1. Introduction

The great quality improvement brought by the use of femtosecond lasers in the field of microprocessing of materials is now well established [1, 2]. Considering a typical averaged ablation threshold of ~100 mJ.cm^-2 in the femtosecond regime, it appeared that the use of >100 μJ-energy pulses was convenient for femtosecond micromachining. 1-kHz repetition rate 1mJ-range femtosecond laser systems based on Ti:Sapphire have then been extensively used both in laboratories and in micromachining companies. Processing durations may benefit from an increase in the repetition rate while maintaining the output energy of the laser source about constant.

In the last ten years, all-solid-state Ti:sapphire femtosecond lasers have been the subject of wide studies to produce multi-watt average power 10-kHz repetition rate pulses in the frame of applications such as x-ray or high harmonic generation. The main difficulty that has to be overcome is the thermal load of the Ti:Sapphire crystals. Indeed, increasing the repetition rate implies an increase in the average pump power. Previous studies conducted at low repetition rate [3] and for both water-cooled and cryogenically cooled Ti:Sapphire crystals [4] show that thermal lensing effects are highly damagable for the developement of high average power amplifiers. Various technical solutions have been considered. Among them, water-cooling is the simplest and has the advantage to be daily running for years in 1-kHz repetition rate sources. A first successful demonstration performed in 1993 [5] was followed in 2000 by the operation of a 10-kHz 7.2-W 30-fs laser source requiring 80 W average pump power and the combination of a regenerative and a 6-pass amplifier [6]. Cryogenic cooling is a very efficient solution that enables to increase considerably the thermal conductivity of Sapphire and to reduce the $\frac{\partial n}{\partial T}$ parameter. Consequently, the amplification geometry is made easier as large pumping diameters are allowed, leading to a high pumping power in a single amplification stage. For instance, 11 W average output power 28-fs duration pulses have been obtained using 100 W average pump power [7]. Moreover, 20-W compressed output power characterized by a beam quality parameter of M²=1.8 were obtained for 180-W pump power [8]. Nevertheless, cryogenic cooling seems not well adapted to industrial micromachining applications as it requires a large mechanical support together with an important liquid nitrogen consumption. Thermoelectric elements were also successfully implemented to cool down the Ti:sapphire crystal [9,10]. Condensation has to be avoided by placing the crystal into an evacuating chamber. Up to 8 W compressed output power were obtained from 80 W pump power [10].

We propose to use the simplest cooling method, that is to say water cooling, to develop a reliable, compact 10 kHz femtosecond laser source whose performances are suitable for industrial micromachining applications.

2. Cavity design

The main difficulty to be overpassed to design a high average-power Ti: sapphire amplifier is thermal lensing. Indeed, in a first approximation, by neglecting stress and thermal expansion effects, the focal length of the induced thermal lens in the Ti: Sapphire crystal may be expressed as :

where r is the pump radius, P_P is the pumping power, n is the refractive index of the crystal, η is wavelength-dependent and is the fraction of the pump power that is dissipated into heat. K is the heat conductivity [11]. In practice, using water-cooling, the focal length of the thermal lens becomes shorter than 10 cm at a repetition rate of 10 kHz.

In the case of V or Z-type cavities where the laser crystal is placed between two concave mirrors in the vicinity of the beam waist, such a short thermal focal length makes impossible to obtain a large mode size in the crystal and to avoid optical damage in the Pockels cell or on the end mirror. Moreover, the mode size in the crystal has to be kept small enough to reduce the dependence on the fluctuations of thermal lensing in the crystal. Consequently, this involves low average pumping power and moderate power extraction.

We developed a new design of cavity, referred to as Laser 1, that enables to overpass these limitations: the Ti:Sapphire crystal is placed in a V-type cavity at a distance of 2 mm from the end mirror whose curvature is adapted to the focal lens of the thermal lens (Fig. 1). Consequently, both mode diameter and size of the stability zone become independent on the thermal lensing. Only the relative position of the stability zone varies and is centered on the value of the thermal focal length.

 

Fig. 1. Schematic of Laser 1.

Download Full Size | PDF

Stability and sensitivity diagrams are plotted in Fig. 2 for cavity parameters detailed in Table 1. The stability curve exhibits a wide stability zone corresponding to a large mode diameter of 290 μm in the crystal. The sensitivity to a cavity misalignment is defined as the inverse of the maximum tilt angle of one cavity mirror that still maintains the laser mode within the pumping area in the crystal. The sensitivity curve keeps well below 1000 rad^-1 nearly all over the stability zone, which ensures an easy alignment of the regenerative cavity.

Table 1. cavity parameters for the design of the regenerative amplifier referred to as Laser 1. See Fig. 1 for schematic.

View Table | View all tables in this article

 

Fig. 2. Stability and sensitivity curves as a function of the inverse of the thermal focal length. The calculations were conducted with parameters of Table 1. A mode diameter of ~300 μm together with a weak sensitivity to a cavity misalignment are obtained over the major part of the stability zone.

Download Full Size | PDF

One last difficulty for the development of high average power Ti:Sapphire femtosecond sources is to get a reliable high average power high beam quality green laser. With the amplifier design depicted in section 2, we point out that a large mode size in the crystal increases drastically the tolerance on the pump laser beam quality, which makes the design of a pumping laser much easier, thus reducing the cost and improving reliability.

3. Experimental set-up

All the Ti:Sapphire crystals involved in these experiments are water-cooled. They are wrapped in an indium foil to optimize the heat contact with a copper mount.

Laser 1 is described in Fig. 3. 20-fs 4-nJ pulses at a repetition rate of 75 MHz from a self mode-locked oscillator (Femtosource) are stretched to 250 ps in an Öffner stretcher using a 1200 lines/mm reflection grating. 3,1-mJ 40-ns pulses at a repetition rate of 10 kHz from an intracavity frequency doubled M²=20 diode-pumped Q-switched Nd:YAG laser (Thalés Laser) are focused down to 2w~350 μm inside a 10×10×15 mm³ Ti:sapphire crystal. The cavity design corresponds to the one detailed in section 2. Optical switching for regenerative amplification is ensured by a water-cooled KD^*P Pockels cell and two dielectric polarizers with reverse wedge. Two 1500 lines/mm reflection gratings are used in the compressor. The beam diameter on the gratings is 1 cm and the overall compressor efficiency reaches 70 %.

 

Fig. 3. Experimental set-up of Laser 1: RR: retroreflector; Ti:Sa: Ti:Sapphire crystal (10×10×15 mm³); G1: 1200 lines/mm grating; G2: 1500 lines/mm grating; M1: flat dichroic mirror; M2: R=8 cm convex mirror; M3: R=50 cm concave mirror; M4: flat mirror; M5: R=275 mm convex mirror; M6: R=550 mm concave mirror; L1: f’=20 cm converging lens; L2: f’=50 cm converging lens; L3: f’=15 cm converging lens; L4: f’=-10 cm diverging lens; L5: f’=50 cm converging lens; TFP: thin film polarizer; KD^*P: KD^*P Pockels cell; FR: Faraday rotator; P: photodiode monitoring.

Download Full Size | PDF

As a comparison, we developed another amplification geometry, referred to as Laser 2, also using water cooling, and based on commonly used laser architectures. Indeed, Laser 2 involves a regenerative followed by a 2-pass amplifier (Fig. 4). The regenerative amplifier consists in a conventional “V-type” cavity in which the brewster-cut Ti:Sapphire crystal (ϕ6×20 mm) is placed between 2 concave mirrors (M2 and M3) in the vicinity of the beam waist. In this geometry, a small mode size (2w~200 μm) is required in the crystal as detailed in section 2. Consequently, the pump power has to be reduced down to 15 W. Moreover, the position of the crystal together with the beam overlap inside the crystal are still critical. Optical switching for regenerative amplication is the same for Laser 1 and Laser 2. 19 W pumping power are then focused down to 2w~330 μm inside a second water cooled Ti:Sapphire crystal (15×15×10 mm³). A second pass in this amplifier is ensured after reflection on a R=5 cm convex mirror whose curvature compensates for converging thermal lensing in the crystal.

 

Fig. 4. Amplification set-up of Laser 2. Ti:Sa: Ti:Sapphire crystal; M1: flat dichroic mirror; M2: R=10 cm concave mirror; M3: R=50 cm concave mirror; M4: flat mirror; M5: R=5 cm convex mirror; TFP: thin film polarizer; KD^*P: KD^*P Pockels cell; FR: Faraday rotator; L1: f’=40 cm converging lens; L2: f’=28.6 cm converging lens; L3: f’=100 cm converging lens; L4: f’=25 cm converging lens; P: photodiode monitoring.

Download Full Size | PDF

4. Results

Laser 1 delivers up to 5 W after amplification and 3.5 W after compression at a repetition rate of 10 kHz, which is, to our knowledge, the highest value ever obtained in a reliable, compact and water-cooled geometry. This value is to be compared to the 3.5 W output power delivered by Laser 2 after amplification (table 2), which keeps close to those described in reference [5] for comparable pumping power. Indeed, the pumping values indicated in Table 1 are measured at the output of the laser, and not on the Ti:Sa crystals. The amplification geometry used in Laser 1 appears to be highly efficient, despite water cooling.

Table 2. Characteristics of the two amplification geometriesreferred to as Laser 1 and Laser 2.

View Table | View all tables in this article

A shot-to-shot rms stability of 1.5% limited by the stability of the pump is obtained after amplification in Laser 1 at 5 W average output power.

Pulse duration, deduced from an autocorrelation trace (Fig. 5), is 60 fs and is still above the Fourier limit corresponding to the spectral width of 20 nm measured after regenerative amplification. This spectral width value results from a spectral narrowing in the regenerative amplifier. Indeed, the 60 nm-wide amplifier input spectrum is affected by the spectral transmission of the amplifier, which is measured to be 35 nm-wide in free-running condition. Anyway, a pulse duration of 60 fs is very well adapted to micromachining applications, as the ultra-short laser-material interaction regime is typically reached below a few hundreds of femtoseconds [1].

 

Fig. 5. Autocorrelation trace of compressed pulses of Laser 1, corresponding to a 60-fs pulse width.

Download Full Size | PDF

 

Fig. 6. Measurement of the M² parameter of Laser 1 output beam running at 3.5 W output power after compression. Beam parameters of M_x ²=1.12 and M_Y ²=1.18 are deduced respectively in both transverse directions X and Y, typical for a high quality beam.

Download Full Size | PDF

The pulse contrast reaches 3000:1. The beam quality is estimated by a measurement of the M² parameter. For doing this, a f=40 cm converging lens is used and a CCD camera is translated in the focal region. The corresponding beam size measurements are plotted in Fig. 6. Beam parameters of M_x ²=1.12 and M_Y ²=1.18 are deduced respectively in both transverse directions X and Y, typical for a high quality beam.

We also mention that Laser 1 is still able to deliver 4.6 W before compression when pumped by a 35-W average power M²=30 green laser, which highlights its ability to be pumped with low beam quality lasers.

 

Fig. 7. Optical microscope photographs of grooves machined in steel by 10 scans with a 60-fs pulse duration and a-b): 1.68J/cm², 1.75mm/s, 1kHz or c-d): 1.84 J/cm², 17.5mm/s, 10kHz. ac): surface of the sample and b-d): bottom of the groove.

Download Full Size | PDF

The ability of this compact water-cooled 10 kHz femtosecond laser source to perform micromachining has been evaluated by machining grooves on steel samples Fe/Cr18/Ni10/Mo3). The samples are translated in front of the focal spot at a translation speed v and two repetition rates are used: v=17.5 mm.s^-1 for 10 kHz and v=1,75 mm.s^-1 for 1 kHz using an electronic Pockels-clock divider. All other experimental parameters are the same: pulse duration of 60 fs, beam parameter M²=1.15, focusing lens of f=25 cm focal length, laser fluence of ~1,7 Jcm^-2 and 10 scans. Optical microscope images in Fig. 7 show the same machining quality. Only the groove width slightly differs, which is due to a small variation in the focal spot diameter and in the laser fluence between 1 and 10 kHz running conditions. A groove width of 85 μm and 100 μm is obtained at 1 and 10 kHz respectively for both repetition rates with similar depth of around 11.5 μm and the processing time is reduced by a factor of 10 when using this compact and reliable 10 kHz source compared to standard 1 kHz femtosecond lasers. Consequently, this source is suitable for high speed industrial femtosecond laser processing applications.

A detailed study of femtosecond processes in the 10-kHz range is currently under progress.

5. Conclusion

We demonstrated a compact reliable water-cooled femtosecond all-solid-state laser running at a repetition rate of 10 kHz. Average power of 3.5 W is obtained after compression using a moderate pumping laser delivering 31 W. The amplification geometry is highly efficient despite water-cooling, and the relatively large mode size in the Ti:Sapphire crystal enables to increase the tolerance on the pump laser beam quality. The pulse duration is 60 fs and the beam is nearly diffraction limited (M²<1.2). We showed that this amplification geometry is more compact and more efficient than conventional designs involving “V or Z-type” architectures followed by multi-pass amplification. The linear lowering of the time process with repetition rate from 1 up to 10 kHz has been demonstrated, opening the way to reliable high speed industrial femtosecond micromachining.