In recent years, the development of high-energy ultrashort pulse lasers in the wavelength range around 2 µm has received strong attention due to numerous applications, among them the generation of high harmonics, extreme UV (XUV) frequency combs, and hard X rays [1–4]. Apart from basic research, several applications in laser-based medical treatment, metrology, and material processing benefit substantially from powerful 2 µm sources [5,6]. Furthermore, high-performance pump sources operating at wavelengths around 2 µm and beyond are an effective driver for generating high-energy, ultrashort pulses in the midwave- and longwave-infrared spectral range via parametric down conversion or terahertz (THz) pulse generation [7–11].

The most attractive optical transitions, for high-power pulsed laser operation in the 2 µm spectral range, occur in the trivalent Ho and Tm ions [5]. Broadband pulses with a wavelength of less than 2 µm are subject to absorption by water vapor in the atmosphere. The latter leads to deterioration of the spatial beam profile and the temporal pulse shape [12]. Holmium-doped laser materials exhibit a gain maximum beyond 2 µm. This relaxes the laser design and enables beam delivery in ambient air, simplifying their adoption to real-world applications. Ho-doped gain materials are characterized by moderate saturation fluence, low quantum defect, a long lifetime of the upper laser level, and a spectral bandwidth sufficient for the generation of ultrashort pulses [5]. Among the plentiful Ho-doped host materials, yttrium lithium fluoride (YLF) and yttrium aluminum garnet (YAG) are most widely used and have proven their suitability for high-energy operation. Aiming at high-energy ultrashort pulse generation, Ho:YLF is beneficial because of its roughly five-times lower nonlinear refractive index compared to Ho:YAG. Furthermore, Ho:YLF exhibits a negative thermal refractive index gradient ${\textit{dn}}/{\textit{dT}}$, whereas the latter has a larger positive value in Ho:YAG. Contrary to Ho:YAG, Ho:YLF is intrinsically birefringent, which is advantageous for reducing the depolarization loss due to thermally induced birefringence at high pump power [13,14].

Here, we report a picosecond (ps) Ho:YLF amplifier system with unprecedented output parameters at kilohertz repetition rates. We demonstrate scaling of the peak power beyond 15 GW in the 1 kHz pulse train, maintaining a brilliant beam quality and excellent stability with a pulse-to-pulse rms of 0.23%. This performance represents the highest peak power of any 2 µm laser amplifier. Compared to our previous Ho:YLF chirped pulse amplification (CPA) setup [15], the new system has been re-designed, and operation parameters modified. Significantly shorter pulses with 2.4 ps duration are achieved by reducing the B integral in the system to ${\sim}{{1}}\;{\rm{rad}}$ at the maximum pulse energy of 52.5 mJ. Figure 1 shows the peak power versus repetition rate reported so far for 2 µm amplifier systems based on Tm- and Ho-doped gain media. A system aiming at similar pulse parameters has been reported in Ref. [16]. It incorporates a cryo-cooled booster amplifier and emits compressed pulses with energy of 220 mJ and 16 ps duration, resulting in a peak power of 13 GW. However, the reported system operates at an order of magnitude lower repetition rate of 0.1 kHz with a pulse-to-pulse stability of 0.8% rms.

 

Fig. 1. Peak power versus repetition rate of ultrashort pulse amplifier systems around 2 µm based on Tm- and Ho-doped gain media. Color coded symbols: Ho:YLF, emission wavelength: $\lambda = {2.05}\;{\rm{\unicode{x00B5}{\rm m}}}$ (red); Ho:YAG, $\lambda = {2.09}\;{\rm{\unicode{x00B5}{\rm m}}}$ (black); Tm-doped yttrium aluminum perovskite (Tm:YAP), $\lambda = {1.94}\;{\rm{\unicode{x00B5}{\rm m}}}$ (green); Tm:fiber, $\lambda = {1.95}\;{\rm{\unicode{x00B5}{\rm m}}}$ (blue). Identical symbol shapes represent the same group. Peak powers were calculated assuming Gaussian pulse envelopes. Data taken from [8,9,15–23].

Download Full Size | PDF

The basic setup of our 2 µm CPA source is shown in Fig. 2. A femtosecond Er:fiber oscillator at 1.55 µm delivers the starting pulses at a 40 MHz repetition rate (Toptica). These pulses are amplified to approximately 8 nJ in an Er-doped fiber amplifier (EDFA) and then launched into a highly nonlinear fiber. The supercontinuum (SC) generated thereby extends from 900–2600 nm with a sub-ps temporal shape. Chirped volume Bragg gratings (CVBG), designed for a center wavelength of 2051 nm (OptiGrate), are used to stretch the seed pulses. Their reflection bandwidth extends over 11 nm (FWHM), and, thus, the CVBGs also serve as bandpass filters for the SC spectrum. The part of the SC spectrum covering the reflection band of the CVBGs is almost flat and shown in Fig. 3(a). The antireflection (AR)-coated CVBGs expose a stretching factor of 39 ps/nm. The seed pulses are pre-amplified in a Tm:fiber amplifier (AdValue) before amplification in Ho-doped laser crystals.

 

Fig. 2. Simplified setup of the 2 µm chirped pulse amplifier. EDFA, Er-doped fiber amplifier; SC, supercontinuum generation; HNLF, highly nonlinear fiber; CVBG, chirped volume Bragg grating; RA, regenerative amplifier; Booster, power amplifier; Ho:YLF, gain media; G, optical gratings.

Download Full Size | PDF

 

Fig. 3. Evolution of the optical spectrum during propagation in subsequent stages of the CPA system: (a) SC seed spectrum; (b) spectrum after CVBGs and Tm:fiber pre-amplifier (blue curve); spectrum of the 12 mJ pulses of the Ho:YLF RA (red curve). Inset: recorded oscilloscope trace with a fast photodiode of the pulse after the Ho:YLF RA [pulse duration (deconvoluted), 270 ps].

Download Full Size | PDF

The Ho:YLF amplifier stages consist of a regenerative amplifier (RA) designed as a ring cavity, and two single-pass booster amplifiers. In-band pumping at 1940 nm is applied for all amplifiers using randomly polarized cw Tm:fiber lasers (IPG Photonics). The 50 mm long end-pumped Ho:YLF amplifier crystals are water-cooled with a coolant flow temperature of 8°C. To benefit from the maximum gain in Ho:YLF at 2050 nm, all crystals are implemented with polarization orientation parallel to the $c$ axis ($\pi$ polarization). Both the RA and booster are enclosed in nitrogen purged boxes to prevent beam distortion due to water vapor absorption at the pump wavelength.

One of the major challenges in high-gain and high-energy amplifier systems is the accumulation of nonlinear phase, typically expressed as B-integral. The main contributions in our system are made by the RA, i.e., the gain medium and the Pockels cell. However, we found that the Tm:fiber pre-amplifier also plays a role, due to the combination of small mode field diameter and weak but relatively short seed pulses. In our previous setup, the 2 µm SC seed pulses were directly amplified in the Tm:fiber pre-amplifier, which came along with a pulse elongation from sub-ps to about 9 ps and a group delay dispersion (GDD) of ${-}{0.85}\;{\rm{p}}{{\rm{s}}^2}$. Its gain spectrum, extending from 2030 to 2110 nm (zero level), has been reported in Ref. [24] (therein Fig. 4). Afterwards, the pulses were stretched to 1 ns by double passing the CVBG. In the present system, the seed pulses are stretched before entering the Tm:fiber pre-amplifier to ${\sim}{0.43}\;{\rm{ns}}$ by a single reflection from the CVBG. Given the reflection band of the CVBG, i.e., from 2045–2057 nm, only this spectral part is amplified in the pre-amplifier. A second reflection upon an identical CVBG after the amplifier results in the targeted seed pulse characteristics for the RA with 0.86 ns duration and 4 nJ energy [25]. The final seed spectrum entering the Ho:YLF RA is very smooth and flat and presented in Fig. 3(b) (blue curve). Thus, the accumulated nonlinear phase in the Tm:fiber pre-amplifier is strongly decreased. Hence, compared to our previous system [15], we reduced the nonlinear phase of the seed pulses and increased their energy, resulting in less gain narrowing, longer chirped pulses, and thus, a further reduced B-integral [26]. After passing an optical isolator, the 40 MHz pulse train is coupled into to the RA, wherein the Pockels cell is simultaneously applied for setting the number of round trips and as pulse picker.

 

Fig. 4. Autocorrelation trace (ACF), measured and simulated, for the compressed 52.5 mJ pulses of the Ho:YLF chirped pulse amplifier at 1 kHz repetition rate.

Download Full Size | PDF

Considering the seed pulse characteristics, the parameters of the RA are carefully adapted to mitigate the common issue of pulse energy bifurcation in such high-repetition rate RAs [25,27]. In parallel, the emitted energy in the saturated single-pulse regime is maximized. The latter has a value of 12 mJ in the 1 kHz pulse train, and the optical-to-optical efficiency is as high as 22%, taking into account the applied cw pump power of 54 W. The pulse energy of 12 mJ is achieved with 19 round trips in the RA. Compared to the setup reported in Ref. [15], gain narrowing is reduced, and the RA output pulses display a roughly 1.5-times longer duration and broader spectrum, similar to the results of a study of an Yb-based RA by Fattahi et al. [28]. The output pulse duration of 270 ps and the 3.5 nm broad spectrum (FWHM) at 2050 nm are shown in Fig. 3(b) (red curves). The spectrum is well fitted by a Gaussian shape, which supports a ${\sim}{1.7}\;{\rm{ps}}$ Fourier-transform limited (FTL) pulse duration. The beam profile is almost perfect with an ${{{M}}^2}$ of 1.2.

After adapting the beam size of the RA output, the pulses are sent to the two single-pass power amplifiers for further amplification, being pumped with 100 W and 120 W, respectively. The single-pass gain of the booster amplifiers amounts to 4.7, resulting in maximum pulse energy of 56 mJ and an average power of 56 W. The behavior of the pulse energy versus pump power is nearly linear with no signs of thermal roll over, i.e., the booster amplifier operates below the saturation level. Thus, the excellent beam profile after the RA is preserved. The other pulse parameters (duration, spectrum, stability) are only marginally affected, i.e., they mostly match to those emitted by the RA [Fig. 3(b)].

In comparison with our previous work [15], the compression scheme is completely exchanged, now enabling exceptional high pulse energies. In our former study, we used CVBGs not only for stretching but also for compression. We observed the onset of depolarization along with an increase of absorption in the compressor CVBG at an input pulse energy of ${\gt}{{20}}\;{\rm{mJ}}$, which set the energy limit in this setup because of temporal and spatial pulse shape distortions. Exchanging the compressor CVBG by dielectrically coated gratings for the 2 µm wavelength range, developed at Fraunhofer IOF, Jena, up to 26 mJ pulse energy were applied, and a compressor throughput efficiency of 77% was achieved [29]. The design of these gratings was further improved by the manufacturer, resulting in excellent performance at 2 µm with a total energy throughput of 93.8% in the compressor of our present CPA system. The dielectric-coated gratings with a size of ${{125}}\;{\rm{mm}} \times {{62}}\;{\rm{mm}} \times {6.35}\;{\rm{mm}}$ exhibit a groove density of 900 lines/mm. The gratings are arranged as Treacy-type compressor providing a GDD of ${-}{{4}} \times {{1}}{{{0}}^4}\;{\rm{p}}{{\rm{s}}^2}$.

The Treacy arrangement temporally compresses the amplified pulses with a measured overall efficiency ${\gt}{{93}}\%$. The recorded intensity autocorrelation trace (ACF) for the compressed pulses are shown in Fig. 4. It exhibits a FWHM of 4.1 ps with an estimated B-integral of ${\sim}{{1}}\;{\rm{rad}}$. The impact of the accumulated nonlinear phase on the temporal pulse shape is simulated by adding it to the spectral phase of the FTL pulse [30]. The best match of the simulated to the measured ACF is shown in Fig. 4. The retrieved pulse shape fits well to the measured ACF, the accompanied weak satellites at 20 ps delay are attributed to slight mismatches in the reflection characteristics of the two CVBGs. The resulting duration of the main pulse is 2.4 ps (FWHM) with an estimated energy content of 85%, implying a record high peak power of 17 GW. The shortening of the pulse duration compared to our previous system [15] from 4.3 to 2.4 ps is mainly attributed to the roughly three-times reduced B-integral in the CPA system.

The compressed 52.5 mJ pulses in the 1 kHz train are distinguished by an excellent pulse-to-pulse stability of 0.23% rms and peak-to-peak fluctuations of 2.1%, as confirmed by the long term stability recording for 2 h (Fig. 5). The thermal equilibrium of the system is attained about 60 min after switching on, accompanied by a ${\sim}{{10}}\%$ rise of the pulse energy. The RA beam quality is nearly unaffected by the following amplification process in the two booster stages. The excellent beam quality of the nearly Gaussian mode at the output of the system is measured to be better than a M-squared parameter of 1.2 (inset Fig. 5).

 

Fig. 5. Long term pulse stability measurement of the 2.4 ps pulses of the warmed-up Ho:YLF CPA at 1 kHz repetition rate. Insets: far-field intensity distribution and histogram of the pulse energy recording.

Download Full Size | PDF

In conclusion, we demonstrate amplification of few-ps pulses in the 2 µm wavelength region beyond 10 GW peak power at a 1 kHz repetition rate. Based on a Ho:YLF CPA, pulses with 52.5 mJ energy and a duration of 2.4 ps are achieved at a central wavelength of 2050 nm. To the best of our knowledge, the achieved peak power of 17 GW is the highest ever reported for any 2 µm amplifier system based on Ho- or Tm-doped gain media (Fig. 1). Furthermore, the 2.4 ps pulse duration is the shortest achieved for Ho:YLF CPA systems.

The water-cooled Ho:YLF amplifier system is further characterized by an excellent stability with a pulse to pulse rms value of only 0.23%. This low noise level is enabled by the high-gain Ho:YLF RA, which is optimized for stable operation in the single-energy regime and emits a record-level pulse energy of 12 mJ. The booster amplifiers further enhance the pulse energy up to a maximum of 56 mJ. Further energy scaling can be achieved by increasing the pump power, since the booster amplifiers show no signs of saturation and only weak thermal lensing, even at maximum pump power. This stable and efficient laser source is a powerful tool and enabling technology for a number of exciting applications, for example, as a pump for the parametric amplification of few-cycle pulses around 5 µm with multi-millijoule energies [29].