1. Introduction

High-power continuous-wave (cw) green light sources have numerous applications such as laser displays, material processing, biological investigations and laser surgery. So far the high power lasers in the green spectral region rely on frequency doubling of near IR solid-state lasers. The realization of efficient doubling has mainly followed three avenues. The first of those is based on second-harmonic generation (SHG) in frequency-locked external enhancement cavities, which has been demonstrated with conversion efficiencies close to 90% at more than 100 W output power in the green [1, 2]. However, these schemes inevitably depend on single-frequency pump sources and proper frequency stabilization of the cavity, which ultimately limits the overall flexibility of such systems and ads to their overall complexity. Although, intracavity-doubled schemes [3], representing the second avenue, are not limited to a single longitudinal mode operation, special measures have to be taken to reduce the intensity noise being introduced by the added non-linear dynamics in the cavity. In contrast to that, the third approach simply employs SHG in quasi-phase matched (QPM) devices, which capitalizes on the possibility of using the highest available non-linearity in the materials and can therefore achieve substantial conversion efficiencies for comparably short crystal lengths. Specifically, output powers of close to 20 W with conversion efficiencies of up to 30% at have been demonstrated in QPM devices, employing single-pass SHG [4, 5]. And even though, multipass [6] and multicrystal [7] approaches showed improved efficiencies of up to 47% and 55%, respectively, the ultimate limit in terms of absolute power is set by the thermal handling capability of the employed nonlinear material. Consequently, considerable research on the subject of which material is best suited for the task has been conducted, where single-pass frequency doubling of the Nd:YAG laser line at 1064 nm has served as a benchmarking experiment to evaluate the performance of different nonlinear materials, namely LiNbO₃(LN), LiTaO₃ (LT) and KTiOPO₄ (KTP) [4–6, 8–16].

A comprehensive summary of those earlier reported results can be found in Fig. 1, which also includes the results for PPKTA, presented in this contribution. Evidently, the initial limitation of photorefractive damage (PRD) and green induced infrared absorption (GRIIRA) in congruent LN isomorphs, could only be partially mitigated by MgO-doping [17], thus restricting the maximum achievable output power for these devices to 3 W. However, considerable effort allowed the growth of stoichiometric LT (SLT), which further decreased the susceptibility of LT-based QPM devices to PRD and GRIIRA, and ultimately allowed to achieve the earlier mentioned powers of nearly 20 W at 532 nm [4, 5]. The major limitation for power scaling in these devices was reported to be the onset of thermal dephasing attributed to linear absorption at the generated visible wavelength, where the good thermal conductivity of SLT (8.8 W/(mK) [15],) proved advantageous for heat dissipation. In any case, these results suggest that nonlinear materials with superior transparency are crucial for further power scaling of QPM devices.

 

Fig. 1 Overview of reported results on single-pass high-power continuous-wave frequency doubling close to or at the Nd:YAG laser line of 1064 nm in QPM devices, ordered by (a) absolute conversion efficiency and (b) normalized conversion efficiency

Download Full Size | PDF

Although, the measured thermal conductivities for KTP isomorphs (2-3 W/(mK)) are lower than for SLT, cw QPM devices based on KTP also demonstrated powers of up to 6 W [13], where similar thermal limitations where observed. However, the prospect of some KTP isomorphs, such as KTA and RbTiOAsO₄, possibly exhibiting lower linear absorption in the visible range than KTP [18], warrants further investigation of these materials for high-power cw QPM devices. Especially, since these materials exhibit better commercial availability than high-quality SLT.

While readily available and used for birefrigent phase-matched mid-infrared optical parametric devices at 1 µm, KTA has so far attracted little commercial interest for SHG devices in the visible range, due to the cumbersome handling of the poisonous As-content in the flux during the crystal growth process and the difficulties in growing single domain crystals. Moreover, birefrigent type II phase-matching is restricted by the short wavelength cut-off around 1075 nm. Nonetheless, periodically poled domain structures in KTA for the visible spectral range are feasible and have recently been demonstrated [19]. Building on these results, this contribution investigates the performance of PPKTA in high power cw single-pass SHG, and explores its potential for future high-power QPM devices for the visible range. In particular, we demonstrate a device delivering 13.6 W output power at 533 nm with a conversion efficiency of 25% in an 8 mm long crystal.

2. Characterization of linear transmission properties in the visible range

The earlier reported data on the linear absorption properties of KTA were obtained using samples from a single vendor [18]. To confirm these results and to gather data with greater general significance, KTA samples from three different vendors from overall four different batches were acquired and their respective linear absorption was measured. All samples were c-cut with dimensions 10x5x1 mm³ and labeled according to their respective measured ionic conductivities (low-conductive – LC; high-conductive – HC; super high-conductive – SHC), see Table 1.

Table 1. Sample conductivities and linear absorption coefficient α at the SHG wavelength of the 1064 nm Nd:YAG laser line for all samples

View Table

The linear absorption of the samples in the visible range was characterized using a Varian Cary 50 absorption spectrophotometer (Agilent Technologies), see Fig. 2.The broadband light emitting from a Xenon lamp was polarized parallel to the c-axis of the crystal and launched along the a-axis, after passing an aperture assuring the crystal samples covered the entire cross section of the probe beam.

 

Fig. 2 Transmission loss measurements with Cary50.

Download Full Size | PDF

Prior to placing the samples in the beam path, reference measurements without the samples were performed. Thus transmission data was obtained, which in turn lead to absorption data by taking Fresnel losses at the crystal surfaces into account. The Fresnel losses were estimated using Sellmeier data from [20]. The reduced light intensity due to the obscured signal path and the required reference measurement lead to a measurement accuracy of the setup of about 0.002 cm⁻¹.

Results of the described measurements are presented in Fig. 3. Apparently, the 4 different samples exhibit considerable differences around the UV absorption edge. While both samples V1L and V1H provided by the first vendor show very low absorption of less than 0.01 cm⁻¹ over almost the entire visible spectral region down to 440 nm, the samples V2H and V3SH display significant absorption above that limit already at 580 nm and 680 nm, respectively. Interestingly, both samples from the first vendor, which differ significantly in conductivity, display superior transmission properties with respect to the other two samples, suggesting that any variations in the speed of the crystal growing process and the accompanied differences in ionic conductivity are not linked with any detrimental absorption in the crystals. It should be noted that the ionic conductivity is primarily associated with hopping motion of potassium ions and is proportional to the concentration of potassium vacancies.

 

Fig. 3 a) Transmission and b) absorption spectra in the E || z polarization direction for all KTA samples.

Download Full Size | PDF

The observed increased absorption in samples V2H and V3SH can therefore likely be attributed solely to impurities, being added either unintentionally during the growth process or intentionally to promote single-domain growth [21]. Nonetheless, the exceptionally low linear absorption at 532 nm earlier measured in [18], could be confirmed for samples V1L and V1H, see Table 1.

3. Experimental setup for high-power SHG experiments

Successful periodic poling could only be achieved in samples from two of the four investigated batches, namely V1L and V2H. In the case of sample V3SH the reversal of the spontaneous polarization did not succeed, since the available poling equipment could not supply sufficient current to compensate for the strong ionic conductivity that screens the polarization switching. In the case of the high-conductive samples from the first vendor (V1H), switching was feasible; however, crystal quality-related irregularities prevented high-fidelity periodically-poled structures. Consequently, SHG generation experiments were restricted to samples V1L and V2H, which achieved normalized conversion efficiencies and effective non-linearities of 1.03%/Wcm, d_eff = 9.7 pm/V and 1.2%/Wcm, d_eff = 10.5 pm/V, respectively. The active poling area for all samples was 8x3.5 mm² along a- and b-direction of the crystals. Due to the existing uncertainties in the available Sellmeier coefficients for KTA, a grating period of 8.49 µm was chosen, which phase-matched room-temperature SHG at 1066 nm, well within the tuning range (1064 – 1073 nm) of the available high-power pump laser. Temperature acceptance bandwidths ΔT_FWHM were measured to be 5.7 K for V1L and 5.6 K for V2H.

Prior to all high-power single-pass SHG experiments, the pump laser was characterized to ensure optimal operation at the previously reported parameters [22], i.e. maximum output power >100 W, spectral width FWHM <60 pm, beam quality factor M² <1.2 and polarization extinction ratio >18 dB. The pump light was focused into the different crystal samples reaching a beam waist diameter of 50 µm, which corresponds to a focusing parameter of ξ = 1.1. For proper thermal contact and heat removal the crystals were wrapped in indium foil and housed in an aluminum holder, which was ultimately mounted in a temperature controlled crystal oven (Eksma CO1). To prevent any detrimental feedback to the pump source, the crystals were slightly tilted with respect to the pump laser beam axis, which became necessary due to the lack of an isolator with sufficient power handling capability. A full setup of the SHG scheme is displayed in Fig. 4.

 

Fig. 4 Setup of the SHG pumping scheme.

Download Full Size | PDF

4. Experimental results and analysis

Since no AR coating was present on the end facets of the crystals, the measured values for both launched pump and green output powers were corrected for Fresnel losses at the crystal surfaces. Hence, all reported values correspond to actual power levels in the chip. To analyze the experimental data, we employed a model based on the numerical solution of spatially resolved coupled wave equations. The model assumes perfect phase matching along the entire crystal and does therefore not take any dephasing effects into account. Pump depletion, the focusing arrangement and linear absorption of pump and SH signal, however, are included in the model. Both numerical and experimental results are displayed in Fig. 5.

 

Fig. 5 Comparison of experimental and numerical results for a) second-harmonic signal vs launched fundamental power and b) second-harmonic conversion efficiency vs launched fundamental power.

Download Full Size | PDF

As the absorption measurements suggested, the experimental results for sample V2H start deviating from the numerical prediction at considerably lower power levels, in comparison to results for sample V1L, due to earlier onset of absorption-induced thermal dephasing. The necessary adjustment of the phase-matching temperature for increasing pump powers is documented in Fig. 6, and is further indication for the delayed dephasing in sample V1L.

 

Fig. 6 Phase-match temperature offset vs generated SH power.

Download Full Size | PDF

As a result, power scaling of up to 13.6 W of green power at 25% conversion efficiency was possible in sample V1L, while power scaling in sample V2H was limited to 5 W at 17% conversion efficiency. For both samples, the upper limit for generated green power was set by the onset of increasing power instabilities of the generated green power, i.e. the peak-to-peak power stability measured over 1 hour for sample V1L increased from 8.5% at 11.4 W to 12% at 13.6 W. Further increase in pump power even leads to isolated peak-to-peak instabilities of larger than 30%, which were determined to be associated with slow response of the oven and temperature controller, which in turn lead to chaotic oscillations of the SHG power. Consequently, correction of the thermal management scheme, such as reducing the device’s aperture size as in [5], promises considerable power improvements.

Furthermore, spectral and spatial properties of the second harmonic beam were characterized using an optical spectrum analyzer and the knife-edge method, respectively. The FWHM spectral width of the green signal was with 75 pm slightly broadened with respect to the pump laser (60 pm), but remained constant at all power levels. Similarly, the characterization of the spatial beam profile was conducted at all power levels, and corroborated that the excellent spatial properties of the fiber-based pump source were preserved in the second harmonic beam. All measured M² – values were below the specified upper limit of the pump lasers M² of 1.2, constituting the generated SH signals suitability as pump source for other laser or parametric processes.

5. Conclusion and outlook

Low power transparency measurements on KTA samples from different vendors were performed to confirm the previously reported correlation between power scaling capabilities of QPM devices and the crystal’s absorption properties in the visible range [5, 13, 16]. Interestingly, considerable differences were observed for the various KTA samples, where increased absorption was in all likelihood related to added impurity content in the respective samples. A fact, that also warrants further investigation of possible performance enhancements of QPM devices based on the cheaper and better studied KTP, since the question of whether improved crystal growth processes can decrease the linear absorption of KTP in the visible range remains open.

Nonetheless, two successfully periodically-poled KTA samples were employed for high-power SHG of green light. A comparison of SHG characteristics for both samples showed the expected behavior with delayed onset of detrimental thermal effects in the sample possessing superior transparency at the SHG wavelength. Ultimately, we were able to generate 13.6 W at 533 nm at a conversion efficiency of 25%, limited by increasing power fluctuations, due to insufficient stabilization of the temperature distribution within the crystal. Although, the reported power levels represent the highest generated green power in a cw QPM device based on any KTP isomorph, we expect substantial improvement in both output power and the conversion efficiency, by optimizing the employed crystal geometry. In particular, longer crystals with smaller apertures should not only improve the small signal efficiency, but also augment heat removal and temperature stabilization within the crystals considerably [5].