1. Introduction

There is a need for efficient and practical laser sources in the 8–14 µm atmospheric transmission window for a range of applications including spectroscopy, remote sensing, and use in military systems. A substantial part of this window is covered by high power CO₂ lasers, but few practical high power sources exist in the lower part of the window below 9.3 µm. Recent progress in quantum cascade lasers has made such devices a possible alternative, but they still offer a limited brightness and power scalability [1]. Another possibility is to employ frequency down-conversion of efficient shorter wavelength sources in nonlinear materials, which has yielded very efficient sources in the 3–5 µm region [2], [3].

One of the most promising nonlinear materials for this purpose is ZnGeP₂ (ZGP) which can be used to cover both spectral regions. Kettridge et.al. [4] have previously demonstrated output powers up to 0.42 W at 8 µm by pumping this material with a 2 µm Tm:Ho:YLF laser, but the conversion efficiency was only 4.2 %, and beam quality data were not reported. Also, the YLF crystal in the pump laser had to be cryogenically cooled to 77 K.

In the present work we use an efficient hybrid Tm:fiber/Ho:YAG 2.1 µm laser to pump a ZGP optical parametric oscillator (OPO) and demonstrate up to 0.95 W output power at 8 µm with 10.7% conversion efficiency and good beam quality. We also discuss some of the factors influencing the efficiency of the device and possibilities for further improvements. The hybrid laser itself has a conversion efficiency of 65% from Tm:fiber to Ho:YAG laser output, and has been used previously to generate more than 5 W from a ZGP OPO tuned to the 3–5 µm region [2]. This provides an opportunity for development of practical high-power laser sources covering both atmospheric transmission windows.

2. Experimental setup

A schematic of the experimental setup is shown in Fig. 1. The 2.1 µm pump source is a Q-switched Ho:YAG laser pumped by a 15 W thulium fiber laser. The laser delivers 9.8 W of average output power at 20 kHz pulse repetition rate. The excellent mode overlap provided by the single mode fiber laser gives the Ho:YAG laser a good beam quality with an M² value of 1.15. A Faraday rotator is inserted between the pump laser and the ZGP OPO to prevent feedback into the pump laser. A half wave plate and a polarizer are used to vary the pump power for the OPO. Due to losses in the optical components the maximum pump power available at the OPO was 8.9 W.

When pumped at 2.096 µm, ZGP is type I phase matched (with e, e, and o polarizations for idler, signal, and pump, respectively) for generation of 2.84 µm and 8.0 µm at an internal angle of 51.2° with respect to the optical axis. This angle is just ~3.5° less than the tuning angle used for conversion to the 3–5 µm range [5], so the same ZGP-crystals can be used for 3–5 µm and 8 µm OPOs, provided that the reflectivity of the AR-coatings is sufficiently low in both wavelength regions. The crystals used in this work were cut for 3–5 µm conversion, and thus had to be rotated by an external angle of ~10° for phase matching at 8 µm.

 

Fig. 1. Experimental set-up.

Download Full Size | PDF

The OPO consists of two identical plane mirrors (M1 and M2) and one or two ZGP crystals. In the case with two crystals they are oriented for walk-off compensation. The mirrors have high reflectivity (94%) for the 2.84 µm signal, a moderate reflectivity of 19% for the 8 µm idler and high transmission (93%) for the pump. The 19% idler reflectivity at the output coupler represents a substantial loss in output power, so there is room for significant improvement with an optimized mirror coating. The roundtrip idler losses, including the crystal coating reflection (2% per surface) and crystal absorption (3% cm^-1), are so high that the OPO is close to singly resonant. The pump was focused to a 480 µm diameter (FWe^-2M) waist close to the input side of the first crystal. Both crystals were 14 mm long and AR coated for all the interacting waves (R<2%).

The signal roundtrip time in the double-crystal OPO was matched to the roundtrip time in the Ho-laser. This reduces the problems that arise from the quasi-periodically varying phase and amplitude in the multi longitudinal mode pump [6]. The effect is most pronounced in doubly resonant OPOs, because of the inherent phase dependency of the amplification process, but simulations indicate that there is also an effect in singly resonant OPOs, which arises from the periodic amplitude variations of the pump [7]. Because of this, and because the OPO is not entirely singly resonant, a round trip time matched OPO was chosen. For comparison a shorter resonator, for which the resonator lengths were not matched, was also investigated.

In the single crystal experiments the resonator length was reduced by a factor of 2 to give a 2-to-1 roundtrip time matching. This is not as effective the 1-to-1 matching, but a small effect can still be seen.

3. Exploiting the Gouy phase shift in walk-off compensating crystal pairs

If a pair of identically cut crystals is used for walk-off compensation in an interaction with two e-waves, the two crystals will have opposite signs of d _eff [8]. This is equivalent to a relative phase shift of π between the crystals, which reduces the gain severely and splits the gain up in two peaks, one on each side of the phase matched wavelength. In the case of ZGP, one crystal should be cut with orientation φ=0 and the other with φ=90° in order to preserve the sign of d _eff. Unfortunately it is difficult to distinguish between the φ=0 and φ=90° direction in a given crystal sample, so suppliers do not usually specify the absolute value of φ. By cutting crystals from the same boule they can provide a pair with 90° difference in the orientation, but cutting this way wastes material from the boule. Another complication is that even if the crystals are cut to preserve the sign of d _eff, the different interacting waves will normally experience different phase shifts in the crystal AR coatings, resulting in a net relative phase shift. In order to compensate for these effects it is desirable to find a method of adjusting the relative phase shift between the crystals, preferably without adding extra components which will add extra complexity and loss.

In this work we use two crystals with the same φ=0 orientation and with an unknown relative phase shift in the AR coatings, meaning that we have an unknown total relative phase shift. In order to try to compensate for this we decided to exploit the Gouy phase shift, which can be varied by adjusting the distance between the two crystals. Because of the small beam diameters and short Rayleigh lengths (below 10 mm for the idler), a substantial Gouy phase shift can be obtained for relatively small adjustments, and the method has the advantage of not introducing extra loss. However, increasing the distance between the crystals will lead to wider beams in the second crystal, which will reduce the gain. When interpreting the results it is difficult to distinguish between the contributions from an improved relative phase and the increased beam diameter, but this adjustment does at least provide an opportunity to avoid the worst-case situation for the relative phase.

To explore the feasibility of this approach we measured the efficiency as a function of the distance between two crystals at full power. The resonator length was kept constant at 50 mm, and the second crystal was moved, thus varying the inter-crystal distance from 2 to 18 mm. The resulting variations in the idler output power are shown in Fig. 2(a), and the output spectra measured at selected positions are shown in Fig. 2(b).

 

Fig. 2. (a) Idler output powers at 8 µm as a function of crystal separation for 8.9W pump power. (b) Selected idler spectra with 2 mm (red) and 8.5 mm separation (green). The error bars show min/max values of slow oscillations in output power due to drift of the double resonance mode clusters.

Download Full Size | PDF

Figure 2(a) shows that there is an increase in output power when the distance between the crystals is increased to 5–10 mm, indicating that the Gouy phase shift can be exploited for improved performance. For larger separations the power decreases, which could be due to an increasing beam diameter and reduced intensity in the second ZGP crystal. The spectra shown in Fig. 2(b) are consistent with a change in the phase shift as the inter-crystal distance increases. At 2 mm separation two peaks are visible in the spectrum, and at 8.5 mm separation there is only one peak, which has shifted towards the center. The fact that the left peak at 2 mm separation is much smaller than the right peak indicates that the effective phase shift is not equal to π, which may be caused by the phase shift in the crystal AR coating. In order to maximize the output power a crystal separation of 5 mm was chosen for the remaining experiments.

The error bars in Fig. 2(a) and all subsequent output curves, indicate the typical min and max value of the slow and periodic oscillations in output power, which we often see in roundtrip-time-matched OPOs. We believe that the source of these oscillations is the varying position of the double resonance mode clusters relative to the gain profile and possibly also relative to some water-vapor absorption lines at the signal wavelength [9]. The position of the clusters depends on the exact optical length of the resonator, which increases slowly during a warm-up period of approximately 1 hour. Some measurement points were recorded at times when the oscillations had become extremely slow, and the full min/max span was therefore not measured.

4. Experimental results

The highest output power in the 8–10 µm range was achieved at 8 µm, as expected from the quantum defect and crystal absorption, which sets in above 8.5 µm. The measured input/output curve at 8 µm is shown in Fig. 3 for both the single- and double-crystal configuration. Using two crystals, 0.95 W was obtained for 8.9 W of pump power incident on mirror M1. This corresponds to a conversion efficiency of 10.7% and a quantum efficiency of 41%. The slope efficiency up to 5 W of pump power is 14%, after which there is a slight reduction in the slope. The most likely reason for this roll-off is a nonlinear absorption of the pump in the ZGP crystals, which we observed in separate pump transmission measurements. Below 5 W of pump power the pump absorption coefficient was constant at 0.035 cm^-1, but it increased to 0.105 cm^-1 at 8.9 W. The mechanism behind this nonlinear absorption is not known. It can be noted that the pump intensity was well below the threshold for optical parametric generation, so any influence from this can be ruled out. The main effects of the increased absorption are higher thermal aberrations as well as a passive loss for the pump, both of which will reduce the efficiency of the conversion process. The fact that the roll-off occurs at the same pump level in the single-crystal experiments indicates that back-conversion is not the main source of the slope reduction. From the OPO with one crystal we obtained 0.8 W of power for this 8 µm tuning.

 

Fig. 3. Input/output curves for single- and double-crystal-OPO tuned to 8 µm idler wavelength.

Download Full Size | PDF

4.1. Tuning

The OPO was tuned by tilting the crystals, and the tuning characteristics are given in Fig. 4. The single-crystal OPO maintained more than 0.45 W of idler power out to 9.6 µm, where it started to drop rapidly. The output from the double-crystal OPO was lower than from the single-crystal OPO at wavelengths above 8.5 µm due to the significant crystal absorption in this spectral region, but it could also be tuned out to 9.8 µm. The reason for the rapid drop in power towards 9.8 µm is that no mixing processes are phase matched for crystal angles below the one that phase matches 9.8 µm (50.6 degrees, see insert in Fig. 4) [10]. For greater angles two distinct pairs of signal and idler wavelengths are phase matched. However, the wavelengths generated depend on the resonator losses, and our mirror coatings favored an idler wavelength below 10 µm. The idler spectral width was 60–70 nm (FWHM) at 8 µm and increased to ~250 nm (FWHM) near 9.8 µm.

 

Fig. 4. Angle tuning data for single- and double-crystal ZGP OPO. Insert: Phase matching calculations for idler wavelengths based on [5].

Download Full Size | PDF

4.2. Matching the roundtrip time

The results shown above were obtained for a mirror separation of 50 mm for the double-crystal OPO and 25 mm for the single-crystal OPO. For the double-crystal OPO the signal roundtrip time in the resonator was matched to the roundtrip time in the Ho-laser, as explained in Section 2. For comparison a mirror separation of 37 mm, for which the resonator lengths were not matched, was also investigated. The input/output curve for the two resonator lengths are shown in Fig. 5. Reducing the resonator length to 37 mm could be expected to increase the efficiency by reducing the build-up time, but in this case the efficiency decreased substantially as shown by the green curve in Fig. 5. This shows that even though the OPO is very close to singly resonant there is a clear performance enhancement when the roundtrip times of the pump laser and OPO are matched. Whether this enhancement is due to the slight double resonance, or whether it is mostly due to the amplitude modulation of the pump, is not clear.

 

Fig. 5. Comparison between the roundtrip-time-matched (RTM) OPO with 50 mm resonator length and a shorter OPO without RTM.

Download Full Size | PDF

For the single-crystal configuration the 25 mm mirror separation corresponded to half of the roundtrip-matched length. This means that the pump phase is the same for every second round trip in the resonator, which gives a smaller efficiency enhancement than the full roundtrip matching [6].

4.3. Beam quality

The beam quality of the double-crystal OPO at full power was estimated by calculating the M²-value from the second order moments of the far field and near field distributions measured using a camera (Spiricon, Pyrocam III), and was found to be 2.7. The roundtrip-time-matched OPO had a very low Fresnel number, which in OPOs normally leads an even better beam quality. However, the combination of large difference between the Rayleigh lengths of the interacting beams and the powerful thermal lens leads to poor mode overlap between the interacting fundamental modes and greater coupling to higher order modes.

4.4. Further optimization

The experimental options were limited by the available mirrors and crystals, so there is room for further improvement. To support the design of the OPO we performed numerical simulations, and they suggest an improved performance if the signal output coupling is increased and/or the crystal length is reduced somewhat. Shorter crystals would reduce any possible back-conversion as well as the absorption losses of all 3 waves, whereas higher output coupling would reduce the relatively high circulating signal intensity and thereby the signal absorption in particular. Another imperfection in the setup is the residual idler reflectivity of the output coupler (R=0.19). The fraction reflected from the output coupler is almost completely lost in two passes through the crystals and reflection from input mirror (R=0.19). The calculated idler power incident on the output mirror is thus 1.15 W, and most of this could be coupled out by a more suitable mirror. Taking these imperfections into account, more than 50% quantum efficiency seems realistic for an improved setup.

5. Conclusions

In summary, we have demonstrated a 0.95 W, 8 µm source with M²=2.7 pumped by an efficient, high power Tm:fiber/Ho:YAG hybrid laser. The conversion efficiency at full power was 10.7%, and the OPO could be tuned out to 9.8 µm. We have shown that the Gouy phase shift can be used to reduce the problem of sign reversal of d_eff that occurs in walk-off compensating configurations with identically cut crystals. We have also seen a significant enhancement of the output power by using a roundtrip-time-matched OPO resonator. With a more optimal choice of crystal lengths and signal and idler output coupling, more than 1.1 W output power and 50% quantum efficiency should be possible.