Ultra-intense, ultrafast lasers^[1] are widely used in strong field physics, such as laser wakefield acceleration^[2,3], high harmonic generation^[4,5] and terahertz radiation^[6]. These lasers are currently generated from a Ti:sapphire-based chirped pulse amplification (CPA) system^[7–9] or nonlinear crystal-based optical parametric chirped pulse amplification (OPCPA) system^[10–12]. Due to the short storage time, Ti:sapphire is typically pumped by frequency-doubled Nd:YAG or Nd:glass solid-state lasers, which results in a bulky and complicated system with high cost. Compared with conventional CPA laser systems, the OPCPA system can obtain a shorter pulse duration and a higher peak power. Its pump laser needs a short pulse duration, a low time jitter, and smooth temporal and spatial distributions because of the intrinsic instantaneous process. Thus, the OPCPA system is more complicated and expensive.

Ytterbium (Yb)-doped solid-state lasers have attracted attention in the last two decades. The Yb-ion has a long storage time, which is suitable for adopting a diode laser as a pump. Ytterbium-doped lasers have been developed to exhibit high pulse energy and power^[13–26]. This kind of laser system can be used for table-top X-ray generation^[27].

The Petawatt Optical Laser Amplifier for Radiation Intensive Experiments (POLARIS) project based on the Yb:glass amplifier, which aimed to achieve a petawatt peak power output with a sub-hertz repetition rate, was started in 1999, and has demonstrated 54 J pulse energy^[13–15]. Meanwhile, the Petawatt Energy Efficient Laser for Optical Plasma Experiments (PENELOPE) project with designed parameters of 150 J/150 fs/1 PW/1 Hz based on the Yb:CaF₂ crystal amplifier is under development^[16]. Based on the Yb:YAG crystal, 1 J pulses with a pulse duration of 5 ps at 0.5 kHz were realized using a cryogenic Yb:YAG active mirror amplifier^[17]. Moreover, 200 mJ 1.1 ps 5 kHz pulses were obtained using thin-disk laser technology^[18]. Using Yb:KYW, a pulse energy up to 27 mJ at 100 Hz was realized with a pulse duration of 560 fs^[19]. A pulse duration down to 182 fs was obtained with an output power of 8.7 W at 500 kHz^[20]. Working at 1 kHz, 4.7 mJ pulses were obtained with a sub-picosecond pulse duration^[21]. Based on the Yb:KGW crystal, a 10 mJ-level pulse energy was achieved at 10 Hz^[22]. In addition, 21 μJ, 270 fs 60 kHz pulses were realized by a hybrid Yb:KGW regenerative amplifier^[23].

We developed a diode-pumped Yb:KGW regenerative amplifier as the front end of the kilohertz terawatt laser system. Broad bandwidth pulses were obtained by applying dual crystals with different orientations of their optical axes^[21,30]. The cavity mode diameter was enlarged to 500 μm to achieve a high energy. The cavity parameters are specially designed to compensate the thermal lensing and preserve good beam quality. Finally, we experimentally obtained 1 kHz, 1.2 mJ pulses with a 9 nm bandwidth based on these considerations. A pulse duration of 227 fs was realized with preliminary compression. The laser fluence on the Yb:KGW crystal was no more than 0.8 J/cm², which was far below the damage threshold. Once the cavity laser energy was over 1.5 mJ, however, the dichroic mirror was damaged because of low performance. We believe that we can further boost the pulse energy to over 2 mJ with new higher-quality dichroic mirrors and by further stretching the chirped pulse duration.

The Yb:KGW laser system has a quasi-three-level behavior. The net gain of the Yb:KGW crystal was not high: several tens of roundtrips were needed to amplify the nanojoule seed to a millijoule level. Thus, the gain-narrowing effect was inevitable in Yb:KGW regenerative amplifiers.

Several methods have been adopted to solve this problem. An acousto-optical programmable dispersive filter can be utilized to modulate the seed’s spectrum to suppress the gain-narrowing effect^[31], but it would add much more cost to this component. The intra-cavity spectral gain curve could be modulated by inserting an etalon that reduces the actual gain at the wavelength with the maximum gain factor^[32,33]. Etalon insertion would lead to additional loss, however, time-to-time optimization would introduce the risk of component damage and instability to the system. A nonlinear regenerative amplifier could directly deliver pulses with a duration as low as 97 fs and a spectral bandwidth of 19 nm^[34,35], whereas the intra-cavity nonlinearity could result in system instability. One simple and robust method of solving this problem is by adopting a dual-crystal cavity design^[21,30,36]. This method was based on the fact that the spectral emission cross-section curves differ from each other when laser polarization is paralleled to the different optical axis of Yb-doped double tungstates^[37–40]. Thus, we adopted a dual-crystal design to compensate for the gain-narrowing effect.

Figure 3(b) illustrates the output spectrum working at the Q-switched mode. The bandwidth in the orthogonal orientation was clearly much broader. The calculated Fourier transform limited pulse duration was as short as sub-160 fs. Under the configuration where the two crystals were placed in the same orientation, with the seed input, the amplified pulse bandwidth was only 3 nm with 180 μJ energy when it broke the dichroic mirrors. In contrast, the seed laser can be amplified to over 1 mJ under the orthogonal orientation configuration.

The combined gain is complex and should be measured under different pumping conditions because of the quasi-three-level system behavior of the Yb-doped gain medium. The blueshift, with one peak from 1037 nm under the same crystal placement orientation to 1035 nm under the orthogonal crystal placement orientation, could result from the reabsorption losses at a long wavelength in the crystal where the laser polarization is parallel to its N_m-axis, making the gain at 1035 nm higher than that at 1037 nm under orthogonal orientation.

The thermal lensing effect can be estimated as^[41] where f_th(r) is the focal length of the thermal lens; K_c is the thermal conductivity of the laser medium; P_heat is the part of the absorbed pump power that converts into heat; dn/dT denotes the thermal-optical coefficient; and s(r) is the fraction of the pump power contained in a beam radius of r.

Previous work showed that the dual-crystal configuration can be used to compensate for the thermal lensing effect^[42,43], where the pulse energy can be boosted up to 6.5 mJ in a Yb:KYW regenerative amplifier working at 1 kHz^[21]. The dual-crystal configuration lowers the thermal load on each crystal while maintaining a high gain, and facilitates the compensation of each one’s thermal lensing effect. Therefore, the laser cavity can work with a large stability range over the thermal lensing effect variation.

The two-pump laser could offer up to 120 W power. Approximately 60% of the maximum pump power was applied in our experiment considering the low performance of the dichroic mirrors. Under this dual-crystal configuration, our cavity supports a 100 W pump power input, as discussed in Section 2.2. High-quality dichroic mirrors will be used in the future, and laser pulses will be stretched to a longer pulse duration. Amplified pulses with an energy above 2.5 mJ can be expected by applying a higher pump power to the regenerative amplifier and making it work at the saturation level. The laser fluence of ~1 J/cm² on the dichroic mirrors is believed to be below the damage threshold of high-quality dielectric-coated mirrors.

Due to the high transparency of our transmission grating, the laser pulses were compressed to 1.2 mJ with a beam diameter of 4 mm corresponding to a total compression efficiency of 80%. Thus, with a further improvement in our regenerative amplifier, 2 mJ compressed pulses can be expected, without damaging any component inside our compressor.

The compressed pulse duration was 227 fs (Figure 9(b)). The wings of the main peak are caused by the uncompensated high-order dispersion. In our system, the stretcher and the compressor share one same grating. Therefore there is little freedom for optimizing the dispersion compensation effect. Sub-200 fs pulses can be expected with a broader bandwidth seed input. Thanks to the dual-crystal configuration, the laser cavity delivers a near diffraction limited laser beam with M² around 1.1 (Figure 9(c)). The output laser would not suffer from beam deterioration when more pump power is applied.

This paper reports on the Yb:KGW dual-crystal cavity. The regenerative amplifier boosted the seed from sub-0.2 nJ to 1.5 mJ. The laser pulses were compressed to 1.2 mJ with a pulse duration of 227 fs. The output laser beam is near the diffraction limit with M² ~1.1, benefiting from the dual-crystal configuration. Moreover, the Q-switched output spectrum of the regenerative amplifier supports a sub-160 fs pulse amplification. Therefore, pulses with energies of up to 2 mJ and duration below 200 fs can be obtained using higher-quality dichroic mirrors and a broader-spectrum laser seed. This regenerative amplifier works as a good pre-amplifier for further laser amplification.

In the future, an investigation on the spectral gain curve of Yb:KGW in the main amplifier will be performed under cryogenic temperature evolution from 77 K to 300 K. The working temperature will be controlled to realize broader gain bandwidth and less shift at the central wavelength of the main amplifier. We plan to boost the pulse energy to above 40 mJ in two-stage multi-pass amplifiers, corresponding to a gain factor of 20. In each amplifier, we will also adopt two cascaded crystals placed under orthogonal orientation to reduce the thermal load on each crystal and compensate for the gain-narrowing effect.