CdSiP₂ Optical Parametric Oscillator Pumped by a Nanosecond Ho:LLF Laser at 2.06 µm with Non-Collinear Phase-Matching

Abstract

We report on a nanosecond-pulsed CdSiP₂ (CSP) optical parametric oscillator (OPO) pumped by a Q-switched 2.06 µm Ho:LiLuF₄ (Ho:LLF) laser in a compact linear cavity with a 10 mm long crystal. Utilizing a non-collinear type-I phase-matching configuration, the OPO achieved an average output power exceeding 3 W in the mid-IR region with good beam quality. The idler and signal waves can be tuned in a wide range of 3.5–4.7 µm by pump tilt and crystal rotation.

1. Introduction

Nonlinear optical (NLO) crystals have long played a pivotal role in mid-infrared (mid-IR) generation, with zinc germanium phosphide crystals (ZnGeP₂ or ZGP) being the workhorse in powerful mid-IR sources for many years [1,2,3,4]. Their properties enable efficient mid-IR generation where direct emission methods are lacking [5]. Recently, cadmium silicon phosphide (CdSiP₂ or CSP) has emerged as a potential alternative NLO crystal with similar mid-IR generation capabilities [6]. CSP is an optically negative uniaxial chalcopyrite, with nonlinear and dispersive properties that allow phase matching for frequency down-conversion with 1–2 µm pumps [7,8,9,10,11,12,13].

Since CSP crystals of a sufficient length and clear aperture have been obtained for the characterization of intrinsic bulk properties and practical implementation in OPO resonators [6], interest in and demonstrations of parametric mid-infrared down-conversion have increased. The band gap of about 2.45 eV (506 nm) has allowed for many demonstrations with widely used and well-established pump lasers emitting about 1 µm with short picosecond pulses [14,15,16,17,18]. Advances in ultrafast parametric sources have been extensively reviewed in [19], and CSP has also been employed in nanosecond-pulsed systems pumped at 1 µm [20]. However, its use for parametric conversion with pump wavelengths significantly longer than 1 µm remains limited, particularly in the range near 2 µm, which remains relatively unexplored.

The first successful realization of a CSP-based OPO was achieved in 2009 [21]. The pump was a room-temperature 1.99 µm Tm:YALO laser. The performance showed a 27% slope, 340 mW output at 2 W pump, and 680 mW threshold. Impurity absorption at the pump wavelength limited its efficiency at this time. Later, a 2–5 µm tunable CSP OPO pumped by a laser source at 1.57 µm demonstrated the highest pulse energy of 10 mJ at a 5 Hz repetition rate near degeneracy around 3 µm [22]. A first attempt to compare ZGP and CSP under similar conditions was made in 2018 with a Tm:YAP ns-pulsed laser (1.94 µm) [23]. A CSP OPO produced 2.5 W with a conversion efficiency of 65%. More recently, a tandem of two CSP crystals in a ring cavity pumped above 2 µm with an 84 W Q-switched Ho:YAG laser (10 kHz, 15 ns pulse width) produced over 25 W of mid-IR output, demonstrating the viability of this material for high average power applications [24]. The emission had a very broad spectrum and a low beam quality ($M^2$-factor > 6).

In this work, we present a CSP OPO pumped by a nanosecond-pulsed Ho:LLF laser operating at 2.06 µm within a compact linear cavity configuration, achieving a total output power exceeding 3 W. We examine the factors limiting further power scaling in the current setup. The short effective interaction length in the non-collinear configuration enables exceptional beam quality at this power level, with an $M^2$ factor approaching unity. Additionally, we analyze the temporal characteristics of the pulses and compare the experimental results with numerical simulations.

2. Experimental Setup

The experimental setup is shown schematically in Figure 1. The pump laser, a 10 kHz Q-switched Ho:LLF MOPA system, is focused to a spot of 1 mm diameter inside the OPO resonator formed by two flat mirrors. The configuration of the pump laser and its integration into the OPO cavity have been previously detailed in our publication concerning ZGP OPOs [25]. Non-collinear phase matching is achieved by tilting the pump beam slightly with respect to the optical axis of the OPO resonator. The input coupling (IC) mirror is highly transmissive for the pump wave and highly reflective for the signal and the idler. The output coupling (OC) mirror is partially transmissive for OPO waves, with $R\approx 50\%$ for the signal, and the drop in reflectivity for the idler wave above 4 µm. The IC and OC mirrors are separated by a physical distance of 40 mm. The transmission data for the OC mirror which provides a quasi-singly resonant operation are shown in Figure 1. The free-running OPO produces a broad emission, and there is still a significant amount of idler feedback in the cavity, which is why it cannot be considered to be truly singly resonant. In fact, the reduced amount of idler light in the cavity results in spectrally stable operation, in contrast to the standard doubly resonant cavity. From the signal and idler separation registered by the pyroelectric camera and the known geometry of the system in combination with the measured wavelengths, the pump tilt and the phase matching angle ${\theta }_{\mathrm{PM}}$ can be deduced.

The CSP sample used for this work was 10 mm long with a clear aperture of 4 × 4 mm² (BAE Systems). It was cut at an angle of ${\theta }_{\mathrm{PM}}={46}^{\circ }$ with respect to the optical axis, which allows for type I phase matching ($e\to oo$). The transmission spectra of the sample were measured before and after the AR coating. The material is transparent in the range of 1–6.5 µm, as shown in Figure 2. The AR coating shows a narrow band transmission peak centered around 2040 nm. This was not optimal for pumping at 2.06 µm, eventually causing considerable back reflection and loss of input pump power. The pump absorption was very low, not measurable in our setup (linear absorption coefficient below $0.01{\mathrm{cm}}^{-1}$). This is consistent with recent advances in CSP crystal growth techniques and progress in suppressing defect-related residual optical absorption [26,27].

3. Results and Discussion

3.1. CSP OPO Conversion Efficiency and Numerical Simulation

The experiment was guided by numerical simulation tools tested before in our ZGP research [28,29]. The model is based on the standard single-frequency approach, propagating beam time slices with the cavity round-trip time length, e.g., [30]. In the crystal, the coupled wave equations are integrated by the split-step method, using fast Fourier transform for diffraction and spatial walk-off [31] and fourth-order Runge–Kutta integration for the mixing relations in the spatial domain. In addition, the effect of non-collinear phase matching is emulated (reduction in interaction length due to oblique propagation of idler and signal in the cavity). The computations for the CSP rely on the Sellmeier dispersion relations, as presented in [9].

In our experiment, the observed conversion efficiency was significantly lower than expected from the numerical simulations. The CSP OPO conversion efficiency data are shown in Figure 3. The conversion threshold was about 1.8 mJ and the slope efficiency, as indicated by the linear fit to the experimental data, was about 0.34. The efficiency roll-over signature was also observed above 3 W of output power. The crystal mounted inside the copper holder was not water-cooled and a significant temperature rise was observed during power scaling, affecting phase matching and leading to a conversion drop. On closer inspection, it was also found that another essential factor affecting the OPO’s performance was the suboptimal coating applied to the crystal surfaces for the pump wavelength. The coating resulted in approximately 10% reflection losses on each crystal surface for the 2.06 µm pump wavelength and approximately 5% losses for the signal and idler waves. When the actual values were fed into the numerical simulations, and a reduced idler feedback at the crystal entrance due to oblique propagation in the plane cavity was introduced, the performance of the CSP OPO was effectively reproduced. Notably, the simulation predicted a total OPO output pulse energy of 0.3 mJ for a 2.6 mJ pump pulse, in close agreement with the experimental results. Further energy scaling is only possible with the improved AR coating. The simulation then predicts more than 1.2 mJ total energy in the OPO pulses for collinear operation (12 W output power).

The slope efficiency in our case was lower than that reported by Cole et al. [23], but the maximum output power exceeded the previous achievement of 2.5 W with 65% optical efficiency. In the most recent CSP OPO demonstration using Ho:YAG for pumping at 2.09 µm [24], the authors reported the conversion threshold at around 15 W at a 10 kHz repetition rate, equivalent to 1.5 mJ, but in their case, they used two CSP crystals (16 mm and 12 mm) inside the ring cavity. In addition, their pump pulse duration was shorter, at 15 ns, so the peak power was higher. It is not possible to make a direct comparison because too many parameters are different, but our high conversion threshold is not far from that in similar reports and is supported by the numerical simulation for a small sample of CSP crystal of only 10 mm.

3.2. Temporal Characteristics of Pulses

Figure 4 gives an overview of the temporal characteristics of the pulses. The first graph in Figure 4a shows the experimental data, where pulse measurements were collected using a HgCdTe photodiode and an oscilloscope. Since the spectral response of the detector is unknown and the amount of light collected by the photodiodes is only fractional, the power levels for the pump and OPO beams have been adjusted to match the total power ratio as measured with power meters. The pump pulse lasts about 24 ns, while the resulting signal and the idler pulses last about 7 ns. In the next graph, Figure 4b, the simulation results for the experimental parameters are shown. This graph also contains a plot of the depleted pump, which visualizes the conversion process. The experimental depletion trace has not been collected, but it is not expected to substantially deviate from the predicted one. The difference in the rising edge can be attributed to the slow response time of the detection system. Remarkably, as the conversion efficiency increases, the simulations predict a faster depletion of the pump and a subsequent convergence of the observed pulses to the pump pulse characteristics. As the pump pulse energy increases, the conversion saturates and longer OPO pulses are expected. This can be seen in Figure 4c,d, which depict the temporal pulse evolution for perfect crystal coating and a pump pulse energy of 3.0 mJ instead of 2.6 mJ, yielding an OPO output of around 0.7 mJ pulse energy (7 W average power) and 1.2 mJ (12 W average power) in the non-collinear and collinear arrangements, respectively.

3.3. Beam Quality

For the beam quality analysis, a 200 mm focal length focusing lens was placed approximately 25 cm from the center of the OPO cavity. The beam was attenuated by reflection from a CaF₂ wedge. For separate analysis, dichroic mirrors were used to filter out one of the beams. The signal and idler beam intensity profiles were recorded at the maximum total output power of 3 W with a pyroelectric camera moved along the propagation direction. The beam diameters were represented as the 4$\sigma$-widths of the measured intensity profiles in the horizontal (x) and vertical (y) directions. How the beam diameter changes as a function of the distance from the focusing lens, forming the caustics, is shown in Figure 5. The relation for the evolution of the second-moment beam radius as given, e.g., in [32], was fitted to establish the $M^2$ factor. The insets in the graphs show the near-field and far-field examples of the beam profiles as recorded by the pyroelectric camera. The outer beam images correspond to the largest beam images recorded for the wings of the caustic, and the middle image is from the focal area.

Fitting to the experimental data confirms that both the signal and the idler exhibit a nearly diffraction-limited beam quality, with $M^2$ factors approaching unity. The data for the idler are noisier due to additional reflections that introduced some measurement artifacts. The idler beam quality is also slightly worse than that of the signal. This is a result of non-collinear propagation and the fact that, unlike the signal, there is no stable cavity mode for the idler beam. The caustics also revealed some astigmatism, which may have been partly introduced by the optical elements used, but might also be related to the non-collinear alignment.

The short crystal and non-collinear phase matching both contribute to a short interaction length during parametric conversion, resulting in a very good beam quality.

3.4. OPO Wavelength Tuning

The conversion efficiency, pulse timing, and beam quality were characterized for fixed signal wavelengths of around 3.85 µm. As shown in Figure 6, we further examined the range of emission wavelengths. The two graphs show examples of signal and idler spectra. The data in the plots were normalized. The spectra were recorded separately by coupling either the signal or idler beam into the monochromator (iHR320 Imaging Spectrometer, Horiba, Kyoto, Japan), then exploring the tuning limits. Therefore, the examples of spectra cannot be related to the phase matching condition and to each other. Overall, the idler and signal waves were tuned in a range of 3.5–4.7 µm by pump tilt and crystal rotation. The tuning was achieved by adjusting the orientation of the crystal and modifying the pump tilt (varying the angle between the crystal optical axis and the propagation direction). The internal separation angle between the signal and idler beams was variable up to $3^{\circ }$. The AR coating of the cavity mirrors, the cavity geometry, and the crystal aperture were the main limitations of the tuning range. Its further extension to the range 3–6 µm would require careful design of the cavity mirror AR coating.

The data in both plots in Figure 6 also contain a spectral peak at 4130 nm. This was obtained by collinear phase matching very close to the point of degeneracy. Because the operation was very unstable and the back reflection from the OPO cavity disturbed the pump laser, it was not possible to study this in detail. In addition, the conversion efficiency was much lower. At a pump pulse energy of about 2.6 mJ, only 0.05 mJ of OPO pulses were recorded. This is due to the OC mirror coating (see transmission in Figure 1). It reflects less than 50% at this wavelength.

We did not observe a significant broadening of the free-running OPO, as was reported by Horton et al. [24]. Our data analysis indicates that the conversion process has not yet reached saturation. This may explain the relatively narrow spectrum. Such behavior has already been observed in the experiments with the ZGP crystal in scenarios of low conversion. At a very low pump energy, only spectral components with a phase mismatch $\Delta k$ around zero appear, and the spectral width should be determined essentially by the narrow spectral width of the pump laser. In our case, the pump is relatively narrow, below 1 nm (relative spectral width below 0.001) [25]. At 3850 nm and 4400 nm for the signal and idler carrier waves, respectively, using a linear approximation to the perturbed energy and momentum conservation relations, with $\Delta k=0$, the spectral widths can be estimated to be of the order of 100 nm or less.

4. Conclusions

In summary, we have demonstrated a CSP OPO emitting more than 3 W in the mid-IR range when pumped by a 2.06 µm laser. The non-collinear phase matching resulted in a good beam quality with a diffraction-limited $M^2$ close to 1. The observed overall efficiency was lower than expected for the 10 mm long CSP sample due to a non-optimal AR coating of the crystal surfaces. We have shown numerically that the CSP sample of this length can be expected to have a much higher output power and emit pulses of the same duration as the pump pulse when the conversion is saturated. Further investigation and power scaling will require re-polishing and re-coating of the samples to remove the limiting factor for this report: the suboptimal AR coating.