1. Introduction

Optical parametric amplification is a well established technique for femtosecond pulse amplification in a very broad range of wavelengths, from the ultraviolet up to the far infrared [1]. For low energy applications, the noncollinear parametric amplification of the white light continuum is typically used and makes generation of pulses shorter than 5 fs possible [2]. The energy of femtosecond pulses amplified in this manner is limited primarily by the parasitic nonlinear effects, e.g. self phase modulation, self-focusing, etc. For high energy applications, the Chirped Pulse Amplification (CPA) technique [3] – a workhorse in the great majority of available femtosecond amplifiers – is combined with the parametric amplification process. Since the first successful demonstration by Dubietis et al. [4], Optical Parametric Chirped Pulse Amplifiers (OPCPA) have paved the way for a new class of femtosecond optical amplifiers. So far, the majority of reported OPCPA setups involved near infrared amplified pulses and green or blue pump lasers. In this spectral range the OPCPA technique is an attractive alternative to the classical laser amplifier technology: a broadband amplification range when used in a noncollinear configuration, a wide range of wavelengths suitable for amplification, a high single pass gain and no heat accumulation in a nonlinear crystal which assures the amplified beam profile of high quality. Tabletop terawatt-class amplifiers pumped by commercial lasers [5] are currently easily accessible. Moreover, setups with peak power in the range of hundreds of terawatts [6] have been demonstrated, with future plans for exawatt powers in the ELI project [7].

Besides intense visible and infrared OPCPA setups there is also a need to produce high peak power ultraviolet femtosecond pulses for example for laser micromachining [8] or to observe an ultrafast photodynamics of molecules [9]. There are a few techniques that can be applied to this end. One approach relies on the direct amplification of ultraviolet pulses in a chirped pulse laser amplifier using a Ce:LiCAF laser crystal [10] pumped with 266 nm. It was shown that such a system can deliver approximately 100 fs pulses with mJ pulse energy. The other scheme is based on the achromatic frequency doubling of the high power femtosecond laser at 800 nm [11], reaching peak powers in the terawatt range. High power fs sources with μJ pulse energy in the UV, using either direct parametric amplification in a NOPA pumped by 267 nm pulses [12] or nonlinear frequency conversion from highly energetic visible NOPA [13–15] were also presented.

Yet another way to generate high energy femtosecond ultraviolet pulses could rely on OPCPA. To the best of our knowledge, only one report on the direct amplification of chirped femtosecond pulses around 400 nm using parametric amplification has been presented so far [16]. The third harmonic (267 nm) from a high energy Ti:Sapphire femtosecond amplifier was used to pump a BBO crystal, while the second harmonic of the same fundamental pulses served as a seed. The overall measured gain was approximately 3500, with the maximum amplified pulse energy of 800 pJ. The authors, however, did not demonstrate recompression of the amplified UV pulses.

Here, we report on a high gain Ultraviolet Optical Parametric Chirped Pulse Amplifier (UV OPCPA) for amplification of ultraviolet pulses, using the fourth harmonic at 266 nm of a commercial ns Nd:YAG laser as a pump beam.

2. Broadband second harmonic generation

Broadband seed pulses in the UV are required for a high power UV OPCPA. Those can be conveniently generated by the frequency doubling of the output of the broadband Ti:Sapphire oscillator operating at around 800 nm. The calculated amplification bandwidth of the parametric amplifier incorporating a noncollinear design (BBO nonlinear crystal and 266 nm pump laser) is approximately 40 nm. In order to be able to take advantage of this bandwidth, one needs a broadband Ti:Sapphire oscillator and an efficient method for broadband Second Harmonic Generation (SHG). To achieve that, we used an achromatic SHG technique [17]. This technique relies on an angular dispersion of the fundamental beam to facilitate simultaneous accurate phase matching for a wide range of frequencies, which leads to an efficient and broadband SHG. In our case, the fundamental beam was delivered by a home-built Ti:Sapphire femtosecond oscillator with an average power of 600 mW at 80 MHz and a bandwidth of more than 100 nm FWHM. In order to introduce the required angular dispersion, the fundamental beam was sent through two SF10 Brewster-cut prisms separated by 83 cm. The prisms form half of a typical prismatic dispersion line [18] in which the beam is laterally dispersed after the second prism (see Fig. 1 .). This configuration provides lateral dispersion in the beam easily controlled by the prisms’ separation and, simultaneously, allows for pulse phase precompensation by a controlled insertion of one of the prisms into the beam.

 

Fig. 1 Schematic representation of the UV OPCPA experimental setup. SF10 – Brewster-cut SF10 prism, FS – Brewster-cut fused silica prism, HWP – broadband half-wave plate, BBO₁ – SHG BBO crystal, BBO₂ – BBO crystal in the multipass parametric amplifier.

Download Full Size | PDF

A proper fundamental beam spectral phase management is essential in the process of a broadband SHG. Not only does it affect the efficiency of the process but also influences the SH spectrum in a complicated manner [19]. After the dispersion line, the beam was focused with a single spherical lens f=75 mm inside a 2 mm long BBO nonlinear crystal (BBO₁, type I, θ=30°). The angular dispersion of the beam was carefully adjusted to fit the phasematching properties of the BBO by the choice of prisms’ material, their separation distance and the focal length of the lens. The quality of the achromatic phase matching is shown in Fig. 2 . – the red line is a plot of sinc²(ΔkL/2) function which represents the spectral filtering present in theSHG process. The waist of the SH beam, which was located inside the BBO₁ crystal was imaged (with magnification of 11.2) using a spherical mirror f=150 mm onto a tip of a Brewster-cut fused silica prism. The magnification of the image was adjusted experimentally to match the angular chirp of the SH beam to the angular dispersion of the prism. After the prism, there was no measurable linear angular chirp and only a small addition of the higher order chirp. The experimental SH spectrum generated using the achromatic configuration is presented in Fig. 2. The power of the IR beam incident on the BBO₁ crystal and the power of the UV beam were 380 mW and 150 mW respectively, which corresponds to 40% conversion efficiency.

 

Fig. 2 Broadband SHG: Measured SH spectrum generated using achromatic configuration in a 2 mm long BBO crystal (blue) in comparison to the SH spectrum generated using the same crystal without achromatic configuration (black). Red – calculated SH conversion efficiency.

Download Full Size | PDF

3. Results and discussion

The pump source for UV OPCPA was a commercial 10 Hz, Q-switched, single longitudinal mode Nd:YAG laser (Continuum, Powerlite 9010). It delivered 500 mJ at the second harmonic frequency. The second harmonic beam was frequency doubled in a KDP crystal to produce 266 nm pulses with energy of 80 mJ and 6 ns FWHM pulse duration. The pump beam was relay imaged onto the input surface of the BBO₂ crystal using a vacuum telescope, to avoid the air breakdown. The time synchronization between the pump and the seed laser was achieved using an electronic delay generator (Stanford Research Systems, DG535), with a ±1 ns time jitter, which is acceptable for 6 ns-long pump pulses.

In order to achieve a reasonable time overlap between the pump and seed pulses we used a broadband, aberration-free stretcher in Öffner configuration with a high stretching ratio. The overall temporal window of the stretcher (time delay between spectral edges of a stretcher corresponding to 385 nm and 425 nm) was 0.75 ns. Due to a poor quality of the routing and stretcher optics, the final energy of the seed beam was 125pJ. In the parametric amplification process, the gain strongly depends on the pump beam intensity. Consequently, one wants to use pump intensities as high as possible and still avoid the pump laser induced crystal damage. For a ns pump pulse duration, the damage threshold of a BBO crystal at 266 nm was measured to be 130-150 MW/cm2 [20]. The damage threshold can be affected by both polishing and coating quality as well as crystal impurities. Using the BBO crystal samples from a single supplier, we experimentally measured the damage threshold to be 200 MW/cm2. This value is a little higher than the previously reported one and we attribute the difference to the different quality of the crystals tested. However, in the UV OPCPA we kept the maximum pump intensity at 120 MW/cm² in order to avoid accidental crystal damage due to possible Nd:YAG laser seeding dropouts.

The initial (unoptimized) settings for the parameters of the amplifier (such as the angle between the pump beam and the optic axis, the angle between the seed and the pump beam, the beams’ diameters, the spatial and temporal overlap) were calculated using the code developed previously [5]. The calculated optimal angle between the pump beam direction and the optic axis was 52.7° while the angle between the pump and seed beams should be adjusted to 6.2°. With this arrangement, a broadband amplification should be possible within a band 40 nm wide around 400 nm measured at the 10% level (see Fig. 3(a) .). We selected a rather long BBO crystal (21 mm, type I, θ=50°) to compensate for a moderate level of the pump beam intensity and thus moderate small signal gain. Even though longer BBO crystals are available, one should bear in mind that the noncollinear configuration results in a geometrical separationof the beams and the effective length of the crystal is limited by the beams’ diameters. In our case the pump beam diameter was 2.2 mm, thus for the angle between the pump and seed beams of 6.2° the calculated maximal interaction length was close to 20 mm. We used a noncollinear multipass configuration [21] with 4 passes through the BBO₂ crystal in order to achieve the overall gain of the order of 10⁶ resulting in multi-μJ output energies. Each pass was adjusted independently to obtain the highest gain using a signal from a fast photodiode as feedback. The total measured gain for the parametric amplifier was 5∙10⁵. To the best of our knowledge, this is the highest value for the OPCPA operating in the ultraviolet reported so far.

 

Fig. 3 OPCPA results: (a) Output pulse spectrum (black) together with seed pulse spectrum (blue), red line - calculated gain profile of the UV OPCPA, (b) Pulse autocorrelation trace (black) measured together with 34 fs FWHM Gaussian fit (red) and the autocorrelation trace calculated from the amplified spectrum assuming flat spectral phase (blue).

Download Full Size | PDF

In a properly designed saturated parametric amplifier, the pump beam intensity distribution is transferred to a signal beam [22]. The 266 nm pump beam profile measured with a CCD camera revealed a regular super-Gaussian profile with no hot-spots visible. However, it is known that lasers with unstable resonators suffer from a complicated spatio-temporal pulse evolution resulting in a “bullet shape” [23] which remains unnoticed when imaged with a time integrating camera. Since our pump laser has an unstable resonator, we expect the pump pulses to display a “bullet-shape” spatio-temporal distribution. Such a distribution is imprinted onto a signal beam profile deteriorating its original Gaussian beam distribution [24]. Thus, a pump laser incorporating a stable resonator design should be considered whenever a high quality beam profile is required.

Amplified pulses with the energy of 60 μJ (23% RMS stability) were sent to a diffraction grating compressor with transmission efficiency of approximately 50%. In order to measure and optimize the UV pulse duration we built a scanning autocorrelator based on a two-photon fluorescence process. The autocorrelation was measured in a 1 mm thick fused silica cuvette filled with a 2,2”'-dimethyl-p-quaterphenyl (BMQ) laser dye solution in cyclohexane. Two replicas of the compressed pulses with a controlled relative delay were focused in the BMQ solution, and the fluorescence with its maximum at 350 nm was recorded with an UV photodiode. The measured autocorrelation trace, together with a Gaussian fit is presented in Fig. 3(b). The pulse duration and the peak power were 24 fs FWHM and 1.25 GW respectively, assuming Gaussian pulse shape. The pulse duration was only 20% longer than a Fourier limit, which is 20 fs indicating a good quality spectral phase management within the UV OPCPA.

As it was already mentioned, prior to the experiment, we performed the numerical modeling of the UV OPCPA. The results of the modeling are only in partial agreement with the experiment. In particular, we found that the experimental gain and bandwidth are significantly smaller than those predicted by the computer code (see Fig. 3(a).). This should be compared with our previous results for the infrared OPCPA [5], which showed a very good quantitative agreement between the numerical modeling and the experiment. Additional experiments with the UV OPCPA led us to the conclusion that the major factor limiting the gain in our setup should be attributed to a thermally induced phase mismatch in the BBO2 crystal. It was most probably caused by the color center formation under strong 266 nm pump laser radiation [25] and subsequently by significant pump beam absorption of about 40% with 10Hz, 80 mJ, 6 ns pulses down-collimated to 2.2 mm diameter. The measured absorption level is significantly higher than that attributed to two-photon absorption. Since the density of the color centers scales linearly with the pump beam intensity the resulting absorption coefficient is proportional to this intensity. Therefore, the effect of this loss mechanism is similar to that of the two-photon absorption and the two phenomena can be described by a single nonlinear absorption coefficient. We measured this effective nonlinear absorption coefficient to be 6.2 cm/GW, which is significantly larger than the two-photon absorption coefficient itself: 0.93 cm/GW [26]. The thermally induced refraction index gradient reduces the effective interaction length, for which the phase matching condition is satisfied. As a result of the thermal mismatch, the total gain and amplification bandwidth are decreased. The other consequence of such a high absorption of the pump beam is its strong attenuation along the crystal, decreasing the parametric gain even further. If our findings concerning the origin of the thermal effects in the BBO crystal are correct, then the performance of the UV OPCPA is mainly limited by the average power of the UV pump beam. Therefore, in principle, one can expect significantly increased parametric gain if the pulse repetition rate is lowered sufficiently to reduce the process of color center formation.

4. Conclusions

We demonstrated a successful parametric amplification of the broadband UV pulses using a multipass configuration and a 266 nm ns pump beam. With a single BBO crystal the overall gain of 2.5·10⁵ was achieved with amplified energy of 30 μJ and peak power of 1.25 GW. The measured overall gain was significantly smaller than anticipated according to the numerical modeling, due to thermal effects and strong nonlinear pump beam absorption. We contribute this effect to the color center formation caused by strong UV pump beam radiation. It results in the effective nonlinear absorption coefficient 7 times higher than a two-photon absorption coefficient itself. However, in principle, it is still possible to limit those parasitic effects by lowering the repetition rate of the pump beam - an approach commonly used for high energy lasers.