1. Introduction

Since random fiber laser (RFL) proposed in 2010 [1], it has led to the quick development of this novel laser source. The laser generation mechanism of RFL utilizes random distributed Rayleigh scattering feedback from optical fibers instead of requiring a well-defined resonant cavity. The spectral properties of the radiation of RFL can be managed by using different gain mechanisms. Raman-gain-based RFLs, whose generation spectrum defined by the Raman gain profile of the fiber used, can obtain many high-order laser outputs [2–4]. Active-gain-based RFLs, whose gain provided by active fiber with different dopants (such as Er, Yb), have varied emission spectrum range for different dopants gain [5,6]. By designing the system structure, the RFLs also excels in wavelength tuning, multi-wavelength output and supercontinuous light generation [7–9]. Therefore, wavelength manipulation is an important research aspect of RFLs. Simultaneous dual-wavelength emission from a RFL is of important scientific interest and has potential for many interesting applications, such as laser communication, laser spectroscopy, medical instrumentation, and holographic interferometry. RFLs have been proven to be a good choice for dual-wavelength laser generation owing to their high stability [10]. In 2016, Aporta et al. presented a tunable dual-wavelength RFL with variable spacing. The tunability of the dual-wavelength lasing is achieved owing to the wavelength selection variability of the point reflectors through strain [11]. In 2017, Xu et al. demonstrated a narrow-linewidth dual-wavelength random fiber laser operating at 1541and 1544 nm based on the semiconductor optical amplifier gain [12]. In 2019, Song et al. reported a dual-wavelength RFL with over 100 nm wavelength interval based on Raman gain which provided by phosphosilicate fiber [13]. In 2020, Zhang et al. reported an all-fiberized linearly polarized dual-wavelength random distributed feedback Raman laser with wavelength, linewidth, and power ratio tunability [14]. By adopting two watt-level bandwidth adjustable optical filters, a spectrum-manipulable dual-wavelength output with nearly a 10 W output power is achieved. The wavelength separation can be tuned from 2.5 to 13 nm. In 2022, Lv et al. demonstrated a switchable dual-wavelength RFL based on random Bragg grating array. The single- and dual-wavelength switching can be realized by adjusting the pump power of 980 nm pump source [15].

Despite the dual-wavelength operation realized using random lasing, designing a stable dual-wavelength all-fiber laser with wide wavelength spacing and high output power remains a challenge. To obtain simultaneous random lasing at two widely separated wavelengths and increase compactness and simplicity of the lasers, the method using Er/Yb co-doped fibers (EYDF) to provide Yb- and Er-ion gain for RFLs is a pretty good choice. Commercial EYDF is generally designed to maximize the energy transfer from the pump absorbing Yb ions to the Er ions to create gain at 1.5 µm. To achieve Yb co-lasing around 1 µm in an EYDF, the fiber length of EYDF must be short enough to allow the Yb optical pumping rate exceeding the inter-ion energy transfer rate between Yb and Er ions, but at the expense of the performance at 1.5 µm. However, short fiber length resulted in an insufficient 1550 nm gain and a considerably unstable 1535 nm laser emission that limited the performance of the dual-wavelength laser. Therefore, to realize dual-wavelength RFL operating at 1064 and 1550 nm, the EYDF needs to be specially designed and fabricated.

In this paper, we propose and experimentally explore an all-fiberized dual-wavelength RFL with bi-directional laser output operating at 1064 and 1550 nm, respectively. The RFL consists of a dual-wavelength RFL seed, a 1 µm power amplifier (PA), and a 1.5 µm PA. In the seed, a 5 m specifically designed EYDF is used to provide Yb- and Er-ion gain for 1064 and 1550 nm wavelengths. The home-made EYDF can achieve simultaneous population inversion of both Er³⁺ and Yb³⁺. Two pieces of passive fibers with 1 and 2 km lengths are adopted to provide random distributed feedback for both wavelengths. Random laser seed with 5.35 W at 1064 nm and 6.61 W at 1550 nm can be obtained, and output power can be scaled to 50 W at 1064 nm and 20 W at 1550 nm through these two PAs. The slope efficiencies of the 1 and 1.5 µm PAs are 70.8% and 28.9%, respectively. The dual-wavelength RFL shows high performances, such as high stability in the time domain and high beam quality with a single transverse mode output. The proposed fiber laser is the first demonstration of a dual-wavelength RFL, with a spectral separation as far as about 500 nm in an all-fiber configuration.

2. Experimental setup

The schematic diagram of the 1064/1550 nm dual-wavelength RFL is shown in Fig. 1. The system consists of a dual-wavelength RFL seed, a 1064 nm PA and a 1550 nm PA. The dual-wavelength RFL seed consists of a 5 m-long homemade EYDF, two high reflectivity fiber Bragg gratings (HR-FBG) centered at 1064 and 1550 nm, a (2 + 1) × 1 signal/pump combiner, two 976 nm LDs, and two rolls of single mode fiber (SMF, Corning, SMF-28E+) with lengths of 1 km (SMF-1) and 2 km (SMF-2). By sharing the same piece of homemade EYDF, the dual-wavelength RFL seed forms two half-opened random laser cavities operating at 1064 and 1550 nm, respectively. The homemade EYDF is cladding pumped by two LDs (Oclaro, 25 W, 976 nm) through a (2 + 1) × 1 signal/pump combiner. Two pieces of SMFs with 1 and 2 km lengths are adopted to provide random distributed feedback for both wavelengths, respectively. These two rolls of SMFs have a core diameter of 8.2 µm with a 0.14 NA. The high attenuation of SMF-28E + at 1064 nm (∼1.5 dB/km) will reduce the efficiency, so we set the fiber length of SMF-1 to 1 km. The 1064 nm HR-FBG and 1550 nm HR-FBG are written in SMFs with full width at half maximum (FWHM) bandwidths of 0.4 and 0.98 nm, respectively. In this experiment, there are no special requirements for the bandwidths of the FBGs. All free fiber ends are angle cleaved around 8° to suppress any parasitic feedback and to ensure that optical feedback was exclusively from Rayleigh scattering.

 

Fig. 1. Schematic of the 1.0/1.5 µm dual-band RFL. EYDF, Er/Yb co-doped fiber; LD, laser diode; ISO, isolator; SMF, single mode fiber; HR-FBG, high reflectivity fiber Bragg grating; YDF, Yb-doped fiber.

Download Full Size | PDF

For further power scaling, the RFL seed light is then injected into the 1064 and 1550 nm PA, respectively. Two 1064 nm isolators (ISOs) with 25 dB isolation are inserted between the 1064 nm seed and 1064 nm PA to enhance the ability to prevent the possible backward signal and protect the components of the RFL seed and make sure all the backward feedbacks in the seed light part come from the Rayleigh scattering. The 1064 nm PA contains a (2 + 1) × 1 signal/pump combiner, an 85 W LD operating at 976 nm, and a piece of single mode Yb-doped fiber (5 m, Nufern, LMA-YDF-10/130-M). A cladding power stripper (CPS) and an angle-cleaved output port are adopted after YDF to strip the cladding light and suppress the unwanted backward reflection. The 1550 nm PA is pumped by an 85 W LD operating at 976 nm through a (2 + 1) × 1 signal/pump combiner. The pump light is injected into a piece of 3.5 m-long double cladding EYDF exhibiting 12/130 µm core/inner clad diameter with 0.20/0.46 NA. The cladding absorption coefficient at 976 nm is about 9 dB/m. A CPS is spliced after the EYDF-1 to strip out the residual pump and cladding light. The output port is angle-cleaved around 8° to avoid the unwanted backward reflection. All the components in the experiment, including EYDFs, YDF, LDs, combiners and FBGs, are placed on a water-cooled heat sink to carry away the heat accumulation. The output power, spectrum, time domain characteristics, beam quality and beam profile are recorded at the output ports with a power meter (Ophir, 150A-BB-26), an optical spectrum analyzer (Yokogawa, AQ6370D), an oscilloscope (Tektronix, TDS 3054C), and a scanning-slit beam profiler (NanoScan, 2s Ge/9/5), respectively.

3. Experimental results and discussion

3.1 Fiber fabrication and fiber characterization

In the EYDF, energy is transferred from the ²F_5/2 state, the excited state of Yb³⁺, to the ⁴I_11/2 level of Er³⁺. Then, energy transfer to ⁴I_13/2 through a quick non-radiative transition forms a population inversion between ⁴I_13/2 and ⁴I_15/2 levels of Er³⁺, and the signal is amplified through stimulated emission. To realize a balanced simulated emission from both rare earth (RE) ions, the energy transfer between Yb and Er ions need to be optimized by controlling the concentrations of the co-doped Al and P. A proper concentration of the co-doped Al and P can also improve the solubility of the glass matrix for Er³⁺ and Yb³⁺ and avoid the clustering effect of RE ions. The Er/Yb fiber preform was fabricated using the chelate gas phase deposition technique in conjunction with the modified chemical vapor deposition system. An all-gas phase doping method was adopted by the system, which has the ability to precisely regulate the gas flow rate and temperature, thus simplifying the deposition process of fiber preforms and reducing the background loss. The Al₂O₃-P₂O₅-SiO₂ ternary system was selected as the matrix material for the preparation of the optical fiber preforms. The doping content of Al and P is 3.8% and 1.8%, respectively. The formation of Al-O-P bonds reduces the phonon energy of the glass matrix to 1135 cm⁻¹ when the doping content Al > P, which is lower than the maximum phonon energy of silica at 1200 cm⁻¹. The lower phonon energy inhibits the non-radiative transition process between the ⁴I_11/2 and ⁴I_13/2 energy levels of the Er³⁺ ions, reducing the rate of energy transfer from the Yb³⁺ ions to the Er³⁺ ions. At the same time, a higher Yb/Er ratio allows Yb³⁺ ions to be evenly and sufficiently surrounded by Er³⁺ ions, forming an effective Er³⁺/ Yb³⁺ ion pair. This helps to achieve better light amplification performance.

Figure 2(a) shows the refractive index profile (RIP) of the homemade EYDF with a core diameter of 10.1 µm and cladding diameter (flat to flat) of 125.2 µm. The core-cladding refractive index difference Δn of the EYDF is 6.62 × 10⁻³ with a corresponding NA of 0.14. The cross-section picture of the EYDF is shown in the inset of Fig. 2(a). It exhibits a symmetrical octagonal cladding to prevent cladding light circulation and thus improve the pump absorption efficiency. To further investigate the homogeneity of the fiber core, the elemental distribution and molar ratio were characterized by EPMA, as shown in Fig. 2(b). The dopant concentrations of Er₂O₃, Yb₂O₃, Al₂O₃, and P₂O₅ were measured to be 0.025, 0.18, 3.80, and 1.80 mol%, respectively. Composition distributions of the homemade EYDF in the core region are presented in Fig. 3. The inner circular area and outer ring in the mapping image of P ions are core region and transition layer, respectively. Al, P, and Yb ions in the central region of the homemade EYDF show lower dopant concentration compared with the outer core region, corresponding to the central dopant concentration “dip” in the fiber core shown in Fig. 2(b). Er ions in the fiber core exhibit more homogenous distribution and no apparent central “dip” can be found. The main reason is that compared to other elements, the doping concentration of Er ions is very low, only 0.025mol%. Under these conditions, the concentration change of Er ions in the fiber core may not be detected.

 

Fig. 2. (a) RIP and cross-section picture of the homemade EYDF; (b) Radial dopant concentration of the homemade EYDF.

Download Full Size | PDF

 

Fig. 3. Images of elemental distribution in fiber core region.

Download Full Size | PDF

To evaluate the absorption coefficient of the homemade EYDF, the cladding absorption spectrum range from 800 nm to 1100 nm was measured using the cut-back method, as shown in Fig. 4(a). Due to the low doping concentration and low absorption cross section of Er³⁺ ions at 976 nm, the absorption peak intensity of Er³⁺ ions in the 9XX nm band are very weak and is completely covered by the absorption peak of Yb³⁺ ions. The absorption coefficient of the homemade EYDF is to be 1.22 dB/m@915 nm and 3.60 dB/m@976 nm, respectively. Figure 4(b) shows the photoluminescence (PL) spectra of the EYDFs co-doped with and without Al₂O₃ excited by a 980 nm laser diode. The PL bands at 1 and 1.5 µm are due to the ²F_5/2→²F_7/2 of Yb³⁺ and ⁴I_13/2→⁴I_15/2 of Er³⁺ ions, respectively. Co-doping with Al₂O₃ will increase the PL intensity of Yb³⁺ ions. As expected, the gain on Yb³⁺ ions band is increased, but the gain on Er³⁺ ions band still predominates, ensuring an adequate 1.5 µm gain with a comparable 1-µm gain using a longer fiber. The co-doping of Al₂O₃ also increases the Stark splitting value, resulting in a significant broadening of the 1.5 µm PL band. Due to the distortion of the Er³⁺ ions ligand field, the energy level band gap (ΔE) between the ⁴I1_3/2 excited state and the ⁴I_15/2 ground state increases and the PL peak appears blue-shifted [16].

 

Fig. 4. (a) Cladding absorption spectrum of the homemade EYDF; (b) PL spectra of the EYDFs co-doped with and without Al₂O₃.

Download Full Size | PDF

3.2 Dual-wavelength RFL performance results and discussion

First, the characteristics of the RFL seed are thoroughly investigated. Figure 5 shows the output power versus the launched pump power for the 1064 and 1550 nm RFL seed. The dual-wavelength RFL seed exhibits two distinct thresholds at both wavelengths (1 W at 1550 nm, 3 W at 1064 nm) depending on the cavity losses and on the gain at 1064 and 1550 nm, including energy transfer between doping ions. Under a maximum pump power of ∼30 W, 5.35 W at 1064 nm and 6.61 W at 1550 nm are measured at the laser output. The random lasing powers at 1064 and 1550 nm do not have linear growth curves, which is owing to the gain competition between these two lasing wavelengths in the power amplification process. But the total power of both wavelengths is linearly increasing with a maximum power of 11.96 W and a slope efficiency of 39.8%.

 

Fig. 5. Output powers versus the pump power for the 1064 and 1550 nm random lasing.

Download Full Size | PDF

The spectral evolution of the 1064 and 1550 nm RFL seeds with the increase of laser power are shown in Figs. 6(a) and (c), respectively. The spectral profiles of the two wavelengths stay smooth and stable when the pump power is above the lasing threshold. These two laser wavelengths exhibit good spectral purity and do not show the unstable 1535 nm laser emission. In our previous experiments using commercial EYDF, short fiber length resulted in an insufficient 1550 nm gain and a considerably unstable 1535 nm laser emission that limited the performance of the dual-wavelength laser, which is shown in Fig. 7. By further increasing the pump power, we can see the continuous broadening of the spectra. To give a complete comparison of the spectral bandwidth evolution versus the pump power, the 3 dB bandwidths of the 1064 and 1550 nm RFL seeds are calculated and given in Figs. 6(b) and (d). For 1550 nm random lasing, the 3 dB bandwidth broadens from 0.72 to 1.41 nm. For 1064 nm random lasing, the 3 dB bandwidth broadens from 0.08 nm to 0.67 nm. This shows obvious spectral broadening as a result of the complicated interaction of a nonlinear Kerr effect [17].

 

Fig. 6. (a) Spectral evolution of the 1550 nm random lasing; (b) 3 dB bandwidth evolution of the 1550 nm seed; (c) Spectral evolution of the 1064 nm random lasing; (d) 3 dB bandwidth evolution of the 1064 nm seed.

Download Full Size | PDF

 

Fig. 7. Unstable 1535 nm laser emission appeared in our previous experiments using commercial EYDF.

Download Full Size | PDF

To further power scaling of 1064 and 1550 nm random lasing, the two seeds are injected into two corresponding PAs, respectively. To ensure a relatively narrow 3 dB bandwidth, we set the pump power of the RFL seed to 16 W. At this time, the 3 dB bandwidth of 1064 and 1550 nm RFL seeds are 0.31 and 1.15 nm, and the output powers of the RFL seeds at 1064 and 1550 nm are 1.72 and 4.39 W, respectively. For 1550 nm PA, the output power versus the pump power is depicted in Fig. 8(a). Given that the EYDF has a high absorption coefficient for 1550 nm light, about 50% of the 1550 nm seed light is absorbed by the EYDF when there is no pump light injection, so the output power from the 1550 nm PA is only 2.4 W. For a given 62.1 W pump light, the 1550 nm PA obtains an ultimate output power of 20.67 W, corresponding to a slope efficiency of 28.9%. The output power of the 1550 nm random laser increases almost linearly with increasing the pump power. The spectral evolution with the increase of laser power is shown in Fig. 8(b). The evolution of the spectral profile shows no obvious spectral broadening. The spectrum at maximum output power ranging from 1000 nm to 1600 nm, as depicted in the insertion graph of Fig. 8(b), can indicate that none of the parasitic lasing at 1535 nm and ∼1 µm are observed during the power amplification process.

 

Fig. 8. (a) Output power versus the launched pump power for 1550 nm random laser; (b) Spectral characteristics of the 1550 nm PA. (c) Output power versus the launched pump power for 1064 nm random laser; (d) Spectral characteristics of the 1064 nm PA.

Download Full Size | PDF

The 1064 nm RFL seed is injected into 1064 nm PA after passing through two 1064 nm ISOs. The use of 1064 nm ISOs leads to an increase in insertion loss. Therefore, only about 1.0 W of 1064 nm seed light is injected into the 1064 nm PA. Figure 8(c) depicts the output power and slope efficiency of the 1064 nm PA versus the pump power. The output power increases almost linearly with the slope efficiency of 70.8%, and the maximal output power under pump power of 70.9 W reaches 50.4 W. The spectral evolution with the increase of laser power is shown in Fig. 8(d). By further increasing the laser power, no spectral broadening is observed.

To give a complete comparison of the spectral bandwidth evolution versus the pump power, the 3 dB bandwidths of the two wavelengths are depicted in Fig. 9. The 3 dB bandwidth at 1550 nm varies between 1.13 and 1.18 nm, whereas the FWHM at 1064 nm varies between 0.29 nm to 0.31 nm. There is no broadening of the 3 dB bandwidths at maximal output power, which is different from the 3 dB bandwidths evolution of RFL seeds.

 

Fig. 9. FWHM variations of 1550 nm PA (a) and 1064 nm PA (b) versus the pump power.

Download Full Size | PDF

The temporal domain traces of 1064 and 1550 nm PAs are measured in the millisecond scale at maximal output powers, as shown in Fig. 10(a). The temporal domain traces, which show typical CW operation of 1064 and 1550 nm PAs, show no significant amplitude fluctuation. To characterize the temporal dynamics quantitatively, the std/mean value defined by the value of standard deviation divided by the mean is applied here. The std/mean values of 1064 and 1550 nm PAs are 0.0558 and 0.0455, respectively. Furthermore, to demonstrate the stability of the whole laser configuration, the long timescale output power is analyzed within a time window of 10 min. As shown in Fig. 10(b), the output powers of 1550 and 1064 nm PAs are stabilized at 20 and 50 W, with extremely small std/mean values of 0.00129 and 0.00265, indicating the excellent stable operation capability of the dual-wavelength RFL in the long timescale. The short- and long-term time domain stabilities are crucial for practical applications.

 

Fig. 10. (a) Temporal dynamics at the maximum output powers in the millisecond scale; (b) Output powers fluctuations in the long timescale (10 min).

Download Full Size | PDF

A beam quality of M²= 1.15 at 50 W output power and 1064 nm wavelength is measured by a Primes laser quality monitor, as shown in Fig. 11(a). Given that the maximum operating wavelength of the Primes Laser Quality Monitor is 1100 nm, the beam quality of the 1550 nm laser cannot be measured. Then, the far field beam profile of the 1550 nm PA with output power of 20 W is measured by a scanning-slit beam profiler (NanoScan, 2s Ge/9/5), as shown in Fig. 11(b). Although the beam profiler can be used to test beam diameters at different positions to obtain the M² value, the lack of a wide range precision displacement device limits our ability to do so. The beam shape indicates that the output laser of 1550 nm PA is a single transverse mode operation with a nearly Gaussian profile.

 

Fig. 11. (a) Beam quality of 1064 nm PA at output power of 50 W; (b) Beam profile of 1550 nm PA at output power of 20 W.

Download Full Size | PDF

4. Conclusion

We have proposed and experimentally demonstrated a 1064/1550 nm dual-wavelength RFL based on EYDF for the first time. We have designed and manufactured a homemade EYDF to solve the problem of insufficient Yb ions gain by optimizing the concentrations of the co-doped Al and P to avoid the total energy transfer from the Yb ions to Er ions. 5.35 W at 1064 nm and 6.61 W at 1550 nm random lasers are obtained. By further optimizing the fiber design and controlling the cavity losses at 1064 and 1550 nm, the random lasing of the two wavelengths will increase linearly. As for 1064 and 1550 nm PAs, the maximum output powers of 50 and 20 W can be achieved with slope efficiencies of 70.8% and 28.9%, respectively. The 3 dB bandwidths of the amplified 1064 and 1550 nm random lasers are 0.31 and 1.18 nm, showing that no broadening of the 3 dB bandwidths occurs during the power scaling. The dual-wavelength RFL showed high performances, such as high stability in the time domain and high beam quality with a single transverse mode output. The results of this experimental study indicate that dual-wavelength RFL operating at 1064 and 1550 nm can be used as excellent light sources for a variety of applications.