1. Introduction

Ytterbium doped fiber lasers operating near a wavelength of 1µm have a broad gain spectrum and high optical-to-optical conversion efficiency that make them attractive for ultrashort pulse generation [1, 2]. The development of the double clad fiber design has lead to an increase in average output power of fiber lasers compared to the use of conventional fiber and the single spatial mode kilowatt average power barrier has already been reached [3, 4]. The realization of high power ultrashort pulse fiber lasers is an area of great interest as well. High power ultrashort pulse oscillator - amplifier chains have been developed [5–7], but an important research question that remains is whether high power/high energy pulses can be obtained directly from a fiber oscillator. This question has been partly addressed by our work. The challenge in such research is that the propagation of ultrashort pulses with high peak power in the fiber cavity is accompanied by nonlinear effects which are responsible for pulse distortion.

The frequency shifted feedback (FSF) modelocking technique was utilized to generate the ultrashort pulses in our laser. The technique itself has previously been studied in detail [8–10] and is an interesting alternative to the use of a semiconductor saturable absorber medium (SESAM) or nonlinear polarization rotation (NPR) as modelocking mechanisms. The FSF technique is especially easy to implement as it does not require any special tuning like matching of the pulse energy saturation for the SESAM-element. It is also polarization insensitive, self-starting and therefore environmentally stable [8–10]. The FSF technique also has the potential to be used for the development of all-fiber single polarization modelocked sources if an all-fiber frequency shifter is incorporated into the laser cavity [11, 12]. The trade-off for these benefits is the fact that a RF source is needed to drive the modelocking element. The big advantage though is that we can obtain controllable Q-switched modelocking (besides normal CW modelocking) by modulating the power output of the RF source, so that the oscillator fiber is effectively used as an amplifier as well.

In previously reported research the FSF modelocked lasers have not been able to generate pulses shorter than a few picoseconds, the average power of FSF lasers did not exceed a few mW and pulse energies were of the order of pJ [13–15]. In this paper we present important new technological results. We use a double clad fiber and much higher pump power than in previous reports. The methodology used is quite similar to that in reference 13, where timebandwidth limited 5 ps pulses were generated. The latter pulses could not be compressed because of their limited spectral width of only ca. 0.3 nm. In contrast to this result we generated linearly chirped pulses with much higher energies in this work, which had a bandwidth of 2.4 nm, and which could compressed to subpicosecond durations.

Although femtosecond fiber lasers have been demonstrated using the FSF technique, the acousto-optic modulator only ensured the self-starting operation of the modelocked regime while the NPR technique was responsible for the pulse formation [16, 17]. Modelocked pulse formation in a FSF laser can be attributed to the interplay between the frequency shifter and the effects of nonlinear self phase modulation (SPM) in the presence of a bandpass filter. The bandpass filter does not have to be an intracavity filter, but can even just be a result of the gain dispersion. An increasing amount of SPM results in the formation of a wider spectrum and in a possible decreased pulse width [9].

The process of pulse generation in a FSF laser has been investigated theoretically in several publications [9, 18, 19]. De Sterke and Steel developed a simplified analytical model of a FSF laser in pulsed operation [20], which predicts that the pulse width t_p has inverse square root dependence on the nonlinear phase shift Φ_NL, i.e. t_p∝(Φ_NL ω ² _f)^-1/2. ω_f is the combined filter bandwidth of the cavity elements. The maximum nonlinear phase shift is defined as Φ_NL=n ₂ k ₀ LI with the nonlinear refractive index n ₂, propagation constant k ₀, the length of the nonlinear medium L and the peak intensity I. Since the peak intensity is the ratio of peak power P_p over the irradiated area π r ² in the nonlinear medium, one obtains that t_p∝r×(n ₂ k ₀ LP _pω ² _f)^-1/2. In a medium with given n ₂ and k ₀, the pulse width can therefore be decreased by either increasing the length of the nonlinear medium and the peak power of the pulses or decreasing the beam radius in the medium.

Rare-earth doped optical fiber lasers offer a small core size as well as considerably longer cavities compared with conventional bulk resonators. With the development of the double-clad design the average power and consequently the achievable peak power of modelocked pulses can be increased by several orders of magnitude compared to traditional core-pumped fiber lasers. Rare-earth doped double-clad fiber is thus the ideal medium to realize an ultrashort pulse modelocked FSF laser. This approach was followed in this work [21, 22].

In this letter self-starting sub-picosecond pulse modelocked operation of an Yb-doped fiber laser using only the FSF technique is demonstrated. Pulse energies up to 120 nJ and peak powers exceeding 40 kW were obtained at a repetition rate of 5.2 MHz in a simple linear cavity design. The laser is tunable from 1074 to 1094 nm, with average output power of up to 870 mW.

2. Experimental set-up

 

Fig. 1. Experimental setup; DM: dichroic mirror, HR @ 1060 nm, HT @ 915 nm; DCF: double-clad fiber ; AOM: acoustic-optic modulator

Download Full Size | PDF

Figure 1 shows the experimental setup of the cavity. A 5W diode with an emission wavelength of 915 nm is used to pump ~20 m of double-clad single-mode Ytterbium-doped fiber with 8° angle cleaved ends. The fiber has a core N.A. of 0.16, a core diameter of 4µm and the beam quality factor was measured with M²<1.1. At the operation wavelengths the fiber has normal dispersion and the dispersion coefficient in the cavity is estimated to -41 ps/nm km. The negative dispersion coefficient prohibits the formation of solitons, which severely limit the achievable peak power.

The laser has an external cavity, which contains an acousto-optic modulator (AOM) excited with a radio frequency (RF) source with a fixed operating frequency of 40 MHz. The measured deflection efficiency of the AOM was up to 40%. The total frequency shift is 80 MHz per roundtrip, because the AOM is used with a linear cavity configuration. The deflected and up-shifted beam is subsequently reflected by a gold coated grating (1200 grooves/mm) used in Littrow configuration, in which the first diffraction order is diffracted back towards the incident direction. The other end of the cavity is formed by a 45° deflecting dichroic mirror and a mirror with 80 % reflection efficiency at the signal wavelength. The output was extracted either directly after the 80% mirror on the launch side of the cavity (Port 1), the non deflected beam of the AOM (Port 2) or the zero diffraction order of the grating (Port 3).

3. Ultrashort pulse modelocked regime

It was found that the Littrow grating configuration was critical for stability: the absence of this intracavity filter lead to multi-pulsing. The grating feedback regime is very robust and the laser can be operated for hours without observing any degradation of the pulsed regime characteristics. Modelocking was obtained for pump powers ranging from 1 W up to the maximum available pump power of 5 W. The pulse repetition rate of 5.2 MHz corresponds to the inverse of the cavity roundtrip time. For pump powers below 1 W, the laser operates in continuous wave (CW) mode. For lower diffraction efficiency of the AOM (20% less than maximum RF power applied), the modelocking becomes less stable, until modelocking stops and the laser operates in CW mode.

 

Fig. 2. Output spectra in modelocked operation available on port 2. The laser is continuously tunable from 1074–1094 nm (a). The spectrum assumes its maximum width of 2.4 nm at a center wavelength of 1078.5 nm (b).

Download Full Size | PDF

The laser’s peak wavelength could be continuously tuned from 1074 nm–1094 nm by rotating the grating around its vertical axis. Fig. 2 shows different examples of output pulse spectra available on port 2. The width of the spectrum changes over the tuning range: while its full width at half maximum (FWHM) is about 1 nm towards the extremes of the range, it broadens to more than 2 nm in the center and assumes its maximum width of 2.4 nm at 1078.5 nm.

The corresponding interferometric autocorrelation for a laser wavelength of 1078.5 nm is presented in Fig. 3(a). The pulses have a large positive linear temporal chirp, which results in a pedestal in the interferometric autocorrelation trace. Using the theoretical expression for the envelope function of the autocorrelation trace of a Gaussian pulse, the pulse duration is determined to ~8 ps. The maximum average output power on a single port (port 2) is 640 mW which corresponds to a maximum pulse energy of 120 nJ.

 

Fig. 3. Interferometric autocorrelation traces. (a): The pulses available directly from the oscillator show a large linear chirp which results in a pedestal in the autocorrelation trace. Fitting with the theoretical envelope function for a Gaussian pulse gives a pulse width of ~8 ps. (b): With a double pass on an external grating the pulses could be compressed to 855 fs (FWHM).

Download Full Size | PDF

We use a double pass external grating to compress the pulses by elimination of the linear temporal chirp. The autocorrelation trace of the compressed pulses is shown in Fig. 3(b). The data is fitted best with a Gaussian pulse with a FWHM of 855 fs. The average power after compression is about 200 mW, which corresponds to an estimated peak power of 40 kW. The time bandwidth product of the compressed pulses is 0.56, which most likely indicates residual nonlinear chirp.

The AOM used in this setup has a preferred polarization state for highest deflection efficiency which is linear and vertical to its base. Therefore the deflected and frequency up-shifted first order beam, which is fed back into the doped fiber, is predominantly vertically polarized. Measurements of the polarization state indicate that in this way polarization stability is obtained on each output port. The laser output is elliptically polarized, with stable vertical polarization component of 93% on port 1.

4. Q-Switched modelocked regime

 

Fig. 4. Temporal evolution of the pulse train (black) and the modulation voltage (red) applied to the AOM in the simultaneously modelocked and Q-switched operation mode. For comparison a part of the corresponding pulse train in the unmodulated modelocked regime (with maximum and constant RF power applied to the AOM) is depicted as a green trace on the same scale (with an offset to enhance clarity). The Q-switched modelocking clearly greatly enhances the maximum peak power. The highest pulse amplitude in the simultaneously Q-switched/modelocked regime corresponds to a pulse energy of ~1µJ.

Download Full Size | PDF

In the previous section we demonstrated that the laser is modelocked when the AOM is driven with a constant RF power and is used as a frequency shifter. By modulating the AOM deflection efficiency the modelocked pulse train is also modulated and the AOM can then be used as a frequency shifter and a modulator simultaneously to induce a controlled Q-switched modelocked operation [13]. The energy per pulse can be significantly increased without using any extra internal or external elements by inducing the Q-switched modelocking. Fig. 5 shows the modulation signal and a close-up of the pulses obtained with this Q-switched modelocked regime. The delay of 1.5 µs between switching to higher modulation voltage and the first visible pulses is due the build up and decay times respectively of the acoustic waves in the AOM. By modulating the AOM deflection efficiency at a repetition rate of 160 kHz we observed that the energy per pulse was about 20 times higher in the Q-Switched configuration than with the unmodulated modelocked regime. This particular modulation frequency gave the highest stable energy output. This phenomenon has not been investigated further yet and will be the subject of further research. The highest pulse amplitude in Fig. 5 corresponds to an energy of ~1µJ directly available from port 2. The pulse energy was measured by using a slow photodetector, which generates voltages directly proportional to the energy in short optical pulses. The pulse energy of the non Q-switched pulse trains could easily be extracted by dividing the average power by the pulse repetition rate, after we made sure that no CW background was present. An oscilloscope trace of the Q-switched modelocking (at the same average output power) yields voltage pulse amplitudes which are roughly 20 times higher than without Q-switching, which means 20 times higher single pulse energy.

5. Conclusion

We have demonstrated a high power and wavelength tunable ytterbium doped fiber pulse laser and have shown that the FSF technique is sufficient by itself to produce subpicosecond pulses. The laser works in both modelocked and Q-switched regime. Average output powers as high as 870 mW, pulse energies up to 1µJ and non Q-switched peak powers up to 40 kW after compression have been achieved while the layout of the cavity is quite simple. All elements of the cavity can be integrated into fiber using fiber Bragg gratings and a fiber integrated AOM, resulting in a highly compact and easy maintainable laser. This makes this specific modelocked fiber laser especially interesting for biomedical or micromachining applications where system simplicity is crucial. Due to the high pulse energy directly available from the oscillator a simple 2-pass bulk amplifier with 20dB gain can be used for amplification to the 100 µJ level which is considered the threshold for micromachining.

New higher power pump diodes with enhanced brightness could increase the energy levels even further and possibly result in even shorter pulses. Intracavity dispersion management can be used to change pulse characteristics and operate the laser in the higher harmonics soliton regime, or even shorter stretched or parabolic pulse operating regions.