We report a 236 nm light source with 20 mW of average power based on critically phase-matched second-harmonic generation in a β-barium borate crystal at room temperature. The fundamental light source was a passively Q-switched 946 nm Nd:YAG laser tunable from 10 – 38 kHz and with a pulse length of 16 ns. In the generation of 473 nm light, periodically poled KTP and BiBO was compared in terms of conversion efficiency and stability.
©2007 Optical Society of America
Ultra-violet light sources find applications in a wide range of areas such as sub-μm lithography, micro-machining, fluorescence detection and imaging, and ophthalmology. Today, light sources used in these areas are mainly excimer lasers, but solid-state lasers are gaining territory because of advantages in efficiency, compactness and low system maintenance cost. However, the output power from solid-state UV lasers is often limited, especially in the shorter wavelength region. This is because the generation of UV radiation from solid-state lasers rely on cascaded frequency conversion, and the nonlinear efficiency is quite low for crystals transparent in the UV spectral region. The latter are primarily crystals from the borate class, e.g. β-barium borate (BBO) and cesium lithium borate (CLBO). Although CLBO has a nonlinear conversion coefficient which is several times higher than that of BBO for wavelengths shorter than 250 nm and has proven to be able to generate UV power in the order of several watts (both in pulsed regime [1, 2] and continuous wave [3, 4]), CLBO is also known for being highly hygroscopic and needs sophisticated mounting in order to protect it from moisture. Moreover, its physical and thermal properties make it unsuitable for AR-coating. For practical reasons, the remaining choice for deep UV generation at the moment is thus BBO.
To enhance the generated UV radiation, external cavities are often used to resonate one of the interacting frequencies [5, 6]. An alternative and simpler approach is the use of pulsed IR sources with high peak-power .
Most of the reported solid-state UV sources emit radiation at 266 nm, but for the fabrication of photonic components such as Bragg gratings and waveguides it is of importance to reach shorter wavelengths. This is because of the higher absorption in glass at wavelengths shorter than 250 nm, which increases the sensitivity for refractive index modification. Gerstenberger et al  reported efficient generation of 236.5 nm radiation using CLBO, but the long-term performance of this UV source is not known.
Here, we report a UV light source based on a frequency-quadrupled, passively Q-switched quasi-three-level Nd:YAG laser, using a straightforward and compact setup. For generation of 236.5 nm radiation, a BBO crystal was used. We also investigate the performance in terms of nonlinear conversion efficiency of several BiBO - and PPKTP crystals of different lengths. Two different phase-matching temperatures were also tried for the PPKTP crystals.
2. Passively Q-switched 946-nm Nd:YAG laser
The experimental arrangement of the passively Q-switched 946-nm Nd:YAG laser is depicted in Fig. 1. We used a simple, linear cavity design to ensure efficient lasing with good beam quality. A fiber-coupled 808-nm diode (Limo GmbH) was used as the pump source, emitting up to 21 W from a 200 μm fibre with a NA of 0.22. The laser crystal was a composite 5 mm long Nd:YAG (doping level 1.1 atm%) crystal with 5 mm long undoped YAG end caps. Its diameter was 3 mm and it was watercooled to a temperature of 15 °C. Both surfaces were AR-coated for 946 nm. Using two plano-convex lenses with focal lengths of 30 mm, a 1/e2-radius beam waist of approximately 100 μm was achieved 2 mm inside the Nd:YAG crystal. The pump absorption ranged between 65% and 77% for diode powers between 7 W and 13 W. A planar mirror with HR coating for 946 nm (AR for 808 nm and 1064 nm) was used as an incoupling mirror, and the output mirror was a curved (ROC = 200 mm) mirror with a transmission of 12 % for 946 nm (AR for 1064 nm).
A compact and practical way of Q-switching solid-state lasers is to use saturable absorbers, and in the 1 μm spectral region Cr4+:YAG is often chosen because of its high damage threshold . In order to achieve high peak powers, a low initial absorber transmission and a sufficiently small laser beam radius are desired, but these demands has to be balanced with the risk of damage on the coatings. In addition, the quasi-three level nature of this laser imposes conditions on the output coupler transmission and the initial transmission of the saturable absorber, because of the reabsorption losses. We investigated these aspects experimentally and made the choice of a 0.5 mm thin uncoated Cr:YAG disc with an initial transmission of 94% that was employed in Brewster’s angle, resulting in a linearly polarized beam with 16 ns-long pulses. For 7 W of absorbed power, the average output power was 0.94 W and the repetition frequency 22 kHz. This corresponds to a pulse energy of 43 μJ and a peak power of 2.9 kW. The beam quality factor M2 was ≤1.5. Increasing the pump power resulted in higher average power, but lower peak power, see Figs. 2(a) and 2(b). A typical pulse train at an average output power around 950 mW is shown in the inset of Fig. 2(a).
Further increasing of the pump power above 13 W degraded the overall performance of the laser in terms of average power, stability and beam quality. The relatively high output coupling of the laser mirror was chosen to minimize the risk of damaging the coatings of the surfaces, since the circulating energy density was high. Nevertheless, for absorbed powers >7W, the load on the (uncoated) surface of the Cr:YAG crystal was apparently so high that it caused damage on the surface. It was assumed that the water absorption around 946 nm was the reason, and to prevent this from occurring, the cavity was purged with N2 gas. This effort did indeed improve the situation and the laser could perform well also at higher pump levels (but still below 13W).
The damage threshold of the Cr4+:YAG crystal is thus the limiting parameter in this laser and consequently in the generation of 473-nm and 236-nm light. As for the average power, it is also limited by the onset of multi-transverse mode operation determined by the available pumping and laser crystal geometry.
3. Second harmonic generation using BiBO and PPKTP
Generation of 473-nm light was obtained by frequency doubling in BiB3O6 (BiBO) cut for type-I phase-matching (ee->o) with θ = 161.7° and φ = 90°, and periodically poled KTP (PPKTP) with two different nominal grating periods, 6.09 μm and 6.03 μm, designed for first-order quasi phase-matching at 25°C and 75°C, respectively. The BiBO crystals had lengths of 10 mm, 5 mm and 3 mm and the PPKTP crystals had lengths of 9 mm, 5 mm and 2 mm. The effective nonlinear coefficient (deff) of the BiBO crystals was 3.34 pm/V. The value of deff for the PPKTP crystals was measured by means of second-harmonic generation using a CW Ti-Sapphire laser, giving 9.7 pm/V < deff < 9.9 pm/V. All PPKTP crystals were uncoated whereas the BiBO crystals had AR-coating for both 946 nm and 473 nm.
To evaluate how different focussing conditions affected the output power and the conversion efficiency of the BiBO crystals, a lens with 40 mm focal length was placed at four different distances from the output mirror of the 946-nm laser, resulting in 10 μm, 15 μm, 20 μm and 30 μm 1/e2-radius beam waists. A dichroic mirror (HR for 946 nm and AR for 473 nm) and a BG39 filter were used to separate the IR from the 473-nm light. The maximum blue output power, measured for an average 946-nm power equal to 1.3 W, as function of focused beam waist is shown in Fig. 3(a) where it can be seen that the longest crystal (10 mm) gives the best results for every focussing condition. As expected from the small crystal acceptance angle (0.65 mrad cm), the generated 473-nm beam was more astigmatic for longer crystals. It should be noted that the optimum beam waist radii according to the Boyd-Kleinman criteria  are 17 μm, 12 μm and 9 μm for crystals of lengths 10 mm, 5 mm and 3 mm, respectively. Thus, it can be concluded that these focussing conditions lies close to those retrieved experimentally, although the latter are somewhat looser. In comparison to the theoretically calculated values for each crystal length, beam waist and 946-nm power (taking pump depletion into account) the experimental results are lower. For each of the three crystals, the maximum average power is around 80% of the theoretical values. Possibly this could be an effect of slightly imperfect fundamental beam quality.
In Fig. 3(b) the peak 473-nm power and the conversion efficiency are shown as function of peak 946-nm power for the 10 mm long crystal with a 20 μm focused beam waist radius. For all fundamental peak power levels, the second harmonic peak powers are lower than expected. Though for the lowest measurement point, the experimental conversion efficiency is within the same range as the calculated (20% compared to 23%), but for increased fundamental peak powers (which do not generally correspond to increased 946-nm average powers, see Fig. 2) the conversion efficiency reaches a saturation level whereas theory predicts an increasing conversion efficiency as function of fundamental power, and for the highest peak power level an efficiency of 32% is expected instead of the 19% experimentally obtained. In fact, the gap between theoretically expected values and experimental data widens as the 946-nm power increases. The reason for this behaviour is not evident, although it can be noted that the relatively small acceptance angle (0.65 mrad for the 10 mm crystal) in combination with a somewhat degraded mode quality of the 946-nm beam for high diode pump power can play a significant role. Whereas the Poynting vector walk-off effect is taken into account in the Boyd-Kleinman theory, the crystal acceptance angle is not and would thus affect the value of the focussing parameter h(ξ, B). Since the acceptance angle bandwidth works as a confined aperture and would thus decrease the efficiency homogeneously for all peak powers, experimental data suggests that there is an additional parasitic effect. We see mainly two candidates for this: nonlinear (absorption) losses and/or dephasing effects due to linear absorption of the blue light leading to heating of the crystal and tuning away from phase-matching. As for other effects, such as photorefraction and greytracking, no evident signs could be seen. In all cases, the power stability was better than 1% and no crystal damage occurred at any power level.
Similar experiments were then carried out for the PPKTP crystals. Crystals with two different phase-matching temperatures, 25°C and 73°C were compared in terms of output power and stability. The high fundamental peak powers in combination with long crystal length can have serious drawbacks because of linear absorption and the photocromic effect in KTP (which gives rise to temperature gradients), causing large instabilities in output power and impaired conversion efficiency. It has earlier been experimentally verified that an elevated crystal temperature can diminish the photochromic effect [10, 11, 12]. These issues will be discussed in more detail below.
In order to avoid crystal damage, rather loose focusing was used, where the 946-nm beam was imaged to form circular beam waist radii of 45 μm, 65 μm and 85 μm, respectively, in the PPKTP crystals using a lens with a focal length of 100 mm. Large instability in the 473-nm average power was observed for both the 9 mm-long crystals under all focusing conditions, and the power level did not exceed 370 mW in any of them. As for the 2 mm-long crystals, the power stability was good, but rather low blue output power levels were measured. An overview of the results achieved is displayed in Table 1.
The highest average 473-nm power and conversion efficiency were achieved for the 5 mm long PPKTP crystal having a poled period of 6.03 μm and a focused beam waist radii of 65 μm at an average 946-nm power of 1.1 W, giving values of 405 mW and 41 %, respectively. The power stability was within 3 %. Harder focussing did not result in larger conversion efficiency but instead increased the power instability. The average second harmonic power and conversion efficiency are plotted in Fig. 4(a) as function of average fundamental power, where the Fresnel losses of the uncoated PPKTP crystal has been taken into account in the calculation of conversion efficiency. The heat-sink temperature was constant throughout the power measurement. An evident saturation in conversion efficiency can be seen. Also, the conversion efficiency for the lowest average powers is appreciably lower than expected. We believe that this is due to a temperature mismatch at low fundamental power. As for the saturation behaviour, a contributing factor is the non-negligible absorption of the visible light, which grows worse for shorter wavelengths . Apart from the 473-nm loss itself, this also causes temperature gradients in the crystal, similar to what has been observed by Pasiskevicius et al . The heat will induce a varying refractive index change in the crystal by the thermo-optic effect, and together with the average temperature variation itself, the phase-matching condition will be distorted. Thus, instability in second harmonic power will follow. Furthermore, the photochromic effect in KTP will cause absorption of the infra-red light, referred to as blue induced infrared absorption (BLIIRA) . This absorption will result in similar effects mentioned above for the visible light absorption, in addition to the decrease in fundamental power available for nonlinear conversion. The combined effect of these phenomena will limit the generated second harmonic power.
In Fig. 4(b) we display the measured phase-matching curve for the 5 mm-long crystal, showing signs of distortion compared to the theoretical sinc2-curve. This is a typical response when temperature gradients are present . From the numerical fit it is deduced that the phase-matching temperature T0 was 73°C and the FWHM temperature bandwidth ΔTFWHM was 7.0°C. The latter value is in fair agreement with the calculated bandwidth of 5.0°C, which would correspond to an effective crystal length of 3.6 mm. For comparison, we made a similar measurement using a CW Ti-Sapphire laser (power level around 400 mW) resulting in the phase-matching curve displayed in Fig. 4(c). This curve is evidently very close to an ideal sinc2-function, apart from the somewhat asymmetrical side-peaks. From the numerical fit a phase-matching bandwidth of 5.2°C was deduced, in excellent agreement with theory. The main reason for the difference in these phase-matching curves is the higher average power of the Q-switched 946-nm laser which leads to more absorbed power, resulting in a larger temperature elevation and shorter effective crystal length.
It should also be mentioned that catastrophic damage occurred in the 2 mm-long, 6.03 μm-poled PPKTP crystal for a focused beam waist radius of 38 μm, which would correspond to a damage threshold of 92 MW/cm2 (fundamental peak intensity, with Fresnel losses taken into account). This value corresponds well to the previously recorded data . It should also be noted that no grey-tracking was observed in any of the PPKTP samples.
In retrospective, a shorter poling period (which would mean a higher phase-matching temperature) might be more effective in terms of avoiding temperature gradients in the crystal and thus improving the power stability. Also, better thermal contact between the heat sink and PPKTP crystal could be a means of improving the power stability.
It can thus be concluded that although BiBO can provide 473-nm light with good power stability and relatively high conversion efficiency at room temperature, PPKTP is still the better choice because of its larger nonlinear coefficient.
4. UV generation
The 473-nm light generated from the 5 mm long PPKTP crystal (Λ = 6.03 μm) was used for UV generation by single-pass frequency doubling in a 3×3×3 mm3 β-BaB2O4 (BBO) crystal cut for type-I (oo->e) phase-matching (θ = 57.5°). The crystal was AR-coated for 473 nm and 236.5 nm. The phase-matching temperature was 23°C and the nonlinear coefficient deff was 1.5 pm/V. A fused silica lens with focal length 40 mm was used to focus the blue light down to a 21 μm 1/e2-radius beam waist in the BBO crystal, which corresponds to loose focusing in the Boyd-Kleinman theory. To measure the 236-nm light, a fused-silica prism and a Schott glass filter (UG5) was inserted after the BBO crystal to separate the UV from the residual blue light. With filter and absorption losses accounted for, the maximum 236.5-nm average power was 19.6 mW. Given that the 473-nm average power was 405 mW, this corresponds to a conversion efficiency of 4.8%. Average UV power and conversion efficiency as function of average 473 nm power is presented in Fig. 5, together with numerical fits. The 473-nm power was increased by turning up the pump power of the 946-nm laser. As is obvious from the figure, no saturation effects occurred at any power levels.
It can be noted that harder focussing (down to 12 μm given by the Boyd-Kleinman criteria) did not increase the output power, which can be explained by the small acceptance angle in the bc-plane in BBO (0.17 mrad cm). For several applications, good beam quality and low beam divergence is of importance. Using a camera with a pyroelectric sensor, we recorded an image of the UV beam, displayed in Fig. 6. Only a few side fringes were detected in the direction of walk-off, as expected, and we estimate that most of the power was contained in the central peak. The far-field diffraction angles were approximately 0.5 mrad × 1.5 mrad (full angle).
5. Conclusions and outlook
In conclusion, we have built a compact and efficient all solid-state UV light source with a wavelength of 236 nm and an average power of 20 mW. The experimental setup is simple and straightforward, using a high peak power passively Q-switched 946-nm Nd:YAG laser with pulse rates between 10 – 38 kHz.
Two different nonlinear crystals, BiBO and PPKTP were compared for the 473 nm second harmonic generation stage. Although BiBO was shown to provide relatively high nonlinear conversion efficiency with good power stability at room temperature, the performance of PPKTP phase-matched at 73°C was better in terms of power and beam quality. A maximum of 405 mW of 473-nm was accomplished from a 5 mm-long PPKTP crystal, corresponding to a conversion efficiency of 41 %. To increase the output power and stability of the visible output from PPKTP, it might be effective to use a shorter poling period, corresponding to a higher phase-matching temperature, to further reduce the photochromic effect. Another optimization step would of course be to increase the peak and average power of the 946-nm laser (and thus scaling up the 473-nm and 236-nm power), but this would require a careful design of the laser cavity to particularly circumvent the risk of damage of the Cr4+:YAG crystal.
The authors would like to acknowledge the support from the Lars Hierta Foundation, the Carl Trygger Foundation and the Helge Axson Johnson Foundation.
References and links
1. T. Kojima, S. Konno, S. Fujikawa, K. Yasui, and K. Yoshizawa, “20-W ultraviolet-beam generation by fourth-harmonic generation of an all-solid-state laser,” Opt. Lett. 25,58–60 (2000). [CrossRef]
2. J. Sakuma, K. Deki, A. Finch, Y. Ohsako, and T. Yokota, “All-solid-state, high-power, deep-UV laser system based on cascaded sum-frequency mixing in CsLiB6O10 crystals,” Appl. Opt. 39,5505–5511 (2000). [CrossRef]
3. J. Sakuma, Asakawa Y., and M. Obara, “Generation of 5-W deep-UV continuous-wave radiation at 266 nm by an external cavity with a CsLiB6O10 crystal,” Appl. Opt. 39,5505–5511 (2000). [CrossRef]
4. J. Sakuma, Y. Asakawa, T. Imahoko, and M. Obara, “Generation of all-solid-state, high-power continuous wave 213-nm light based on sum-frequency mixing in CsLiB6O10,” Appl. Opt. 39,5505–5511 (2000). [CrossRef]
5. L. Chang, S. Wang, and A. Kung, “Efficient compact watt-level deep-ultraviolet laser generated from a multi-kHz Q-switched diode-pumped solid-state laser system,” Opt. Commun. 209,397–401 (2002). [CrossRef]
6. M. Oka, L. Liu, W. Wiechmann, N. Eguchi, and S. Kubota, “All solid-state continuous-wave frequency-doubled Nd:YAG laser,” IEEE Quantum. Electron. 1,859–865 (1995). [CrossRef]
8. J. Dong, “Numerical modeling of CW-pumped repetively passively Q-switched Yb:YAG lasers with Cr:YAG as saturable absorber,” Opt. Commun. 226,337–344 (2003). [CrossRef]
9. G. Boyd and D. Kleinman, “Parametric interaction of focused Gaussian Light Beams,” J. Appl. Phys. 39,3597–3639 (1968). [CrossRef]
10. V. Pasiskevicius, H. Karlsson, F. Laurell, R Butkus, V. Smilgevicius, and A. Piskarskas, “Highly efficient optical parametric oscillator in red spectral region with periodically poled KTP,” Opt. Lett. 26,710, (2001). [CrossRef]
11. S. Wang, V. Pasiskevicius, and F. Laurell, “Dynamics of green light-induced infrared absorption in KTiOPO4 and periodically poled KTiOPO4,” J. Appl. Phys. 96,2023–2028 (2004). [CrossRef]
12. J. Hirohashi, V. Pasiskevicius, and F. Laurell, “Picosecond blue light-induced infrared absorption in single-domain and periodically poled ferroelectrics,” Submitted to J. Appl. Phys.
13. G. Hansson, H. Karlsson, S. Wang, and F. Laurell, “Transmission measurements in KTP and isomorphic compounds,” Appl. Opt. 39,5058–5069 (2000). [CrossRef]
14. V. Pasiskevicius, S. Wang, J. Tellefsen, F. Laurell, and H. Karlsson, “Efficient Nd:YAG laser frequency doubling with periodically poled KTP,” Appl. Opt. 37,7116–7119 (1998). [CrossRef]
15. S. Spiekermann, F. Laurell, V. Pasiskevicius, H. Karlsson, and I. Freitag, “Optimizing non-resonant frequency conversion in periodically poled media,” Appl. Phys. B. 78,211–219 (2004). [CrossRef]
16. S. Wang, Fabrication and characterization of periodically-poled KTP and Rb-doped KTP for applications in the visible and UV (Doctoral Thesis in Physics, Stockholm, Sweden, 2005), http://www.laserphysics.kth.se.