Optical parametric oscillators (OPOs) are commonly used sources for coherent light, providing every wavelength in a wide spectral range from visible to mid-infrared, often those that are not directly provided by lasers. Depending on the beam parameter requirements of the target application, a large variety of customized OPO designs have been put into practice applying different kinds of pump lasers and nonlinear crystals. In the mid-infrared spectral region, orientation-patterned gallium-arsenide (OP-GaAs) pumped by 2 μm lasers has shown very good performance in a wide wavelength and beam parameter range [1–4]. In some aspects GaAs exceeds the properties of other well-known crystals like ZGP, GaSe, or AGSe and it has all the advantages of quasi-phase-matching. The effective nonlinearity, the transparency range, and the thermal and mechanical characteristics make it the crystal of choice for 2-μm-pumped OPOs emitting at wavelengths above 10 μm.

On the other hand, OP-GaAs crystals have a drawback—a limited aperture. Since the crystals are fabricated by epitaxial growth [5–8], the usable thickness is limited to about 1.5 mm so far, depending on the length of the poling period. Additionally, the threshold of laser-induced damage limits the maximum fluence and intensity, respectively, applied to the crystal surface. Starting from this, one can find a reasonable point of operation through a detailed simulation of the process. Recent publications about OPOs are comparable to this one in some points, even though the combinations of spectral and energetic characteristics are different. Feaver et al. [9] reported on a short linear OPO emitting at a wavelength of around 10.6 μm with a conversion efficiency of up to 3.6% at a 700 μJ pump pulse energy. Clément et al. [10] demonstrated a single-frequency nested-cavity OPO pumped by a microlaser with output energy in the range of 2 μJ at a 10.3 μm wavelength. Higher conversion efficiencies up to 38% (pump to signal + idler) were shown by Hildenbrand et al. [11], though the signal and idler wavelengths were in the 3–5 μm range.

The system described herein (Fig. 1) comprises a single-longitudinal-mode (spectral width $<100\mathrm{MHz}$) thulium fiber laser with a wavelength of 1.95 μm, a repetition rate of 50 kHz, and a 150 ns pulse duration as the pump source. The beam quality factor ${\mathrm{M}}^2$ is 1.1 in horizontal and 1.3 in vertical direction. A combination of optical isolator and waveplate both protects the pump laser from back reflections and allows for continuous power tuning without any beam parameters having to be changed. The maximum average power measured at the position of the nonlinear crystal is 12 W.

 

Fig. 1. Experimental setup of the signal-resonant OPO: WP half waveplate; M1, M2 curved mirrors; M3, M4, M5, M6 plane mirrors.

Download Full Size | PDF

The OPO is set up in a bow-tie geometry, and the OP-GaAs crystal is mounted in a temperature-controlled oven, which is attached to a five-axis table in the middle between two curved mirrors, M1 and M2. The five-axis table is used to find the optimum position and angle of the beam path inside the crystal. With respect to advanced stability requirements, this system can also be set up with the alignment of the two laser beams, but for laboratory purposes the chosen way is more comfortable. The temperature is set to 25°C with an accuracy of $<0.1\mathrm{K}$ with a thermo-electric device. At this point of operation the idler output wavelength is 10.6 μm. The signal wavelength is 2.39 μm.

The crystal is 40 mm long and 5 mm wide. The orientation-patterned layer with a period length of 74.5 μm is about 1.3 mm thick. Both facets are antireflection coated for all three waves. Losses of the pump beam due to absorption and reflexes at the facets were measured to less than 3.5%. If reflection losses of about 0.5% per facet are assumed, the absorption is about $0.00625{\mathrm{cm}}^{-1}$. Losses for signal and idler were not measured but are supposed to be in a comparable order of magnitude. The pump beam is focused into the OP-GaAs crystal in a single pass. Mirrors M1 and M2 are antireflection coated for the pump and the idler wave and high-reflection coated for the signal wave. Both surfaces of these mirrors have spherical curvatures with radii of 150 mm. Mirror M3 is highly reflective for the signal, and mirror M4 is an output coupler with a reflectivity of 70% for the signal. Mirror M5 is highly reflective for the pump wave and high-transmission coated for the idler wave to separate the two beams. M6 is an uncoated ZnSe wedge that reflects a part of the idler beam for characterization purposes. Mirrors M3, M4, M5, and M6 have plane substrates. The generated idler wavelength is monitored with a wavemeter (HighFinesse WS6-200 IR-III), the pulse duration is measured with a photovoltaic multiple junction detector (Vigo PVM), the idler power is measured with a Coherent LM3, and for the spatial measurements a Spiricon PyroCamIV is applied. In general, a pointing laser with a wavelength similar to the signal wave can be applied through a mirror M4. This helps align the OPO cavity. In future work, a narrow bandwidth seed laser will be applied for spectral stabilization and control.

The design objective for the OPO is to obtain a high conversion efficiency with a sufficient margin to the damage threshold of the optics, especially the OP-GaAs crystals. Additionally, the idler pulse length should be about 100 ns. This aim is achieved by a numerical simulation of the OPO performance, analyzing the influence, and the interplay of relevant design parameters. These parameters are cavity length, pump spot and eigenmode size, pump pulse length, roundtrip losses and output coupling, effective nonlinearity, and absorption. The model is based on an existing software tool developed by Fraunhofer ILT for optical wave propagation and nonlinear light matter interaction described, e.g., in [12]. For the OPO simulation, the pump pulse is cut into time slices, which are propagated through the OP-GaAs crystal. Due to the parametric gain, additional electromagnetic fields at the signal and the idler wavelength arise from quantum noise. During several roundtrips through the cavity their intensity rises until the pump pulse ends.

By varying all of the influencing factors, the performance of the OPO is simulated in the entire parameter space. Besides the search for the desired point of operation, all of the correlations and relative connections of the parameters are calculated.

For example, in Fig. 2 the fluence of all three waves inside the crystal versus signal beam radius is shown for different degrees of output coupling of the signal. In this case it is calculated for a 300 μJ pulse energy and a 500 mm cavity length. From the literature value of $1\mathrm{J}/{\mathrm{cm}}^2$ at 10 ns [9], we estimated the laser-induced damage threshold (LIDT) to be $3.8\mathrm{J}/{\mathrm{cm}}^2$ for 150 ns pump pulses, assuming that the dependency of the pulse length $\tau$ is proportional to ${\tau }^{1/2}$ due to thermal heat dissipation. In order to ensure an LIDT margin factor of 5 (i.e., $0.76\mathrm{J}/{\mathrm{cm}}^2$), this is given for signal beam radii between 123 and 170 μm depending on the output coupler. With these combinations, different idler pulse lengths are generated and different conversion efficiencies are obtained. With special regard to acceptable fluences, high conversion efficiency, and a desired idler pulse length of 100 ns, the best tradeoff found by simulation is 500 mm for the cavity length, 70% reflectivity of the output coupler, and 165 μm signal beam radius. The pump spot radius was 155 μm.

 

Fig. 2. Simulation of the OPO operation point. Fluence versus signal beam radius for different signal output couplers.

Download Full Size | PDF

The design of the cavity allows the cavity length to be varied without influencing the resonator stability too much, which is also the case when the focal length of the thermal lens of the crystal is changed. Furthermore, varying the cavity length in a range of $\pm 10\%$ allows the signal beam waist to be changed in the crystal by about $\pm 7\%$ in our experiments.

By balancing the whole parameter set (mode size, reflectivity of the signal output coupler, cavity length, and pump beam parameters) experimentally, deviations of a few percent from the simulations are found. However, these deviations are small and the relative dependency of the performance on the parameters is correct.

Figure 3 shows the measured average idler power versus pump power (black line/triangles) and its corresponding conversion efficiency (blue line/stars). The plotted idler power was measured behind M5 (without M6, see Fig. 1). The pump power was measured at the crystal position behind M1. The maximum idler power was 812 mW at 12 W pump power. The pulse energy at a 50 kHz repetition rate was 240 μJ for the pump pulse and 16.2 μJ for the idler pulses. This corresponds to an optical-optical conversion efficiency of $P_{\text{idler}}/P_{\text{pump}}=6.76\%$ and a quantum conversion efficiency of 36.8%. To our knowledge this is the highest value for 2-μm-pumped OP-GaAs OPOs with an output wavelength above 10 μm.

 

Fig. 3. Average idler power and corresponding efficiency versus pump power. Alignment was carried out at 12 W pump power.

Download Full Size | PDF

The alignment of the cavity and the mode matching of the pump and signal were carried out at maximum pump power. Due to thermal effects resulting from absorption inside the crystal, the alignment of the cavity—leading to maximum efficiency—is not identical for different operation points. Therefore, the last point at maximum pump power is slightly higher than the extrapolation of the slope. With this alignment, the OPO threshold is reached at a pump power of 7.4 W. When aligned at a lower pump power, the threshold drops to 6.7 W. Measuring the average signal power behind the signal output coupler (M4) gives a value of 2.95 W at maximum pump power. With this, the overall optical-optical efficiency is 31.3%.

As there are neither spectral filters nor any active control of the cavity length yet, the signal and idler output wavelengths are not stable. The center wavelength varies from pulse to pulse by several 100 MHz. Even though the spectral width of each pulse might be narrow, the average output is broad and longitudinally multimode.

With a PyroCamIV, the idler beam quality ${\mathrm{M}}^2$ was determined to be 1.43 in a horizontal and 1.63 in a vertical direction (see Fig. 4). Besides this, Fig. 4 shows the idler beam cross section (inset). The shape is round and Gaussian in good approximation. Due to the 80 μm large pitch of the camera pixels, this is an estimation of higher values because the determination of the beam radii at the focus position is inexact. A comparison with the knife-edge method in a horizontal direction showed a slightly smaller value of 1.38.

 

Fig. 4. Idler beam quality ${\mathrm{M}}^2$ ($1/{\mathrm{e}}^2$) and idler beam profile.

Download Full Size | PDF

The length of the idler pulses is about 100 ns (see Fig. 5). As expected, this is shorter than the pump pulse, and the rising slope is steep (10% to 90% in 13.2 ns), while the falling edge follows the pump pulse.

 

Fig. 5. Pump and idler pulse length.

Download Full Size | PDF

The pulse length shown in Fig. 5 was measured at maximum power. In general, the shape and the length of the idler pulses depend on the conversion efficiency, i.e., on the reflectivity of the signal output coupler, the cavity length, and the beam diameters as mentioned above. In the case of special demands for pulse length or shape, this has to be taken into account in the design phase.

As described by Kuo et al. [13] and Kieleck et al. [14], the orientation of the crystallographic axis has a strong influence on the conversion efficiency with respect to the polarization direction of the pump, signal, and idler waves. Theoretical models predict that the effective nonlinearity has a maximum when the pump polarization is parallel to the [111] crystallographic direction, while it is the same for the [110] and [001] directions. However, we observed a global maximum idler output when pumping parallel to the [110] direction (see Fig. 6), and another local maximum for the [001] direction while the minimum was measured for a 45° angle. The 35° angle, which corresponds to the [111] direction [13] did not show any special points of interest. This is comparable to the results of Kieleck et al. [14].

 

Fig. 6. Average idler power versus pump polarization angle and corresponding crystal axis.

Download Full Size | PDF

Further measurements of the signal and idler polarization or with the circular polarized pump were not carried out. Possible explanations for the discrepancies in theory concerning the optimum polarization direction are that the performance of the optical coatings differs for s- and p-polarized light or there was stress-induced birefringence in the GaAs crystal. Future investigations will pay heed to this issue.

In conclusion, we developed an optical parametric oscillator based on OP-GaAs as a nonlinear crystal, pumped by a 1.95-μm-wavelength fiber laser, and emitting pulses with a wavelength of around 10.6 μm (tunable by temperature tuning). The average output power of the idler was 812 mW, which corresponds to a quantum conversion efficiency of 36.8%. The overall optical-optical efficiency is 31.3%. The beam quality ${\mathrm{M}}^2$ is 1.4 in a horizontal and 1.6 in a vertical direction with a round and Gaussian beam shape. Concerning the influence of the pump beam polarization, we observed another relationship between efficiency and polarization direction than that predicted by the theory, but it was comparable to the results of other groups.

Future work will concentrate on maximizing the output power and seeding the OPO process as well as actively controlling the cavity length for single-frequency operation. At that stage, detailed measurements of the spectral characteristics and the long-term stability will be carried out.