Spectrum-agile hundred-watt-level high-power random fiber laser enabled by watt-level tunable optical filter

Random fiber lasers (RFLs), which operate via random distributed feedback (RDFB), have attracted increasing attention since their first demonstration in 2010.¹⁾ Unlike the traditional fiber laser with a well-defined resonator, the RFL is free of a precise cavity, making the system simpler in structure and more economical.^2,3) Moreover, RFLs based on Raman gain offer numerous advantages, e.g., they have low background spontaneous emission, have small quantum defects, and are free of the photodarkening effect.^4,5) In recent years, intensive investigations on RFLs concerning the power performance and spectral and temporal properties have been conducted, including power scaling,^6–9) linewidth narrowing,^10–12) wavelength tuning and expansion,^13–16) and pulsation operation.^17–19) In addition, RFLs have found a wide variety of applications in remote sensing and telecommunication,^20–22) for frequency conversion in second-harmonic generation,²³⁾ as the pump source for mid-infrared lasers and supercontinuum light sources,^5,24–26) and as a stable seed for power amplification.²⁷⁾

The spectral manipulation of fiber lasers is highly demanded for certain practical applications, such as pumping the optical parametric oscillator (OPO) and frequency doubling, where the optical efficiency and the output wavelength are closely associated with the linewidth and the central wavelength of the pump source, or acting as the seed of a high-power fiber master oscillator power amplifier, where the intensity of the amplified spontaneous emission is strongly related to the central wavelength, while the modal instability during high-power amplification has a close connection to the linewidth of the seed.²⁸⁾ In this case, a stable seed with wavelength and linewidth tunability is beneficial for precisely selecting the seed with the most suitable spectral properties for high-power amplification, as well as for paving the way for more practical applications.

Previously, we experimentally demonstrated the spectral manipulation of an RFL, achieving a wavelength tuning range of ∼20 nm, a linewidth tuning range up to 1.4 nm, and an output power of ∼23 W.²⁹⁾ However, the power level cannot satisfy the requirements for some practical applications, such as OPO, whose generation threshold is usually dozens of watts. In addition, there is no theoretical verification as to whether the spectral manipulation is possible in the case of a higher output power. Therefore, it makes sense to investigate the spectral manipulation property for a high-power RFL in theory and in experiment. In this study, we theoretically investigated the output-power evolution and power distribution of a forward-pumped RFL, proving the possibility of spectral manipulation in the high-power regime through a low-power handling device. Consequently, we developed a hundred-watt-level high-power RFL with both wavelength and linewidth tunability by utilizing a watt-level tunable optical filter (TOF). In contrast to typical situations, the term "high-power" in this manuscript means that the power ranges from tens of watts to >100 W, while the term "low-power" refers to "watt-level."

To determine the optimal parameters of the laser scheme prior to experiments, we used the classical steady-state light propagation equations to theoretically investigate the output-power evolution and the longitudinal power distribution of an RFL. The schematic setup for the numerical simulation is displayed in Fig. 1, comprising a classical half-opened cavity including a piece of passive fiber that produces RDFB, as well as a fiber mirror, which can be a fiber Bragg grating, a fiber loop mirror (FLM), or other fiber point reflectors providing point feedback. The pump light is injected through a wavelength-division multiplexer (WDM). Considering that a backward-pumped half-opened cavity may lead to a pulsed output owing to the gain modulation,¹⁹⁾ we applied the forward-pumped structure.

Zoom In Zoom Out Reset image size

The model for the numerical calculations is as follows:^2,6,30)

$$\begin{align} \pm\frac{dP_{0}^{\pm}}{dz} &= -\frac{\lambda_{1}}{\lambda_{0}}g_{\text{R1}}(P_{1}^{+} + P_{1}^{-} + 4h\nu_{1}\Delta\nu_{1}B_{1})P_{0}^{\pm} \\&\quad- \alpha_{0}P_{0}^{\pm} + \varepsilon_{0}P_{0}^{\mp} \end{align} \tag{ 1 }$$

$$\begin{align} \pm\frac{dP_{1}^{\pm}}{dz} &= -\frac{\lambda_{2}}{\lambda_{1}}g_{\text{R2}}(P_{2}^{+} + P_{2}^{-} + 4h\nu_{2}\Delta\nu_{2}B_{2})P_{1}^{\pm} - \alpha_{1}P_{1}^{\pm} \\ &\quad+ \varepsilon_{1}P_{1}^{\mp}+ g_{\text{R1}}(P_{0}^{+} + P_{0}^{-})(P_{1}^{\pm} + 2h\nu_{1}\Delta\nu_{1}B_{1}) \end{align} \tag{ 2 }$$

$$\begin{align} \pm\frac{dP_{2}^{\pm}}{dz} &= g_{\text{R2}}(P_{1}^{+} + P_{1}^{-})(P_{2}^{\pm} + 2h\nu_{2}\Delta\nu_{2}B_{2})\\ &\quad - \alpha_{2}P_{2}^{\pm} + \varepsilon_{2}P_{2}^{\mp} \end{align} \tag{ 3 }$$

$$\begin{align} B_{j} &= 1 + \cfrac{1}{\exp\biggl[\cfrac{h(\nu_{j-1} - \nu_{j})}{k_{\text{B}}T}\biggr]-1}\quad (j = 1,2), \end{align} \tag{ 4 }$$

where h is Planck's constant, k_B is the Boltzmann constant, and T is the fiber temperature. B denotes the population of the photons, which introduce noise from spontaneous Raman scattering. ν is the wave frequency, and Δν is the bandwidth of the Stokes wave. α, g_R, and ε denote the signal loss, Raman gain coefficient, and Rayleigh backscattering coefficient, respectively. λ₀, λ₁, and λ₂ represent the central wavelengths of the pump light, the first-order Stokes light, and the second-order Stokes light, whose powers at different positions are represented as P₀, P₁, and P₂, respectively. Assuming that the pump power is injected from the left side, the boundary conditions can be described as follows:

$$\begin{align} P_{0}(0) &= P_{\text{in}} \end{align} \tag{ 5 }$$

$$\begin{align} P_{1,2}^{+}(0) &= R_{\text{L1,2}}P_{1,2}^{-}(0) \end{align} \tag{ 6 }$$

$$\begin{align} P_{1,2}^{-}(L) &= R_{\text{R1,2}}P_{1,2}^{+}(L), \end{align} \tag{ 7 }$$

where P_in is the input pump power, and R_L1,2 and R_R1,2 denote the reflectivity of the left and the right end, respectively. Considering that the fiber mirror is a set of fiber devices with insertion losses, in the later experimental research, R_L1 and R_L2 are set as 0.62 and 0.36 respectively, according to an exploratory experiment. Even though the reflectivity of the fiber mirror is not high, the longitudinal power distribution of the lasing is mainly in the forward direction, and the power at the reflector end of the fiber is low.³¹⁾ The values of the other parameters used for the numerical calculation are listed in Table I.

We simulated the evolution of the output power and the power distribution along the passive fiber. As depicted in Fig. 2(a), the generation threshold of the random lasing is calculated to be 63 W, and the maximum output power reaches >117 with a pump power of 139 W, which confirms the feasibility of the hundred-watt-level high-power output with a 160-m-long short-cavity RFL. The theoretical longitudinal power distribution with a 140-W pump power is plotted in Fig. 2(b). The position that we are most concerned about is the highly reflective end (z = 0), where we intend to place the low-power handling TOF. Hence, for the convenience of analyzing the power performance near z = 0, the power distribution is shown within a range of 10 m. It can be seen that the powers of the forward and backward 1st-order Stokes light are ∼1.1 and ∼1.7 W, respectively, at z = 0. The inset of Fig. 2(b) presents the power distribution along the whole fiber. With the pump light propagating forward, the random lasing that comes from the spontaneous Raman noise can be amplified to become increasingly intensive. The power of the random lasing reaches the maximum value around 125 m, where the forward second-order Stokes light begins to be amplified. At the end of the passive fiber, random lasing of >114 W can be obtained. Therefore, from the viewpoint of numerical calculation, it is possible to achieve the spectral manipulation of a hundred-watt-level high-power RFL by inserting a watt-level TOF into the highly reflective end of the half-opened cavity.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The experimental setup is schematically shown in Fig. 3. The RFL has a classical half-opened cavity, including a piece of polarization-maintaining (PM) passive fiber and a circulator-based PM FLM at one end. The maximum insertion loss of the circulator from one port to another is 0.4 dB. A linearly polarized 1,055-nm ytterbium-doped fiber laser, which can export a maximum power of 133.4 W, acts as the pump source. The pump light is injected through the 1,050-nm port of a PM WDM and propagates along the following 160-m-long PM passive fiber with a core/inner cladding diameter of 10/125 µm. The 1,110-nm port of the PM WDM is spliced with a PM tapper at a coupling ratio of 0.1/99.9 and insertion loss of 0.22 dB. The PM tapper is utilized to monitor the input and output characteristics of the PM FLM. By inserting a bandwidth-adjustable TOF with power handling of 3 W and insertion loss of 0.8 dB into the circulator-based PM FLM, the central wavelength and the linewidth of the output spectrum can be tuned independently. Moreover, considering the insertion loss of the tapper, circulator, and TOF, the reflectivity of the 1st-order Stokes R_L1 is estimated as 0.625, which is slightly higher than the value used in the simulation. For convenience, we introduce ports 1, 2, and 3 to represent the output port of the RFL and the backward and forward 0.1% ports of the PM tapper, respectively. In addition, all the free end facets are cleaved with an angle of 8° to suppress the undesired backward reflection.

Zoom In Zoom Out Reset image size

First, we investigated the spectral and power performance of the RFL without employing the TOF. The output spectrum from port 1 with a 133.4-W pump power is displayed in Fig. 4(a), wherein three spectral components are observed, corresponding to the 1,055-nm residual pump light, the 1,112-nm first-order Stokes light, and the 1,170-nm second-order Stokes light. Their proportions are calculated to be 0.7, 99.1, and 0.2%, respectively, through numerical integration based on the spectral data. When we concern the dependence of the evolution of the output spectrum from port 1 on the pump power the conversion from single-peak spectrum to double-peak spectrum has been observed, which has been studied and well explained in the previous research.³²⁾

Zoom In Zoom Out Reset image size

The spectral broadening attributing to the nonlinear effects such as cross-phase modulation (XPM) and self-phase modulation (SPM) has also been observed.³³⁾ Furthermore, by measuring the total output power and performing numerical integration based on the spectral data at different pump powers, we can obtain the powers of the residual pump light and the 1st-order Stokes light at different pump powers, as illustrated in Fig. 4(b). The experimental lasing threshold is measured to be ∼62 W, and the maximum output power of the first-order Stokes light is approximately 112.5 W, corresponding to a high optical efficiency of up to ∼84.2%. Moreover, the results of the numerical simulation, which are represented by the solid lines, fit well with the experimental data, indicating that the theoretical model and the parameters that we employed are feasible and effective.

By fixing the passband of the TOF at the minimum value of 0.7 nm and adjusting the transmission wavelength, the central wavelength of the RFL can be tuned continuously from 1,095 to 1,115 nm, i.e., with a tuning range of ∼20 nm. The normalized spectra from port 1 are shown in Fig. 5(a), and Fig. 5(b) shows the spectra from port 3 in log coordinates, demonstrating how the TOF influences the spectra after the FLM, as well as the final output of port 1. The tuning range of the central wavelength is limited by the unevenness of the Raman gain profile. When the operating wavelength is tuned below 1,095 nm or above 1,115 nm, the spontaneous Raman noise centered at ∼1,106 nm increases rapidly, resulting in the difficulty of the signal lasing. Therefore, the powers of the 1st-order Stokes light from port 1 vary drastically when the central wavelength is tuned with a full pump power. As shown in Fig. 5(c), the maximum output power of ∼109 W can be achieved at 1,110 nm, and an output power exceeding 100 W can be achieved from ∼1,102 to ∼1,113 nm. In addition, the wavelength whose output power decreases by 1 dB from the maximum value covers approximately 16.5 nm (from ∼1,097.5 to ∼1,114 nm); thus, the corresponding available Raman gain with a relatively high efficiency is calculated to be ∼3.8 THz (from ∼11 to 14.8 THz), which is comparable to that of the reported tunable RFLs with only Raman gain.^13,34,35) Figure 5(d) shows the variation of the output powers from port 3 with respect to the central wavelength. Port 3 is the 0.1% sample of the light transmitted through the circulator and the TOF; the maximum power is measured to be ∼0.12 mW at 1,107.5 nm and the average power at different central wavelengths is calculated to be ∼0.105 mW. Taking account into the losses induced by the tapper, the circulator, and the TOF, the power inside the FLM should be hundred-milliwatt-level, which is far smaller than the power-handling limit of the TOF. Thus, we experimentally showed that it is feasible to achieve the hundred-watt-level high-power output and spectral tuning of an RFL by employing a watt-level power-handling TOF. Increasing the output power of the spectrum-agile RFL may give rise to obstacles such as high-order Stokes waves and the limited power handling of the tunable filter.

Zoom In Zoom Out Reset image size

Furthermore, we investigated the linewidth tunability of the hundred-watt high-power RFL. The central wavelength is fixed at 1,110 nm, which yields the highest output power. By widening the passband of the TOF, the output spectrum from port 1 can be gradually broadened, as shown in Fig. 6(a). The minimal FWHM linewidth that we can achieve is ∼2.1 nm, by adjusting the passband of the TOF to the minimum value of 0.7 nm, while the maximum FWHM linewidth is ∼5.5 nm with a passband of >14 nm. The inset image in Fig. 6(a) shows the FWHM linewidth as a function of the passband of the TOF. The FWHM linewidth exhibits nearly linear growth with a passband less than 14 nm and becomes saturated at ∼5.5 nm when the passband of the TOF is increased. The minimal FWHM linewidth is limited by the spectral broadening owing to the nonlinear effects such as SPM and XPM. The spectral evolution with the pump power when the passband is fixed at 0.7 nm is shown in Fig. 6(b), the inset of which shows the dependence of the FWHM linewidth on the pump power. The output spectrum with a pump power of ∼70.6 W exhibits an FWHM linewidth of ∼0.29 nm, which is significantly smaller than the passband of the TOF, whereas with the increase of the pump power, the FWHM linewidth increases rapidly to >2 nm owing to the nonlinear spectral broadening. In addition, the gain competition due to the unevenness of the Raman gain profile is responsible for the saturation of the maximum FWHM linewidth. Figure 6(c) shows the linewidth-tunable spectra from port 3 when the passband of the TOF varies from 0.7 to 10 nm, which indicates how the bandwidth-adjustable TOF filters the spectrum. As a result, the spectrum after the TOF exhibits a relatively flat top with sharp edges, and with the passband of the TOF widening, the central wavelength of the spectrum remains nearly unchanged, whereas the intensity gradually decreases because of the decrease of the spectral power density. The output powers of the first-order Stokes light from port 1 and the powers from port 3 are shown in Fig. 6(d) and the inset, respectively. It follows that tuning the linewidth has little effect on the output power of the first-order Stokes light, and the minimum and maximum output power are ∼105.8 and ∼111.3 W, respectively, showing a fluctuation of <5%. On the other hand, the power from port 3 increases regularly with the passband of the TOF, which is understandable because a wider passband allows more power to be transmitted. The maximum power from port 3 is ∼0.19 mW, indicating that the power inside the FLM is hundred-milliwatt-level, which is far smaller than the power-handling level of the TOF.

Zoom In Zoom Out Reset image size

We also analyzed the linewidth tunability of the high-power RFL at different central wavelengths. The colored area of Fig. 7 shows the spectral tuning range of the FWHM linewidth, as well as the central wavelength. As mentioned previously, the minimum value of the FWHM linewidth is limited by the nonlinear spectral broadening, while the maximum FWHM linewidth that we can achieve is confined by the gain competition. Moreover, the tuning range of the FWHM linewidth is closely related to the central wavelength. Near the peak region of the Raman gain profile (approximately 13.2 to 14.6 THz), the FWHM linewidth can cover a relatively wide range, especially at 1,110 nm, which is located between the two peaks of the Raman gain profile. When the passband of the TOF is widened and the pump power is increased, the peaks near 1,106 and 1,112 nm oscillate simultaneously owing to the nearly equal gain and then couple together accompanying nonlinear spectral broadening, finally resulting in a relatively large FWHM linewidth up to ∼5.5 nm. However, with the central wavelength deviating from the peak region of the Raman gain profile, the tuning range of the FWHM linewidth becomes significantly narrower. For example, the linewidth at 1,115 nm is difficult to tune because of the weak Raman gain. Thus far, we have experimentally realized the spectral manipulation of a hundred-watt-level high-power RFL through a watt-level TOF.

Zoom In Zoom Out Reset image size

We have reason to think that for achieving the spectral manipulation of a high-power fiber laser, it is difficult and unnecessary to directly filter the output spectrum. The spectral manipulation can also be achieved by utilizing a low-power device, as long as the longitudinal power distribution meets the requirement. In this sense, adjusting the filter is equivalent to changing the boundary conditions of the laser cavity, through filtering and amplifying the weak signal, and finally the high-power output can be manipulated.

In conclusion, we theoretically and experimentally demonstrated the spectral manipulation of a hundred-watt-level high-power RFL by employing a watt-level TOF based on high-fidelity theoretical verification and careful system-parameter design. As a result, the maximum output power reached approximately 112.5 W with an optical efficiency up to ∼84.2%. The central wavelength continuously covers the range of 1,095–1,115 nm, while the tuning range of the FWHM linewidth, which is closely related to the central wavelength, can be tuned within a range of ∼1.1 to ∼2.7 times the minimum linewidth. At 1,110 nm, where the highest output power is obtained, the tuning range of the FWHM linewidth is 2.1–5.5 nm. These results are beneficial for the investigation of the inherent characteristics and applications of the RFL and introduce new possibilities for the spectral manipulation of high-power fiber lasers.

Acknowledgments