High-power ultrafast thin-disk multipass amplifiers for efficient laser-based manufacturing

3.1 Amplification of picosecond pulses

Figure 5 shows the 2-stage thin-disk multipass amplifier scheme that was used for the experiments reported in [15]. A slightly modified TruMicro 5050 delivering a linearly polarized beam with 6.5 ps long pulses (FWHM) at an average power of 105 W and a repetition rate of 300 kHz was used as a seed laser. The beam quality factor M ² of this seed beam was measured to be below 1.2. Thin disks of Yb:YAG with a radius of curvature of 20 m, a doping concentration of 10–11%, and thicknesses of approx. 170 and 120 µm were used for the first and the second multipass amplifier, respectively. Single-pass configuration (see Figure 1a) with an array of 60 individually adjustable mirrors was employed in both amplifiers. This allows for a total of 30 reflections of the beam at the HR coated rear side of the thin-disk laser crystal on its path through each of the amplifiers. The thin-disk laser crystals were both pumped at a wavelength of 969 nm i.e. on the zero-phonon line of Yb:YAG. An average output power of 435 W was attained after the first amplifier stage. The overall performance of the two-stage amplifier system is shown in Figure 6a. A maximum output power of 2050 W, corresponding to an extraction efficiency of 44% was achieved. At a repetition rate of 300 kHz, this corresponds to a pulse energy of 6.8 mJ.

Figure 5:

Experimental setup composed of two cascaded thin-disk multipass amplifiers. A slightly modified TruMicro5050 laser delivering 6.5 ps long pulses with an average power of 105 W at a repetition rate of 300 kHz is used as the seed laser.

Figure 6:

(a) Power and extraction efficiency at the exit of the second thin-disk multipass amplifier and (b) caustic in the horizontal and vertical planes and selected intensity profiles of the amplified beam at 2050 W of output power.

The beam quality factor M ² of the beam exiting the second amplifier stage was measured to be 1.57 and 1.36 in the horizontal and vertical plane, respectively. The measured beam caustic is shown in Figure 6b. Using an auto-correlator the pulse duration at the maximum output power was measured to be 7.7 ps (FWHM) leading to a peak power of approx. 815 MW.

3.2 Amplification of femtosecond pulses

A modified Tangor HP laser from the company Amplitude was used as a seed laser for the experiments presented in [16] and in the following. It delivers up to 127 W of average output power in the first diffraction order of the internal acousto-optic modulator (AOM) at a pulse repetition rate ranging from 40 MHz down to 500 kHz. The beam propagation factor M ² amounts to 1.25 and the pulse duration was measured to be 475 fs (FWHM). The Tangor allowed for an operation in burst mode.

To allow pulse on demand and fast modulation of the pulse train on the workpiece, polarization multiplexing was implemented (see Figure 7) by using both beams emitted from the 0th and first diffraction order of the internal AOM as seed beams. The half-wave plate introduced in the beam path of the 0th order beam allowed the rotation of its polarization by 90° with respect to that of the beam from the first diffraction order. Both beams are then recombined and injected collinearly into the TDMPA using a TFP. Beam shaping optics (spherical mirrors) are used to adapt the properties (beam diameter and divergence) of the seed beams (0th and first orders) to the diameter of the pump spot on the amplifying disk. An Yb:YAG thin-disk crystal with a doping concentration of 10–11% and a thickness of approx. 120–130 µm was used for the present experiment. The disk with a diameter of 17 mm was glued on a diamond heat sink and mounted inside a pumping unit allowing up to 24 passes of the pump radiation through the disk. Zero-phonon pumping of the Yb:YAG disk at 969 nm was implemented using a 2 kW spectrally stabilized (with a volume Bragg grating) laser diode. The radius of curvature of the disk was set to 20 m by design (see Section 2). The seed beam was geometrically folded 40 times over the Yb:YAG laser disk by means of an array of 80 adjustable mirrors. A TFP placed at the exit of the TDMPA allowed the separation of the beams originating from the 0th and first diffraction order of the seed laser. The beam from the 0th order is directed to a beam dump whereas the beam from the first order is directed to the workpiece and used for processing. The motivation to feed both beams through the amplifier is to ensure constant amplification properties by avoiding any impact on the gain when modulating the pulse sequences used for materials processing.

Figure 7:

Setup of the TDMPA delivering sub-ps pulses at an average power of >1 kW and enabling modulation of the pulse train. The blue and red colors indicate the paths of the beams emitted from the 0th and the first diffraction port of the seed laser. AOM, acousto-optic modulator; Comp, pulse compressor; HWP, half-wave plate; TFP, thin-film polarizer; RMP, retro-reflecting mirror pair.

The measured output power and extraction efficiency of this laser and amplifier system are given in Figure 8a. An output power of 1058 W was achieved with an optical efficiency (seed power being subtracted) of 46%, thus corresponding to an energy per pulse of 2 mJ at a repetition rate of 500 kHz. The measured beam caustic is shown in Figure 8b. The beam propagation factor M ² was measured to be 1.48 and 1.33 in the horizontal and the vertical plane, respectively. The pulse duration was measured to be 825 fs, thus corresponding to a peak power exceeding 2.2 GW.

Figure 8:

(a) Measured power and extraction efficiency of the TDMPA at a repetition rate of 500 kHz and (b) measured caustic of the emitted beam at 1059 W of average output power.

No significant difference in the extracted output power was observed when operating the system at different repetition rates or changing between single-pulse or multi-pulse burst modes. However, as shown in Figure 9, spectral broadening was observed for the single-pulse mode. This indicates the potential to compress the pulses down to a pulse duration of <100 fs. Further exploitation of this effect will be the subject of future investigations.

Figure 9:

Measured spectra of the seed and the amplified beam.

To be able to use the developed laser system for material processing, its power and beam pointing stability were important criteria and were given particular attention during the development and implementation of the laser system. An excellent performance was achieved thanks to advanced designs of the optomechanical components used in the TDMPA. Temperature-controlled aluminum cast breadboards together with highly stable mirror holders and a water-cooled mirror array were custom-designed and fully qualified before their final integration. By this, a highly stable system was achieved as shown by the time-dependent power measurement illustrated in Figure 10 at a power of ∼1 kW. The measurements lasting over several hours reveal a relative standard and peak to valley deviation of 0.63 and 6%, respectively. Furthermore, the beam propagation factor M ² was measured at the beginning and the end of the experiment. It remained unchanged over the whole measurement time. The angular beam pointing variation was measured to be <30 µrad (rms) and <100 µrad (peak to valley) in both the planes of the major and minor axes of the measured deviation ellipses.

Figure 10:

Long-term measurement of the power stability of the emitted laser beam at an average power of ∼ 1 kW.

3.3 Second and third harmonic generation

Ultrafast lasers emitting in the visible (green) and the ultraviolet spectral range offer great opportunities for many applications where the wavelength plays an important role in the light–matter interaction. The possibility to fabricate smaller structures, as well as the enhanced linear- and nonlinear absorption, can be beneficial for material processing applications such as surface structuring, drilling, micro-welding, and processing of transparent and ceramic materials [17], [18], [19], [20]. The present section is therefore devoted to our latest achievements of high powers at the wavelengths of 515 and 343 nm that were obtained by second and third harmonic generation of the infrared (IR) beams emitted from the TDMPA described above. Lithium-Triborate (LBO) crystals were used for this purpose. The obtained results are summarized in Table 1 [12], [13], [14], [15], [16]. The THG setup is depicted in Figure 11. For the SHG the second nonlinear crystal was removed and the filtering optics was adapted to separate the IR from the green beam. Furthermore, the beam shaping optics was also adapted for optimum SHG conversion.

Figure 11:

Schematic of the THG setup.

4.1 High-quality and high-throughput laser milling of silicon and metals

Burst operation was used for the experiments presented in the following. The energy of the burst was evenly divided over five pulses within the burst (5 PPB). The pulses within the burst were separated by 22.7 ns. The focal diameter on the workpiece was measured [22] to be 90 ± 5 µm.

The benefit of processing silicon with bursts [23] was confirmed in a first experiment conducted with five pulses per burst at a low average power of 57 W and an energy per burst of 114 µJ, corresponding to a peak fluence of 0.7 J/cm². The beam was moved over the sample’s surface with a scanning speed of 10 m/s, which corresponds to a pulse overlap of 78%. The line hatching distance was set to 20 µm. In a further experiment, 950 W (corresponding to an energy per burst of 1900 µJ) was used on the workpiece. In order to maintain a low peak fluence of 0.7 J/cm², the beam diameter on the surface of the sample was set at 372 µm by shifting the focus position to 17 mm below the sample surface. Furthermore, the scanning speed was increased from 10 m/s to 24 m/s to maintain a moderate pulse overlap of 87% and avoid excessive melt and crack formation [24]. The corresponding surface structures as measured by scanning electron microscopy (SEM) for moderate and high average power are shown in Figure 12a and b), respectively.

Figure 12:

SEM images of the laser-milled surface at (a) moderate average power of 57 W and (b) high average power of 950 W d _f  = beam diameter n the workpiece, v _s  = scanning speed.

As shown from the SEM analysis of the samples obtained in both experiments, a similar surface quality covered with nanoparticles and LIPSS is obtained. The measured roughness S _a is low for moderate and high average power with 0.5 and 0.4 µm, respectively. With the increased average power of 950 W, a significantly higher material removal rate of 216 mm³/min was achieved when compared to the 11 mm³/min obtained at an average power of 57 W.

The high-throughput processing was demonstrated, at an even higher power of 1010 W by milling of chamfer geometry on the workpiece corresponding to a peak fluence of 0.74 J/cm². Figure 13 shows the topography of the processed sample as measured with a laser scanning microscope (LSM, Olympus, OLS4000). The measurement reveals a low roughness value of S _a  = 0.4 µm in the upper zone of the chamfered sample (region I) and S _a  = 0.6 µm in the lower zone (region II).

Figure 13:

LSM measurement of a chamfer milled into a silicon wafer. Process parameters: laser wavelength = 1030 nm, power = 1010 W, inter-burst repetition rate = 500 kHz, energy per burst = 2020 µJ, number of pulses per burst = 5, peak fluence = 0.74 J/cm² [21].

The flatness of the tilted surface was investigated by evaluating the peak-to-valley height after filtering the LSM measurement with a cut-off wavelength of 250 µm. The tilted surface is smooth with a low peak-to-valley height of 5.7 µm over the large area of 3.5 × 3.5 mm². Although the ablated depth per scan is about 3 µm, no steps were detected along the offset parallel processed lines. The energy-specific ablated volume of 3.8 µm³/µJ during laser milling with the maximum power of 1010 W corresponds to a high material removal rate of 230 mm³/min. To the best of our knowledge, this is the highest material removal rate reported so far for laser milling of silicon with ultrafast lasers, while at the same time achieving high surface quality with S _a  ≤ 0.6 µm and no surface defects with diameters exceeding 1 µm. In comparison to previously reported results for laser milling of silicon achieving low surface roughness, this is 740 times higher removal rate than that reported in ref. [24] and a 443 times higher material removal rate than that reported in ref. [25]. By this one can conclude that with the combination of high average power and sub-ps pulses with an appropriate process strategy (burst mode, low peak fluence, high feed rate, etc. …) one can achieve a significant increase in laser milling throughput of silicon while maintaining high surface quality.

The potential of high throughput and simultaneously high quality in milling with 1 kW ultrafast lasers is not limited to silicon with the developed processing strategy described above. Application on widely used metals such as stainless steel, copper, and aluminum (see Figure 14) was conducted and showed comparable results to silicon with a fine ripple structure on the surface and roughness S _a in the range of 1 μm at record-breaking removal rates of 126, 174 and 180 mm³/min for copper, aluminum, and stainless steel, respectively.

Figure 14:

High quality and high throughputs laser milling of (a) stainless steel, (b) copper, and (c) aluminum using ultrashort laser pulses and 1.01 kW of average power.

Additionally, micrometer size ripple structures were generated on the surface by the laser milling process. These ripples allow for potential applications to adjust the structural color, the wetting behavior, cell growth, or friction behavior of the processed surface [19]

4.2 High quality and high throughput laser milling of diamond

The same laser and processing machine (described in Section 4.1) were also used for the ablation of PCD. The usable power on the workpiece was slightly lower (990 W instead of 1010 W) due to a slight misalignment of the TDMPA. These investigations are reported in detail in ref. [26]. Bursts containing three pulses were used for the experiments which showed an increase of the ablation rate by a factor of 140 as compared to traditional mechanical processes and by up to 70 times when compared to the use of state of art 100 W-class fs lasers. A 14% reduction of the initial surface roughness S _a was also achieved by using the 3-pulses burst and can be further improved by the use of higher numbers of pulses per burst (>5). Last but not least no phase transformation, as revealed by Raman spectroscopy analysis of the processed samples, was observed when processing at ∼1 kW of average power [26].