1. Introduction

High power ultrashort pulse lasers gain increasing importance as tools in fields as varied as micro and nano processing, bio-medicine and metrology, to name a few. As applications drive the need for higher average power and energy per pulse, laser development is aiming to satisfy this need. Generally, power scaling is acknowledged to reach furthest with both the thin-disk and the fibre-amplifier approach which achieve remarkable levels of output power with ps or sub-ps pulses [1-5], albeit with non-trivial effort. Fibre based systems, for instance, use parabolic or chirped puls amplification (CPA) and several amplification stages employing large mode area fibres to control nonlinearities in order to achieve average powers in excess of 100W at tens of MHz [1,3]. High power femtosecond pulse generation from an oscillator directly have so far been restricted mainly to the thin disk approach [e. g. 2,4,5], with average powers >60W and pulse energies in excess of 10µJ.

In comparison, bulk laser technology has recently been used to amplify fs-pulses from the few-hundred mW level to 77W without CPA using only a single Yb:YAG slab amplifier [6]. It is predicted that this amplification scheme is potentially scalable to several hundred watts, indicating that bulk technology is by no means fundamentally limited. The classical approach of using bulk laser crystals in high power femtosecond oscillators with diffraction limited beam quality is typically constrained by various effects. Firstly, the average power which can be obtained from a typical end-pumped crystal is limited by thermo-mechanical and –optical effects (cracking, thermal lens). Secondly, the typically small gain cross sections of fs-laser materials require reasonably tight modes in the medium leading to limitations in the pulse energy due to large nonlinear phase shifts, particularly in the soliton regime, requiring extensive amounts of negative dispersion. The most promising and advanced laser materials for fs-generation in a diode-pumped oscillator available to date are Yb-doped crystals (e. g. YAG, KGW, KYW, etc.). Limitless power scaling using these materials in an end pumped oscillator configuration is additionally hampered by a range of parameters such as diode brightness, reabsorption as well as available crystal sizes and doping concentration. An example for a laser design in which these parameters have been somewhat optimised using a single laser crystal was shown in [7], where an output power of 10W was achieved using soliton mode-locking.

Power scaling using multiple laser crystals in a resonator was described by Eggleston [8]. In summary, one can overcome the thermal effects met with a single crystal by distributing the load over multiple rods in a periodic resonator structure. Additionally, multiple gain elements increase the small signal gain and hence the extraction efficiency of the resonator for standard doping levels and diode brightness. In this work we have followed this approach in a double Yb:KYW crystal setup, allowing the extraction of more than 24W cw in a diffraction limited beam using two pump diodes with up to 30W power each. Using a semiconductor saturable absorber mirror (SESAM) both soliton and positive dispersion mode-locking regimes were explored resulting in 14.6W and 17W output power respectively with pulsewidths around 450fs in both cases.

2. High power laser head, cw operation

A schematic of the laser design is shown in Fig. 1. Essentially, there is a symmetric short cavity with a length of 440mm, terminated by M5 and containing the two 5% doped, n_g-cut Yb:KYW crystals of 2mm length out of focus. An output coupler, OC 1, with a reflectivity of 90% is used in this configuration. Extensions to this resonator are trivial and are made with the purpose of including a SESAM, dispersion compensating mirrors GTI 1 and 2 for soliton mode-locking as well as further output coupling OC 2. The latter is merely an aid to achieve positive dispersion in the case of chirped pulse mode-locking. By design this component has a positive dispersion of +250fs² and a reflectivity of 95%. In this case the reflectivity of OC 1 was increased to 99%. Both Yb:KYW crystals were pumped by up to 30W at 981nm from fibre coupled laser diodes with a core size of 200µm and using an imaging ratio of 1:2. The crystals were mounted in copper holders and cooled to a temperature of 18°C using Peltier elements.

 

Fig. 1. Layout of the laser with LD 1 and 2 fiber coupled laser diodes (30W @ 981nm, 200µm core), L collimating and focussing lenses, M1 and M2 dichroic mirrors, XTAL 1 and 2 Yb:KYW crystals (l=2mm, ng-cut, 5% doped), M3 and M4 curved mirrors with R=200mm, M7 and M9 are curved mirrors with R=600mm and M8 is a flat turning mirror. M5 is the end mirror of the short cavity. GTI 1 and 2 are Gires-Tournois type dispersive mirrors with -500fs² negative dispersion. OC 1 had a reflectivity of 90% when operating the symmetric short cavity laser head on its own (M5 in place). For the soliton mode-locked long cavity OC1 had a reflectivity of 85%. In the case of pos. dispersion mode-locking OC 1 was increased to 99% whilst an additional output coupler OC 2 with 95% reflectivity and +250fs² dispersion was placed in the long cavity.

Download Full Size | PDF

The continuous wave P-curve of the short resonator is shown in Fig. 2. With a pump current of 42A (P_pump=28.6W per LD), an output power of 23W was extracted from OC 1 whilst the beam quality was still well below M² < 1.1, as measured with Coherent ModeMaster. At higher pump power (30W per LD) the M² degraded to values slightly worse but still smaller than 1.2 with >24W output power which is, to our knowledge, the highest power achieved from a Yb:KYW bulk laser to date, comparable with that of a thin disk head.

At this power the laser had an efficiency of 40% with respect to the total emitted pump power and nearly 50% with respect to the absorbed pump power. The slope efficiency of 51% is deduced from a linear fit as depicted in Fig. 2. The laser emission was polarized along the n_p-direction of the crystal indicatrix with an extinction ratio of >20dB. At the highest pump power the laser operated at a wavelength of 1044nm.

 

Fig. 2. Output power of the short resonator in CW operation as a function of the total pump power emitted by both fibre coupled pump diodes. The slope efficiency is 51%.

Download Full Size | PDF

3. Soliton mode-locking

In order to operate the laser in the soliton regime we extended the cavity length to both accommodate a SESAM and provide anomalous dispersion in the form of Gires-Tournois-type dispersive mirrors GTI 1 and 2 (confirm Fig. 1). This resonator had a length of nearly 1.9m, corresponding to a repetition rate of f_rep=79.8MHz. Furthermore, output coupling from OC 1 was set to 15% whilst OC 2 was a highly reflecting mirror. The SESAM used to start and stabilize mode-locking had a modulation depth of around 1% and there were 6 bounces on each dispersive mirror (GTI 1 and 2), resulting in a total dispersion of -12000fs² per round trip. Under these conditions the laser was mode-locked self-starting and with a single pulse per round trip for pump powers between 34W and 50W. At 50W of pump the shortest pulses were emitted at an output power of P_out=14.6W, compared to 10W achieved in [7]. At this working point the pulsewidth was measured using an intensity autocorrelator. Assuming a sech² soliton pulse shape the deconvolved pulsewidth is τ_FWHM=450fs. Figure 3 depicts the measured autocorrelation function with the inset showing the corresponding optical spectrum. With a spectral width of Δλ=2.6nm and a pulsewidth of τ_FWHM=450fs one calculates a timebandwidth product of 0.322, which is essentially at the theoretical limit.

 

Fig. 3. Intensity autocorrelation of the soliton mode-locked laser close to the double-pulsing limit (P_pump=50W, P_out=14.6W). The deconvolved pulsewidth is 450fs, assuming a sech²(t) pulse. The inset shows the associated optical power spectrum with a FWHM of 2.6nm, corresponding to a time-bandwidth product of 0.322.

Download Full Size | PDF

Although higher output power was possible, particularly in the double pulsing regime, the slightly reduced efficiency of the mode-locked laser mainly stems from the losses on the GTI-mirrors which were estimated to be around 3% per round trip. Also, in order to lower the nonlinear phase shift per round trip and to lower the intensity on the SESAM, output coupling was increased to 15% as compared with 10% in cw-mode. The operation of this laser was basically straight forward with mode-locking reliably self-starting and no degradation observable on the SESAM. In spite of operating without cover, the laser displayed RMS fluctuations of below 1% as sampled by a fast photodiode and digital oscilloscope as well as a thermal power meter.

4. Positive dispersion mode-locking

Mode-locking of cavities containing broadband laser media like Ti:sapphire or Yb-doped crystals and glasses in the presence of positive cavity dispersion has recently received increasing attention [11–17]. The basic proposition behind this scheme is to reduce the nonlinear phase incurred during fs-soliton propagation through intra-cavity elements, including air in long cavities, by lengthening the cavity puls through the action of normal dispersion. Depending on parameters such as gain bandwidth, SESAM modulation and recovery time, amount of positive dispersion and nonlinearity, the pulses from such an oscillator are typically chirped and orders of magnitude longer than a soliton of similar energy and spectral width which implies potential for energy scaling. Theoretical and experimental studies were carried out to gain a more fundamental understanding of the physics of such lasers [e. g. 15–17].

For our experiments with the double-crystal oscillator, in order to achieve stable operation, we introduced +500fs² of positive dispersion per round trip from a 5% output coupler (OC 2) used as a folding mirror (confirm Fig. 1). The positive dispersion of this component was merely a by-product of its particular design but was used here to our advantage. The GTI multi-bounce was not present in this configuration and there was an additional 1% output coupler (OC 1). We have used the same SESAM as with the soliton laser. Lacking the availability of a transmission grating with suitable line density and power handling capability at the time of experimentation, we used output 1 (approximately one tenth of the total laser power) to compress the pulses in a Tracey-type compressor made from a gold reflection grating with 1200 lines/mm. Although this is not ideal, it serves the purpose of showing the compressibility of the pulses. Highly efficient (95%) transmission gratings have become available leading to less than 20% loss in the compression of the total laser power without discernible thermal effects [e. g. 18]. Figure 4 shows the spectrum of the pulses at 17W output power (a) and the uncompressed as well as compressed pulse autocorrelation (b).

 

Fig. 4. (a) Pulse spectrum of positive dispersion mode-locked oscillator. Total output power P_out=17W before compression (P_pump=50W). (b) Corresponding autocorrelation of the uncompressed pulse with insets showing the autocorrelation of the compressed pulse. Assuming a sech²(t) pulse, the compressed pulses have a width of 470fs.

Download Full Size | PDF

The shape of the power spectrum is typical for chirped pulses and the autocorrelation before compression is around an order of magnitude longer than that of a soliton laser with similar pulse energy and spectral width. Using a grating separation of 100mm, corresponding to a calculated negative dispersion of around -0.8ps², the pulses were compressed to a minimum autocorrelation width of 740fs. Assuming a sech²(t)-pulse shape this would translate to a pulse-FWHM of 470fs. However, pedestals are to be expected on the compressed pulse, indications of which can be seen when comparing the measured autocorrelation with that of an ideal sech²(t) (confirm inset of Fig. 4b).

The pulse energy of the positive dispersion mode-locked oscillator is not substantially higher than that of the soliton regime. This is not surprising since a utilisation of the energy scaling potential of this scheme would require a lengthening of the cavity and a suitably adapted dispersion management and SESAM.

6. Conclusions

We have presented experimental results from a high power laser containing two Yb:KYW bulk crystals as gain media. Operated in cw the laser delivered up to 24W in a TEM₀₀ mode which was limited by the available pump power and which is, to our knowledge, the highest power achieved from a Yb:KYW bulk laser to date, comparable with that of a thin disk head. When the laser was mode-locked using a SESAM, it produced 14.6W and 17W in the solitonand positive-dispersion-regime respectively. The pulsewidths in both cases were around 450fs. Power scaling using another pair of crystals in a periodic continuation should be fairly straight forward and we estimate that a cw-power in excess of 50W is possible this way. Furthermore, preliminary experiments using the above laser in a regenerative amplifier configuration have shown output powers of more than 15W at 500kHz repetition rate, which will be reported elsewhere.