1. Introduction

A variety of applications of laser oscillators providing ultra-short microjoule pulses at MHz repetition rates have led to the progress of different oscillator concepts [1, 2, 3]. Especially applications such as laser micromachining, nonlinear spectroscopy, or laser surgery profit from these sources [4, 5, 6]. Recently, we demonstrated a solitary mode-locked Yb:KY(WO₄)₂ (Yb:KYW) thin disk oscillator with cavity dumping generating 3 µJ femtosecond pulses at a 1 MHz repetition rate [1, 7]. Concerning energy scalability this system is currently limited by the B-integral from several microjoules of intra-cavity pulse energy, where the contributions of both the resonators ambient air (or gas, i.e. He) and the electro-optical-modulator crystal dominate and limit the solitary dispersion management for the current system. In order to scale down the nonlinearities and to reduce the total required negative group-delay-dispersion (GDD) for even higher pulse energies it is necessary to either significantly reduce the nonlinear propagation length or to operate the oscillator with less peak power. The latter approach can be realized with a positive dispersion concept. Beside various fiber approaches such as [8, 9] this method has already been successfully demonstrated with Ti:sapphire Kerr-lens mode-locked oscillators [3, 10, 11] (‘chirped-pulse oscillator’), where the heavily chirped pulses from the oscillator are externally compressed below fifty femtoseconds. Moreover, theoretical investigations have been carried out to gain more knowledge about the fundamental physics of such lasers [12, 13]. In this paper we report on what is to our knowledge the first passively mode-locked Ytterbium-based oscillator with cavity dumping and positive dispersion. The material properties of Yb:KYW such as the low quantum defect (in our case λ_pump/λ_laser≈0.96) along with its high thermal conductivity compared with Yb:glass, as well as its broad band emission cross section compared with Yb:YAG make this material particularly suitable for the generation of femtosecond high energy pulses. Among these characteristics it is commercially available and ideal for direct diode pumping. Surprisingly we found that the spectral bandwidth of the positively chirped pulses is even broader than in the solitary regime [14] (unlike Ti:sapphire systems [10]), and femtosecond pulses are obtained after external compression.

2. Experimental set-up

As in the solitary set-up from [14] a 1 mm long Yb:KYW-crystal is used as active gain medium. The crystal was used in a n_g-cut geometry and is doped with 5 % Ytterbium. The laser is polarized in parallel to the N_m-axis in order to ensure the highest emission cross section at a center wavelength around 1030 nm. Figure 1 depicts the experimental set-up. The active medium is end-pumped at 980 nm by an internally collimated multi-emitter bar which is capable of delivering slightly more than 30 W of output power. The laser cavity is designed with an effective laser mode radius of approximately 125 µm along with the pumping radius beeing slightly larger for transversal single-mode operation. For pulsed operation we stretched the resonator to a total length of 8.64 m (repetition rate of 17.35 MHz) with a Herriott-type multipass cell [15] to make the configuration more compact. The entire footprint is about 90 cm x 50 cm. The oscillator is sealed by a housing which is purged with dry air to prevent the hygroscopic BBO-crystals from degradation. The distance between the Herriott-cells curved mirrors (radius of curvature of 1 and 2 m) equals approximately 760 mm with four reflections on each mirror and a total beam path of around 6 m. The Pockels-cell consists of two 18 mm long BBO-crystals. The high voltage controller applies a high voltage gate to the BBO crystals each eighteenth round trip (resulting in 1 MHz pulse repetition rate), and the output coupling ratio that emerges at the thin film polarizer (TFP) can be varied by the voltage amplitude. The high-voltage switch of the Pockels-cells was limited to a maximum dumping frequency of 1 MHz for the required voltage amplitude. The saturable absorber mirror at the end of the linear cavity is responsible for the passive mode-locking. For stable mode-locking operation it was necessary to focus tightly on the absorber. The fluences on the absorber exceed the range of 6 mJ/cm ² at a maximum intra cavity pulse energy of up to 4.7 µJ. At this power level the SAM requires water cooling. By means of a pump-probe scheme we measured a relaxation time for the SAM of a few picoseconds. We approximated a total material dispersion of +4250 fs² per round-trip for all participating optics. From here, the magnitude of the net intracavity dispersion could be easily reduced by type and number of negative dispersive mirrors which were located between the pockels-cell and the SAM. For this purpose two different types of GTI-mirrors (Gires-Tournois-mirrors) were used with a GDD of - 250 fs² and - 500 fs² respectively.

 

Fig. 1. Schematic of the laser set-up, L1: focussing lens (25 mm), M1: HR 1030 nm/AR 980 nm, M2: representative for dispersive mirrors of different numbers and type, EOM: electro-optical-modulator (BBO-Pockels-cell), TFP: thin film polarizer, SAM: saturable absorber mirror, Herriott-type cell: 4 reflections per mirror (not shown).

Download Full Size | PDF

3. Results and discussion

With the SAM the laser was operated in stable cw-mode-locking regime. The process was self-starting, and without any dumping the intra-cavity pulse energy was in the region of 3.5 µJ at an absorbed pump power of approximately 5.3 W. To sustain stability it was necessary to raise the pump power simultaneously while increasing the dumping ratio to compensate for the dumping losses. This was not necessary in the case of the solitary oscillator [14] which was less sensitive. We observed stable mode-locking with dumping ratios up to 45 % with an absorbed pump power of around 8 W. At that point the maximum output energy was beyond 2 µJ which is a factor of 1.4 higher compared to the similare solitary version. We operated the oscillator in four different dispersion configurations, namely 250 fs², 750 fs², 1250 fs² and 2250 fs² for the total GDD per round trip. The resulting power spectra are compared in Fig. 2 on a logarithmic scale. The rectangular shape is very typical for chirped-pulse oscillators [3, 10, 11]. Starting from a large dispersion of 2250 fs², the spectral bandwidth becomes significantly broader with decreasing net-dispersion, while the available output power remained constant for all four different dispersion values. The solitary oscillator provided bandwidths of only 3.3 nm, here we observed the broadest bandwidth (FWHM) of 8 nm at 250 fs². With further decreased GDD no stable pulsing was achieveable. To verify single pulse operation the pulse train from the oscillator was observed with an RF-spectrum analyzer (see Fig. 3) as well as with a long range (130 ps) autocorrelator. With a dispersion of +250 fs² the duration of the chirped output pulses was measured to be in the range of 3 ps (4.4 ps autocorrelation width). We measured a rmsnoise below 1 % and a beam parameter M² of less than 1.1. The contrast ratio between the dumped pulses and the background exceeded 500:1.

 

Fig. 2. Optical power spectrum of the output pulses for the different dispersion regimes.

Download Full Size | PDF

 

Fig. 3. The RF-spectrum of the laser oscillator reveals stable pulsing.

Download Full Size | PDF

For external pulse compression the laser operated at +250 fs² where the broadest spectrum enables the shortest available pulse duration. The given spectral bandwidth below 10 nm yields a required compressor dispersion of several ten thousand fs². Thus a grating compressor was prefered to a prism configuration mainly because of the by far shorter apex distance. Another conceivable alternative would be a grism compressor as described in [16]. In this case however two fused silica transmission gratings with 1250 lines per millimeter were employed (see e.g. [17]). Since the optimum apex distance between both gratings was calculated to be too short (about 3 cm) for our set-up, we inserted two cylindrical lenses each with a focal length of 300 mm to enable a larger separation. The losses of the compressor configuration are approximately 50 %. They derive from both the gratings (20 %) and the uncoated lenses. The compressed pulses were analyzed with a noncollinear intensity autocorrelator based on second harmonic generation in a 1 mm thin BBO crystal at type I critical phase matching. Calculating the pulse duration directly from the power spectrum we obtain a Fourier limit of 360 fs (490 fs autocorrelation width). The measured autocorrelation signal is shown in Fig. 4 and reveals a FWHM of 570 fs resulting in a pulse duration of approximately 420 fs. We attribute the residual chirp to misalignments of the rather complex compression set-up.

 

Fig. 4. Background free intensity autocorrelation trace after compression, shortest measured FWHM-value of 570 fs.

Download Full Size | PDF

4. Conclusion

In conclusion, we demonstrate a passively mode-locked Yb:KYW oscillator with cavity dumping operated in the positive dispersion regime. The output energy exceeded 2 µJ at 1 MHz repetition rate, and we observed a spectral bandwidth as high as 8 nm with compressed pulse durations in the 400 fs range. On first sight, these strong results are counterbalanced by the requirement of a grating compressor. Nevertheless, in direct comparison to the solitary oscillator we were able to demonstrate a significant enhancement of the pulse energy due to the chirped pulse operation. This study paves a possible way for scaling up the energies of directly diode pumped solid-state laser oscillators. Using this laser as a seed source for a post-amplification stage we will profit both from the high seed energy as well as from the large bandwidth.

This work was funded by the Bundesministerium für Bildung and Forschung (BMBF) under contract 13N8723.