1. Introduction

Attribute to the evident advantages of fiber lasers over their bulk counterparts, such as simple design, high stability, and low alignment sensitivity, fiber-based mode-locked lasers as an alternative source of the ultrafast optical pulses have been intensively investigated, and various applications have been proposed, such as optical communication, supercontinuum generation, nonlinear optics, and optical sensors. By adjusting the cavity structure with appropriate dispersion management and exploiting different types of devices and techniques, such as semiconductor saturable absorber [1], nonlinear polarization rotation technique, carbon nanotube [2], Graphene oxide [3, 4], and pulse modulated intensity modulator [5], many types of mode locking have been proposed and experimentally validated in mode locking pulse formation in fiber laser, such as conventional soliton [6, 7], stretched soliton [8, 9], similariton [10, 11], and dissipative soliton [12–16]. However, overdriven nonlinear effect in the optical fiber may limit the scaling up of the pulse energy. For instance, soliton splitting and multi-pulsing may occur as pulse peak power arises in soliton mode-locked fiber laser [17, 18].

In recent years, a new mechanism of generating high-energy pulses from mode-locked fiber lasers has been developed and investigated experimentally, which is called Dissipative Soliton Resonance (DSR). The signature characteristics of this mode locking regime are rectangular pulse waveform and non-wave-breaking with increasing pump power. The corresponding physical mechanism has been revealed by solving the cubic-quintic Ginzburg-Landau equation [19–22]. Several techniques of mode locking have been utilized to achieve high energy wave-breaking-free rectangular pulse by DSR at different wavelength band, such as the technique of nonlinear polarization rotation [23–26], SESAM-based mode locking [27], figure-eight fiber lasers [28–30], and Graphene oxide based mode locking [31]. However, almost all of them are passively mode locking. As far as we know, there is no report on rectangular wave-breaking-free high energy pulse generation from actively mode locked fiber laser.

Here, we present nanosecond scale high energy rectangular pulse generation in an actively mode locked Yb-doped fiber laser based on an intra-cavity intensity modulator and a polarization controller. The pulse dynamics exhibit several characteristics that are similar to DSR mode locking, while several unique experimental results are presented. At fundamental mode locking, the laser can generate rectangular pulse with duration tunable from 890 ps to above 124 ns. The pulse width and pulse energy can be tuned flexibly and independently. The effect of adjusting modulation signal parameters on the output pulse temporal dynamics is presented in detail.

2. Experimental setup

The experimental setup is shown in Fig. 1.In this structure backward pump scheme is used with a maximum power 500 mW 976 nm laser diode as pump source. The pump light is coupled into a section of 1.5 meters piece of Yb-doped fiber (YDF) through a 980/1064 nm wavelength division multiplexer (WDM). The Yb-doped fiber has a nominal absorption coefficient of 250 dB/m at 975 nm and a group velocity dispersion (GVD) of −35 ps/nm/km at 1060 nm. An optical filter with 1 nm 3 dB-bandwidth is located after the amplifier to select operating wavelength and filter the spectrum. A polarization-sensitive LiNbO3 Mach-Zehnder intensity modulator (10 GHz bandwidth) is employed in the cavity as the modulator to achieve active mode locking. The pulse generator can generate duration tunable electrical rectangular pulse with ~30 ps rising/falling time, as narrow as 80 ps in duration and as high as 6.25 GHz in repetition rate. The generated electrical pulse signal is amplified by a wideband amplifier to drive the modulator (the term “modulation pulse width/duration” in this article means the electrical rectangular applied on the modulator). A polarization controller is applied before the modulator to achieve effective polarization controlling. A section of around 80 meters single mode fiber (Corning HI 1060) is incorporated between the polarization sensitive modulator and the polarization controller to increase cavity length and enhance nonlinear effects. The fiber has a GVD of −30.8 ps/nm/km and the total cavity dispersion was estimated to be 1.5 ps². The following experimental results indicate that the polarization controller plays a key role in pulse formation. A polarization-insensitive optical isolator is utilized to ensure the unidirectional propagation of the light in the cavity. The optical pulses in the cavity are coupled to the output port through a 30/70 coupler. The output pulses are monitored by an optical spectrum analyzer (measurement scale of 600~1700 nm, 0.02 nm minimum resolution), and converted to electrical signal by a high speed photodetector (25 GHz bandwidth), then analyzed by a 20 GHz real-time oscilloscope and a radio frequency spectrum analyzer (13.5 GHz bandwidth).

 

Fig. 1 Schematic of the actively mode-locked Yb-doped fiber laser. WDM: 980/1060 nm wavelength-division multiplexer. YDF: Yb-doped fiber. PC: polarization controller. MZIM: Mach Zehnder intensity modulator. OSA: optical spectrum analyzer.

Download Full Size | PDF

3. Experimental results and discussion

With proper setting-up of the modulation signal parameters (pulse width, repetition rate), and proper adjustment of the polarization controller, the laser can operate in stable and self-starting mode locking regime. Figure 2 shows a typical output pulse train observed. In this case, the electrical pulse duration of the pulse generator is set at 10 ns and the repetition rate is tuned to fit the cavity frequency c/nL. The pump power is 250 mW and the average output power is 8.2 mW. Figures 2(a) and 2(b) present the temporal characteristics of the pulse train observed in the oscilloscope in the range of 50 ns and 10 μs respectively. The pulse waveform is a rectangular shape with sharp edges and flat top, and the full width at half maximum (FWHM) of the pulse is measured to be 9 ns. The fundamental repetition rate is 2.49 MHz, corresponding to a cavity length of 83 m. The radio frequency (RF) spectrum observed in the radio frequency spectrum analyzer is shown in Fig. 2 (c) and the inset therein with span of 80 kHz and 100 MHz, and resolution bandwidth of 100 Hz and 1 kHz, respectively. It is revealed that the signal-to-noise ratio (SNR) is suppressed better than 70 dB, indicating an ultra-high temporal stability. The optical spectrum of the pulse train is shown in Fig. 2(d). The spectrum has a quasi-triangular shape with a narrow bandwidth on the top and broad bandwidth at the base part. The 3 dB spectrum width is 0.12 nm and the 20 dB bandwidth is 1 nm.

 

Fig. 2 (a) The single waveform of the mode-locking pulse observed in the Oscilloscope; (b) Oscilloscope trace the mode-locking pulse train; (c) RF spectrum of the output pulses with a span of 80 kHz and 100 MHz (inset figure); (d) Optical spectrum of the output pulse.

Download Full Size | PDF

To be mentioned, in addition to the intra-cavity intensity modulator as the mode-locker, the polarization controller also performs a critical role in the pulse formation. Proper adjustment of the polarization controller is required to achieve stable mode locking regime. Otherwise, the pulse waveforms come to splitting and breaking with numerous fluctuating minor pulse peaks, as demonstrated in Fig. 3(a).Proper adjustment of the polarization controller can transform the waveform back to rectangular shape. In comparison to the stable mode locking state presented in Fig. 2, Fig. 3 presents the unstable wave-breaking state in time domain and frequency domain under the same conditions except for the state of the polarization controller. As we can see, with improper polarization controlling, the pulse waveform become splitting and excess noise appears in the frequency domain alongside the mode locking frequency.

 

Fig. 3 The wave-breaking state observed in (a) the Oscilloscope and (b) the RF spectrum analyzer.

Download Full Size | PDF

Furthermore, with the adjustment of the modulation pulse width and proper adjustment of the polarization state, stable pulse train with widely duration-tunable rectangular waveform can be generated from the laser. Figure 4 shows the evolution of the output pulse waveform and spectrum with an increasing modulation pulse width at a fixed pump power of 250 mW. As seen in Fig. 4, with the modulation pulse width increasing from 1ns to 10ns, in Fig. 4(a) and 10 ns to 100 ns, in Fig. 4(b), the output pulse width (FWHM) changes from 890 ps to 9 ns(in Fig. 4(a)) and 9 ns to 90 ns (in Fig. 4(b)). The output pulse width increases monotonically with the widths of the electrical driving pulse, but slightly narrower. Figure 4(c) shows the evolution of the optical spectrum with different modulation pulse width (from 10 ns to 100 ns). As the modulation pulse width increases from 10 ns to 100ns, the base parts of the spectra narrow down accordingly.

 

Fig. 4 Evolution of the output pulse shape (a) 1-10 ns, (b) 10-100 ns, and (c) optical spectrum under modulation of different pulse widths.

Download Full Size | PDF

For the purpose of better revealing the pulse characteristics, Fig. 5(a) further shows the experimentally measured pulse width and pulse energy versus the modulation width at a fixed pump power of 250 mW. As can be seen from Fig. 5(a), the pulse width increases monotonically to 90 ns when the modulation width changes from 1 ns to 100 ns and the pulse energy increases from 0.8 nJ to 7 nJ. With the fixed pump power, the pulse energy increases rapidly in the region of 1 ns to 20 ns, but increases slowly in the region of 30 ns to 100 ns, which manifests that narrow modulation durations restrict the pulse energy scaling in the cavity. As the modulation width is increased, the restriction is diminished, and the increasing of pulse energy gradually slows down.

 

Fig. 5 (a) Output pulse width and pulse energy versus the modulation width at the fixed pump power of 250 mW; (b) Output pulse energy versus the pump power level at the modulation width of 10 ns, 20 ns, 50ns and 100 ns, respectively.

Download Full Size | PDF

Figure 5(b) shows the output pulse energy versus the pump power level at the modulation width of 10 ns, 20 ns, 50 ns and 100ns, respectively. As we can see, in general, the pulse energy increases linearly and monotonically with pump power level. The threshold power and slope efficiency alter with modulation width. Longer modulation width corresponds to higher threshold power and higher power slope efficiency. However, as the pump power increases, the pulse duration of the output pulse stay almost unchanged, which is mainly determined by the modulation width. The highest pulse energy 19.8 nJ with pulse duration of 124 ns can be obtained under pump power of 500 mW and 140 ns modulation width. Higher pulse energy can be expected with the more available pump power.

At the pump power of 250 mW, when the modulation width exceeds 100 ns, the pulse waveform become splitting and fluctuating, just like the case shown in Fig. 3(a). In this case, more pump power is needed to recover the steady rectangular non-breaking pulse train. Actually, there is an upper-limit of modulation width to ensure stable pulse formation at a certain pump power level. At lower pump power, the upper-limit of modulation width will decrease, because long duration and low pump power lead to low pulse peak power, and when the pulse peak power is too low to meet the requirement of pulse formation, pulse waveform come to break. In other words, the modulation width and pump power should be set in pairs to ensure wave-breaking-free.

As we know, the time domain instability can be ascertained by observing the radio frequency spectrum of the pulse train. For a stable and amplitude-even pulse train with repetition frequency f, there will be a single line at f on the frequency spectrum. If the pulse train turns unstable, noise figure or additional lines would appear alongside the frequency f, resulting in a degradation of SNR. At a certain pump power, there is a gradual reduction in SNR of the output pulse as the modulation width is increased. For instance, at the pump power of 250mW, SNR decreases 9 dB as the modulation width increases from 20 ns to 100 ns, as we can see in Figs. 6(a) and 6(b) (the resolution bandwidths are all set at 100 Hz). If we take the 10 dB reduction in SNR from the SNR of 20 ns modulation width as the criterion of the upper-limit of modulation width, the upper-limit of modulation width increases with an increasing pump power, as shown in Fig. 6(c).

 

Fig. 6 The RF spectrum of pulse trains with modulation width of 20 ns (a) and 100 ns (b) under the pump power of 250 mW; (c) The evolution of the upper-limit of the modulation width and the output width with respect to increasing pump power level.

Download Full Size | PDF

The mode locking regime stated above is mainly characterized with two critical identities: one is rectangular pulse waveform and the pulse durations can be widely tunable under the control of the modulation signal, and the other is the non-wave-breaking high pulse energy output under strong pump light. The characteristics of the laser are largely depended on the modulation pulse widths, the states of the polarization controller and the pump power level. To be specified, the pulse duration mainly depends on the modulation width, while it is relevant to pump power level in some cases. The pulse energy and pulse peak power are dependent both on the pump power level and the modulation width. The modulation width affects the pulse energy by restricting the pulse energy under narrow modulation width, while, the pump power determines the pulse energy in a linear way.

In regard to passive mode locking fiber lasers, there are plenty of recent works with similar features, mainly the fore-mentioned DSR mode locking. However, as far as we know, there are no similar experimental observations reported in actively mode locked Yb-doped fiber laser. Comparing with DSR mode locking, the similarities and differences co-exist. The shared identities are duration-tunable rectangular waveform and non-wave-breaking feature. However, the pulse duration and the pulse energy of DSR mode locking is solely determined by pump power. The pulse duration and energy basically monolithically increase with pump power. By contrast, in our laser, there is one more parameter, namely the modulation width, the pulse duration and pulse energy can be determined independently. The pulse peak power is variable with pump power and modulation width, while, the peak power of DSR mode locking stays almost unchanged as the pump power increase. The independent tunability of the pulse duration, energy and peak power may be a great advantage, even necessity in some applications.

Since the polarization controller performs a critical role in pulse formation and stability of the mode locking, it is presumed that the combination of polarization controller, polarization dependent modulator and the nonlinear-enhancement fiber between them forms an artificial saturable absorber based on nonlinear polarization rotation effect. The modulator acts as an active mode locker as well as a polarizer. Hence, the combined action of intra-cavity loss modulation based on the intensity modulator and nonlinear polarization rotation controlled by the polarization controller form the unique mode locking behaviors stated above.

However, in consideration of the distinctive characteristics of this mode-locked fiber laser, it is imprudent to ascribe the mode locking mechanism to DSR. Since there is no related experimental and theoretical research reported before, the detailed explanations of the physical mechanism behind this mode locking phenomenon may require further experimental and theoretical investigations.

Moreover, it is observed that the evolution of output pulse presents some intriguing variations when the modulation frequency is detuned around the cavity frequency as shown in Fig. 7, where the modulation width is fixed at 10 ns. As we can see, detuning of the modulation frequency results in narrowing down of the output pulse width and variation of the pulse shape. With a slight positive detuning, the optical pulse is formed as quasi-rectangular shape with a slower rising edge, while the falling edge is kept unchanged. On the contrary, with a slight negative detuning, the falling edge gradually slows down with an increasing detuning value, while the rising edge is kept. For a larger detuning, the optical pulse shape continues deforming from the rectangular shape, with one of the rising and falling edge turning slow. A detuning of larger than ± 50 kHz in 10 ns-modulation will result in absence of mode locking behavior.

 

Fig. 7 Evolution of the pulse shape with different detuning frequency. (a) Positive detuning; (b) Negative detuning.

Download Full Size | PDF

4. Conclusion

We present a new method of generating flexible nanosecond scale rectangular wave-breaking-free pulse from actively mode locked Yb-doped fiber laser, which is based on an intra-cavity intensity modulator and a polarization controller. The laser is featured with flexible and independent tunabilities in pulse width, pulse energy and pulse shape as well as the capability of high energy pulse output. Many applications may benefit from the superior temporal tunability of this kind of pulsed fiber laser.

Comparing with fore-reported rectangular wave-breaking-free pulse generation from passively mode locking, our fiber laser exhibits many distinct advantages in several aspects. For instance, it is independently controllable in peak power and pulse energy and it is easy to be synchronized with pulse shaping devices for an expanded area of applications. Besides, since the laser dynamics exhibit some novel characteristics, it will help us to understand the mechanism of active mode locking in a new perspective.