promising avenue for future power scaling of rapidly tunable single-frequency sources in the mid-infrared. This article is a revised and expanded version of a paper entitled “Amplification of a Single-Frequency Interband Cascade Laser around 3240 nm using a Dysprosium-doped Fluoride Fiber”, which was presented at the “2025 Conference on Lasers and Electro-Optics Europe European Quantum Electronics Conference, Munich, Germany, 24 June 2025” [21].

2. Materials and Methods

The all-fiber architecture of the amplifier is similar to the one simulated recently by Annunziato et al. [22]. The experimental setup is shown in Figure 1. A DFB-ICL from Nanoplus (Gerbrunn, Germany) (DFB-300600, T0-66 cubic mount) with a collimated output is used as a seed for the laser amplifier. It outputs up to 11 mW of power and is tunable from 3239 nm to 3245 nm. The seed output beam size is reduced using two ZnSe lenses in order to pass through an optical isolator (IO-4-3200-WG, Thorlabs, Newton, NJ, USA) used to prevent any optical power from being redirected to the DFB-ICL in case of parasitic lasing from the dysprosium-doped fiber amplifier. Due to losses in the optics and the isolator, 81% of the seed power reaches the injection lens.

The aspherical injection lens is made of ZnSe (anti-reflection coated, f = 25 mm) to properly inject the DFB-ICL signal into the single-mode core of an undoped double-clad fluoride fiber on which a fuseless pump combiner was made in-house [23]. Note that a cladding mode stripper (CMS) was added at the input of the system to ensure that the seed signal is injected only in the fiber core of the combiner. The pump combiner is used to couple the light from a 90 W 980 nm laser diode into the cladding of the undoped fiber. The latter is then spliced to the Er³⁺ fiber laser emitting around 2910 nm, i.e., the green fiber in Figure 1, and has an architecture similar to the one reported in [24]. This Er³⁺ fiber laser, which serves as a pump source for the dysprosium fiber amplifier, is made out of a 3.5 m, 7 mol.% erbium-doped fluoride fiber that is mode matched with both the undoped fiber and the dysprosium-doped fiber. All splices were performed using a Vytran GPX-3000 (Thorlabs, Newton, NJ, USA) equipped with an iridium filament. Two fiber Bragg gratings (FBG) at 2910 nm acting as input and output couplers of the pump fiber laser cavity were written through the fiber coating using 800 nm fs-pulses [25]. The use of FBGs, as well as the pump combiner, allows for the amplifier system to be all-fiber, making it more robust and the alignment steadier.

The choice of 2910 nm as the pump wavelength was made in order to prevent parasitic lasing of the dysprosium. The obvious drawback is that this wavelength is away from the absorption cross-section peak of the dysprosium-doped fluoride fiber located near 2820 nm [26], leading to weaker pump absorption. However, for this in-band pumped dysprosium system, the emission cross-section peaks around 2860 nm, which means that it naturally emits preferentially near this wavelength, which could cause parasitic lasing at such a wavelength if there is any feedback from the cleaves or splices. This is especially true with a strongly pumped dysprosium fiber, as in this experiment, since the signal reabsorption is strongly diminished at shorter wavelengths. Pumping at 2910 nm thus helps to prevent parasitic lasing around 2860 nm. One can compensate for the weaker absorption by choosing a longer gain fiber, making 2910 nm a safer pump wavelength without significant drawbacks.

The 2910 nm erbium fiber laser is then used to core-pump a 5 m-long dysprosium-doped fiber (2000 ppm), which amplifies the 3240 nm seed through in-band pumping. This fiber and the erbium-doped and undoped fibers are commercially available from Le Verre Fluoré (Bruz, France).

The actual maximum seed power injected into the dysprosium fiber was only 3.6 mW, corresponding to an injection efficiency of 41%. This was partly due to the presence of two fusion splices (10% loss each). Note that the signal absorption by the erbium ions is negligible at the DFB-ICL wavelength of 3240 nm, so that it is essentially not absorbed over the 3.5 m long erbium-doped fiber, with background losses of 28 dB/km (~0.1 dB losses for the 3.5 m fiber laser). Background losses of dysprosium fiber are 85 dB/km (~0.4 dB losses for the 5 m amplifier). The previous background loss data were provided by the manufacturer at 3500 nm, outside the dysprosium absorption band. Both fiber tips, i.e., the entrance of the pump combiner and the output of the dysprosium fiber, were cleaved at an angle of ~5–6 degrees to prevent any parasitic lasing of the amplifier.

The output power of the amplifier was measured for different pump and seed powers with a XLP12-3S-H2-D0 (Gentec-EO, Québec City, QC, Canada) thermopile powermeter. A thin-film interference filter, fabricated in-house, was used to transmit the 3240 nm wavelength, filtering out the residual pump and thermal noise. The transmission of this filter was actually 92% at 3240 nm and 0.44% at 2910 nm.

3. Results

Figure 2a shows the amplified 3240 nm signal as a function of the pump power for different seed powers. The output power increases exponentially at lower pump power, i.e., when the fiber is improperly pumped and strong signal absorption is present. It becomes linear at higher pump power and starts to saturate over 11 W of pump power, when residual pump begins to exit the system, as shown in Figure 3.

Accordingly, Figure 2b shows the unsaturated linear increase of the 3240 nm output power vs. the seeded one. Note that saturation starts to be observed at higher pump power, slightly beyond 2.5 mW of injected seed power.

The maximum output power achieved is 30 mW, with a rather low pump-to-signal efficiency of 0.4%.

Figure 4 shows an experimentally obtained map of the measured gain of the amplifier for different values of both the injected seed and the pump powers.

One sees that, for our 5 m-long active fiber, the gain appears to be maximum around 2.2 mW of seed. This is slightly more obvious at high pump power, although it remains a relatively small variation around a maximum gain of ~10 dB.

Figure 5a shows the normalized output spectrum of the amplifier for different pump values at the maximum seed power (3.63 mW). It shows a very narrow spectrum with an FWHM measured to be 0.05 nm for every pumping level (including no pumping). This corresponds to the minimum resolution of the optical spectrum analyzer (OSA) used for the measurement (Yokogawa AQ6375L, 0.05 nm resolution, Yokogawa Electric Corporation, Tokyo, Japan). The bandwidth is thus narrower than the resolution value. Note that the DFB-ICL seed bandwidth provided by the manufacturer is ≤0.1 pm (≤3 MHz).

Figure 5b shows the output spectrum for a constant pump power of 10.3 W and different seed powers, for a fixed temperature. We can see that the system is easily tunable through current modulation of the seed source, which is typical for DFB-ICL systems, allowing for rapid tuning of their output spectrum up to the tens of kHz range. The change in amplification due to the tuning wavelength, for an equivalent seed power, is negligible in this spectral range.

The output spectrum was also measured outside the spectral range shown in Figure 5a,b. No parasitic lasing or ASE was observed from 2800 nm to 3400 nm.

The beam quality was measured. The knife-edge technique, along with a ZnSe lens having a 100 mm focal length, was used to calculate the beam width at different distances from the beam waist. The calculated value of the beam width was then used to fit the beam divergence and thus calculate the M² value.

We found a M² value of 1.17, which is close to the theoretical limit of a diffraction-limited beam, as expected for a single-mode fiber output.

4. Discussion

The low efficiency and power of the system are to be expected considering the limited gain of about 10 dB at the seed wavelength of 3240 nm, quite far from the maximum gain of this Dy³⁺ transition occurring near 2860 nm, combined with the low average power of a few mW provided by the seed laser. Propagation losses in the fiber caused by the presence of scattering defects in the gain fiber, the strong signal reabsorption, and the residual (unused) pump power are some of the factors that could be limiting the gain provided by the amplifier.

It is interesting to compare these results to the simulated results reported by Annunziato et al. [22]. Their simulated setup is similar to the one presented here, except they used 5 W of pump power at a different wavelength (2820 nm instead of 2910 nm) and a different fiber length (3 m instead of 5 m). The absorption cross-section at 2910 nm of this dysprosium transition is about 40% weaker than at 2820 nm. The 2 m longer gain fiber thus somewhat compensates for the weaker absorption. Although direct comparisons are difficult because they also did not use the exact fiber we used, the results are of the same order of magnitude as our results shown in Figure 4. They found, for a similar seed power, 5 dB of gain at 3240 nm for a 5 W pumping power, while we achieved 5 dB of gain around 7 to 8 W of pump power. The difference can be explained in part by the different pump wavelength and fiber length, but could also be explained by different propagation losses and a different fiber geometry.

They also suggested that the amplifier gain changes very little with seed power with a signal around 3240 nm, in agreement with our measurements. This explains, in part, the quite low efficiency of our amplifier, but it also means that, with a stronger seed, the overall efficiency of the system and maximum output power can be significantly improved. With small seed powers, 10 dB of gain means that the signal is only amplified on the order of magnitude of tens of milliwatts, for multiple watts of pump, thus yielding a poor efficiency. Using a stronger seed (in the tens to hundreds of milliwatts), the signal would be amplified up to the watt level, with a much better pump-to-signal efficiency. This is a promising result because it means that adding more stages to the current amplifier would significantly increase its conversion efficiency and output power.

It should be mentioned that the gain is currently limited, in part, by the length of the dysprosium-doped fiber. As shown in Figure 3, pump absorption starts to saturate for our maximum pump power, leading to unabsorbed pump power. A longer fiber, provided that more pump power is also available at 2910 nm, would thus allow us to increase the small signal gain. A higher doping concentration could provide an amplifier with a shorter optimal gain fiber length, thus potentially lower overall propagation losses; however, this could lead to unwanted thermal effects or, if the concentration is too high, to concentration quenching.

As discussed before, a promising avenue to explore for power scaling this architecture further would be to add successive amplifier stages to the current setup. We believe that our results, as well as the simulations in [22], show that the easiest way to increase the output power and overall efficiency of the system is to increase the seed power. Combined with an active fiber length that would be adjusted for the maximum available pump power, this would allow for significantly higher output powers, since the current system would be used as a 30 mW seed for that second stage of the amplifier, and so on. This approach is very feasible, since the second stage would be an identical copy of the first, simply joined by a splice. This same type of splice was performed twice while building the first stage. With a second combiner and an erbium fiber laser as a pump, the setup would thus remain all-fiber.

In a previous work, we showed that laser sources emitting around 3240 nm, with good beam quality, are interesting for advanced spectroscopy techniques like, in the case of our previous experiment, active imaging of methane and water vapor using a fiber laser as an illumination source [10]. However, data analysis following the field work was more complicated than it would have been with a narrow-linewidth laser source, since the broader linewidth of the fiber laser needs to be accounted for. This wavelength range and the wavelength range around 3270 nm are very interesting for methane spectroscopy since they align with strong methane absorption lines that do not overlap with other atmospheric gases’ spectral lines. Many other demonstrations were performed for the detection of methane at these wavelengths, but they mostly used DFB-ICL [9,27,28] for their narrower linewidth and fast tunability. However, they suffer from a limited output power, with the current record holder, to our knowledge, emitting 24 mW [13] around 3250 nm. While the output power of our current system is only slightly higher, we believe it has greater and faster power scaling potential, with the added benefit of a single-mode fiber couple output. It should be noted that dysprosium is currently the most available doping ion for fiber amplification at this wavelength [5]. While holmium could be an option as it was reported to lase around 3220 nm in fluoride glass fibers [29], this system was so far much less developed compared to dysprosium. Very few single-pass, mid-infrared dysprosium amplifiers exist, and our system is, to our knowledge, the highest gain reported for such an amplifier [20]. Oscillators have reached significantly higher outputs, such as reported in [5,6,10,30], but the system presented here has a significantly narrower linewidth than the previous fiber lasers, while having an easier, faster, and more extensive tuning system. This shows interest for future uses of this amplifier in methane sensing, provided the output power can be improved.

5. Conclusions

In conclusion, we report a mid-IR dysprosium-doped fiber amplifier seeded with a DFB-ICL around 3240 nm. It is, to our knowledge, the first fiber amplifier to use this type of seed in the mid-infrared. The amplification stage is all-fiber, constructed using a dysprosium-doped fluoride fiber as the gain fiber, which is core-pumped by a 2910 nm erbium-doped fiber laser. The amplifier yields ~10 dB of gain with an output power of 30 mW and a pump-to-signal efficiency of 0.4%. The output power and gain are limited by the length of the dysprosium fiber in combination with the available pump power, leading to a significant amount of unabsorbed pump exiting the amplifier at high powers. The laser output spectrum maintains the very narrow linewidth of the DFB-ICL seed with a FWHM below the OSA resolution of 0.05 nm. The beam quality was also investigated, resulting in a M² of 1.17, confirming the single-mode operation of the system. We believe this shows a promising path for future power scaling of single-frequency, single-mode, and rapidly tunable mid-IR sources. This is especially true in the 3–4 µm window, which is an interesting region for spectroscopy, but where progress has been slower for laser development.