Modern laser-based experiments need ultrashort pulse durations at increasingly higher average powers. Stable-high repetition rate amplifier systems increase the statistical relevance of the experiment and allow faster data acquisition times. Available Ti:sapphire amplifiers require complex cooling to reach tens of watts [1,2]. In comparison, optical parametric chirped-pulse amplification (OPCPA) [3,4] has the advantage of low thermal load in the amplification medium and requires no active cooling. Most importantly, pump amplifiers for OPCPA are available up to kilowatts of average power with short pulse duration down to $\sim 0.5\mathrm{ps}$ [5]. Such high average powers can, however, cause heating due to residual linear absorption of the nonlinear crystals [6]. It is thus imperative to have a good strategy for selecting high quality crystals. Previously demonstrated high power OPCPA systems use a Ti:sapphire oscillator to provide OPCPA and pump amplifier seed [7–9]. While Ti:sapphire oscillators are still used extensively, a white-light generation (WLG) based OPCPA seed source derived from the pump source is a very elegant and inexpensive solution [10–12]. Most systems deploy relatively short pulses around 300 fs from Yb-doped amplifiers to drive the WLG [10,11,13], but it has been shown that it can also be generated with $\sim 1\mathrm{ps}$ pulses [12], which are available from high-power pump sources, such as Innoslab [12] or thin-disk amplifiers [14]. The objective of the present work is to operate a MHz-class high-power noncollinear OPCPA with high conversion efficiency, seeded by WLG with 1 ps WLG pump pulses. The heat load was reduced by selecting a nonlinear crystal with low absorption at 515 nm. Further, the OPCPA should offer flexibility in terms of bandwidth and pulse duration tuning [15].

The OPCPA pump source was an InnoSlab amplifier with a flexible repetition rate front-end as seeder (as described in [16]), operated at 3.25 MHz repetition rate. The average power used for the experiment was 140 W [central wavelength $\lambda c=1030\mathrm{nm}$, pulse duration ${\tau }_{1030}\sim 1.1\mathrm{ps}$, inset of Fig. 1(a), ${\tau }_{AC,1030}=1.55\mathrm{ps}$]. The corresponding beam quality factors are $M_x^2=1.2$ and $M_y^2=1.1$ in the two cartesian axes [Fig. 1(a)]. The InnoSlab amplifier was operated without chirped-pulse amplification (CPA), which simplifies the design and improves the overall stability. As shown in the schematic diagram [Fig. 1(b)], part of the amplifier output, 8 W was split (beam-splitter, BS) to drive the WLG. For OPCPA pumping, the remaining $\sim 130\mathrm{W}$ were frequency-doubled to 60 W at 515 nm in a 8 mm lithium triborate (LBO) crystal using noncritical phase matching (SHG); this setup is comparable to [17]. The OPCPA-pump beam was focussed into a $6\times 6\times 6{\mathrm{mm}}^3$ beta-barium borate (BBO) crystal with an $f=500\mathrm{mm}$ lens (OPA). The beam waist in the BBO crystal was $w_0=87\mathrm{\mu m}$ with a Rayleigh length of $z_R=46.7\mathrm{mm}$ and a pump intensity of $65\mathrm{GW}/{\mathrm{cm}}^2$. $55\pm 1\mathrm{W}$ of power were effectively used for OPCPA pumping.

 

Fig. 1. OPCPA pump amplifier properties and setup. (a) Caustic measurement of beam radius $w_{x,y}$ at different spatial positions along the propagation axis. The black boxes show selected spatial profiles of beam and focus. Inset: autocorrelation measurement (black) and Gaussian fit (red) of 1030 nm pump pulses from a 140 W Yb:YAG InnoSlab amplifier with 1.1 ps pulse duration. (b) Schematic diagram of the system, pumped with a 140 W Yb:YAG InnoSlab amplifier. After the beam-splitter (BS), $\sim 130\mathrm{W}$ are used for second harmonic generation (SHG) to achieve $\sim 60\mathrm{W}$ for pumping optical parametric amplification (OPA). A fraction of the InnoSlab output (8 W) is focussed into a yttrium aluminum garnet (YAG) crystal with an $f=178\mathrm{mm}$ lens (L1) to drive the white-light continuum (WLG). After a $f=100\mathrm{mm}$ refocussing lens (L2) and a high-pass filter (F, $\lambda <1000\mathrm{nm}$), the broadband WLG beam stretched in an SF11 prism pair, amplified (OPA) and compressed in 55 mm fused silica.

Download Full Size | PDF

The pump beam fraction used to drive the WLG (8 W) corresponded to a pulse energy of 2.46 μJ. This beam was focussed with an $f=173\mathrm{mm}$ lens (L1, $\mathrm{NA}=0.0057$) in a YAG crystal [13]. The effective length of the filament was about 6 mm and ended approximately at the exit facet of the YAG crystal, as shown in Fig. 2(a) (inset). A typical WLG spectrum is shown in Fig. 2(a) (gray line) with a bandwidth of almost 500 nm. After refocusing with an $f=100\mathrm{mm}$ lens (L2), a 1000 nm edge-pass filter (F) was used to filter out the 1030 nm WLG driver pulses. The energy content in the white-light continuum was $48.6\pm 0.15\mathrm{nJ}/\mathrm{pulse}$ (25.8 nJ between 620 and 1000 nm). In earlier experiments, a stable WLG operation was achieved for longer than one day using $<400\mathrm{fs}$ pulse durations at 4 MHz repetition rate [10]. With the 1 ps pulses in this work, the YAG crystal had to be continuously moved (about 20 μm per 15 min) perpendicularly with respect to the beam propagation axis to prevent material degradation. For stretching the WLG pulses to about 700 fs (at $1/e^2$ intensity) for OPCPA seeding, a negative chirp was introduced by an SF11 prism pair (prism stretcher) with an apex distance of $L_P=236\mathrm{mm}$. This reduced the effective length of the OPCPA seed path, compared to a fused-silica prism stretcher. The utilization of a prism stretcher best serves the emphasis of this work, which is to demonstrate flexible tunability of the central wavelength and the spectral bandwidth of the amplifier. For compression after the amplified pulses, a 55 mm fused silica slab was used after the BBO.

 

Fig. 2. Operation modes of a single high-power OPCPA stage. (a) Broadband amplified signal (OPA, red) with $\mathrm{\Delta }\lambda =238\mathrm{nm}$ at 5% intensity maximum; white-light generation spectrum (WLG, grey line). Inset: side view of the filament within the YAG crystal for WLG. (b) Selected tunable narrow-band spectra between 700 and 900 nm, with $\mathrm{\Delta }\lambda =54\mathrm{nm}$ bandwidth at 10% intensity maximum. (c) Autocorrelation (black line) at 800 nm [orange spectrum in (b)], corresponding 29.1 fs FWHM pulse duration; calculated autocorrelation (gray line) for 26 fs Fourier-limited pulse (gray dotted line). Inset, amplified signal beam profile.

Download Full Size | PDF

The OPCPA results are shown in Fig. 2. The seed-beam waist was chosen to be twice as large as the pump beam for good spatial overlap. First, the OPCPA was optimized for broadband amplification. The output average power was $11.4\pm 0.1\mathrm{W}$ (pulse energy of 3.5 μJ), corresponding to a pump-to-signal conversion efficiency of $\eta =20.7\%$. The corresponding spectrum is shown in Fig. 2(a) (red line), supporting a bandwidth-limited pulse duration of 6 fs FWHM. Then, the tunable operation was tested by varying the central wavelength between 700 and 900 nm. Corresponding spectra are shown in Fig. 2(b). In this configuration the OPCPA output power was reduced to 4 W, because the stretching was optimized for broadband pulses rather than for narrowband pulses. Figure 2(c) shows an autocorrelation measurement of the compressed pulse at a central wavelength of 800 nm [orange spectrum from Fig. 2(b)], corresponding to a pulse duration of 29.1 fs, which is within 10% of the Fourier-limited pulse duration (gray lines). The corresponding beam profile with an ellipticity of 0.87 is shown in the inset of Fig. 2(c). Theoretically, a beam quality with $M^2\approx 1.2$ is possible at conversion efficiencies $\eta >20\%$ [18].

At high-pump average-power, heat accumulation within the nonlinear crystal limits the conversion efficiency. Thermo-optic refractive index changes affect phase-matching in the presence of a temperature gradient. In BBO, the strongest absorption is expected from the high-power pump wave at 515 nm and idler components above 2 μm wavelengths [6,18]. The idler absorption can be avoided by choosing the lowest seed signal cut-off at around 690 nm. Thus, the residual linear absorption of the pump wave can be considered as the main heating source. A BBO crystal with a low linear-absorption coefficient at 515 nm was selected from a number of uncoated crystals from different companies. The absorption coefficients of these crystals were measured using common path interferometry [19]. The measured absorption coefficients of crystals at 515 nm were as high as several hundreds $\mathrm{ppm}/{\mathrm{cm}}^{-1}$. The crystal used for the experiment had the lowest absorption coefficient of $12.76\pm 7.09\mathrm{ppm}/\mathrm{cm}$. This volumetric absorption measurement is shown in Fig. 3(a). Figure 3(b) shows a finite element analysis of optical pump-power absorption in BBO crystal. This simulation shows the dependence of maximum temperature, $T_{\text{center}}$, and temperature gradient, $\mathrm{\Delta }T$ (approximately within the waist $w_0$ of the pump beam), inside the crystal on absorbed optical pump power. The working point for this experiment can be narrowed down to the shaded area in Fig. 3(b). The largest temperature gradient in the crystal was $<0.1\mathrm{K}$. This is not critical for phase matching. The simulation suggests that an increase of at least one-order of magnitude of pump power would be possible.

 

Fig. 3. Absorption in BBO crystal. (a) Measured absorption at 515 nm within the volume of a $6\times 6\times 6{\mathrm{mm}}^2$ BBO crystal (common-path interferometry). (b) Calculated maximum temperature $T_{\text{center}}$ and gradient $\mathrm{\Delta }T$ (center-boundary) in dependence on absorbed pump-beam power. OPCPA stage was operated in the shaded area. Inset, temperature distribution ($z$. beam propagation axis), calculated with finite element analysis of the experimental configuration.

Download Full Size | PDF

In summary, a sub-30 fs, wavelength-tunable, 3.25 MHz repetition rate OPCPA is presented. The system is pumped by a 140 W InnoSlab amplifier and seeded by white-light continuum pulses generated in a 20 mm YAG using a fraction of the 1 ps pump pulses. This system avoids a broadband Ti:sapphire seed oscillator and a CPA scheme in the pump amplifier, thus representing a major step toward the availability of compact and stable ultrashort short-pulse amplifiers at average powers above 10 W. For the wavelength-tunable OPCPA operation, a compact SF11 prism pair is used. Alternatively, for sub-10 fs pulse durations, a chirped-mirror stretcher can be used [11]. The necessity to improve the quality control for low-absorption nonlinear, optical crystals for high power applications is pointed out. The intent of this work is to improve the building blocks for a high-power amplifier system and increase the repetition rate for laser experiments that require fast data acquisition.