Ultrashort laser pulses in the ultraviolet (UV) spectral region are desired for many applications. For example, in laser material processing, the short UV wavelength allows for smaller structures to be fabricated compared to infrared lasers which are widely used in industry today [1]. In addition, short wavelengths allow for higher absorption, which potentially enables a much higher throughput when machining large bandgap materials.

Apart from this, fundamental research desires powerful UV lasers with shortest possible pulse durations for studies on smallest spatial and temporal scales. Furthermore, it is well known that short wavelength driving lasers can significantly enhance the conversion efficiency of high harmonic generation (HHG) into the extreme ultraviolet (XUV) region, compared to the commonly used near-infrared (NIR) lasers. This is mainly due to the enhanced single-atom response, but also to the strong atomic dispersion of noble gases at short wavelengths [2]. By driving HHG with the second harmonic of a high-power femtosecond fiber laser (515 nm wavelength), a record high average power of 1 mW has recently been achieved within a single harmonic at $\sim 22\mathrm{eV}$ photon energy [3]. Although UV driving wavelengths are highly desired to further enhance the XUV photon flux, there are no suitable high average power UV lasers available yet. Particularly, the combination of short and energetic femtosecond laser pulses, good beam quality, and high average power is highly demanding for the laser architecture. Normally, the UV spectral region is reached by frequency conversion of NIR lasers in nonlinear crystals. Although thin-disk [4–6], slab [7], and fiber lasers [8–10] today provide femtosecond pulses with up to 1 kW of average power, the highest reported average power of femtosecond UV lasers has not exceeded a few watts yet [11,12].

Recently, 234 W of average power have been achieved at 343 nm with $\sim 6\mathrm{ps}$ pulses by generating the third harmonic of a Yb-YAG thin-disk laser [5]. However, the strongly distorted beam profile observed in these experiments indicates thermo-optical effects in the employed lithium triborate (LBO) crystals at this average power level. Please note that thermo-optical effects at high power operation have already been reported and investigated for optical parametric amplifiers in the visible (VIS) and NIR [13], and for frequency doubling to the green (515 nm) [14]. Due to even higher absorption coefficients in the UV spectral region, these effects are expected to place a severe limitation on the average output power of UV lasers. Here we report on a 343 nm femtosecond laser system delivering pulses as short as 110 fs with up to 51 W of average power at 10 MHz repetition rate. In addition, 100 W of average power have been achieved with 730 fs pulses at a 3.5 MHz repetition rate. In both cases, an excellent, nearly diffraction-limited beam quality, has been achieved ($M^2<1.4$).

The experimental setup is depicted in Fig. 1. A Yb-femtosecond-fiber chirped-pulse amplification system (FCPA) delivers $\sim 290\mathrm{fs}$ pulses and up to 620 W of average power at 1030 nm wavelength. It is based on the system described in [9,10] and utilizes a coherent combination of eight parallel fiber amplifiers. A half-wave plate, combined with a thin-film polarizer placed in front of the experiment, controls the average power of the incident fundamental laser beam.

 

Fig. 1. Schematic of the experimental setup. A fiber-chirped-pulse amplification system (FCPA) delivers femtosecond pulses at 1030 nm wavelength. A half-wave plate (HWP) and a thin-film polarizer (TFP) are used to attenuate the fundamental laser beam. An additional HWP is used to rotate the linear polarization plane in front of the two crystals for SHG and THG.

Download Full Size | PDF

Beta barium borate (BBO) has been selected as the nonlinear medium due to its high nonlinear coefficient and damage threshold. For frequency conversion, the laser beam size is reduced to $\sim 1\mathrm{mm}$ diameter by means of a Galilean-type lens telescope. Subsequently, the collimated beam is sent through a first BBO ($\theta =23°$, type 1) for second-harmonic generation (SHG) and a second BBO ($\theta =32°$, type 2) for third-harmonic generation (THG). The thickness of the nonlinear crystals is limited by the phase-matching bandwidth and set to 0.5 mm for the SHG crystal and 0.4 mm for the THG crystal. The polarization of the transmitted/attenuated beam is rotated with a second half-wave plate placed in front of the nonlinear crystals. The optical axis of the SHG crystal is kept in the plane parallel to the optical table, while the polarization of the fundamental and the orientation of the THG crystal are rotated perpendicular to the beam to optimize the THG efficiency. In addition, the angle between the fundamental beam and the optical axis (phase-matching angle) of each crystal is adjusted to obtain the highest conversion efficiency.

The generated green (515 nm) and UV (343 nm) light is separated from the infrared (1030 nm) via several dielectric mirrors. Subsequently, the average powers of all three waves are measured. In addition, the spectra and temporal pulse profiles are characterized, and a SPIRICON M-200 beam profiler is used to measure the beam quality factor $M^2$. Since no significant differences in the $M^2$ values for $x$- and $y$-axis have been observed, we report the average value $M^2=(M_x^2+M_y^2)/2$ throughout the manuscript.

First, experiments were performed with 290 fs pulses at 10 MHz repetition rate with up to 62 μJ pulse energy resulting in a maximum peak intensity of $\sim 50\mathrm{GW}/{\mathrm{cm}}^2$ at the nonlinear crystals. Using an AR-coated 0.3 mm thick BBO, up to 42 W of average power have been generated at 343 nm. However, the beam quality degrades from $M^2<1.3$ at low average power to $M^2\sim 4.6$ at 42 W.

This thermal beam quality degradation, which can be seen clearly in the left inset of Fig. 2 (measured intensity profile at the focus), is due to thermally induced beam distortions caused by absorption of the generated third harmonic radiation, in combination with the low thermal conductivity of BBO (${\kappa }_{<\mathrm{BBO}}=0.08\dots 0.8\mathrm{W}/\mathrm{Km}$, depending on the crystal orientation) [15]. Of course, the quality of the crystal and the coatings determines the thermal load and, hence, the beam degradation threshold power.

 

Fig. 2. Beam quality versus average power of the third harmonic generated with 290 fs IR pulses at 10 MHz repetition rate. Insets: measured intensity profiles at the focus.

Download Full Size | PDF

In a next step, the 0.3 mm thick BBO was replaced by a sandwich structure composed of a 0.4 mm thick BBO with a 1 mm thick sapphire plate attached to the front and rear surfaces. A schematic drawing of this structure is shown in Fig. 3. Due to the high thermal conductivity of sapphire ($\kappa {<}_{\text{sapphire}}=42\mathrm{W}/\mathrm{Km}$ [16]), these plates serve as heat spreaders and have been attached to the BBO crystal by direct bonding [14]. Both sapphire plates have been coated with an AR-coating for the relevant wavelengths ($R<1\%$ for 1030, 515, and 343 nm) at the surfaces to air, while the contact surface to the BBO was uncoated (reflection loss per surface: $R<0.1\%$ for 1030, 515, and 343 nm).

 

Fig. 3. Schematic drawing of the hybrid structure for THG.

Download Full Size | PDF

With this hybrid structure, up to 51 W of average power could be achieved at 343 nm with a measured beam quality factor of $M^2=1.2$. The beam profile at the focus is shown in the inset on the right side of Fig. 2.

The pulse duration of the UV pluses has been characterized by a cross-correlation of the generated UV and the remaining IR pulses. The measured cross-correlation is displayed in Fig. 4 and indicates a half-width of 310 fs.

 

Fig. 4. Dots, measured cross-correlation between the infrared pulses and the generated ultraviolet pulses at 51 W of average power; line, Gaussian fit; inset, measured spectrum of the generated third harmonic.

Download Full Size | PDF

The duration and shape of the remaining IR pulses after frequency conversion have been measured to be very similar to the driving IR pulses delivered by the laser itself (290 fs, Gaussian-like). Assuming a Gaussian profile for both, the duration of the generated UV pulses can be determined to 110 fs. The measured spectral bandwidth was 1.9 nm FWHM (see inset Fig. 4), which corresponds to a Fourier limit of $\sim 60\mathrm{fs}$.

Note that the conversion efficiency from IR to UV in this configuration was only 8%. The reason is temporal walk-off of between the IR and the SH pulse within the first and second BBO ($93\mathrm{fs}/\mathrm{mm}$ [17]), but also within the 1 mm sapphire located between them ($138\mathrm{fs}/\mathrm{mm}$ [18]). Indeed, by chirping the IR pulse to longer durations, a higher efficiency was achieved. This is mainly caused by the better temporal overlap of the IR and the SH pulse within the second BBO crystal. At the same time, the laser repetition rate has been reduced step-wise, while maintaining a maximum average power of 620 W and a peak intensity of $\sim 50\mathrm{GW}/{\mathrm{cm}}^2$. At 3.5 MHz repetition rate, 177 μJ pulse energy, and 990 fs IR pulse duration, 100 W of average power could be achieved in the UV, which corresponds to an overall IR to UV conversion efficiency of 16.1%.

In this configuration, the beam quality has been analyzed again for different average power of the generated UV pulses. The results are presented in Fig. 5. With the employed sapphire/BBO/sapphire stack, the beam quality remained excellent (see the inset on the right side of Fig. 5), and the measured $M^2$ increased only slightly from $M^2=1.2$ at low power to $M^2=1.35$ at 100 W average power. In contrast, when the stack has been replaced by the 0.3 mm thick BBO crystal, strong thermal effects have been observed at average powers higher than 40 W. The corresponding beam profile measured at the focus at 55 W is displayed in the left inset of Fig. 5.

 

Fig. 5. Beam quality versus average power of the third harmonic generated with 990 fs IR pulses at 3.5 MHz repetition rate. Insets: measured intensity profiles at the focus.

Download Full Size | PDF

 

Fig. 6. Dots, measured cross-correlation between the infrared pulses and the generated ultraviolet pulses at 100 W of average power; line, Gaussian fit; inset, measured spectrum of the generated third harmonic.

Download Full Size | PDF

The measured cross-correlation between UV and IR at 100 W of average power is shown in Fig. 6. It indicates $\sim 730\mathrm{fs}$ pulse duration, assuming Gaussian pulses in the UV. In conclusion, we reported on a high average power femtosecond laser at 343 nm delivering femtosecond pulses at up to 100 W of average power.

This was enabled by generating the third harmonic of a 620 W femtosecond fiber laser in two BBO crystals. Sapphire heat spreaders attached to the nonlinear crystal via direct bonding have efficiently mitigated thermal effects in the BBO crystals. Thus, a nearly diffraction-limited beam quality was maintained up to the highest power level. This approach has the potential to improve thermo-optical properties of high average-power frequency converters within the whole spectral range from the UV up to the IR.

In the future, even higher UV average powers can be expected from kW average power IR driving lasers [5,7,8], and the conversion efficiency could be improved by controlling the temporal walk-off between the interacting light waves. In addition, dispersion management and subsequent pulse compression (e.g., with chirped mirrors [19]) will enable few-10 fs pulses with high average power in the UV. Furthermore, the pulse energy of the UV pulses can easily be scaled up to a mJ [10] with larger beam diameters and crystal apertures. Thus, multi-100 W, mJ-class, few-10 fs UV lasers are in reach. Such lasers would be ideal not only of drivers for micro-machining applications, but also for UV-driven strong field processes in fundamental science.