1 Introduction

Fiber lasers emitting multiple wavelengths are of significant interest recently due to their potential applications in optical communications, instrument testing, sensing system and signal processing. With its high and broadband gain in the communication window, erbium-doped fiber (EDF) is commonly used as gain media in generating the wavelengths simultaneously. However, the EDF-based lasers are not stable at room temperature because of gain-mode competition in the EDF. Many techniques have been proposed to alleviate the competition such as cooling EDF in liquid nitrogen [1], exploiting nonlinear polarization rotation [2], utilizing polarization hole burning of laser cavity [3], employing hybrid gain medium of Brillouin and EDF [4] and inducing four-wave mixing (FWM) effect [513]. Of these techniques, the use of FWM is superior in terms of design’s simplicity and flexibility. Deposited graphene [5, 6], bismuth-based EDF [7] and highly nonlinear fiber (HNLF) [813] are the medium that are commonly used for the induction of FWM. Deposited graphene and bismuth-based erbium-doped fiber laser (EDFL) are more compact in size, but they require special fabrication and deposition process which could make the design more complex. Stable multiwavelength EDFL based on FWM in HNLF was firstly reported by Liu et al. [8]. They successfully generated dual-wavelength EDFL with a highly nonlinear photonics crystal fiber as a stabilizer. Their design, however, lacks multiwavelength functionality in which the lasing wavelength and its spacing are fixed. Han et al. [9] then proposed spacing-tunable multiwavelength EDFL, but the control of wavelength spacing is discrete and not continuous. In their subsequent attempt, a multiwavelength EDFL with continuous wavelength spacing was demonstrated [10]. By modifying the chirp ratio of a fiber Bragg grating using a special mechanical apparatus, continuous tunability of 0.033 nm/mm was achieved. The design comes with drawbacks though. While the wavelength spacing tunability is continuous, it is limited from 0.32 to 0.81 nm and requires a special apparatus of two translation stages and a sawtooth wheel for operation which makes the design more complex. Moreover, the design is not switchable from multiwavelength to single wavelength. It is therefore desirable if we can have an EDFL scheme which has wide, continuous and simple-to-operate tunability and also switchable between single and multiwavelength.

In this paper, on the basis of FWM effect in HNLF, we propose and demonstrate a ring EDFL that has not only continuous but also wide tunability, obtained with a relatively simple method. The complex operation and limitation of tunability are rectified by the use of tunable bandpass filter (TBF) in selecting the oscillating wavelength. Besides wide, continuous and mechanical-free tunability, the proposed scheme is also switchable between single and dual-wavelength. The switching mechanism is done by adjusting reflected power coming from a dual-selective element. In this scheme, the single-wavelength EDFL can be continuously tuned from 1,542.07 to 1,564.85 nm and continuous wavelength spacing from 0.52 to 22.78 nm can be achieved for the dual-wavelength EDFL.

2 Self-stability of FWM on dual-wavelength EDFL

When an optical field is applied to an optical fiber, it can create fiber nonlinearities as a result of nonlinear response of bound electrons in the material. FWM is one of the nonlinearities in the optical fiber. In FWM, a new wave which is called idler in the literature emerges when waves mix in the fiber. The efficiency of generating the new wave, however, depends on the phase matching of the waves which is very much reliant on the dispersion of the fiber. The output power of the idler at frequency \(\omega_{ij\,k} = \omega_{i} + \omega_{j} - \omega_{k}\) can be written as follows [14]:

$$P_{{\omega_{\,ij\,k} }} \left( L \right) = \frac{{1,024\pi^{6} }}{{n^{4} \lambda^{2} c^{2} }}\left( {D_{X} } \right)^{2} \frac{{P_{i} \left( 0 \right)P_{j} \left( 0 \right)P_{k} \left( 0 \right)}}{{A_{\text{eff}}^{2} }} \cdot {\text{e}}^{ - \alpha L} \cdot \frac{{\left( {1 - {\text{e}}^{ - \alpha L} } \right)^{2} }}{{\alpha^{2} }}\eta$$
(1)

where P i (0), P j (0), P k (0) are the input optical powers of the waves in which i, j and k denote three different waves that beat each other, D X is the degeneracy factor, n is the refractive index of fiber, c is the light velocity in vacuum, α stands for fiber losses, A eff is the effective area of fiber, L is the fiber length, and η is the FWM efficiency which can be written as follows:

$$\eta = \frac{{\alpha^{2} }}{{\alpha^{2} + \left( {\Updelta \beta } \right)^{2} }}\left[ {1 + \frac{{4\exp \left( { - \alpha L} \right)\sin^{2} \left( {\Updelta \beta L/2} \right)}}{{\left( {1 - \exp \left( { - \alpha L} \right)} \right)^{2} }}} \right]$$
(2)

with

$$\Updelta \beta = \beta \left( {\omega_{i} } \right) + \beta \left( {\omega_{j} } \right) - \beta \left( {\omega_{k} } \right) - \beta \left( {\omega_{{i{\kern 1pt} j{\kern 1pt} k}} } \right)$$
(3)

where β is the propagation constant.

In EDFLs, FWM in the optical fiber is exploited in stabilizing the output power of the lasers. The stability mechanism is illustrated in Fig. 1 where the lengths of vertical arrows represent the power of the spectrum. Figure 1a shows the effect of gain-mode competition on the scheme without the balance of FWM self-stability process. As the scheme does not have a stabilizer, the two lasers are always in a fierce competition for erbium ions in the EDF. Consequently, the stronger laser conquers the erbium ions, and this causes the suppression of other laser and subsequently leads to the loss of the channel. On the contrary, with the presence of a FWM stabilizer in the scheme, the dual-wavelength EDFL is stable and no input laser is lost in this case as shown in Fig. 1b. The EDFL stability actually comes from the balancing act between gain-mode competition and multiple FWM processes in the scheme [11]. The gain-mode competition increases the power discrepancy between the two lasers but the magnitude is then reduced by the multiple FWM processes. Let denotes that the power of lasers at ω 1 and ω 2 as P 1 and P 2, respectively, and the power of generated signal and idler at ω 3 and ω 4 as P 3 and P 4, respectively. From Fig. 1b and Eqs. (1)–(3), the aforementioned powers can be written as follows: \(P_{{{\kern 1pt} 1}} = P_{{\omega_{\,1} }} + \Updelta P_{1}\) where \(\Updelta P_{1} = P_{{\omega_{\,212} }}\) is the power variation at ω 1, \(P_{{{\kern 1pt} 2}} = P_{{\omega_{\,2} }} + \Updelta P_{2}\) where \(\Updelta P_{2} = P_{{\omega_{\,121} }}\) is the power variation at ω 2, \(P_{{{\kern 1pt} 3}} = P_{{\omega_{\,112} }}\) and \(P_{{{\kern 1pt} 4}} = P_{{\omega_{\,221} }}\). There are three cases that we can consider for the FWM self-stability mechanism; (1) If P 1 > P 2, then Δ(P 1) < 0 and Δ(P 2) > 0, so \(P_{\,1} \to P_{\,2}\) (the laser at ω 1 transfers its power to that at ω 2), (2) If P 2 > P 1, then Δ(P 2) < 0 and Δ(P 1) > 0, so \(P_{2} \to P_{1}\) (the laser at ω 2 transfers its power to that at ω 1) and (3) If P 1 = P 2, then Δ(P 1) = Δ(P 2) = 0, so no exchange of power between laser at ω 1 and laser at ω 2. During the FWM self-stability mechanism, two degenerate FWM \(\left( {2\omega_{\,1} = \omega_{\,3} + \omega_{\,2} ,\;2\omega_{\,2} = \omega_{\,1} + \omega_{\,4} } \right)\)and a single nondegenerate FWM \(\left( {\omega_{\,1} + \omega_{\,2} = \omega_{\,3} + \omega_{\,4} } \right)\) involve in the process. As the contribution of nondegenerate case is a factor of 103–104 less than that of degenerate case, the impact of nondegenerate FWM on the stability of lasers can be ignored [12]. In the degenerate FWM processes, two photons of frequency ω 1 (or ω 2) are annihilated to create one photon of frequency ω 2 (or ω 4) and another photon of frequency ω 3 (or ω 1). In other words, the degenerate FWM produce pairs of waves that are identical down to the level of individual photons. Hence, it follows that the power increment (or reduction) at ω 1 is approximately equal to \(P_{{\omega_{\,112} }}\) or \(P_{{\omega_{\,121} }}\), i.e., \(\Updelta \left( {P_{1} } \right) \approx P_{{\omega_{\,112} }} \approx P_{{\omega_{\,121} }}\). Likewise, the power increment (or reduction) at ω 2 is approximately equal to \(P_{{\omega_{\,212} }}\) or \(P_{{\omega_{\,221} }}\), i.e., \(\Updelta \left( {P_{2} } \right) \approx P_{{\omega_{\,212} }} \approx P_{{\omega_{\,221} }}\) [12].

Fig. 1
figure 1

Effect of gain-mode competition on dual-wavelength EDFL for scheme a without the balance of multiple FWM processes, b with the balance of multiple FWM processes

3 Experiment and operating principle

Figure 2 shows the experimental setup of the switchable single- and dual-wavelength EDFL with continuous wavelength spacing tunability. The proposed EDFL consists of a commercial IPG Photonics erbium-doped fiber amplifier (EDFA) for light amplification, a polarization controller for optimizing FWM efficiency through polarization, a 500-m-long HNLF, a dual-wavelength selective element for tuning the dual wavelength, a 90/10 coupler which extracts 10 % of oscillating light in the cavity as the output and a circulator to ensure unidirectional oscillation besides connecting the dual-selective element to the ring cavity. The dual-selective element is composed of a 50/50 coupler, two variable optical attenuators (VOA) to adjust reflected power coming into the ring cavity, two tunable TBFs for selecting two wavelengths, a 95 % reflectivity mirror and a circulator which acts as a mirror by connecting its port 1 to port 3 of the circulator. The mirrors are used for reflecting the tuned wavelengths back to the ring cavity. The EDFA has gain bandwidth ranging from 1,542 to 1,565 nm with output power up to 30 dBm. The TBFs have FWHM bandwidth of 0.23 nm and is tunable over C-band region. The HNLF used to help stabilize the EDFL is a 500-m-long dispersion-shifted fiber which is manufactured by a Furukawa company OFS. The HNLF has nonlinear coefficient of 11.5 W-1 km-1 and zero-dispersion wavelength at 1,556.5 nm. Its dispersion and dispersion slope at wavelength 1,550 nm are −0.1 ps/(nm-km) and 0.015 ps/(nm2-km), respectively. The laser output and reflected oscillating wavelength are recorded and monitored by an optical spectrum analyzer with resolution of 0.015 nm.

Fig. 2
figure 2

Experimental setup of the switchable and continuously tunable EDFL

The proposed scheme is switchable between single- and dual-wavelength EDFL simply by varying the reflected power coming into the cavity. Varying the reflected power in this scheme is done by adjusting the attenuation of VOAs in each arm of the dual-selective element. When the reflected power of one arm is more than that of another arm, the scheme acts as a single-wavelength EDFL since only a wavelength has enough power to circulate in the cavity, while the other wavelength is suppressed as a result of insufficient gain. When the reflected powers of both arms of the dual-selective element are almost the same, the scheme then acts as a dual-wavelength EDFL. The principle behind the stability of the dual-wavelength laser is based on the FWM effect in HNLF. It is known that homogeneous line broadening of EDF causes gain-mode competition which then leads to the laser instability. With degenerate FWM processes in the HNLF where the energy from the higher power waves is transferred to the lower power waves, the gain-mode competition in the EDF is reduced. This makes the dual-wavelength laser stable.

4 Results and discussions

Figure 3a shows the tunability of the single-wavelength EDFL at EDFA output power 11.95 dBm. The proposed EDFL can act as a single tunable laser by adjusting the attenuation of both VOAs such that the reflected power from one arm is more than that of the other arm of the dual-selective element. Consequently, the weaker wave is suppressed and the stronger wave lases in the cavity. It can be seen in the figure that the tunability spans from 1,542.07 to 1,564.85 nm, limited by the gain bandwidth of EDFA which is from 1,542 to 1,565 nm. The tunability of the laser can be improved further if we have an EDFA with wider gain bandwidth and TBFs with wider tunability range. Besides wide bandwidth, the tunability of the single-wavelength EDFL is also continuous. This is due to the continuous tunability of TBFs. As depicted in Fig. 3b, the peak power fluctuations of the single-wavelength EDF laser at wavelengths 1,542.11, 1,550.02, 1,558.05 and 1,564.9 at 2 min interval for 30-min duration were also recorded. The peak power fluctuations here are defined by the difference between the maximum and minimum peak power recorded during the duration. The peak power fluctuations are <0.5 dB, indicating no significant power variation. Besides the power fluctuations, we also investigate the side-mode suppression ratio (SMSR) of the wavelengths. The SMSR is higher than 41.33 dB for all the tuned wavelengths.

Fig. 3
figure 3

a Tunability of single EDFL, b its peak power fluctuations at 2-min interval for half an hour

The effect of gain-mode competition on the dual-wavelength EDFL without and with the balance of FWM self-stability mechanism is then investigated experimentally. In this measurement, the lasing wavelengths were tuned at 1,558.98 and 1,562.1 nm and the attenuation of both waves was adjusted such that the reflected powers from both arms in the dual-selective element are almost the same. These two frequencies are denoted as ω 1 and ω 2, respectively, for the following discussion. For the experiment without the assistance of FWM as a stabilizer, we take out the HNLF from the experimental setup shown in Fig. 2. As illustrated in Fig. 4a, b, the two lasers compete between each other for the EDF gain. Consequently, the weaker laser is suppressed, leaving only a laser oscillating in the cavity. On the contrary, when we insert the HNLF into the experimental setup to help stabilize the EDFL, the competition for the EDF gain is balanced. Figure 4c, d illustrates how the dual-wavelength EDFL is stabilized by multiple FWM processes. The effect of gain-mode competition is compensated by multiple FWM processes in the HNLF that shuns the stronger laser to dominate the erbium gain. Two degenerate FWM processes of \(2\omega {\kern 1pt}_{1} = \omega_{3} + \omega_{2}\) and \(2\omega {\kern 1pt}_{2} = \omega_{1} + \omega_{4}\) involve in the multiple FWM processes where ω 3 and ω 4 are the two new waves that emerge along with the lasers at the HNLF output. Benefiting from the self-stabilizing effect of FWM in the HNLF, the EDFL is stable with peak power variation <1 dB.

Fig. 4
figure 4

a Output spectra of dual-wavelength EDFL without the stability assistance of FWM at 2-min interval for half an hour, b its peak power fluctuations during the scanning, c output spectra of dual-wavelength EDFL with the stability assistance of FWM at 2-min interval for half an hour, d its peak power fluctuations during the scanning

Figure 5 illustrates the output spectra of the dual-wavelength EDFL with different wavelength spacing at EDFA output power of 11.95 dBm. By adjusting the wavelength of TBFs, the laser with continuous wavelength spacing can be achieved. In this measurement, the wavelength of a TBF was fixed at 1,564.85 nm and the wavelength of another TBF was tuned from 1,542.07 to 1,564.33 nm. It can be seen in the figure that the maximum wavelength spacing is 22.78 nm, while the minimum is 0.52 nm. The maximum spacing is limited by the gain bandwidth of EDFA which ranges from 1,542 to 1,565 nm, while the FWHM bandwidth of TBF of 0.23 nm limits the minimum spacing of the dual-wavelength EDFL. The resulting wavelength spacing is not only wide but also continuous as a result of the use of TBFs, which have continuous tunability. The output stability during the whole tuning range of the wavelength spacing is also monitored. As shown in Fig. 5, the output spectra are stable with peak power fluctuations of <1 dB.

Fig. 5
figure 5

Continuous wavelength spacing tunability and stability of the dual-wavelength EDFL with a spacing of 0.52 nm, b its peak power fluctuations at 2-min interval for half an hour, c spacing of 14.79 nm, d its peak power fluctuations during the scanning, e spacing of 22.78 nm, f its peak power fluctuations during the scanning

The proposed EDFL can be improved in some aspects. Because of the 0.23 nm FWHM bandwidth of TBFs, the EDFL is not a single longitudinal mode (SLM) laser but a multilongitudinal mode [8, 12]. As the total cavity length of the EDFL is in the order of hundred meters, the longitudinal mode spacing is expected to be in megahertz or kilohertz. Such the narrow mode spacing, however, can be filtered by the incorporation of a linewidth-narrowing saturable absorption filter inside the cavity in order to realize a SLM laser [15]. Another room for improvement is to increase the number of lasers generated as the proposed EDFL can only generate two lasers at maximum. A higher number of lasers can be obtained if we introduce more reflected wavelengths (powers) in the scheme by, for example, placing a fiber Bragg grating before the TBF to reflect more wavelengths into the cavity [13].

5 Conclusion

In summary, we have proposed and experimentally demonstrated a stable and widely tunable single- and dual-wavelength EDFL. The stability results from the multiple FWM processes in the HNLF that mitigate the mode competition in the EDF. The use of TBFs in our proposed scheme for the tuning mechanism allows for a wide, continuous and easy-to-operate tunability. The single EDFL can be continuously tuned from 1,542.07 to 1,564.85 nm, and the wavelength spacing of the dual wavelength ranges continuously from 0.52 to 22.78 nm. The performances of the tuning range can be improved further if we use smaller FWHM bandwidth of TBF and wider gain bandwidth of EDFA. As a result of self-stabilizing effect of FWM in HNLF, the laser is stable with its output peak power fluctuations of <1 dB for both single- and dual-wavelength EDFL. By adjusting the reflected power from the dual-selective element through variable optical attenuators, the EDFL is switchable between single and dual wavelengths.