Advances in High-Power, Ultrashort Pulse DPSSL Technologies at HiLASE

Abstract

The development of kW-class diode-pumped picosecond laser sources emitting at various wavelengths started at the HiLASE Center four years ago. A 500-W Perla C thin-disk laser with a diffraction limited beam and repetition rate of 50–100 kHz, a frequency conversion to mid-infrared (mid-IR), and second to fifth harmonic frequencies was demonstrated. We present an updated review on the progress in the development of compact picosecond and femtosecond high average power radiation sources covering the ultraviolet (UV) to mid-IR spectral range at the HiLASE Center. We also report on thin-disk manufacturing by atomic diffusion bonding, which is a crucial technology for future high-power laser development.

1. Introduction

Compact, kW-class, near-infrared pulsed lasers are becoming highly important for both science and industry. In particular, picosecond and sub-picosecond lasers have turned out to be indispensable tools in laser material processing. They are beneficial primarily for their highly accurate and special processing of many materials, such as the micro-patterning of large surfaces [1,2,3], or the processing of carbon-reinforced plastics [4]. Specifically, Yb:YAG lasers have lately reached a magic value of an average output power of 1 kW while their pulse energy exceeded 1 J [5], both in burst mode [6] and continuous pulsing [7,8], although often with a pulse length exceeding 10 ps.

Rapid progress is also expected in mid-infrared and UV spectral regions. The growth of large volume conversion crystals and the advent of new kinds of nonlinear crystals enables efficient conversion of high average power and high energy picosecond near-infrared (near-IR) pulses with a wavelength of 1 µm to deep ultraviolet (UV) [9,10] or mid-infrared (mid-IR) spectral regions [11,12]. This broadens the application potential of near-infrared ultrashort-pulse lasers in high-tech applications. For example, high-power picosecond laser pulses of fundamental and harmonic (up to fifth) frequencies will enable high-speed, high-precision micro- and nano-structuring of large surface areas on various materials with flat-top beams using diffraction optics and interference techniques [13]. Ultrashort laser pulses in the mid-IR spectral range can also open ways for 3D direct laser writing inside biological samples, polymer materials, and narrow-bandgap semiconductors which are opaque for visible and near-IR laser light [14]. Furthermore, tunable spectral sources from near-IR to mid-IR at enhanced laser power can improve the sensitivity of infrared spectroscopy in applications for monitoring trace gases in atmosphere and biomedicine [15]. Besides industrial applications, near-IR picosecond pulses are also essential, for example, in THz waves generation [16,17], attosecond science [18], or femtosecond optical parametric chirped pulse amplification (OPCPA) pumping [19].

Near-IR thin-disk lasers also proved their ability to generate high average power short pulses, which are suitable for laser produced plasma-based (LPP) [20], free electron laser-based (FEL) [21,22], or inverse Compton scattering-based (ICS) [23] extreme UV radiation sources. It helps, for example, to construct new stations for semiconductor lithography at 13.5 nm or 6.7 nm and to support evolution in electronics. An important requirement for application in lithography is stability of all laser parameters and compactness in the actual application field. Intensive study has been performed at the HiLASE Center to confirm the high beam quality thin-disk laser with good pointing stability [24,25]. Higher conversion efficiency from pump pulse to amplified pulse energy is the key to realizing these characteristics. Mid-IR, compact, high average power diode-pumped sources based on Ho-doped materials also have potential to directly pump powerful CO₂ lasers, which would make the CO₂ lasers more compact for lithographic stations [26].

In this paper, we report on progress in the development of ultrashort pulsed laser technology and laser components development since 2015, when a first review paper was published [27]. The technology of Yb:YAG thin-disk lasers Perla A, B, and C emitting in near-infrared ranges has matured, and goals have been updated to finally exceed 1 kW of average power and 1 J of pulse energy in following years (Figure 1). Nowadays, up to 500 W of average power can be generated. Together with the frequency conversion of these lasers and the construction of a new mid-IR disk laser beamline, the spectral range from UV (200 nm) to mid-IR (10 µm) in picosecond and sub-picosecond temporal range should be covered (Figure 1). This progress, in development since 2015, is summarized in

Table 1. Comparison of progress in typical average output power, pulse energy, pulse repetition rate, pulse length (full width at half maximum (FWHM)), and beam quality of Perla thin-disk laser platforms achieved since 2015 [27].

Year	Perla A	Perla B	Perla C	Perla D
2015	Not available	45 mJ; 1 kHz; 45 W	1 mJ; 1 kHz; 90 W; 1.9 ps	Not available
2017	100 mJ; 10 Hz proof-of-principle experiments, under design	45 mJ; 1 kHz; 45 W; <1 ps, M2 < 1.3, Also new: 50 W; 10 kHz burst	9 mJ; 50 kHz; 450 W; or 4.5 mJ; 100 kHz; 450 W; 1.3 ps, M2 < 1.5–2; RMS stability < 1.3%	Under design

and

Table 2. Comparison of progress in typical average output power, pulse repetition rate, and conversion efficiency of frequency up- and down-conversion stages pumped by the picosecond Perla C thin-disk laser platform since 2015 [27]. Mid-IR OPG: optical parametric generation producing mid-infrared pulses; Mid-IR OPA: optical parametric amplification amplifying mid-infrared pulses.

Year	Second Harmonic 515 nm	Fourth Harmonic 257.5 nm	Fifth Harmonic 205 nm	Mid-IR OPG + OPA 1.5–3 µm
2015	11 W; 100 kHz; 40% conversion efficiency	2 W; 100 kHz; 18% conversion efficiency	Not available	Under design
2017	40 W; 100 kHz; 56% conversion efficiency	6 W; 100 kHz; 18% conversion efficiency	0.8 W; 100 kHz; 20% conversion efficiency	1.7–2.6 µm tunability (both signal + idller waves); 8.5 W signal (<2 µm); 5 W idler (>2 µm); 77 kHz

.

1.1. Perla Picosecond Thin-Disk Laser Platform

The development of compact thin-disk-based kW-class thin-disk lasers started at HiLASE several years ago. In-house developed thin-disk platforms [28] at HiLASE recently received the name Perla. Several thin-disk Perla platforms with various pulse parameters exist (Figure 1). The most progressive is the continued construction of the high repetition rate platform Perla C, which is important for the majority of industrial picosecond laser applications including applications in lithography. The Perla C system based on a regenerative amplifier aims for 1 kW of average power at a variable repetition rate 50–1000 kHz [29,30]. Nowadays, the system can be operated up to 500 W.

The design of the Perla C is based on a common CPA (Chirped Pulse Amplification) method. A low pulse energy (nJ) Yb-doped fiber oscillator producing several picosecond-long chirped pulses with a 20-nm bandwidth (full width at half maximum (FWHM)) is followed by fiber preamplifiers, two stage thin-disk regenerative amplifiers, and a highly dispersive chirped volume Bragg grating-based (CVBG) pulse compressor (chirp rate 220 ps nm⁻¹, full bandwidth 4 nm). The design specification of this system is targeted to 5 mJ of pulse energy at 100 kHz repetition rate; however, upon requirement, the system was modified for 50 kHz operation up to 500 W of average power. Prospectively, the adjustability of the repetition rate from 50 kHz to 1 MHz at 0.5 kW of average power should be enabled by the installation of a new high-voltage Pockels cell driver. The research trial also aimed to upgrade the system to the level of 1-kW of average power by keeping the repetition rate unchanged, i.e., 100 kHz. Thin disks used in the regenerative amplifiers are diamond-bonded Yb:YAG single crystals (220 µm thick, 7 at. % doping) continuously pumped at the zero-phonon-line wavelength (968.8 nm) [24,25].

The optical setup of the first-stage regenerative amplifier [31] includes a standing-wave cavity, thin-disk laser head with a pump spot diameter of 2.8 mm, a 500 W continuous wave (CW) pump diode laser, and a dual beta barium borate (BBO) crystal (each 8 × 8 × 25 mm³) and a Pockels cell to enable the fast switching on and off of the system. The maximum output pulse energy from this first regenerative amplifier before compression is 1.2 mJ (average output power of 120 W) at a pump power of 430 W in a nearly diffraction-limited beam (M² ≈ 1.4). Optical-to-optical efficiency therefore reaches 29%. A maximum efficiency of 33% is reached at 1 mJ pulse energy, i.e., 100 W of average power. Compressed pulses were obtained with the CVBG compressor (net-efficiency 85%) and the duration of 1.3 ps.

The second-stage amplifier is seeded by a fraction of the uncompressed output of the first stage (≈20 W, 200 µJ). The thin disk is pumped by a CW, 968.8-nm fiber-coupled diode laser on a pump spot with a diameter of 5.2 mm. The disk is located in a ring cavity with two V-passes through the disk per one roundtrip, and the footprint area of the amplifier is only 100 × 60 cm². A dual Pockels cell equipped by BBO crystals with dimensions of 10 × 10 × 25 mm³ for each of the crystals placed in in-house developed holders is operated at a half-wave voltage of 10-kV and a repetition rate of 50 or 100 kHz. A maximum CW output power of 565 W was obtained at 1.21 kW pump power, and the optical-to-optical efficiency was 47%. In seeded operation with the input pulse energy of 0.2 mJ, obtained pulses were 4.5 mJ at 100 kHz (450 W average output power) with an extraction efficiency of 43%, as shown in Figure 2a. Pulses compressed by a CVBG compressor with a net-efficiency of approximately 85% have a 1.4-nm bandwidth (FWHM) and a pulse duration of 1.8 ps (FWHM). In the 50-kHz, regime we generated 9 mJ pulses (450 W, uncompressed) with 1130 W of pump power, i.e., optical-to-optical efficiency reached an excellent value of almost 40%. The repetition rate of the laser system is planned to be tunable from 50 kHz to 1 MHz after completing an appropriate Pockels cell driver. In the 50-kHz regime, a several-hours-long stable operation was demonstrated with an average output power of 320 W (6.4 mJ) in fundamental transverse mode, as shown in Figure 2b. The laser was operated for 10 h at 330 W with a power fluctuation root mean square as low as 1.2% (root mean square value (RMS)). Because of rising compressor depolarization losses at full output power, dispersion detuning, and beam deformation caused by high CVBG temperature, the optimization of the pulse compression and output pulse shape is ongoing. The performance of the Perla C laser system is studied in the entire range of its repetition rates and optimized pulse compression, and a concept for a 1-kW, sub-picosecond upgrade is included in the present research subjects.

1.2. Perla A and Perla B Platforms

Both the Perla A and Perla B platforms are CPA systems, which aim to high energy pulses. The Perla B was redesigned [32] in order to provide a burst of pulses and better fit experimental requirements. Intra-burst repetition rate was set to 10 kHz, and a burst repetition rate of 100 Hz with optional switching back to original configuration operating continuously at a repetition rate of 1 kHz was described in detail in References [24,27]. In the burst mode, each 1-ms long pulse burst consists of ten 1–2-ps long pulses. The pulses are generated from an Yb-doped fiber oscillator (nJ pulse energy), stretched to 1.5 ns by a chirped-fiber Bragg grating stretcher, and preamplified in a system of fiber preamplifiers to approximately 10 µJ. A 100-µJ pulse burst is after the preamplification seeded to a thin disk-based regenerative amplifier and amplified to 10 mJ per pulse, i.e., 100 mJ per burst. The amplifier is quasi-continuous wave (QCW) pumped (1 ms pulse length) by fiber-coupled laser diodes at zero-phonon line wavelength (968.8 nm), and is switched on and off by a pair of 10 × 10 × 22 mm³ BBO crystals. The output beam is nearly diffraction-limited, and the M² parameter measured for continuous pulsing at 1 kHz was better than 1.3.

Finally, the amplified pulses will be boosted up to 600 mJ per burst in a 10-pass thin disk-based multipass amplifier. The amplifier is pumped at 940 nm by a fiber-coupled laser diode module (2 kW). Both the amplification stages employ a diamond-bonded Yb:YAG thin-disk with a thickness of 220 µm, 7 at. % doping concentration, and an 8-mm clear aperture; however, the first regenerative amplifier uses only 4 mm of the diameter of the aperture. Prospectively, the output could be boosted by adding a booster amplifier equipped by large size thin disks with an energy extraction of 1.5 J or 3 J per burst, depending on the disk size and pump module available [32]. Pulse compression of the pulse burst is realized by an multi-layer dielectric (MLD)-based grating compressor. Reaching sub-1-ps pulses is expected, as it was already demonstrated on the 45-mJ pulses. A demonstration of the multipass amplifier at full power is planned in 2018.

The Perla A is a new platform based on cryogenically-cooled disk laser technology. Laser operation of quasi-three-level Yb:YAG gain media at low temperature brings many advantages such as higher thermal conductivity, which is almost 4× higher at 100 K for 1 at. % Yb-doping then the room temperature thermal conductivity, lower thermo-optical coefficient (at 100 K, approximately one eighth of its original value at room temperature), and lower thermal expansion coefficient (at 100 K, one third of the room temperature value) [33]. These properties significantly improve the laser cooling efficiency and reduce effects like thermal lensing, which usually prevent lasing at high power and room temperature operation. On the other hand, low temperature operation shrinks the emission bandwidth of Yb:YAG more than six times to approximately 1.5 nm (FWHM) [34], and the achievable pulse length is longer than for room temperature operation. It normally approaches 10 ps [5].

Development of the cryogenic Perla A platform has started this year and the system is still under design. The platform front-end will consist of an Yb-doped fiber oscillator providing µJ pulses, and Yb:KYW bulk, room-temperature-operated preamplifier reaching output pulse energy of 10–20 mJ at 1 kHz repetition rate, due to the prospective upgrade of the whole system to 1 kHz. Yb:KYW was chosen because of the good overlap of its peak wavelength and the wavelength of cryogenically-cooled Yb:YAG, whose peak emission shifts to shorter wavelengths as the temperature decreases. The pulses from the Yb:KYW amplifier will be seeded to a cryogenically-cooled Yb:YAG single slab four-pass amplifier with ceramic Yb:YAG gain medium with saturable absorber cladding and will boost the pulse to 100 mJ. Final, power amplification is planned by a new design of a liquid nitrogen-cooled Yb:YAG thin-disk laser, which is still under technical consideration. The expected pulse energy is about 1.2 J in a diffraction-limited beam and 100 Hz repetition rate with a potential of a 1-kHz upgrade. Due to emission bandwidth shrinking at low temperatures, bandwidth-limited pulse length prolongs and will probably reach 10 ps after compression. Behavior of the single slab gain medium in a closed-loop He cryostat for the Perla A preamplifier was already studied at low temperatures, both experimentally and theoretically, in Reference [35]. Commissioning of the system at full power is scheduled by the end of 2019.

1.3. Atomic Difusion Bonding (ADB)

Critical components for thin-disk laser development are high-quality, efficiently cooled thin disks. Since the disks themselves are very thin and fragile, they are carried by a heatsink with high thermal conductivity. The bonding layer between the gain medium and the heatsink must be highly thermally conductive, as well as mechanically rigid in order to withstand high operating temperatures and mechanical stresses caused by thermal gradients during the laser operation. Thin disks are often bonded to a metallic heatsink by indium-based metallic solder. Such a connection is not perfectly homogeneous; the thermal conductivity of such a contact is insufficient for kW class lasers, and creates a bottleneck for efficient heat extraction. Soldering also generates residual mechanical stresses in the gain medium. Therefore, the usual approach is to use a special epoxy bonding to a diamond heatsink. We employed another technique for the first time, so-called Atomic Difusion Bonding (ADB) [36], which is an epoxy-free process based on the recrystallization of the contacting layers, which thus allows the formation of a perfect thermal bridge [37].

The first demonstration of the ADB process was realized by the bonding of Yb:YAG to an Al heatsink (Figure 3). Nanocrystalline Au films were deposited on both Yb:YAG and Al heatsink surfaces with Ti underlayers, as shown in Figure 3. The Ti underlayers enhanced the adhesion strength of the deposited films on the bonded surface and controlled the crystallographic orientation of the deposited films. Au-layer-coated surfaces were bonded at room temperature under high vacuum environment and pressure application. During this procedure, the Au film recrystallizes and creates a single metallic layer. ADB with extremely thin Au films ensures better thermal management, mechanical tolerances, and prevention of the out-gassing of materials in a vacuum or high temperature environment. The strength of the connection depends on film thickness, surface cleanliness, and the quality of surface polishing. Since the thickness of the nanocrystalline layer can be controlled within a certain percentage of variation, such a uniform bonding interface offers the fabrication of a laser active medium with tight tolerances on thickness. Unlike thermal diffusion bonding or soldering, no annealing or curing at high temperatures is required, making it possible to work with materials of different coefficients of thermal expansion.

We experimentally compared the potential of commercially available indium-soldered Yb:YAG thin disks and ADB bonded Yb:YAG/YAG composite ceramics thin disks [38]. It was proven that the peak surface temperature of ADB disk pumped at 280 W (4.5 kW cm⁻² for the incident pump beam) is 57 °C lower than the temperature of the soldered disk under identical operating conditions. By a thin-disk surface deformation measurement setup developed at HiLASE [37], we also demonstrated the different mechanical behaviors of the disk when pump radiation is applied. The CW output power reached in the proof of principle experiment was 100 W and 177 W from the soldered and the ADB disk, respectively. A new series of ADB disks with optimized gain medium and bonding procedure parameters with metallic and dielectric heatsinks for kW lasers is now under preparation.

2. Mid-IR Picosecond Pulse Generation

Sources at wavelengths in the mid-IR range (between 2 and 8 μm) are under development due to their important applications [11,39], such as minimally-invasive neurosurgery and plastic and polymer processing. Scientific applications, for instance high harmonic generation [40], dielectric laser acceleration [41], or vibrational spectroscopy [42], also benefit from the use of a mid-IR source.

A high average power wavelength tunable picosecond mid-IR source based on optical parametric generation (OPG) and amplification (OPA) in a nonlinear optical crystal is being developed. In this parametric down-conversion process, a powerful pump beam of the shortest wavelength generates signal and idler beams of longer wavelengths. In the wavelength conversion setup (Figure 4), the Yb:YAG thin-disk laser delivers the pump beam of 100 W average power at a repetition rate of 77 kHz, wavelength of 1030 nm, and pulse width of about 2 ps. Part of this fundamental beam pumps an OPG in a 10-mm long, periodically poled lithium niobate crystal (PPLN). The generated wavelength is determined by the PPLN’s poling period and temperature. Tunability of the signal wavelength between 1.46 μm and 1.95 μm was achieved. The corresponding idler wavelengths were in the range of 2.18 μm to 3.50 μm. The signal beam up to 80 mW was generated at 2 W of pump power, when the double pass of the beams through the PPLN crystal was employed.

The main part of the pump beam is delivered into an OPA stage, in which the signal beam from the OPG stage is amplified. The stage consists of two 10-mm long potassium titanyl phosphate (KTP) crystals arranged in the walk-off compensating arrangement. Signal and idler tunability between 1.70–1.95 μm and 2.18–2.62 μm, respectively, was demonstrated by changing the phase-matching angle in KTP crystals. The signal beam was amplified up to 8.5 W at 42 W pump. The power of the generated idler was up to 5 W.

Improvements of the system for the achievement of higher output powers and broader tuning ranges were proposed. These are based on higher pump power, another OPA stage, and the use of Potassium Titanyle Arsenate (KTA) crystals, which have a broader transparency range in comparison to KTP crystals. Thermal effects caused by residual absorption [43] of the beams in nonlinear optical crystals at high average power will be studied as well.

Besides the OPG/OPA-based mid-IR source of radiation, a six-year project on the development of a Ho-doped picosecond thin-disk laser system emitting near 2.1 µm is starting at HiLASE to support OPA in the 4–6 um spectral region (Perla D in Figure 1). In the first phase, a 100-W regenerative amplifier with a 10-kHz repetition rate and a fully Ho-doped fiber-based front-end will be constructed. In the second phase, we are planning to boost pulse energy in order to generate 1-J picosecond pulses and 1 kW of average power. All of the Ho-based systems will be pumped by Tm-doped CW fiber lasers, which are under construction now. Detailed technical concept is still under consideration. Direct applications for this kind of laser system besides OPA pumping are modifications of semiconductors, processing of plastics, laser-induced damage threshold measurement, etc.

3. Harmonics Generation

Various applications benefit from higher photon energy in visible, UV, and DUV (Deep Ultraviolet) wavelength ranges, which can better match the absorption of materials such as copper, carbon fiber-reinforced plastics, glasses, etc. When the picosecond pulses are considered, the laser-matter interaction is specified by a very small heat-affected zone. Therefore, the pulses of 100 W, 100 kHz, 1030 nm first Perla C amplifier were converted into harmonic frequencies in order to fulfill the requirements of user experiments. For the harmonics generation, we use borate crystals LBO (lithium triborate), BBO (beta-barium borate), and CLBO (cesium lithium borate) due to their suitable properties, such as relatively high nonlinear coefficient, high transparency, and commercial availability [44]. The second harmonic (515 nm) of the fundamental beam is generated in the LBO crystal (Type I phase-matching (PM), 8 × 8 mm aperture, 10 mm long, antireflective (AR) coated on both sides for 1030 nm and 515 nm, routine conversion efficiency of 56%) at the output powers up to 40 W (for 70 W input), and it is further is used for the generation of the third and fourth harmonics in cascade conversions (Figure 5). The power of the third harmonic—achieved by 2ω + 1ω sum frequency generation in an LBO crystal (Type II PM, AR/AR@1030 + 515 + 343 nm, 8 × 8 mm, 5 mm long, conversion efficiency of 35% related to the second harmonic), placed just behind the first LBO, without any optimization—is 6.7 W, and will soon be a subject of the next system upgrades. The fourth harmonic (257.5 nm) in the DUV range is produced either in the BBO (8 × 8 mm, 3 mm long, with AR/AR@515 + 257.5 nm, conversion efficiency of 14% related to the second harmonic) or CLBO (6 × 6 mm, 6 mm long, uncoated) crystals (both Type I PM) by the second harmonic generation of a 515-nm beam. The output power was up to 6 W at the conversion efficiency of about 18% (related to the second harmonic, neglecting the Fresnel reflections), while the CLBO crystal resulted as a better convertor showing higher output power without a trend for saturation in efficiency [9]. The fifth harmonic (206 nm) is realized by 4ω + 1ω sum frequency generation in a CLBO crystal (Type I PM, 12 × 12 mm, 4 mm long, uncoated, conversion efficiency of 20% related to the fourth harmonic) at the output power of 0.8 W [45]. The ovens with the crystals for the DUV radiation are placed in a metal box filled with argon so as to protect the crystal surfaces against moisture and DUV light-generated ozone. The CLBO crystals are always kept at a recommended temperature of 150 °C to prolong their lifetime [46]. The present values of the harmonics outputs derived from the Perla C beam (first regenerative amplifier only) are the values after a long-term operation of the system and do not represent the fully optimized fundamental beam (e.g., higher orders of pulse dispersion were not fully compensated).

When the 1-kW, 1-MHz upgrade of the thin-disk amplifiers is accomplished, it will be used for harmonic generation as well. The second and third harmonic aims to reach 500 W and 400 W, respectively. Up to 100 W average power should be generated in the DUV. The increased thermal issues, which normally lead to phase mismatch and conversion efficiency reduction, will be decreased by the sandwich structure of a transparent heat spreader and nonlinear crystal or actively cooled multi-plate design of the harmonic generation stage.

4. Conclusions

Development of reliable and compact kW-class thin-disk-based picosecond lasers continues at the HiLASE Center. We demonstrated Perla thi- disk platforms for the delivery of high repetition rates and high energy ultrashort pulses. The Perla C regenerative amplifier routinely generates pulses up to almost 500 W of average power. The application potential of Yb:YAG Perla beamlines is broadened by conversion to harmonic frequencies from second to fifth ones, and by OPG and OPA to mid-IR. Lately, we have also started the development of a cryogenically-cooled thin-disk laser beamline for Yb:YAG or Ho-based pulsed lasers with a goal to generate a 1-kW picosecond beam at 1.03 and 2.1 µm, respectively. In our development, we employ several unique solutions such as CVBG compressors for high average power pulses or atomic diffusion bonding for the preparation of in-house made Yb:YAG thin disks.