1. Introduction

Driven by a diverse range of potential applications including micromachining, spectroscopy and nonlinear imaging, research in ultrafast fiber lasers operating at 1 μm has undergone a period of explosive growth in the past few years. These sources are intrinsically associated with many desirable qualities compared to their solid-state counterparts such as compactness and low manufacturing cost. However, product-scale adoption of mode locked fiber lasers in industrial environments is hindered by their sensitivity against external perturbations. Indeed, thermal and mechanical stress, ever present in said environments, are known to influence the birefringence properties of fiber cavities, potentially resulting in unpredictable degradation of device performance unless the cavity is constructed exclusively out of polarization maintaining (PM) fiber components. [1–4]

Conventionally mode-locked fiber lasers have been constructed following either of two predominant design principles: (i) in all-anomalous dispersion cavities the intracavity pulse propagates unchanged as a fundamental soliton due to balancing of dispersion with Kerr nonlinearity [5], whilst (ii) in dispersion-managed cavities elements of both normal and anomalous dispersion are utilized to compensate for each other [6]. However, in 2006 Chong et al. contested these ideas by reporting on a cavity consisting exclusively of all-normal dispersion (ANDi) elements [7]. In this configuration, a spectral filter is used to “reset” the field at the end of each roundtrip so as to compensate for the large dispersion-induced positive chirp acquired during the roundtrip [8–11]. With regards to the one micron regime, the possiblity of building such ANDi-oscillators yields particularly profound implications due to the intrinsic dispersion of silica being normal at 1μm. Indeed, because ANDi lasers do not require any form of complicated intracavity dispersion compensation schemes they offer a significantly simpler and more cost-efficient design for mode-locked fiber oscillators [12–16].

A particular operation regime of ANDi-lasers is that of a giant chirp oscillator (GCO) [17]. In a GCO the laser cavity is purposefully lengthened to introduce a very large net normal dispersion. Oscillators operating in this regime deliver temporally broad pulses with low-repetition rate but high-energy and a very large linear chirp. Although many different GCO configurations have been demonstrated in the past few years [17–21], to the best of our knowledge all of the reported architectures have employed either nonlinear polarization evolution (NPE) [17–19] or saturable absorbers (SAs) [20, 21] as the mode-locking mechanism. However, both NPE and SAs possess known drawbacks that limit their applicability in industrial applications: (i) NPE requires adjustable control over polarization which restricts its use to non-PM cavities whereas (ii) SAs suffer from long-term reliablity problems and power handling issues.

In this Paper we demonstrate what to our knowledge is the first GCO mode-locked with a nonlinear amplifying loop mirror (NALM). Compared to a previous NALM-locked ANDi-design demonstrated in our group [16], the laser reported here delivers nearly two orders of magnitude larger pulse energies (> 10 nJ) without sacrificing the compressibility of the pulses into the sub-500 fs regime. Because the all-fiber cavity is constructed using exclusively PM elements the laser is robust and environmentally stable. In fact, to our knowledge this design constitutes the first all-PM-fiber GCO and therefore the first GCO that can be considered truly environmentally stable. We also present extensive numerical simulations that highlight the versatility of the design, illustrating how simple cavity modifications allow for the output pulse energy and duration to be adjusted to suit different applications.

2. Cavity configuration and numerical simulations

Figure 1 shows a schematic of the proposed cavity configuration. As with our earlier design [16], the cavity consists of two distinct modules: (i) the main laser module containing a segment of low-doped Yb-fiber pumped by a laser diode at 976 nm, an isolator, an output coupler (OC) and a band-pass filter centered at 1030 nm (1.7 nm passband) and (ii) the NALM-module which contains a segment of highly-doped Yb fiber and a segment of single-mode fiber (SMF). The NALM module is connected to the main module using a 55/45 coupler and our OC extracts 80 % of the intracavity power, in contrast to the 20 % extraction used in our earlier design. In our present configuration we also introduce a long segment of SMF between the gain fiber and the NALM. In fact, as we shall see below, it is precisely this segment of SMF that allows for the oscillator to operate in the GCO regime and whose length, L_SMF, plays a key role in determining the output pulse energy and duration. At this point we wish to stress that even in the absence of the long SMF segment the configuration here differs from our earlier design: extensive numerical simulations revealed that accessing the GCO regime requires the SMF segment in the NALM to be significantly shorter (∼ 2 m) from that used in our earlier design (∼ 10 m). Despite this modification we have observed no indications that the current configuration would not share the remarkable stability of our earlier design which has now accumulated nearly 6000 hours of continuous operation with no interruptions in mode-locking.

 

Fig. 1 Schematic of the cavity. BPF, bandpass filter; LD, laser diode

Download Full Size | PDF

In order to assess the characteristics of the proposed configuration we first use rigorous numerical simulations based on the generalized nonlinear Schrödinger equation to model the laser. Our modeling includes the Raman effect, full dispersion profile, self-steepening and the initial condition corresponds to white noise (see Ref. [16] for details). The gain dynamics along both active fibers are modeled for given pump diode powers (which we fix here as 180 mW) using an analytical three-level model [22]. In the simulations we vary the length of the SMF segment separating the active fiber from the NALM so as to directly investigate its role on the output pulse characteristics. Then, for each distinct value of L_SMF, we iterate the cavity until the field converges to a stable solution.

In Fig. 2(a) we plot the energy and r.m.s. duration of the simulated output pulses as a function of L_SMF. Several observations can be made. We see how the output energy initially increases but begins to saturate at L_SMF ≈ 30 m before abruptly transitioning into a regime of linear growth. This transition is also apparent in the evolution of the output spectrum shown in Fig. 2(b), suggesting that at this SMF length the oscillator transitions from one mode-locking regime to another. It is precisely this regime, achieved for L_SMF > 30 m, that corresponds to the desired GCO-regime, in which the output pulses are characterized by high energy, long temporal duration and a large linear chirp. To see this explicitly, we show in Fig. 3 how the output pulse temporal characteristics change for increasing L_SMF. We can see from Fig. 3(a) that as the SMF length increases the output pulses become ever broader, yet retain an almost constant temporal shape. This is also confirmed in Fig. 3(b,c) where we plot the temporal profile in more detail for selected L_SMF. Here we also plot the frequency chirps associated with the output profiles and observe linear variation for both cases.

 

Fig. 2 (a) Output pulse energy (dark blue) and temporal r.m.s. width (black) as a function of L_SMF. (b) Shows the output spectrum for varying L_SMF. Black and white arrows in (a) and (b) denote the transition to the GCO regime, respectively.

Download Full Size | PDF

 

Fig. 3 (a) Output pulse temporal profile as a function of L_SMF. (b) and (c) show the temporal profiles (black) and corresponding chirps (dark blue) in more detail for L_SMF = 70 m and L_SMF = 150 m, respectively.

Download Full Size | PDF

It is remarkable how the output pulse energy can be up-scaled simply by increasing the length of the SMF segment, with energies exceeding 10 nJ achievable using modest lengths [see Fig. 2(a)]. To gain more insight, we fix L_SMF = 50 m and show in Fig. 4(a) and 4(b) the temporal and spectral evolution, respectively, of the pulse inside the cavity over one roundtrip. Note that here we only show the clockwise propagation inside the NALM for simplicity. We can clearly observe from Fig. 4(a) how the amplified pulse undergoes significant stretching in the SMF before being re-amplified in the NALM. This notion allows us to interpret the long SMF segment to act as an intracavity stretcher between the two active fibers, allowing for the second amplification to proceed without inducing detrimental wave-breaking effects. Another remarkable feature, observed from Fig. 4(b), is the amount of spectral shaping during a single roundtrip. Indeed, we can readily see how the 1.7 nm spectrum at the beginning of the roundtrip is broadened to nearly 15 nm during propagation.

 

Fig. 4 Simulation results showing the evolution of the (a) temporal and (b) spectral profiles during a single roundtrip in the cavity with L_SMF = 50 m.

Download Full Size | PDF

3. Experimental results

Encouraged by our numerical results we constructed an all-fiber all-PM cavity similar to that shown in Fig. 1. Each fiber segment was carefully spliced using a PM splicer and the laser output was diagnosed with a fast photodiode, optical and radio-frequency (RF) spectrum analyzers and a commercial frequency-resolved optical gating (FROG). We also used a grating pair with 1200 lines/mm to externally recompress the output pulses. In order to experimentally investigate the role of the SMF separating the main gain fiber and the NALM, we performed our experiments with two distinct lengths of PM-980, namely 50 m and 100 m. In both cases the laser could easily be mode-locked by simply adjusting the pump diode drive currents in the main module as well as in the NALM module. In fact, the possibility to influence the NALM’s mode-locking characteristics by simply tuning its pump diode represents a major benefit in obtaining stable mode-locking. Because SAs are void of such a degree of freedom, we believe the NALM to be a superior mode-locker in the current system. When mode-locked, the laser emits a stable train of pulses with an approximately 60 dB signal-to-noise ratio in the RF spectrum, is insensitive to thermal and mechanical stress and can be operated without interruptions for extended periods of time. From the RF spectrum we also inferred that the laser was emitting uniformly spaced pulses at the fundamental cavity repetition frequency which was further verified using a fast oscilloscope.

Our experimental results are summarized in Fig. 5 where we plot the oscillator output spectrum (a,d) and temporal profiles retrieved from a FROG trace before (b,e) and after (c,f) grating-pair compression for both used SMF lengths [in (a–c) L_SMF = 50 m and in (d–f) L_SMF = 100 m]. We also list the performance characteristics of the oscillator and corresponding pump diode powers in Table 1. For completeness, we also show in Table 1 the laser characteristics when the long SMF segment is removed altogether. We can see from Fig. 5 that for both SMF lengths the output spectrum as well as the oscillator temporal shape display qualitatively similar features with a broad spectrum spanning nearly 15 nm and a long temporal pulse duration. For the shorter segment of SMF the energy of the output pulses is 10 nJ while 16 nJ is reached when a longer segment of SMF is used. In contrast, although mode-locked operation is achieved even in the absence of SMF, the pulse energy is limited to 2.4 nJ. These values are in excellent accordance with the numerical predictions presented above. Whilst the temporal and spectral characteristics at oscillator output are qualitatively similar for both SMF lengths significant discrepancies can be observed in the compressed pulse profiles. Indeed, for L_SMF = 50 m the pulses can be compressed down to 250 fs with negligible pedestals whereas for L_SMF = 100 m compression is limited to approximately 350 fs with nearly 20 % of energy contained in the pedestal. We believe this is due to third order dispersion accumulated in the SMF that cannot be compensated with the used grating-pair. Of course, due to the imperfect compression the pulse peak power does not reach its full potential for the longest SMF length. In fact, we find that with our current compression scheme (double-pass grating pair with a ∼ 25% overall efficiency) the largest peak power is obtained when using L_SMF = 50 m as seen in Table 1.

 

Fig. 5 Experimental results. In (a,d) we plot the output spectrum, in (b,e) the oscillator temporal profile retrieved from a FROG trace and in (c,f) the temporal profile post-compression. In (a–c) L_SMF = 50 m, in (d–f) L_SMF = 100 m and the insets in (a,d) show the spectrum on a linear scale.

Download Full Size | PDF

Table 1. Laser performance characteristics and pump powers. P_m, main pump power; P_N, NALM pump power; f_rep, repetition rate; P_avg, average output power; Δτ_osc, oscillator FWHM duration; Δτ_compr, compressed FWHM duration; P_p, compressed peak power.

View Table

4. Conclusion

We have reported what to our knowledge is the first environmentally stable GCO. Owing to the use of a NALM we have been able to construct a robust all-PM all-fiber cavity that is insensitive to thermal and mechanical stress while preserving the ease of mode-locking characteristic of NPE. We have used extensive numerical simulations to reveal the versatility of the design, and shown how the output pulse characteristics (energy, duration) can be controlled by varying the amount of SMF separating the gain fiber and the NALM. Experiments conducted with different SMF lengths confirm these predictions, with energies as high as 16 nJ observed with a ∼ 100 m long cavity. We anticipate our results to represent a significant step towards making mode-locked fiber lasers a viable alternative to their solid-state counterparts in many industrial applications where environmental stability is a stringent device requirement.