We report on a diode-pumped double-slab Yb:KYW regenerative amplifier with combined gain spectra and intracavity dispersion compensation. At repetition rates > 20 kHz an average power of up to 5 W after compression was generated, resulting in a maximum pulse energy of 250 µJ. Pulse durations of around 190 fs were achieved by using a GRISM compressor with a transmission efficiency of 90%.
©2010 Optical Society of America
High energy femtosecond laser systems are of interest for material processing applications, especially for ultra precise micro- and nanomachining of sensitive and transparent materials. Their ultra short pulse duration allows for cold ablation processes with reduced thermal load even compared to picosecond lasers, thus being applicable to a larger variety of materials [1,2]. The majority of femtosecond lasers used so far is represented by high energy Ti:sapphire-based chirped pulse amplifier (CPA) systems which are commercially available for more than 10 years . However, up to now they have not been adopted to industrial applications due to their rather complex and cost intensive structure in combination with high maintenance and operation costs.
For applications requiring a high average power and a comparably low pulse energy, high repetition rate fiber-based CPA femtosecond lasers can be used. They can be easily scaled in average power but are still limited to pulse energies of a few µJ at short pulse durations of around 200 fs. However, material processing applications often require pulse energies of some tens up to some hundreds of µJ. Recently, great progress has been made in the development of high energy directly diode-pumped femtosecond amplifier systems. They offer advantages like higher repetition rates, minimized system complexity and costs, significantly increased wall-plug efficiency and reliability. Typical diode-pumped amplifier systems based on Ytterbium-doped monoclinic double tungstates, such as Yb:KYW and Yb:KGW, generate pulses with energies in the desired regime and durations of around 300-400 fs [4,5]. However, most of these systems do not support significantly shorter pulse durations due to the limited gain bandwidth of the active medium. Furthermore, they typically show a tendency to longer pulse durations at higher pulse energies resulting from the effect of gain-narrowing during amplification . Shorter pulses can only be realized in combination with different spectral broadening techniques like pulse shaping or spectral filtering , which are often connected with additional losses or an increased system complexity. Another concept is nonlinear pulse compression which implies a limited variability of the operation parameters of the laser system along with some additional compromise regarding residual distortion of the spectral shape and phase characteristics .
A promising method to reduce gain narrowing and to enhance the effective gain bandwidth is to combine gain media with separated gain maxima and overlapping broadband gain, which has been realized in oscillators first , and which we have recently demonstrated with a thin-disk based regenerative amplifier .
In this paper we present the transfer of this approach to a double-slab regenerative amplifier. As laser medium we used Ytterbium doped potassium yttrium monoclinic double tungstate oxide, Yb:KY(WO4)2 or short Yb:KYW . Owing to its crystal structure this material exhibits three optical axes each showing different optical properties. The optical axes Nm and Np are the ones with the highest maximum emission cross-sections in the wavelength region around 1030 nm. Their fluorescence peaks are separated by more than 10 nm but the bandwidth of both of them is greater than 20 nm. These features make it possible to combine the gain of both optical axes to achieve a broad gain spectrum supporting pulse durations of below 200 fs.
In our regenerative amplifier presented within this paper the gain combination was realized by incorporating two Yb:KYW crystals with an appropriate orientation into the resonator. In contrast to a setup with a single pump source and a fixed combination ratio of Nm- and Np-gain as defined by the number of passes per crystal axis, this approach offers the possibility to adjust the ratio of Nm- and Np-gain by means of the applied pump power on each of the crystals. Furthermore, we used a fiber-based wavelength-tunable seed front-end and a fiber-based pulse stretcher to address the full bandwidth of the regenerative amplifier and to minimize system complexity. A highly efficient grating-prism (GRISM) compressor allowed for higher order dispersion compensation . Additionally, the intracavity dispersion has been fully compensated to overcome the need for readjustment of the compressor when amplifier parameters are changed.
2. Experimental setup
Our laser system consisted of a fiber-based femtosecond seed front-end, a double-slab regenerative amplifier with combined gain spectra and a GRISM compressor. The Yb-based front-end that was composed of a mode-locked fiber oscillator, a fiber-based pulse stretcher with about 100 m of standard single mode fiber and a pre-amplifier, provided seed pulses with an energy of 8 nJ at a repetition rate of 28.4 MHz. The seed spectrum was centered at a wavelength of 1033 nm and had a full width at half maximum (FWHM) of 11.8 nm. At 1% of the peak intensity the spectrum covered a range from 1021 nm to 1045 nm which was sufficient for seeding both gain spectra of the regenerative amplifier. The seed pulses had a chirped duration of 75 ps ± 5 ps as measured with a high-speed photodetector with 12 ps FWHM impulse response. The chirped pulses were compressed with our GRISM compressor, which allowed for simultaneous compensation of second and third order dispersion and showed a transmission efficiency of 90%. A standard intensity autocorrelator was used in the background free mode of operation for temporal characterization of the ultra short pulses. Both optical spectrum and intensity autocorrelation function (ACF) of the compressed seed pulses are shown in Fig. 1 .
Compression of the chirped seed pulses led to an intensity autocorrelation function with a FWHM of 229 fs. To deduct the compressed pulse duration from the measured autocorrelation function the appropriate deconvolution factor was calculated from the optical spectrum in the conventional way via a Fourier transformation (FT) assuming a zero phase. It was determined as the ratio between the FWHM of the calculated autocorrelation function of the FT and the FWHM of the FT itself. For all the following experiments presented within this paper the widths of the measured autocorrelation functions and the corresponding calculated pulse durations will be given. Of course, this temporal characterization of the pulses does not imply a precision to a single femtosecond due to the limited precision of the instrument and due to the unknown phase. Since the deviation of the measured autocorrelation width from the calculated autocorrelation width of the Fourier-transform-limited (FTL) pulse was generally well below 15%, the assumption that has been made for the calculation of the pulse duration is considered to be justifiable.
In case of the seed pulses the corresponding deconvolution factor of 1.37 led to a duration of the compressed pulses of around 170 fs. The Fourier-transform limit, as calculated from the optical spectrum, was 154 fs. The deviation of the calculated pulse duration compared to the FTL pulse duration of about 10% can be attributed to nonlinear phase contributions from the front-end that could not be compensated with this compressor.
The setup of the double-slab regenerative amplifier with combined gain spectra is shown as a schematic in Fig. 2 .
The resonator contained two Ng-cut Yb:KYW crystals with a doping concentration of 5% that were 1.65 mm long, 5 mm wide and 1 mm high with facets parallel to the Nm-Np plane. The crystal facets were anti-reflection coated for both pump and laser wavelength and formed a wedge of 1.15° in order to prevent etalon effects. Two different types of crystals were used, one with Nm and one with Np along the 5 mm dimension of the crystal. Owing to the intracavity thin film polarizer the resonator mode was polarized horizontally along the Np-axis of the first and the Nm-axis of the second crystal in order to combine their different gain spectra within the amplifier. The pump light at a wavelength of 981 nm was provided by two fiber coupled diode lasers with a maximum CW output power of 30 W each. Short pump fibers (95 mm length, 200 µm core diameter, NA = 0.22) were used to minimize depolarization effects in order to efficiently address the Nm-axis of each Yb:KYW crystal for maximum absorption. The pump beams were collimated and focused through dichroic mirrors to a pump spot diameter of around 420 µm in both cases. In combination with the crystal design this setup resulted in an absorption of approximately 75-85% of the launched pump power so that sufficient bleaching of the gain media was ensured, thus minimizing reabsorption processes during amplification. The difference of the maximum emission cross sections of the optical axes Nm and Np could be compensated by an appropriate pump power ratio in order to achieve an amplified pulse spectrum with approximately equal contributions from both crystals at high pulse energies. Therefore the lower maximum emission cross-section of Np required a higher pump power compared to the Nm-crystal. However, it has to be taken into account that the optimum combination ratio also depends on other factors like intra-cavity losses and the number of round-trips within the amplifier.
An optical switch composed of a thin film polarizer (TFP), a quarter-wave plate and a BBO Pockels cell was used to inject the seed pulses into the regenerative amplifier and to eject the amplified pulses. The cavity was designed for a round-trip time longer than the rise time of the Pockels cell and for a beam diameter of around 1.5 mm inside the Pockels cell to prevent the BBO crystal from damages and to minimize nonlinearities. All cavity mirrors were highly reflecting for the laser wavelength of around 1030 nm and designed for near zero dispersion. The amplified output pulses were separated from the input beam via a Faraday isolator and guided to the GRISM compressor to be dechirped to ultra short pulse durations.
3. Experimental results
The regenerative amplifier was operated at a total pump power of 45 W, whereof 27.5 W were launched at the Np-crystal and 17.5 W were launched at the Nm-crystal. This combination was chosen in order to provide a broad amplified spectrum supporting pulse durations of < 200 fs over a wide range of output powers. A further increase in pump power for the Nm-crystal led to a slight reduction of average power at the point of maximum energy extraction which may be attributed to the influence of thermal lensing. However, no limitation due to bifurcation  was observed.
Owing to the double-slab concept the amplification with Nm and Np occurred more independently from each other as compared to our thin-disk setup based on a single gain medium . In combination with the different maximum emission cross-sections of the two optical axes this resulted in a characteristic spectral evolution as a function of the number of round-trips within the amplifier. This evolution was investigated at a repetition rate of 100 kHz and the corresponding results are shown in Fig. 3 . As the accumulated dispersion increases with the number of resonator round-trips the GRISM compressor was readjusted for the shortest pulse duration at each state of operation. Pulse energy and output spectra were measured after compression.
At a low number of resonator round-trips the optical spectrum was dominated by the Nm-gain at a center wavelength of 1030 nm resulting from the higher initial gain (Fig. 3a). For a higher number of round-trips the Np-gain with the lower emission cross-section increased. After 30 round-trips a state of operation was reached where the optical spectrum showed a characteristic “M-shape” with two equal maxima at about 1030 nm and 1038 nm and a spectral FWHM of 12 nm (Fig. 3b). Further increase of the number of round-trips led to an optical spectrum that was dominated by the Np-peak (Fig. 3c). In spite of this spectral evolution the width of the measured autocorrelation function did not vary significantly (256 fs to 279 fs). Therefore the calculated pulse duration was always below 200 fs. With a compressor efficiency of 90% the maximum average power after compression was 5.0 W corresponding to a pulse energy of 50 µJ at a repetition rate of 100 kHz.
The characteristic spectral evolution as described above does not show any significant advantages or disadvantages over the thin-disk setup that we have previously presented. However, the spectral shape also couples to the temporal shape of the uncompressed pulses as a result of the temporal distribution of wavelengths within the chirped pulses. In case of the amplification to high pulse energies it might be more favorable to operate the amplifier close to the “symmetric” spectrum. This would lead to a lower peak intensity of the uncompressed pulses inside the amplifier as compared to a state of operation with a highly asymmetric spectrum at the end of the amplification process.
In order to overcome the need for readjustment of the compressor after a variation of the amplifier parameters, several of the initially applied conventional highly reflecting cavity mirrors were replaced by dispersive mirrors to compensate for the intracavity dispersion. Final dispersion adjustment was performed by using two Quartz plates resulting in a total additional dispersion of around −4000 fs2 per round-trip. With this combination an optimized compensation of the intracavity dispersion (Pockels cell, laser crystals, TFP) was achieved. To demonstrate the benefit of a zero dispersion cavity the amplifier was operated at a repetition rate of 20 kHz and the compressor was optimized only once for the shortest pulse duration at a pulse energy of 120 µJ (24 round-trips).
The output pulse energy was scaled over a wide range by a variation of the number of resonator round-trips which resulted in a spectral evolution that was equivalent to Fig. 3a-3c. The corresponding pulse energy of the compressed pulses is shown in Fig. 4a . The width of the autocorrelation trace, which was measured without readjusting the compressor, the calculated pulse duration and the corresponding Fourier-transform limit are shown in Fig. 4b.
The pulse energy increased almost linearly up to a value of 194 µJ (30 round-trips) and then converged to the maximum value of 250 µJ (42 round-trips). Owing to an optimum compensation of the intracavity dispersion the pulse duration was nearly constant even for significant changes of the output pulse energy from 12 µJ to 250 µJ and without any readjustment of the compressor. A slight increase of the FTL pulse duration, as well as the width of the measured autocorrelation trace of the compressed pulses, indicates a minor impact of residual gain-narrowing. The autocorrelation width was in the range of 249 fs to 283 fs which corresponded to a calculated pulse duration of around 190 fs, being 9% above the FTL pulse duration on average. This deviation may be attributed to nonlinear phase contributions from the front-end that could not be compensated with this compressor.
For different repetition rates from 20 kHz to 120 kHz the maximum pulse energy and the corresponding characteristics of the compressed pulses were measured. In Fig. 5a the maximum energy of the compressed pulses versus the repetition rate is shown. Figure 5b shows the corresponding autocorrelation width, the calculated pulse duration and the Fourier-transform limit of the compressed pulses. Within this range of repetition rates the maximum average output power nearly stayed constant at a value of 5.0 W corresponding to a pulse energy of 42 µJ at a repetition rate of 120 kHz and a pulse energy of 250 µJ at a repetition rate of 20 kHz. The autocorrelation width did not change significantly and was in the range of 277 fs to 285 fs which corresponded to a calculated pulse duration of around 195 fs. The Fourier-transform limit was 189 fs on average. These results show that with this setup a pulse duration of slightly below 200 fs can be achieved up to the maximum output power of 5.0 W over a wide range of repetition rates.
Figure 6 shows the optical spectrum (a) and the corresponding intensity autocorrelation function (ACF) of the compressed pulses together with the calculated FTL ACF (b) at the maximum pulse energy of 250 µJ at a repetition rate of 20 kHz.
At the maximum pulse energy, which corresponds to 42 round-trips within the amplifier, the spectrum was dominated by the Np-peak at a wavelength of 1038 nm. The FWHM of the measured autocorrelation function was 282 fs. This corresponded to a calculated pulse duration of 196 fs which was 9% above the calculated FTL pulse duration.
4. Conclusion and outlook
A diode-pumped double-slab Yb:KYW regenerative amplifier with combined gain spectra has been demonstrated. The combined gain allowed for the amplification of sub 200 fs pulses while significantly reducing the impact of gain-narrowing. At the optimum combination ratio of both gain contributions a characteristic “M-shaped” output spectrum with a spectral FWHM of about 12 nm was achieved.
To address the full bandwidth of the regenerative amplifier and to achieve a minimum system complexity a fiber-based seed front-end and a fiber-based pulse stretcher were used. A GRISM compressor with a transmission efficiency of 90% allowed for simultaneous compensation of second and third order dispersion. Additionally, intracavity dispersion compensation was applied to the regenerative amplifier in order to overcome the need for readjustment of the compressor when amplifier parameters are changed. This led to a nearly constant pulse duration with a maximum variation of around 10% even for significant changes of repetition rate and pulse energy.
At a repetition rate of 120 kHz an output pulse energy of 42 μJ corresponding to an average power of 5.0 W was achieved after compression. With a deconvolution factor of 1.43, as calculated from the pulse spectrum, the dechirped pulse duration was 197 fs. Reducing the repetition rate to 20 kHz led to an output pulse energy of 250 μJ after compression. The corresponding calculated pulse duration was 196 fs.
Significantly higher pulse energies should be feasible at lower repetition rates. However, scaling up the pulse energy would require further stretching of the seed pulses to durations well above 100 ps to reduce nonlinearities and to operate well below the damage threshold of the intracavity components. This will be subject of further investigations.
In comparison to recent publications with a similar double-slab setup but without combined gain spectra  we achieved a comparably low efficiency concerning the launched pump power and the resulting output power. We believe that this may be attributed to a relatively low-quality surface polish of the laser crystals on the one hand, which was identified by a microscopic analysis of surface damages after preliminary experiments. Furthermore, the average power of the system might be limited by thermal lensing. This effect can most likely be reduced by an improved resonator design similar to that in  and an optimization of the laser crystals regarding their doping concentration and length. Thus, we are confident to be able to achieve output powers in the range of 10 W and pulse energies well above 250 µJ with a similar setup by replacing the laser crystals and by increasing the chirped pulse duration of the seed input.
The main advantage of the double-slab concept over the thin-disk setup that we have presented previously is a reduced system cost resulting from the less complex pump configuration and the lower production cost for the active media. Up to now the availability of appropriate Yb:KYW thin-disks is still poor. Due to the anisotropy of this material it is very difficult to manufacture high-quality thin-disks with low curvature and low astigmatism which is the prerequisite to achieve a performance as presented in , thus resulting in a low production yield. However, thin-disk based setups are still unrivaled in terms of power-scalability.
This work was funded by the German Federal Ministry of Education and Research within the FULMINA Project under contract no. 13N8722.
References and links
1. A. Ostendorf, G. Kamlage, U. Klug, F. Korte, and B. N. Chichkov, “Femtosecond versus picosecond laser ablation,” Proc. SPIE 5713, 1–8 (2005). [CrossRef]
2. P. Engelhart, S. Hermann, T. Neubert, H. Plagwitz, R. Grischke, R. Meyer, U. Klug, A. Schoonderbeek, U. Stute, and R. Brendel, “Laser Ablation of SiO2 for Locally Contacted Si Solar Cells With Ultra-short Pulses,” Prog. Photovolt. Res. Appl. 15(6), 521–527 (2007). [CrossRef]
3. S. Backus, C. G. Durfee, M. M. Murnane, and H. C. Kapteyn, “High power ultrafast lasers,” Rev. Sci. Instrum. 69(3), 1207–1223 (1998). [CrossRef]
4. A. Beyertt, D. Nickel, and A. Giesen, “Femtosecond thin-disk Yb:KYW regenerative amplifier,” Appl. Phys. B 00, 1–6 (2005).
5. H. H. Liu, J. Nees, and G. Mourou, “Directly diode-pumped Yb:KY(WO(4))(2) regenerative amplifiers,” Opt. Lett. 27(9), 722–724 (2002). [CrossRef]
6. P. Raybaut, F. Balembois, F. Druon, and P. Georges, “Numerical and Experimental Study of Gain Narrowing in Ytterbium-Based Regenerative Amplifiers,” IEEE J. Quantum Electron. 41(3), 415–425 (2005). [CrossRef]
7. C. P. J. Barty, G. Korn, F. Raksi, C. Rose-Petruck, J. Squier, A.-C. Tien, K. R. Wilson, V. V. Yakovlev, and K. Yamakawa, “Regenerative pulse shaping and amplification of ultrabroadband optical pulses,” Opt. Lett. 21(3), 219–221 (1996). [CrossRef] [PubMed]
8. M. Larionov and A. Giesen, “50-kHz, 400-μJ, sub-100-fs pulses from a thin disk laser amplifier,” Proc. SPIE 7193, 1–8 (2009).
9. S. Han, W. Lu, B. Y. Sheh, L. Yan, M. Wraback, H. Shen, J. Pamulapati, and P. G. Newman, “Generation of sub-40 fs pulses from a mode-locked dual-gain-medis Nd:glass laser,” Appl. Phys. B 74(9), s177–s179 (2002). [CrossRef]
10. U. Buenting, H. Sayinc, D. Wandt, U. Morgner, and D. Kracht, “Regenerative thin disk amplifier with combined gain spectra producing 500 microJ sub 200 fs pulses,” Opt. Express 17(10), 8046–8050 (2009). [CrossRef] [PubMed]
11. M. C. Pujol, M. A. Bursukova, F. Güell, X. Mateos, R. Solé, J. Gavaldà, M. Aguiló, J. Massons, F. Díaz, P. Klopp, U. Griebner, and V. Petrov, “Growth, optical characterization, and laser operation of a stoichiometric crystal KYb(WO4)2,” Phys. Rev. B 65(16), 165121 (2002). [CrossRef]
12. S. Kane and J. Squier, “Grism-pair stretcher–compressor system for simultaneous second- and third-order dispersion compensation in chirped-pulse amplification,” J. Opt. Soc. Am. B 14(3), 661–665 (1997). [CrossRef]
14. A.-L. Calendron, K. S. Wentsch, J. Meier, and M. J. Lederer, “High power and high energy Yb:KYW regenerative amplifier using a chirped volume Bragg grating,” in The Conference on Lasers and Electro-Optics (CLEO)/The International Quantum Electronics Conference (IQEC) (Optical Society of America, Washington, DC, 2009), CFD2.