Numerous scientific and technological applications demand high-energy and high average power lasers as enabling tools. In particular, optical parametric chirped-pulse amplifiers (OPCPAs) providing sub-femtosecond pulses [1,2], THz generation [3,4], spectral broadening followed by pulse compression [5–7], and high harmonic generation (HHG) [8] enormously benefit from a reliable laser source with output pulse durations below 1 ps. All these applications require very similar laser parameters to enable high efficiency and performance: pulse durations shorter than 1 ps and good beam quality. However, pulse energy requirements could vary from tens to hundreds of millijoules at kilohertz repetition rates. As a result, the average power could reach challenging kilowatt levels, which makes a thermo-optical design of the lasers crucial. The classical engineering response is to increase the gain medium cooling surface–volume ratio, employ Yb³⁺ as an active laser dopant, owing to its low quantum defect, and the development of industrial-grade high-brilliance laser diodes. This results in ytterbium-based room-temperature laser technologies, such as thin-disk, InnoSlab, or coherently combined fiber systems. The thin-disk and InnoSlab approaches involve changing the geometry of the gain medium to increase the cooling surface while reducing the overall volume. Although this makes it possible to achieve an average power of kilowatts and an output of hundreds of millijoules, it requires highly engineered multi-pass construction of the laser pump and laser pulse propagation design [9,10]. Unfortunately, further steps to increase output energy and average power may eventually bring those technologies to their limits and excessively complicate the laser design. In coherently combined fiber laser systems, where the heat is distributed along the fiber, non-linear effects limit the stretched pulse energy to the order of hundreds of microjoules per large mode area fiber; currently, 12 large-core fibers have now been coherently combined to increase power to 10.4 kW at 80 MHz [11]. So far, the maximum energy achieved with this technology is 10 mJ [12]. Although coherent fiber combining technology has reached an average output power level of tens of kilowatts, the next step in increasing the output pulse energy beyond tens of millijoules requires further comprehensive development, with significant challenges in fiber combination and production [13].

Operating ytterbium lasers at cryogenic temperatures further improves the thermo-optical strength of the host and increases emission cross section and gain at the expense of reduced gain bandwidth [14–16]. This allows comparable output parameters to be obtained, with significantly simplified laser geometries. Unfortunately, the best-effort Yb:YAG high-energy cryogenic amplifier has an output pulse duration of around 5 ps, owing to strong spectral narrowing at cryogenic temperatures. Despite a more than eightfold drop in the thermo-optic effect, owing to the positive thermo-optic coefficient of Yb:YAG, the realization of more than 50 W of average output power in rod-type geometries at cryogenic temperatures is difficult. This leads to the introduction of more advanced Yb:YAG amplification elements with composite disks, complicating the overall amplifier structure [17,18]. One possible step for further development of cryogenic ytterbium lasers is to find another host crystal with more acceptable laser characteristics. Recent research shows a different gain material, Yb:YLF, with more than 10 nm gain bandwidth and a negative thermo-optic effect at cryogenic temperatures, making it an attractive alternative to reach sub-picosecond pulses at high average output power levels [19,20]. Yb:YLF is a uniaxial crystal, which exhibits different spectral and thermo-optic properties for E//a and E//c axes, opening up exciting new horizons for cryogenic laser development. Previous efforts in developing cryogenic cooled Yb:YLF amplifiers demonstrate a high average power regenerative amplifier, reaching 250 W with a maximum of 20 mJ pulse energy or a low repetition rate (10 Hz) multi-pass amplifier outputting 300 mJ [21,22]. Here, we aim at a multi-stage amplifier laser system applicable as a reliable kilohertz laser source for THz generation and OPCPA pumping. We illustrate, schematically, the 100 W average power cryogenic Yb:YLF laser system in Fig. 1. The full-fiber front end is composed of an Yb-doped femtosecond fiber oscillator, employing all-normal-dispersion (ANDi) mode-locking [23], followed by a non-linear parabolic pulse amplifier after a Gaussian spectral filter. The output is stretched by chirped fiber Bragg gratings (CFBGs). The stretcher constructed of two gratings, the first CFBG has a chirp rate of 306 ps/nm and the second one has a chirp rate of 100 ps/nm, and the latter has tuning capability. These CFBGs are designed to compensate for the dispersion of a 1760 lines/mm grating Treacy compressor operated at a 60° incidence angle, and have 7.5 nm reflection bandwidth centered at 1019 nm.

 

Fig. 1. 100 mJ cryogenic Yb:YLF laser system. BS, beam-stabilization unit; FR, Faraday rotator; HWP, half-wave plate; LN, liquid nitrogen Dewar; OI, optical isolator; PC, Pockels cell; QWP, quarter-wave plate; Regen, regenerative amplifier; TFP, thin-film-polarizer. Inset: Crystal geometry

Download Full Size | PDF

The front end provides 20 nJ pulses with 2.5 nm full width at half maximum (FWHM) centered at 1018 nm stretched to 1 ns imaged to 2.2 mm beam size into a bow-tie regenerative amplifier through a Faraday isolator. The pulse enters the amplifier cavity by reflecting from a TFP and is picked by a water-cooled beta barium borate (BBO) PC (rise/fall time 3.5 ns/8 ns) and a half-wave plate. The 3 m long cavity consists of two concave dichroic mirrors (DMs) and two flat mirrors with high reflectance (HR); this arrangement ensures stability of the resonator for a wide range of thermal lensing under high pumping conditions at different repetition rates. We employ a 2 cm long 1% Yb:YLF rod crystal with an aperture of 10 mm × 15 mm with a 3 mm undoped cap adhesive-free bonded on each optical face. The caps reduce surface deformation under strong end-pumping. The crystal is metal soldered from the top to a tungsten-copper heat-spreader and thermally connected through a compressed metal gasket to the cooling plate of a Dewar (inset of Fig. 1). Using an a-cut Yb:YLF crystal allows us to exploit both E//a and E//c axes for amplification, depending on the selected polarization. The Dewar is cooled to 78 K by boiling liquid nitrogen. A 2 kW, 960 nm industrial fiber-coupled diode module pumps the crystal with a gated 250 µs pump pulse duration, with an absorption efficiency of 95%. The driving current of the laser diode is optimized to 220 W of absorbed power, corresponding to 20 mJ output energy at 1 kHz repetition rate. The p-polarized pulse travels through the cavity exposed to the E//a gain of the Yb:YLF crystal, which helps to keep almost all the seeder’s initial bandwidth. The PC switches the pulse from the cavity by reflection on a second TFP after 80 round trips. The beam is magnified in a Galilean telescope to 3.6 mm diameter and directed to an 8-pass amplifier; owing to pristine (M²= 1.1) beam quality, no relay imaging is necessary. The output energy stability corresponds to 5% rms, and the active beam-stabilization system controls beam pointing. The measurement was made at an output energy of 20 mJ over 6 h.

The p-polarized beam enters the 8-pass amplifier through a Faraday rotator, a pair of TFPs, and a HWP. Then, after a couple of reflections on a z-fold mirror pair, it goes through a second 1% Yb-doped YLF crystal with similar specifications, installed into an identical Dewar cooled by liquid nitrogen. A vacuum Keplerian 2f telescope images the beam from the center of the crystal to a back-reflecting mirror, which returns the beam to the system with a slight (0.5°) angular detuning. The beam returns to the crystal center, bypassing the input mirror, and is transferred to an alternative path with a compensation Galilean telescope (f = −75 cm and f = 100 cm), which counteracts the thermal lens in the Yb:YLF crystal. After double-passing the telescope and a quarter-wave plate (QWP), the beam retroreflects to the system. We use the compensation telescope not only to counteract the thermal lens in the cavity but also to gradually increase the beam size through the propagation up to 3.9 mm to avoid high fluences close to the damage threshold. Then, completing four passes, the s-polarized beam returns to the entrance of the amplifier but reflects from the TFP to return to the system for an additional four passes. Finally, the amplified beam leaves the system, owing to the action of a Faraday rotator and a TFP. The crystal is pumped by a 3 kW industrial laser diode unit, with a pumping pulse duration of 250 µs.

The driving current is tuned to reach 390 W absorbed power with an absorption efficiency of 95% when the pump beam is double-passed. The dependence of output pulse energy on absorbed pump pulse energy is shown in Fig. 2.

 

Fig. 2. Measured variation of output laser pulse energy as a function of absorbed pump energy at 1 kHz. Inset: Near-field beam profile at 102 mJ.

Download Full Size | PDF

We extract 102 mJ from the 8-pass amplifier, corresponding to 20% of optical to-optical efficiency. The near-field beam profile shows a 97% Gaussian intensity distribution (Fig. 2). The M² measurement exhibits 1.21 in the horizontal and 1.12 in the vertical direction (Fig. 3). Given that the output beam of the regenerative amplifier has an M² of 1.1, we can explain the apparent decline of the beam quality coefficient in the horizontal direction by the horizontal angular detuning in the Keplerian 2f telescope. The small off-axis propagation caused the spherical aberration and deteriorated the beam quality. The beam intensity distribution in the planes along the M² scan remains Gaussian (insets to Fig. 3). We believe that the main reason for the obtained good beam quality at such high pulse energies and high average power levels is the negative thermo-optic effect (dn/dT) of Yb:YLF. The Yb:YLF crystal at cryogenic temperatures has a negative dn/dT of −0.5 × 10⁻⁶K⁻¹ and −1.8 × 10⁻⁶K⁻¹ at E//a and E//c, respectively, which helps to balance other thermal lensing effects, such as mechanical stress and bulging.

 

Fig. 3. M² measurement of beam quality of compressed 100 mJ pulse. Insets: Beam profiles at different caustic positions.

Download Full Size | PDF

The polarization multiplexing sequence of the 8-pass amplifier is as follows: the first two passes of the laser pulse with E//a gain through the Yb:YLF crystal and E//c gain in the subsequent four passes; the last two passes are completed with E//a gain. Peak emission cross sections (ECSs) of Yb:YLF E//a (0.75 × 10⁻²⁰ cm²) and E//c (2.4 × 10⁻²⁰ cm²) at cryogenic temperature are spectrally shifted to each other. This not only maintains the output spectrum of the regenerative amplifier, but also extends it somewhat into the longer wavelength region (Fig. 4). A more than threefold difference in gain between the E//c and E//a passes results in a significant imbalance in energy extraction, but the approximately 20 mJ extracted during E//a amplification has a visual impact on the final compressed pulse duration. At the same time, we note that using both gain axes helps us to simultaneously benefit from good E//c energy extraction and broadband E//a gain. The order of polarization multiplexing can be changed by including an additional HWP after the TFP. We note an interesting lack of dependence of the energy and spectrum of the amplified pulse on the order of polarization multiplexing. The spectral bandwidth of the output pulse is only 2 nm FWHM, which is the result of the limited and not optimally distributed bandwidth of the seed source.

 

Fig. 4. Measured optical spectra at (blue) output of regenerative amplifier and (red) output of final amplifier and normalized ECSs of Yb:YLF for (green) E//a and (yellow) E//c axes at 78 K.

Download Full Size | PDF

The amplified laser pulse is compressed in a single-grating Treacy compressor, with a high-efficiency multi-layer dielectric grating (1760 lines/mm) at a 60° angle of incidence with a throughput of 89%. We choose the single-grating layout, owing to its ease of alignment and tuning. The HR coated return 90° retro-reflector and roof mirror are installed with 10 µrad precision. The 980 fs FWHM pulse duration was measured by second-harmonic autocorrelation with a sech² fit, presented in Fig. 5. We attribute the 15% pedestal to the unfortunate fact that we use a CFBG at a slightly offset wavelength where the group delay lacks apodization. Hence, we assign the pedestal to higher-order dispersion, which cannot be compensated by compressor or stretcher tuning.

 

Fig. 5. Compressed pulse duration measured by second-harmonic autocorrelation; sech² fit corresponds to pulse duration of 0.98 ps FWHM. Inset: Measured third-order autocorrelation.

Download Full Size | PDF

The measured third-order autocorrelation (inset to Fig. 5) reveals high symmetry and confirms the overall good dispersion compensation up to the third order.

In conclusion, we have demonstrated a chirped-pulse amplification laser system based on cryogenically cooled Yb:YLF rod-type amplifiers. The system delivers more than 100 mJ of laser pulses with pristine beam quality at a repetition rate of 1 kHz. We have compressed the output pulse to 980 fs in a Treacy compressor. The laser system is planned for use as a pumping source for THz generation, OPCPA pumping, and multi-pass cell pulse compression experiments, where the level of technology is approaching pulse energies above 100 mJ [24]. The results achieved give confidence in the significant scalability potential of cryogenic Yb:YLF laser technology. However, there are two main factors limiting the system development. First, the observed threshold for laser-induced damage under vacuum cryogenic conditions is about 3 J/cm² for a 1 ns pulse, which limits the extraction of the amplifier at low repetition rates. The second issue is the rod geometry, which can only effectively absorb 600 W of average pump power, reaching its cooling capacity limit. As a solution, we are currently preparing a new seeder source for the system, providing 10 nm spectral bandwidth, centered at 1016 nm. This seed pulse at the same 0.4 ns/nm chirp rate will have a duration of about 4 ns and allow for high fluences and more efficient extraction. In addition, this bandwidth is necessary to utilize the full Yb:YLF gain and can lead to compressed output pulse duration of about 300 fs. Detailed thermal simulation shows that changing the crystal geometry to a slab will extend the pump’s absorbed power limit to 1 kW average power. Overall, we hope to achieve at least a threefold increase in output energy and average power in future studies.