We report on a ZnGeP2-based optical parametric oscillator (OPO) with 22 W of output power in the 3-5 µm range and a beam quality factor M2 ≈1.4. The OPO uses a novel V-shaped 3-mirror ring resonator that allows two passes of the beams through the same nonlinear crystal. The pump is a 39 W hybrid Tm:fiber laser/Ho:YAG laser.
©2010 Optical Society of America
High power mid-infrared (3-5 µm) laser sources have important applications in defense and remote sensing, and there are extensive research efforts aimed at improving output power, efficiency, and beam quality of such sources.
The best established route to laser radiation in this spectral range is frequency down-conversion of near infrared (1-2 µm) laser radiation in optical parametric oscillators (OPOs). Applying a 2 µm pump source has several important advantages compared to the more usual 1 µm sources: First, when converting from 1 µm to the mid-infrared in a single-stage OPO, most of the converted energy is transferred to a near-infrared signal beam. The conversion efficiency to the desired mid-infrared idler beam is limited by the ratio of the pump and idler wavelengths. With a 2 µm pump source this ratio is about twice as high, and for broadband applications, such as some defense applications, it may even be possible to have both the signal and idler wavelengths in the useful spectral range. Second, pumping at 2 µm allows the use of ZnGeP2 (ZGP) and orientation patterned GaAs (OP-GaAs), both of which have too high absorption at 1 µm to be pumped there. A detailed comparison with crystals that can be pumped at 1 µm is outside the scope of this paper, but it can at least be said that ZGP and OP-GaAs are very attractive because they have high nonlinearity, good transmission in the mid-IR and good thermal and mechanical properties. ZGP is also commercially available in high quality. OPOs with almost 60% conversion efficiencies to the mid infrared have been demonstrated with both materials [1,2]. ZGP is the most mature one of these two, and output powers over 14 W from a single OPO based on ZGP was reported in 1999 . A 2007 report on efforts to improve crystal quality and processing of ZGP  mentioned more than 30 W power from an OPO, but without giving any details. OPOs based on OP-GaAs have similar conversion efficiency to those with ZGP, but because of the limited aperture size, the highest output reported from an OP-GaAs OPO is still only 2.85 W . In the present work we obtain more than 20 W of output power and excellent beam quality from a ZGP OPO with novel and very simple ring resonator geometry.
Given the excellent performance of these nonlinear materials, the most challenging component of the system can be the 2 µm pump laser. Thulium (Tm) and holmium (Ho) both have useful laser transitions near 2 µm. Tm has high absorption at ~800 nm with an efficient cross-relaxation process to a stable upper level of a 2 µm transition, but this transition has a low cross-section making it less suitable for Q-switched operation. Ho, on the other hand, has a 2 µm transition with a high cross-section, but lacks absorption on wavelengths where high power laser diodes are available. However, Ho does absorb wavelengths that can be generated by Tm, so by combining Tm and Ho, it is possible to achieve both efficient diode pumping and Q-switched operation. Having the Tm-ions in a fiber laser can be advantageous because most of the heat comes from the quantum defect in the Tm laser system, and in a fiber it can be easily removed and does not distort the beam. A CW Tm-fiber laser at 1.9 µm can then be used to pump a bulk Ho:YAG or Ho:YLF laser emitting at 2.1 µm. The bulk Ho-laser allows Q-switched operation with high pulse energies (mJ-level), which is suitable for pumping a ZGP-based OPO.
A second challenge in making a high power mid-infrared OPO source is to avoid reflection from the OPO into the pump laser. An OPO with a standing wave cavity inevitably leads to such feedback, so some kind of isolation will be required. The need for isolation is particularly strong for an OPO with double-pass pump, which reflects the unconverted pump beam back towards the laser. Unfortunately, Faraday rotators available at 2 µm have lower damage threshold and power handling capacity than those for 1 µm. The obvious solution is to use a ring resonator, and this is commonly applied, especially in CW pumped systems. However, in ns-pulsed systems it is important to keep the round-trip time small in order to achieve rapid build-up of the signal, and this can be difficult with a ring resonator. Another problem with ring resonators is that angle tuning the crystals misaligns the resonator unless the lateral shift of the beams is eliminated by using two identical crystals turned in opposite directions. In this paper we demonstrate a simple and novel V-shaped 3-mirror ring OPO, where the ring is in the non-critical plane of the crystal, as shown in Fig. 1 . This allows two passes of the beams through the same crystal. Tuning the crystal translates the output beam in the tuning plane, but if the output mirror is plane, this does not cause misalignment.
2. The V-shaped 3-mirror ring resonator
In the present work, the nonlinear material is ZGP, but the design can also be used with other materials, as discussed below. The two identical mirrors on the right, Min, couple the pump beam in and out of the resonator. These mirrors are placed close together in order to make the arm L3 short. This makes the angle between the two other arms in the ring small enough to allow both of them to pass through the same nonlinear crystal.
The V-shaped ring resonator shares some advantages with a standing wave resonator with double-pass pump: The two-pass pump eliminates the passive backward pass, where the beams are exposed to thermal lens effects without any gain, and where signal and idler back-convert in a doubly resonant OPO. The second pump pass also allows the use of a shorter non-linear crystal, which leads to shorter build-up time and smaller thermal effects. The V-shaped resonator has some additional advantages: First, the ring resonator eliminates the strong pump feedback. Second, since the two beam passes in the crystal do not overlap, the damage threshold, in terms of the maximum allowable pump fluence, should be significantly higher than for the standing wave resonator, and the lens effect induced by the thermal load is also effectively reduced. The beam only experience the lens induced by approximately half the thermal load per pass, while in the normal double-pass configuration the beam experiences the total thermal lens power in both passes. For the beam separation advantages to take full effect, the separation in the crystal (Δ in Fig. 1) must be at least comparable to the beam diameter (FWe −2M). This implies that the angle between the ring resonator arms and distance (L1) from the out-coupling mirror to the crystal cannot be too small.
A complicating factor for any doubly resonant OPO with double-pass pump is that the effective gain depends on the relative phase shift of the beams between the two passes through the crystal,Eq. (1.1)) is equivalent to reversing the sign of the nonlinear coefficient in the second pass. The effect of sign reversal and phase shifts between crystals has been studied in the context of walk-off compensating crystals , and a relative phase shift of π leads to zero gain for the exactly phase matched frequencies. In practice, an OPO will shift to a frequency where phase mismatch partially cancels Δψ r and the net gain is maximised. From the gain equations in Ref. 5 it can be shown that in most relevant situations this optimized double-pass small signal gain will be larger than the single-pass gain, even for the for the worst possible relative reflection phase (Δψ r = π).
Compared to other ring resonators, the V-shaped 3-mirror ring has the advantages of short round-trip time and that a single ZGP crystal can be used for both passes. The crystal is tuned by rotating it in the critical plane, which is perpendicular to the plane of the ring. Our tuning parameter is the angle α between the crystal Z-axis and the plane of the ring. The angles between the resonator legs (inside the crystal) and the tuning plane (which we take to be the XZ-plane of the crystal) are denoted βj, where j = 1 or 2. We denote the polar angles of the two resonator legs with respect to the crystal Z-axis by θj and φj, where5], so it is not reduced significantly by a small φ. In the setup described in Section 3 we have βj = 2.7° and α = 53°. Using Eq. (1.2) this gives φ = 3.4°, which means that the effective nonlinear coefficient is reduced by less than 1%. If the thermal load is different in the two legs, it may be desirable to tune them differently to optimize the performance. We have not tried this, but if necessary such tuning may be achieved by making the β angles asymmetric and exploiting the small variation of θj with β.
The V-shaped resonator is not restricted to uniaxial crystals: If the legs of the V are symmetric with respect to a principal plane of a biaxial crystal, they will have the same phase mismatch also in that case. It is more complicated for triclinic and monoclinic crystals, where the principal axes change slowly with wavelength, but even in such crystals there are planes with approximate symmetry for phase mismatch.
3. Experimental set-up
3.1 Tm-fiber-laser-pumped Ho:YAG laser
The experimental set-up is shown in Fig. 2 . The Ho:YAG laser is pumped by a commercial Tm-doped fiber laser (Model TLR-70-1908 from IPG Photonics). This fiber laser delivers 70 W of power in a single mode fiber at 1907 nm, and it can be temperature tuned over a range of about 1 nm. The wavelength overlaps an absorption peak in Ho:YAG. The Ho-laser is based on a design described in an earlier work , but the pump power has been scaled up by a factor 5. To accommodate this power scaling, some changes had to be made to the previous design: The holder for the laser rod is water-cooled, the pump spot size is greater, a convex resonator mirror partially compensates for the thermal lens, and a λ/4-plate is inserted to reduce the effect of thermally induced birefringence.
The output from the fiber laser is focused through a 45°-mirror (M2) into a 700 µm diameter spot (FWe−2M) in the 25 mm long Ho:YAG rod, which has 0.72 at. % Ho+3-ions. The pump is reflected by mirror M1 for a second pass through the rod. These conditions give us more than 95% pump absorption during lasing, and only a moderate up-conversion loss. The unabsorbed pump is returned to the fiber laser, which is stated to tolerate −10 dB feed-back. No significant power instability was observed from this feed-back. The HR-mirror (M1) has 100 mm convex radius of curvature. Between the HR-mirror and the laser rod we have inserted a λ/4-plate with a principal axis aligned parallel with the laser’s polarization. This reduces the resonator loss resulting from stress induced birefringence in the laser rod . The pump input mirror (M2), also acts as a polarizer for the laser mode. It has high reflectance for the s-polarization and 50% reflectance for the p-polarization, which is enough to ensure linearly polarized operation. The plane output coupler (M3) has 50% reflectance. The laser was Q-switched by a 19 mm long acousto-optic Q-switch at a pulse repetition rate of 45 kHz. A 100 µm thick fused silica plate was inserted in the laser resonator as a low finesse etalon in order to select one of two otherwise simultaneously lasing transitions. We selected the 2090 nm laser line because this gave somewhat higher output power.
If the round-trip time in the laser can be matched to the OPO round-trip time, or some simple fraction or multiple of it, within the coherence length of the laser, the conversion efficiency of the OPO will be increased . To accommodate 2-to-1 round-trip matching (the roundtrip time in the laser is twice that of the OPO) for an acceptable OPO resonator length, the laser resonator was made as short as possible and had a physical length of only 63 mm.
3.2 V-shaped ZGP OPO
The angle between the two long arms in the resonator was 17° outside the crystal, and the crystal was placed 5 mm from the output coupler (M6 in Fig. 2). This gave a beam separation of 1.5 mm, which was more than adequate to avoid beam overlap on the crystal surface. We have observed significantly different performance with different ZGP crystals, even when comparing crystals from the same supplier. In this OPO, we got the best results with an 8.5 mm long crystal obtained from Eksma Optics (Vilnius, Lithuania). The crystal had 6 by 8 mm aperture and was clamped to a water cooled aluminum block, which was kept at 15 °C.
The long arms in the resonator had a physical length 25 mm, and the short resonator arm, L3 in Fig. 1, was only 6 mm long. The output from the Ho-laser was focused to a 600 µm spot (FWe−2M) on the ZGP crystal. Mirrors M4 and M5 had both more than 98% reflectance for s-polarized signal and idler, and more than 98% transmittance for the p-polarized pump. The output coupler (M6) had 99% reflectance for the pump and ~50% reflectance for both signal and idler, and the OPO was thus doubly resonant.
The alignment of the V-shaped 3-mirror ring was relatively easy. After establishing the desired angle on the output coupler, we observed the pump beam reflections from M4 and M5 and adjusted them to overlap in the near-field and far-field (see Fig. 2). After this overlap was established, a small adjustment in the vertical direction was needed to fine tune the resonator.
We do not know the relative phase shift on mirror M6 nor in the AR coatings on the crystal. This unknown phase may have a small negative effect on the efficiency of our set-up, as discussed in Section 2. Simulations of the set-up indicate that the maximum output from the OPO will vary with 10-15% from the worst to the best reflection phase of M6.
4. Results and discussion
4. 1 Ho:YAG laser results
The pump spot size is critical for the performance of the Ho-laser. A too small pump spot size leads to poor beam quality and loss due to thermal birefringence. As the spot size increases the conversion efficiency drops due to the three-level nature of the laser transition in Ho. We also observed that secondary pulses appeared when the pump beam was wide. We believe that these pulses are high order modes from the poorly depleted area around the fundamental resonator mode. After some optimization we were able to extract more than 42 W of output power from the Ho-laser, with less than 1% of the power in the secondary pulses and an M2-value of approximately 1.7. The beam quality was estimated by rotating slit measurements of the beam diameter in several positions after a focusing lens. The measured input/output characteristics are shown in Fig. 3 . The pulse length at full power was 24 ns (FWHM). After a few days operation the output from Ho-laser dropped to 39 W, and the pulse length increased to 32 ns (FWHM), which has been used in the OPO experiments reported here. We have not found the cause for this, but the laser has not degraded further during several weeks of frequent use. The available pump power incident on the OPO, after lenses and power adjustment optics, was 37.7 W.
4.2 ZGP OPO results
The measured input/output curve is shown in Fig. 4 , together with the corresponding conversion efficiency. With 37.7 W pump power we obtained 22 W of average power in the 3-5 µm range. The conversion efficiency was 58%, and the slope efficiency was 75%. The output slope showed no sign of roll-off at high power, which indicates that the high efficiency can be maintained at even higher average pump-power. The result was obtained with a round-trip time that was matched to half the round-trip time in the Ho-laser. When we reduced the path length in the OPO with 6 mm we observed a 5 W increase in the threshold power, but the slope efficiency did not change. Effectively, the round-trip time matching increased the efficiency from 50% to nearly 60%.
A doubly resonant OPO pumped by a single-frequency beam can oscillate only on pairs of signal and idler frequencies that satisfy energy conversion and are close to cavity resonances for both the signal and idler. A typical output spectrum contains a few clusters of longitudinal modes where this criterion is satisfied . A multi-longitudinal-mode pump usually smears out these spectral features, but when the mode spacing of the signal and idler is equal to the mode spacing of the multi-mode pump, which is the case when the round-trip times are matched, the spectrum resembles that obtained with a single-frequency pump . Our case, with 2-to-1 round-trip matching, in which the pump has half the mode spacing compared to the 1-to-1 case, is equivalent to having two pump beams with modes shifted half a mode spacing with respect to each other. This creates two interleaved cluster patterns, which effectively double the number of clusters within the acceptance bandwidth of the OPO.
The cluster effect can clearly be seen in the measured output spectrum shown in Fig. 5 . The OPO runs on two pairs of signal and idler clusters around 3.9 µm and 4.5 µm, and they drift slowly with the thermal expansion of the resonators. For the signal clusters we measure a cluster distance of 1.35 THz. However, when we calculate the distance using the available Sellmeier equations we get a distance of 2.0 THz . The reason for this discrepancy is not known, but the cluster positions are very sensitive to small variations in the group indices for the signal and idler.
The round-trip time matching allows the OPO resonator to have relatively long air gaps without reducing the conversion efficiency. The 39 mm of air gaps and the narrow pump beam in our OPO, makes the Fresnel number small (~0.5). This is very beneficial for the beam quality of the OPO. We obtained the beam quality from waist- and far-field measurements using a pyro-electric camera (PyroCam III, SPIRICON Inc.) and measured an M2-value of 1.4 at full power. The far-field distribution directly from the OPO is shown in Fig. 6 .
We have described and demonstrated a novel and very simple OPO ring resonator which is suitable for high power operation and high beam quality. This design was used to realize a ZGP OPO with 22 W of output power in the 3-5 µm range and a beam quality of M2 ≈1.4. The OPO was pumped by a Tm-fiber-laser-pumped 2.1 µm Ho:YAG laser, and the slope of the OPO output remained linear up to the highest available pump power of 37.7 W, indicating that even further power scaling should be possible.
References and links
2. C. Kieleck, M. Eichhorn, A. Hirth, D. Faye, and E. Lallier, “High-efficiency 20-50 kHz mid-infrared orientation-patterned GaAs optical parametric oscillator pumped by a 2 microm holmium laser,” Opt. Lett. 34(3), 262–264 (2009). [CrossRef] [PubMed]
3. E. C. Cheung, S. Palese, H. Injeyan, C. Hoefer, J. Ho, R. Hilyard, H. Komine, and J. Berg, “High power conversion to mid-IR using KTP and ZGP OPOs,” in Advanced Solid State Laser, M. M. Fejer, H. Injeyan, and U. Keller, eds., (Optical Society of America, Washingthon DC, 1999), pp. 514–517.
4. P. G. Schunemann, “Advances in Mid-IR Materials,” in Conference on Lasers and Electro-Optics/Quantum Electronics and Laser Science Conference and Photonic Applications Systems Technologies, (Optical Society of America, 2007), pp. CThL3.
5. V. G. Dmitriev, G. G. Gurzadyan, and D. N. Nikogosyan, Handbook of nonlinear optical crystals, (Springer-Verlag, 1997).
6. R. Hua, S. Wada, and H. Tashiro, “Principles and limitations of a quarter-wave plate for reducing the deploarization loss from thermally induced birefringence in the Nd:YAG lasers,” Opt. Commun. 175(1-3), 189–200 (2000). [CrossRef]
7. G. Arisholm, E. Lippert, G. Rustad, and K. Stenersen, “Effect of resonator length on a doubly resonant optical parametric oscillator pumped by a multilongitudinal-mode beam,” Opt. Lett. 25(22), 1654–1656 (2000). [CrossRef]
8. R. C. Eckardt, C. D. Nabors, W. J. Kozlovsky, and R. L. Byer, “Optical parametric oscillator frequency tuning and control,” J. Opt. Soc. Am. B 8(3), 646–667 (1991). [CrossRef]
9. D. E. Zelmon, E. A. Hanning, and P. G. Schunemann, “Refractive-index measurements and Sellmeier coefficients for zinc-germanium phosphide from 2 to 9 µm with implications for phase matching in optical frequency-conversion devices,” J. Opt. Soc. Am. B 18(9), 1307–1310 (2001). [CrossRef]