Ultra-short pulse laser sources operating in the mid-infrared region have important application prospects in fields such as molecular spectroscopy, laser surgery, bio-diagnostics, and material processing [1]. The 2.8-µm Er-doped ZBLAN fiber laser has achieved significant progress in the last decade. A 2.8-µm continuous wave Er-doped ZBLAN fiber laser with an all-fiber structure achieved a maximum power of 41.6 W [2]. Mode-locked fiber lasers operating at 1, 1.5, and 2 µm have been well developed. In contrast, mode-locked fiber lasers that operate in the mid-infrared region remain a challenge because of the slow development of optical fiber devices and free-space optical components. In 2015, Duval et al. realized the first 2.8-µm mode-locked Er-doped ZBLAN fiber laser based on nonlinear polarization rotation (NPR) [3]. Subsequently, Hu et al. demonstrated 497-fs pulses using a similar NPR mode-locking method [4]. Moreover, 2.8-µm mode-locked Er-doped ZBLAN fiber lasers based on real saturable absorbers such as semiconductor saturable-absorber mirrors (SESAMs) [5,6], graphene [7], and black phosphorus [8] have also produced good results. Er-doped ZBLAN fiber amplifiers have been used to boost the 2.8-µm pulse seed to the level of a few watts, accompanied by pulse nonlinear compression [9–11]. Although these 2.8-µm Er-doped fiber oscillators and amplifiers have achieved impressive results, expensive and bulky free-space optical components were adopted in these systems, which reduced their reliability and limited their applications.

Alternatively, mid-infrared pulses can feasibly be achieved through optical nonlinear frequency conversion. For instance, the soliton self-frequency shift (SSFS) is a widely used to obtain ultra-short pulses out of the gain range of rare-earth-doped fibers. Some researchers demonstrated 2-µm ultra-short pulses based on the SSFS in highly nonlinear fibers pumped by 1.5-µm pulses [12–15]. The SSFS in thulium-doped fiber amplifiers (TDFAs) has also received considerable attention for obtaining ultra-short pulses beyond 2 µm [16–18]. Limited by the transmission loss of silica fibers, it is difficult for redshifted solitons to exceed 2.5 µm [19]. Although some results were achieved in the mid-infrared SSFS by pumping fluoride fibers with high energy near-infrared ultra-short pulses [20–22], the free-space structure limits their applications. Cascading a piece of high GeO₂-doped silica fiber [23–26] or fluoride fiber [27] to the pigtail fiber of a TDFA has realized an SSFS of ∼2.8 µm. Once a high-performance 2.8-µm pulse seed is obtained, it seems natural to amplify it in an Er-doped fiber amplifier. Recently, Wu et al. realized an all-fiber structured 2.8-µm picosecond pulse amplifier seeded by the SSFS from a mode-locked thulium-doped fiber laser with a maximum power of 1.25 W and a pulse energy of 26.02 nJ for the first time [28]. The pump combiner is a critical component for realizing an all-fiber structure. Schäfer et al. fabricated the first fluoride-fiber side-pump combiner, which spliced an angle polished multi-mode fluoride fiber to a double-cladding fluoride fiber as the pump pigtail fiber [29]. Magnan-Saucier et al. reported another technique for side-pumping fluoride-based double-cladding fibers, in which a long-tapered multi-mode silica fiber with a waist diameter of 12 µm was wrapped on a double-cladding ZBLAN fiber [30]. In Ref. [28], Wu et al. adopted this method but only realized a coupling efficiency of 31%. A higher coupling efficiency of combiners is expected to be achieved using the end-pump technique as has been demonstrated in silica fiber pump combiners.

In this Letter, we demonstrate an all-fiber 2.8-µm ultra-short MOPA system seeded by the SSFS of 2-µm ultra-short pulses. A mode-locked thulium-doped fiber laser was used as the pulse seed. A 2.8-µm pulse seed was obtained by a cascaded SSFS in a TDFA and a passive fluoride fiber (PFF). Then, the 2.8-µm pulse seed and a 976-nm pump laser from a multi-mode laser diode (LD) were coupled to a piece of Er-doped ZBLAN fiber via a novel home-made high-efficiency end-pump silica-fluoride fiber combiner (EPSFFC). With a relatively low pump power, gain narrowing leads to spectral narrowing of the amplified 2.8-µm pulses. With an increase in the 2.8-µm pulse energy, nonlinear soliton self-compression and spectral broadening were observed. A pulse with a pulse energy of 42 nJ and a pulse width of 115 fs was achieved with a maximum output power of 3.42W.

A schematic of the all-fiber 2.8-µm laser system is shown in Fig. 1(a). An NPR mode-locked thulium-doped fiber laser was adopted as the pulse seed for the system, as shown in Fig. 1(b). The mode-locked pulse train was amplified in a TDFA. The thulium-doped fibers had a core/cladding diameter of 10/130 µm (SM-TDF-10P/130-HE, Nufern). Then the amplified pulses were injected into a piece of 5-m PFF with a core/cladding diameter of 6.5/125 µm (ZFG SM [1.95] 6.5/125, LVF). A piece of ∼20-cm SM-1950 fiber with a core/cladding diameter of 7.5/125 µm was spliced between the SMF-28 fiber and the PFF to decrease the splice loss. The SM-1950 fiber was spliced to the PFF using an arc discharge fiber fusion splicer with a loss of ∼1.5 dB at 2 µm. The redshifted soliton and 976-nm pump laser were coupled to a 5-m Er-doped ZBLAN fiber (ZFG DC [2.50] (Er³⁺ 70000) 15/240 × 260/290, LVF) via a home-made EPSFFC. A micro-image of the EPSFFC is shown in Fig. 1(c). A PFF was spliced to a passive double cladding ZBLAN fiber with core/cladding diameter of 14/250 µm (ZFG DC [2.20] 14/250/290, LVF) as the signal fiber. A Vytran GPX-3000 fiber fusion splicer was used to splice these fluoride fibers. As the passive double cladding ZBLAN fiber had a much larger diameter than the PFF, there was a shoulder at the splice point. Therefore, there was a spare space at the splice point to accommodate a fiber tip. Meanwhile, a piece of multi-mode silica fiber with a core/cladding diameter of 105/125 was tapered to the outer diameter of ∼45 µm. The tapered silica multi-mode fiber was cleaved at the taper waist. Then the tapered fiber end clung to the spare space at the splice point. These fibers were then fixed on a piece of sapphire glass sheet using low-refractive index UV curing glue and packaged into a metal box. The EPSFFC had a coupling efficiency of ∼85% for the 976-nm multi-mode LD. An endcap composed of a 200/250 ZBLAN fiber was spliced to the output end of the Er-doped ZBLAN fiber.

 

Fig. 1. (a) Schematic of the all-fiber 2.8-µm MOPA system. (b) Schematic of the NPR mode-locked thulium-doped fiber laser. (c) Micro-image of the EPSFFC.

Download Full Size | PDF

With a pump power of 380 mW, stable fundamental repetition rate mode-locking could be achieved easily. The mode-locked output spectrum had a central wavelength of 1933 nm and a 3-dB bandwidth of 10.1 nm, as shown in Fig. 2(a). A stable pulse train detected by an oscilloscope is shown in Fig. 2(b). The measured radio frequency (RF) spectrum around the fundamental repetition rate is shown in Fig. 2(c). The RF signal was located at 75.27 MHz and exhibited a signal-to-noise ratio of 75 dB. The autocorrelation trace and fitting curve of the output pulse are shown in Fig. 2(d). The pulse width was ∼637 fs, assuming a sech shape.

 

Fig. 2. Output characteristics of the mode-locked thulium-doped fiber laser. (a) Spectrum. (b) Pulse train. (c) RF spectrum around the fundamental rate. (d) Autocorrelation trace and fitting curve.

Download Full Size | PDF

Figure 3 shows the output spectra of the TDFA for various output powers. With an increase in the 793-nm pump power, high-efficiency SSFS occurred in the TDFA and silica pigtail fiber of the isolator. The Raman soliton shifted to ∼2064 nm with a total output power of 186 mW, as shown in Fig. 3(a). The soliton comprised ∼99.2% of the total power calculated using the spectral integral. The Raman soliton was redshifted to ∼2198 nm with an output power of 406 mW. A further increase in the 793-nm pump power resulted in the emergence of the second Raman soliton. When the output was 424 mW, the first Raman soliton had a central wavelength of ∼2249 nm, as shown in Fig. 3(c). A further increase in the pump power resulted in drastic amplification and SSFS of the second Raman soliton, whereas the first soliton exhibited a small central wavelength redshift and some bandwidth narrowing. When only one soliton appeared, its longest central wavelength was located at ∼2250 nm, as shown in Figs. 3(a)–3(d).

 

Fig. 3. Output spectra of the TDFA with various output powers.

Download Full Size | PDF

The following PFF had a zero-dispersion wavelength of ∼1.65 µm provided by the manufacturer. In the PFF, the Raman soliton redshifted further to the mid-infrared region. Figure 4 shows the output spectra with various 793-nm pump powers. The central wavelength of the most redshifted Raman soliton increased from ∼2380 nm to ∼2800 nm when the 793-nm pump power was increased from 2.93 W to 3.78 W. Nevertheless, the central wavelength of the most redshifted Raman soliton decreased with a further increase in the pump power. This was because the further increase in the 793-nm pump power resulted in a decrease in pulse energy and bandwidth of the first Raman soliton output from the TDFA.

 

Fig. 4. Output spectra of the PFF with various 793-nm pump powers.

Download Full Size | PDF

In the following experiments, we maintained the 793-nm pump power at 3.78 W. The output power of the pulse seed from the Er-doped ZBLAN fiber was 7 mW. The large loss mainly resulted from the splicing point between the signal fiber of the EPSFFC and the PFF, and between the PFF and the double cladding ZBLAN fiber because of the large mode mismatch. The absorption of the Er-doped ZBLAN fiber also resulted in some absorption of the pulse seed at 2.8 µm. Despite the large loss, the power of the 2.8-µm pulse seed was still sufficient to be amplified in the Er-doped ZBLAN fiber amplifier as proved in the following experiments. The output power of the Er-doped ZBLAN fiber with respect to the launched 976-nm LD pump power is shown in Fig. 5. In the output power measurement, a piece of long-pass filter with an edge of 2055 nm was used to filter out the residual pump. The long-pass filter had a transmission of ∼88% at 2.8 µm. A maximum output power of 3.42 W was obtained with a 976-nm LD pump power of 20.1 W. The slope efficiency was ∼17.2% and no obvious decrease in efficiency was observed. Further increasing the efficiency of the EPSFFC and optimizing of the Er-doped fiber length will be promising for increasing the slope efficiency.

 

Fig. 5. Output power of the Er-doped ZBLAN fiber with respect to the launched 976-nm pump power.

Download Full Size | PDF

The output spectra of the Er-doped ZBLAN fiber amplifier at various output powers are shown in Fig. 6. The spectrum intensity at ∼2.8 µm was significantly enhanced with an increase in the launched 976-nm pump power. Initially, the spectrum had a considerably narrower bandwidth than that of the seed. For instance, with an output power of 725 mW, the amplified spectrum had a central wavelength of ∼2784 nm and a 3-dB bandwidth of ∼11 nm. This can be attributed to the gain-narrowing effect in the Er-doped ZBLAN amplifier. The amplified spectrum around 2.8 µm broadened continuously with increasing pump power, especially in the long-wavelength region. With the maximum output power of 3420 mW, the spectrum had a 3-dB bandwidth ranging from 2774 nm to 3016 nm. Simultaneous nonlinear spectral broadening and temporal pulse compression were accompanied by pulse amplification. Meanwhile, asymmetrical spectral broadening was observed, which is more distinct at high powers. We believe that stimulated Raman scattering led to a stronger gain in the long-wavelength region, which is a symptom of SSFS. An obvious SSFS is expected with a higher output power.

 

Fig. 6. Output spectra of the Er-doped ZBLAN fiber amplifier at various output powers.

Download Full Size | PDF

The measured autocorrelation traces and their sech fitting curves for the amplified 2.8-µm pulses at various output powers are shown in Fig. 7. The 2.8-µm pulses generated in the PFF should had an estimated pulse width of hundreds of femtoseconds. However, its autocorrelation trace could not be directly measured because of its low power. The measured pulse width was ∼6.3 ps with an output power of 189 mW, as shown in Fig. 7(a). The broadening of the pulse width mainly resulted from the large dispersion of the passive double-cladding ZBLAN and Er-doped ZBLAN fibers. With an increase of the output power, the pulse width was gradually compressed accompanied by spectral broadening. The pulse width was compressed to 225 fs with an output power of 991 mW and a broad pedestal could be observed, as shown in Fig. 7(c). The autocorrelation trace had two side pulses on the pedestal with the output power of 2134 mW, as shown in Fig. 7(d). The pedestal degraded significantly, and the pulse width was compressed to ∼115 fs with an output power of 3420 mW. Spectral broadening occurred in the amplifier mainly resulting from self-phase modulation and significant positive chirp was introduced in this process. Meanwhile, the negative dispersion of the active fiber compensated the positive chirp resulting in compression of the pulses. The pedestal and side pulses, which are sensitive to the characteristics of the seed pulse such as pre-chirp, pulse width, and spectral bandwidth, can be attributed to the nonlinear chirp introduced in the spectral broadening and high-order dispersion of ZBLAN fiber [31]. We believe that the pedestal and side pulses were prevented because of a better chirp compensation in the case of an output power of 3420 mW. A typical pulse train and its RF spectrum of the amplified 2.8-µm pulse are shown in Figs. 7(f) and 7(g), respectively. The amplified 2.8-µm pulse train naturally had the same repetition rate of 75.2 MHz with the seed oscillator. Meanwhile, the signal-to-noise ratio was ∼55 dB, which indicates some degradation compared with the seed oscillator. Adopting an all polarization maintaining structure is promising for increasing the signal-to-noise ratio of the amplified 2.8-µm pulse.

 

Fig. 7. (a)–(e) Autocorrelation traces (black) and their sech fitting curves (red) of the pulses at the output of the Er-doped ZBLAN fiber amplifier at various output powers. (f) Pulse train and (g) RF spectrum with a resolution bandwidth of 100 Hz for the amplified 2.8-µm pulse.

Download Full Size | PDF

In conclusion, we experimentally demonstrated a compact 3.42-W all-fiber femtosecond 2.8 µm MOPA system. A mode-locked thulium-doped fiber laser was used as the pulse seed for the laser system. A high efficiency Raman soliton redshift was realized in a TDFA. The Raman soliton was continuously redshifted to ∼2.8 µm in the PFF. A novel home-made EPSFFC was designed and fabricated to couple the 2.8-µm seed and 976-nm pump laser to the Er-doped ZBLAN fiber. Intense spectral broadening and soliton self-compression took place in the amplifier, and a 3.42-W 2.8-µm pulse with a width of ∼115 fs was achieved. This research paves the way for practical applications of the mid-infrared ultrafast fiber lasers. Higher power and larger pulse energies are expected by further optimizing the system.