The development of high-repetition-rate ultra-short laser sources providing extremely intense pulses, especially in the ultraviolet (UV) region, is of great importance in microscopy, ultrafast spectroscopy, quantum optics, and scientific research^[1–3]. The short wavelength with high single-photon energy allows rapid absorption and a small focal spot, potentially enabling high machining accuracy and patterning ability even in wide bandgap materials^[4].

In recent times, extensive research has been conducted on nonlinear frequency up-conversion of picosecond lasers generating high-power UV^[5,6] and blue light^[7,8] through third harmonic generation (THG) and second harmonic generation (SHG), respectively. Compared with picosecond lasers, femtosecond lasers have much narrower pulse duration and higher peak intensity and can be powerful tools in laser precision machining^[9,10]. Recently, there have been several reports on the development of femtosecond UV laser sources based on nonlinear frequency conversion in femtosecond IR lasers^[11–14]. However, the temporal walk-off effect caused by group velocity mismatch will reduce the overlap of three interacting waves in the sum-frequency generation (SFG) crystal, decreasing conversion efficiency. Traditionally, a delay line is used to control the temporal overlap after the unconverted pump, and second harmonic (SH) beams are separated using a dichroic mirror^[15,16]. In this way, the THG results of up to 22% near-IR-to-UV conversion efficiency are obtained based on an Yb fiber laser with 5 W average power at 260 fs pulses^[15]. The most common drawbacks of such a delay line system lay with their bulky size and complicated system design. Besides, the delay line containing separated and recombined beams is difficult to align and sensitive to minor disturbances in the environment.

Alternatively, introducing one or multiple compensation plates (CPs) with relative anomalous dispersion enables tabletop UV laser systems with simple series configuration and high conversion efficiency. The potential of CPs can be seen by the impressive 800 to 270 nm THG efficiencies of 5.3%^[17] and 8.48%^[18] with Ti:sapphire laser systems. However, the repetition rates of high-energy femtosecond Ti:sapphire lasers are limited to a few kilohertz, which is not suitable for material processing applications^[19–21]. In addition, the bulky design and high price make them less competitive compared with fiber lasers. In particular, the rapid development of Yb-doped femtosecond fiber lasers has brought about a new generation of high-power, compact, stable, and cost-effective laser sources for scientific research and industrial applications^[22–24]. However, to the best of our knowledge, there are no studies to date on compact and practical UV lasers based on femtosecond fiber lasers with CPs.

In this article, we present a cascaded THG scheme with a CP of β-BaB₂O₄ (BBO) crystal based on a home-made Yb-doped femtosecond fiber laser. The thickness and cutting angle of the CP are optimized to reverse both the spatial and temporal walk-off. By optimizing the parameters of the CP and incident light intensity, the efficient generation of UV output is achieved with a power as high as 2.23 W and a near-IR-to-UV conversion efficiency of ∼23%. To the best of our knowledge, this is the first experimental report on a UV source at 345 nm based on an Yb-doped femtosecond fiber laser with the CP providing output power >2W at 1 MHz. The tabletop UV source has a power stability of 0.67% RMS and almost no power degradation over 6 h. By contrast, the maximum UV output power with the complex traditional delay line is 1.9 W due to the loss of two dichroic mirrors and multiple reflections, and the power stability is 1.32% RMS with a clear drift at a rate of approximately 0.73% per hour.

Figure 1 illustrates a schematic diagram of the experimental arrangement. The system employed a home-made Yb-doped femtosecond fiber laser based on the chirped-pulse amplification technique, producing maximum output pulse energy of 10 µJ. The laser can be operated from 100 kHz to 1 MHz through an acousto-optic modulator, which implies that the maximum average power is 10 W at a 1 MHz repetition rate. Figure 2(a) shows a spectral width of 14.6 nm (1/e2) centered at 1032.6 nm. The pulse width is approximately 420 fs. Figure 2(b) shows the M2 factors (M2<1.2), and the inset shows the near-field beam profile of the pump laser. The optimization of THG was performed at a 1 MHz operational frequency.

The input power to the nonlinear crystals was controlled using a half-wave plate (λ/2) and a thin-film polarizer (TFP). The maximum pump power incident onto the SHG crystal was 9.7 W after reflections of the TFP and plano mirror. A 3 mm long and 5mm×5mm aperture LiB3O5 (LBO) crystal cut at θ=90° and φ=13.2° in the optical x−y plane for critical phase matching was chosen as the SHG crystal to match industrial applications. The pump beam was focused at the center of the SHG crystal using an achromatic lens, F1, with a focal length f1=200mm when a trade-off between the conversion efficiency and damage threshold of the SHG crystal was considered. The achromatic lens was optimized to minimize the chromatic aberration across a broad bandwidth, and then the high beam quality of the SH beam was obtained. The measured beam waist diameter of the fundamental beam is ∼112 µm. We achieve a maximum SHG conversion efficiency of 42% corresponding to 4.1 W average power at 517 nm in the LBO, as shown in Fig. 3(a). The inset in Fig. 3(a) shows the SH beam profile at the maximum average output power, and the measured ellipticity value is ωminor/ωmajor=0.93. Figure 3(b) shows the measured spectrum of the SH beam with a spectral bandwidth of 5.5 nm (1/e2). The measured beam quality factors are Mx2=1.48 and My2=1.44 [Figs. 3(c) and 3(d)].

For the SFG of the unconverted pump and the SH beam into UV radiation at 345 nm, we used a 3 mm long and a 4mm×4mm aperture BBO crystal cut at θ=32.1° for type I(o+o→e) critical phase matching due to its high effective nonlinear coefficient deff (deff=2.01pm/V, SNLO software package) value. To ensure the phase matching of type I, a dual-wavelength waveplate (DW) was used to rotate the polarization of the fundamental light individually to coincide with the polarization of the SH beam. The collimation (F2) and focusing (F3) lenses were placed in front of the BBO crystal to achieve optimum generation conditions, which satisfied the 4f condition to minimize the aberration. To efficiently produce UV light, a CP was used to reverse both the spatial and temporal walk-off between the pump and SH pulse after SHG.

As shown in Fig. 4(a), the CP can be used to reverse both the spatial and temporal walk-off due to its birefringence effect. To obtain a good compensation effect, each component’s spatial and temporal walk-off should be theoretically calculated. First, the walk-off angle of LBO is 8.0 mrad, and the walk-off distance is Δd=3×8.0=24µm. There is no spatial walk-off between the pump and the SH beams in the type I SFG crystal. Second, the theoretical delay time induced by each component between the central wavelengths of the fundamental and the SH beams is given in Table 1^[25]. The focal lengths of F2 and F3 were both 100 mm. Hence, to control the spatial and temporal overlap, the CP was designed to compensate for the spatial deviation of the SH beam (Δd=24µm) and the delay time before the BBO crystal. Traditionally, the CP is a birefringent crystal that is cut with an appropriate length and angle, such as BBO and calcite^[26]. We chose BBO here because of its higher damage threshold and better surface quality. Figure 4(b) shows the group refractive index of SH (blue curves) and fundamental (red curves) beams as a function of cutting angle θ in the optical x−z plane of BBO. When the SH and fundamental beams are e and o waves, respectively, the SH pulses transmit faster than the fundamental beam in the BBO plate designed at a certain angle greater than 38.37° (purple-shaded region). The temporal and spatial compensation of a 1-mm-thick BBO crystal with different cutting angles θ is illustrated in Fig. 4(c). Hence, by optimizing the thickness L and cutting angle θ, the CP can compensate for the required space and time delay.

To obtain optimum SFG efficiency, we used CPs with different compensation capabilities (Table 2). Each CP is distinguished for different delay times because the temporal walk-off dominates before SFG. Then, the UV output power for different CPs with f2=100mm was measured at the maximum pump power, as shown in Fig. 5(a). Different focusing lenses (F3) were used to optimize the incident light intensity inside the BBO for maximizing the output UV power. Three lenses with focal lengths of 100, 125, and 150 mm were tested. Obviously, the output UV power is the highest using f3=100mm. The lens system with f2=f3=100mm is 1:1, resulting in pump and SH beam waist diameters of ∼112 µm and 97 µm, respectively. If not compensated, that is Δt=0, the measured UV output power is 1.25 W. The output powers at 345 nm, PUV, of CP1 and CP2 are 1.9 and 2.1 W, respectively. The UV output power increases with Δt until the maximum output power is obtained using CP3. Subsequently, the output power of the UV pulses begins to decline due to overcompensation. Figure 5(b) shows the average output power and conversion efficiency for CP3 as functions of pump power. The UV power and UV efficiency increase with the increase of the pump power, resulting in a maximum UV power of 2.23 W, corresponding to a single-pass near-IR-to-UV conversion efficiency as high as 23%. The inset in Fig. 5(b) shows the UV beam profile whose measured ellipticity value is 0.78 at maximum average output power, and Fig. 5(c) shows its corresponding spectrum. To further investigate the effect of these CPs, we measured the UV power for different temporal walk-off induced by different lenses, as shown in Fig. 5(d). The system of f2=f3=125mm has the same spot size in the BBO as f₂ = f₃ = 100 mm. The center thicknesses of the f=100mm and f=125mm focal length lenses are 2.2 and 3.3 mm with 223.5 and 335.3 fs delay times, respectively. A maximum UV power output of 2.15 W was achieved by the system of f2=f3=125mm using CP4. Despite the temporal walk-off of the system being different, high output UV power can still be obtained with the CP.

By contrast, we reproduced the experiment by using the conventional delay line with f2=f3=100mm. Ideally, the tunable delay line could provide a continuously variable delay time, and thus higher UV output power could be achieved. Unfortunately, the maximum UV output power (efficiency) is 1.9 W (19.6%) due to the loss of two dichroic mirrors and multiple reflections. Figure 6(a) shows the output power at 345 nm and conversion efficiency as functions of pump power. The inset in Fig. 6(a) shows the UV beam profile at the maximum average output power, and the measured ellipticity value is 0.84. The elliptic UV beam profile can be attributed to the spatial walk-off of the beam and thermal distortion caused by two-photon absorption in the BBO crystal. Finally, we measured the power stability of the UV source. The tabletop UV laser system was operated for 6 h in an open environment. As Fig. 6(b) shows, there is almost no power degradation over 6 h by the CP method. The periodic fluctuation of the UV output power may be due to the periodic change of the ambient temperature because the measurement was done without any temperature control. The power stability is 0.67% RMS and 3% peak-to-peak power fluctuation. By comparison, the delay line’s power stability is 1.32% RMS and 7% peak-to-peak power fluctuation, with a clear drift at a rate of approximately 0.73% per hour. The decrease in power stability compared with the CP case can be attributed to complicated system design containing separated and recombined beams.

In conclusion, we have designed a compact and efficient THG scheme with a simple CP of BBO crystal based on a femtosecond Yb fiber laser source. The thickness and cutting angle of the CP are optimized to reverse both the spatial and temporal walk-off. By optimizing the parameters of the CP and incident light intensity, a maximum UV output of 2.23 W is obtained at a repetition rate of 1 MHz when the compensation time induced by the CP is −712 fs, corresponding to a single-pass near-IR-to-UV conversion efficiency as high as 23%. The tabletop UV laser system has a power stability of 0.67% RMS in 6 h with almost no power degradation. By contrast, the maximum UV output power with the complex traditional delay line is 1.9 W due to the loss of two dichroic mirrors and multiple reflections, and the power stability is 1.32% RMS with a clear drift at a rate of approximately 0.73% per hour. In the future, even more compact THG structures based on the CP method without collimation and focusing lenses can be expected from hundreds of microjoules (µJ) pulse energy IR driving lasers. The method shows remarkable potential for applications in stable, compact, and efficient UV generation.