1. Introduction
The development of new, more affordable, reliable and efficient methods for generating short and ultrashort coherent pulses in various spectral ranges is an actual task of laser science. Recent advances in this direction are related to the discovery of physical mechanisms that enable the shaping of regular pulse trains in fiber and hybrid lasers without resorting to traditional modulation or saturable absorption techniques. The novel, emerging approaches rely upon the particular temporal dynamics of such physical mechanisms as frequency comb generation in high-quality nonlinear microcavities [
1,
2,
3,
4], frequency-restricted instabilities of various natures [
4,
5,
6,
7], self-pulsing due to active-ion quenching and energy up-conversion in heavily doped rare-earth fibers (so far elaborated for Tm-doped fibers) [
8,
9,
10,
11], and quasi-synchronous gain modulation in rare-earth-doped fibers (so far elaborated for Yb-doped fibers) [
12,
13,
14].
The quasi-synchronous gain modulation technique features simplicity, reliability, energy efficiency and scalability, compared with other fiber-based approaches. Unlike the conventional synchronous pumping approach (employed so far only in Raman-based [
15,
16] and Tm-based fiber laser systems [
17,
18]), the novel technique does not require a complicated powerful laser source of short or ultrashort pulses for pumping an active medium and generating pulses similar in duration to pump pulses. The quasi-synchronous gain modulation can be implemented with direct continuous-wave diode pumping, which makes this solution more compact, affordable, and energy efficient.
The underlying physics of pulse shaping via the quasi-synchronous gain modulation were originally discovered and described for an effectively two-level gain medium, namely an Yb-doped fiber pumped at 0.98 µm [
12,
13,
14]. This discovery enabled mode-locked pulse generation to be triggered by the relatively slow and weak modulation of the pumping radiation power. The necessary condition for shaping narrow pulses via this method is a slight overrating of the gain modulation frequency compared to the conventional synchronous pumping approach. This condition provides maximum gain for the narrow leading edge of the generated light waveform and gain depletion for the rest of this waveform, thereby limiting the gain–loss balance to a narrow time frame.
In this work, we explore, for the first time, the feasibility of the above pulse shaping mechanism in lasers with a three-level gain medium, namely an Er-doped fiber pumped at 0.98 µm. We demonstrate that, despite the very different energy level diagram and much longer (10 times) upper-state lifetime of the Er-fiber-based laser gain medium [
19,
20], the quasi-synchronous pump power modulation enables the shaping and stationary generation of a regular train of high-energy nanosecond pulses in Er fiber lasers, similar to the Yb fiber lasers. Moreover, we show that this method is applicable not only under the conditions of normal intracavity dispersion (typical of Yb all-fiber lasers), but also under the conditions of anomalous dispersion (more typical for Er and Tm all-fiber lasers). This extends the spectral coverage via the considered method and validates its generality.
2. Methods
The experimental setup layout is shown in
Figure 1. To comply with the conceptual simplicity of the method, we used the simplest ring laser configuration with a minimum number of intracavity elements. The active fiber was a core-pumped highly doped Er fiber (Liekki Er80-8/125). The length of this fiber was as short as 0.4 m. The fiber-coupled hybrid device, TIWDM, integrates a tap, an isolator, and a wavelength-division multiplexer. This device allows only unidirectional lasing in the cavity ring and facilitates the injection of pumping radiation at 980 nm from an external source into the core of the Er fiber.
The signal passband through the entire set of intracavity passive elements was limited to the wavelength range of 1530–1565 nm, owing to their technical characteristics.
The pump source was a fiber-coupled laser diode (LD) whose output optical power can be electronically controlled (in the range of 0 to 0.7 W) through a modulation input of the LD current driver. This current driver with a modulation bandwidth of 0.74 MHz formed the electronic bottleneck in the pump modulation system. In fact, the fast and deep arbitrary modulation of heavy current in high-power pumping LDs is a generally known technical issue that is more complicated for more powerful LDs. Thus, the studied physical mechanism that enables the shaping of relatively short laser pulses by relatively slow and weak pump power modulation definitely possesses a practical value.
To reduce the fundamental pulse repetition rate and accordingly increase the per-pulse energy, the cavity was extended by a 2.3 km long non-zero dispersion-shifted fiber (NZDSF, True Wave Classic). The resulting cavity length corresponded to a fundamental pulse repetition rate of about 88.9 kHz. To avoid the destructive (destabilizing and energy-limiting) impact of excessive nonlinear phase incursion [
21] and stimulated Raman scattering on pulsed lasing [
22], we followed the principle of power management proposed in [
23] for high-energy ultra-long fiber lasers. By providing a high output coupling ratio of 90% at a proper intracavity position, we minimized the peak power in the ultra-long passive fiber span. This inhibited both the reaching of the stimulated Raman scattering threshold [
24] and the accumulation of the excessive nonlinear phase by generated pulses. A moderate degree of self-phase modulation and four-wave mixing due to Kerr nonlinearity can even be a stabilizing factor for stationary pulse generation in an anomalous dispersion fiber laser, as discussed in [
9]. We validated the method for both anomalous and normal dispersion by replacing the NZDSF (which had a negative group velocity dispersion: β
2 < 0) with a dispersion-compensating fiber (DCF, Sumitomo Electric, β
2 > 0) at the end of our study.
The upper inset in
Figure 1 is a sketchy illustration of the expected pulse shaping via the asymmetrical amplification of the generated pulse in the active fiber. This effect was originally achieved in Yb-doped fiber lasers [
12,
13,
14] via the quasi-synchronous sine-wave modulation of the pump power when the modulation frequency
fmod was slightly exceeded relative to the intrinsic fundamental pulse repetition rate
f0 (determined by the cavity optical length
Lopt and defined as the frequency spacing of the cavity longitudinal modes):
where
,
, and
c is the speed of light.
Such frequency-overrated modulation ensures maximum gain recovery at the arrival of the leading edge of a generated waveform and gain depletion for the rest of the waveform (as qualitatively shown in
Figure 2). The resulting slight narrowing of the waveform is repeated after every round trip. Ultimately, this leads to the stationary generation of narrow pulses shaped from the initial intensity-modulated continuous wave.
The distinctive features of this method (compared to gain switching) are that the modulated pump power is always kept above the lasing threshold, the steady-state lasing is sustained continuously, and the gain modulation depth is very small (≤1%).
Herein, we explored the feasibility of similar pulse shaping in lasers with an Er-fiber-based gain medium, whose energy level diagram (the intracavity inset in
Figure 1) and upper-state lifetime differ from those of Yb-fiber-based laser gain media. Despite these differences, we anticipated the feasibility of this effect in Er fiber lasers, assuming the applicability of the two-level approximation from early theoretical studies of the Er ions kinetics and population inversion dynamics under different pumping conditions [
19,
20]. These studies suggest that the involvement of the short-lived level E3 in the pump energy transfer has a negligible effect on the dynamics of population inversion and laser gain if the pump power at 980 nm is below 1 W. We also relied on promising simulation results obtained from our early numerical model of effectively two-level lasing [
12], after readjusting the lifetime of the upper laser level to match that in Er fiber lasers.
3. Results
For achieving quasi-synchronous gain modulation in the studied Er fiber laser, the current and, consequently, the optical power of the pumping LD were modulated by a direct-current-biased frequency-tunable sine-wave signal from the programmable RF AWG. The modulated pump power oscillated between 0.1 and 0.7 W, always remaining above the lasing threshold of 90 mW.
Starting at an arbitrary kilohertz-scale modulation frequency, we typically observed continuous-wave (CW) lasing with a rather weak and slow intensity modulation that tended to follow the modulating sine wave. The modulation frequency fmod was then smoothly tuned to approach the intrinsic fundamental pulse repetition frequency of f0 ≈ 88.9 kHz, which was determined by the cavity optical length.
Downward approaching to
f0 made the initial, weakly and slowly (1/
fmod ≈ 11.3 μs) modulated CW laser radiation gradually evolve into a sequence of narrowing pulses, as evidenced in
Figure 3a. The monotonic pulse narrowing, induced by adiabatically decreasing the modulation frequency towards
f0, did not affect the average laser power (~16 mW) and ended at the frequency of
flow ≈ 88.97 kHz, with a pulse duration of about 750 ns. The achieved duration was 15 times shorter than the modulation period (1/
fmod ≈ 11.3 μs), as can be seen from the measured time traces in
Figure 3b.
Further lowering the frequency led to pulse breaking, presumably due to the excessive nonlinear phase incursion caused by the rise in the peak power. When the modulation frequency fell into the range of flow ÷ f0, the laser switched from stable single-pulse generation (per intracavity round trip) to unstable multi-pulse generation. Below f0, the laser returned to weakly and slowly modulated CW lasing. Backward (upward) frequency tuning made it possible to reverse the evolution of lasing with a slight hysteresis, leading to longer pulses.
Once triggered, the generation of a regular pulse train with the pulse repetition frequency corresponding to
flow ≈ 88.97 kHz remained stable for an unlimited time.
Figure 4a shows a lengthy discrete series of time traces collected at 1 min intervals over the course of 1 h, by means of a real-time oscilloscope and a fast photodiode (PD1). This observation provides evidence of the regularity and stability of the pulse train. The oscillograms show that the generated pulses have high contrast (~99%), and that the fluctuations in their peak power are very small (<1%). The achieved pulse generation was also very stable in terms of the average laser power, as evidenced in
Figure 4b. This figure shows a lengthy discrete series of the average power measurements at 1 min intervals over the same 1 h course. The mean and standard deviation of these measurements were calculated as 16.051 mW and 0.014 mW, respectively.
We also measured the radiofrequency (RF) spectrum of the generated pulse train and analyzed it using criteria similar to those used for mode-locked laser radiation. As seen in
Figure 5a, the recorded RF spectrum has the characteristic comb-like structure with a relatively high signal-to-noise ratio, reaching 75 dB in the vicinity of the pulse repetition frequency. Thus, the RF spectrum bears similarity to the typical RF spectra of mode-locked laser radiation.
Additionally, we measured the optical spectral properties of the generated pulse train. As seen in
Figure 5b, the recorded optical spectrum has a central wavelength of 1532 nm. Its full width at half maximum (FWHM) is about 2 nm. Overall, the measured spectral characteristics indicated that the output pulsed laser radiation had a relatively high spectral purity: there was no noticeable contribution from amplified spontaneous emission, secondary lasing wavelengths, or Raman scattering.
At the final stage of our study, we validated the feasibility of the method also for the case of normal intracavity dispersion. To provide such dispersion, we replaced the primary NZDSF (β2 < 0) with a dispersion-compensating fiber (DCF, Sumitomo, β2 > 0). The DCF had a shorter length (1.25 km) compared to that of the NZDSF. Therefore, the fundamental pulse repetition frequency f0 increased up to nearly 164.6 kHz. Intracavity optical losses also increased, as the intrinsic and splicing losses of the DCF were higher than those of the NZDSF. Despite these changes, the lasing dynamics with quasi-synchronous gain modulation turned out to be quite similar to those initially studied.
Similar to the previous case, approaching
f0 from above caused the initial, weakly and slowly (1/
fmod ≈ 6.1 μs) modulated CW laser radiation to gradually evolve into a sequence of narrowing pulses, as evidenced in
Figure 6a. The monotonic pulse narrowing, induced by adiabatically decreasing the modulation frequency toward
f0, did not affect the average laser power (11 mW) and ended at the frequency
flow ≈ 164.65 kHz, with a pulse duration of about 500 ns. The achieved duration was 12 times shorter than the modulation period (1/
fmod ≈ 6.1 μs), as shown in the measured time traces in
Figure 6b. Further frequency lowering transformed the single-pulse generation into pulse burst generation when the modulation frequency fell into the range of
flow ÷
f0. Below
f0, the laser returned to weakly and slowly modulated CW lasing.
Once triggered, the generation of a regular pulse train with the pulse repetition frequency corresponding to
flow ≈ 164.65 kHz remained stable for an unlimited time. As seen in
Figure 7a, the RF spectrum measured for such a pulse train has a characteristic comb-like structure with a relatively high signal-to-noise ratio, reaching nearly 75 dB in the vicinity of the pulse repetition frequency; this is similar to the RF spectrum in the previous case. The measured optical spectrum of this output pulsed radiation (
Figure 7b) has nearly the same central wavelength as in the previous case. However, the spectrum is narrower: its FWHM is about 0.7 nm. This can be explained by the fact that the lower peak power causes weaker nonlinear effects, which are responsible for spectral broadening.
In both cases, the optical spectra had a maximum intensity at wavelengths as close to the characteristic wavelength (1530 nm) of the maximum gain of erbium fibers [
19,
20] as allowed by the passbands of the intracavity elements.
4. Discussion
Thus, the quasi-synchronous gain modulation was shown to be applicable in shaping the regular trains of nanosecond laser pulses in Er fiber lasers, both with anomalous and normal intracavity dispersion. However, the achieved degree of pulse shortening with respect to the sine-wave modulation period varied depending on the intracavity conditions. We believe that the laser performance can be further optimized by adjusting these conditions, including the cavity length, loss and chromatic dispersion.
The observed evolution of lasing dynamics upon modulation frequency tuning and the resulting stationary pulse generation in both cases (anomalous and normal dispersion) correspond to the characteristic pulse-shaping mechanism via the quasi-synchronous gain modulation, which was originally revealed in Yb fiber lasers [
12,
13]. This can be concluded from a qualitative comparison of the measured temporal variations in the laser and pump powers (
Figure 3 and
Figure 6) with those in
Figure 2 and the corresponding figures in [
12,
13]. Moreover, the actual values of the modulation frequency that provided stationary generation with the minimal pulse duration in both dispersion cases satisfied the condition expressed in Equation (1). Indeed, with
fmod set to be equal to
flow (see
Figure 3 and
Figure 6), the overrating factor of the quasi-synchronous gain modulation falls into the expected characteristic range of
(see Equation (1)).
Distinguished by its remarkable simplicity (implementation requires the minimum possible set of ring laser elements), the method allowed a highly stable regular train of high-energy nanosecond pulses to be obtained. To the best of our knowledge, our method has enabled, for the first time, the stationary generation of nanosecond pulses with an energy of up to 180 nJ to be obtained at a wavelength of about 1.53 µm in Er all-fiber lasers without the employment of any active intracavity modulators or saturable absorbers (neither artificial nor material [
25,
26]). The use of these devices for high-energy pulse generation via Q-switching [
27] or mode-locking [
28] not only complicates the design of fiber lasers, but also reduces their reliability and may impose additional limits on the laser performance. The known approaches to pulse generation through intracavity instabilities in Er fiber lasers [
5,
7] do not result in the same high pulse energy and pulse train stability as our method.
Given the demonstrated efficiency of this method in Yb and Er fiber lasers, it would be desirable to explore its feasibility also in Tm-doped fiber lasers operating near 2 μm. Tm fiber lasers have been gaining extensive attention due to the optimal suitability of their wavelength and performance in a wide range of applications, including those in biomedicine and surgery, remote sensing, material processing, and spectroscopy [
29,
30]. Tm fiber lasers can be pumped by commercially available fiber-coupled diode lasers with a wavelength of about 790 nm. Therefore, the quasi-synchronous pump power modulation required for our method can be performed in a Tm fiber laser in the same simple way—by modulating the current of the pumping laser diode. At the same time, a Tm-fiber-based laser gain medium has a particular diagram of involved energy levels and some features in active ions kinetics and population inversion dynamics [
31,
32,
33], which distinguishes this medium both from Yb- and Er-fiber-based laser media. Thus, the
3F4 →
3H6 radiative transition that is used for lasing in Tm fiber lasers at a wavelength of 2 µm exhibits a two-component decay of the upper state, with characteristic times of 0.54 ms and 0.24 ms [
31], which are notably shorter than the lifetimes of the upper laser levels in Yb and Er fiber lasers. Nevertheless, such a shortened lifetime can be considered as still sufficient for our method since the interpulse periods of pump energy accumulation in the active fiber were about 0.01 ms or less in the demonstrated lasers. Energy transfer in a Tm-fiber-based gain medium also includes a cross relaxation cooperative process between adjacent Tm
3+ ions, which additionally contributes to population inversion upon pumping [
32,
33]. To determine whether this process can noticeably affect the performance of the quasi-synchronous pumping method, a separate theoretical study with numerical simulation is required, which is our upcoming research task. So far, we may only assume that the appropriate tuning of quasi-synchronous pump power modulation results in the shaping of a regular train of nanosecond high-energy pulses in a Tm fiber laser operating near 2 µm, similar to the demonstrated Yb and Er fiber lasers.
5. Conclusions
The results obtained in this study confirm that the quasi-synchronous gain modulation is a fairly universal and efficient method that enables the stationary generation of high-energy nanosecond pulses in various fiber lasers operating in different spectral ranges, without resorting to traditional fast modulation or saturable absorption techniques. We have applied this method, for the first time, in Er fiber lasers operating near 1.5 µm. We have shown that, in these lasers, this method is nearly as reliable and effective as in the previously employed Yb fiber lasers operating near 1.1 µm, despite the significant differences in the laser energy level diagrams, the lifetimes of the upper laser levels, and the characteristics of intracavity chromatic dispersion. Thus, the physical mechanism of pulse shaping through the quasi-synchronous gain modulation and asymmetrical amplification of generated waveforms was shown to be reliably reproducible in various rare-earth-doped fiber lasers with varied cavity parameters. The required pulse generation dynamics were achieved in all cases through the proper modulation settings.
Thus, the explored method facilitates the stationary generation of high-energy nanosecond pulses at different wavelengths, with different repetition rates, and additionally enables fine electronic control over the duration and peak power of the generated pulses.