Widely tunable (2.6–10.4 μm) BaGa4Se7 optical parametric oscillator pumped by a Q-switched Nd:YLiF4 laser

We report on the first BaGa4Se7 nanosecond optical parametric oscillator pumped by Q-switched Nd:YLiF4 laser at 1053 nm. The oscillator exhibits a pump threshold energy as low as 0.25 mJ. Mid-infrared (MIR) idler wave tuning from 2.6 μm to 10.4 μm is demonstrated with an angle-tuned type-I (o-ee) y-cut sample, highlighting the superior performance of this novel large bandgap chalcogenide nonlinear crystal to generate tunable coherent radiation over its full MIR transparency range (0.47–18 μm). The phase-matching data are used to identify the most accurate dispersion relations of this compound among three recently published Sellmeier equations. Damage threshold measurements yielded values as high as 2.04 J cm−2 at 100 Hz pulse repetition rate, one of the largest among existing MIR χ(2)nonlinear materials.


Introduction
Widely tunable optical parametric oscillators are attractive laser sources for mid-infrared (MIR) nonlinear optics applications, because of the scarcity of direct-emitting conventional laser sources in this spectral region (2-20 μm). While oxide-based periodically-poled nonlinear compounds belonging to the ferroelectric crystal family (mainly LiNbO 3 , LiTaO 3 and KTiOPO 4 ) are commercially available, their IR transparency does not exceed 5 microns and therefore they cannot be used for deeper MIR applications. For instance the 8-14 μm range corresponding to one of the atmospheric transparency window is interesting for LIDAR or countermeasure systems while coherent sources in the strongly water-absorbing 6-7 μm range can be used for medical (surgery) applications. More generally tunable coherent MIR laser sources are useful spectroscopic tools to probe rovibrational molecular fingerprints with various applications in trace gas sensing such as environmental monitoring, detection of hazardous species, life science or medical applications such as real-time breath analysis for early disease diagnosis [1][2][3][4][5].
Currently very few nonlinear chalcogenide compounds possessing both wide transparency range above 5 μm and large energy bandgap are available. The bandgap wavelength requirement of λ g <500 nm is necessary to avoid two-photon absorption (TPA) when they are pumped with popular 1 μm range solid-state emitting lasers such as Nd 3+ lasers or Yb 3+ fiber lasers. Only a handful of ternary compounds satisfies the requirement of wide bandgap and extended MIR transparency range, with proven capabilities for down-conversion to the longwave MIR (λ>5 μm) using 1.064 μm pump optical parametric oscillation (OPO) [6]. Among these latter ones -whose use in nanosecond OPOs have been widely proven owing to their commercialization-one can quote the benchmark wurzite-type ternary chalcopyrite AgGaS 2 (AGS) belonging to the A II B IV C 2 V semiconductor group and, to a lesser extent, the much more limited transparency range ZnGeP 2 (ZGP) belonging to the A I B III C 2 VI group. During more than a decade silver gallium sulfide (AgGaS 2 ) has been the only OPO material with the longest demonstrated idler wavelength [7].
Since the mid 90's, some novel ternary lithium containing chalcogenides (LiInS 2 , LiInSe 2 , LiGaS 2 , LiGaSe 2 ) produced with large enough single-crystal size have emerged [8,9], leading to various nanosecond OPO demonstrations covering the mid-IR [10][11][12] despite their rather low nonlinearity (d il <10 pm V −1 ) compared to ZGP or the difficult-to-grow mercury thiogallate, HgGa 2 S 4 (HGS) [13,14]. The recently re-discovered II-IV-V 2 chalcopyrite compound CdSiP 2 (CSP) with its high nonlinearity (d 36 =92 pm V −1 [6]) could appear as an alternative to ZGP, however its low bandgap (λ g ∼539 nm) combined with limited MIR transparence (λ6.5 μm) almost restrict its use to the non-critically phase-matched generation of ∼6 μm radiation when pumped at 1064 nm [15]. Alternative to birefringence phase-matched compounds, orientation-patterned III-IV semi-conductors (orientation-patterned GaAs and GaP) with high nonlinearity have been recently engineered, however their low bandgap energy does not allow 1 μm pumping not only because of TPA but also due to strong absorption resulting in tremendous thermal load [16]. OP-GaAs and OP-GaP require thus pumping above ∼2 microns and have limited aperture (<1 mm) due to quasi-phase-matching limitations [17,18]. The continuous coverage of the MIR spectrum from 5 to 20 μm requires thus the continuous synthesis of novel nonlinear birefringent ternary chalcogenide compounds, combining both wide energy bandgap and extended MIR transmission.
Recently two new promising ternary chalcogenide compounds have been successfully synthesized in large size using the Bridgman-Stockbarger technique: BaGa 4 S 7 (BGS) [19][20][21] and BaGa 4 Se 7 (BGSe) [22,23]. Both sulfide and selenide compounds exhibit wide bandgap energy (E g =3.54 eV with λ g =0.350 μm for BGS; E g =2.64 eV with λ g =0.469 μm for BGSe) and extended MIR transparency (up to 12 μm for BGS and from 0.47 to 18 μm for BGSe at 0% transparency level). The two compounds are biaxial, offering hence an extended phasematching capability over their 0.3 cm −1 level transparency range (BGS: 0.545-9.4 μm; BGSe: 0.776-14.7 μm). However despite their similar chemical composition, BGS crystallizes in the mm2 orthorhombic point group as the isostructural LiGaS 2 while BGSe crystallizes in the monoclinic m point group (offering thus an extended phase-matching capability than BGS), with the dielectric axes x and z aligned respectively with the b and c crystallographic axes [21,22]. Their nonlinear coefficients are however about twice smaller than that of AgGaS 2 (d 36 =13 pm V −1 [24]) but due to their largest bandgap both compounds exhibit much higher damage threshold at 1064 nm (1.2 J cm −2 for 12 ns pulse and 10 Hz repetition rate [19] and 3.7 J cm −2 for 14 ns pulse at 100 Hz [20]). The first BGS nanosecond OPO pumped by a Nd:YAG laser at 1064 nm was demonstrated by Tyazhev et al [25], in which the idler wave could be tuned from 5 to 7.3 μm, and delivering a maximum 0.5 mJ pulse energy at 6.2 μm. From this study and comparing with the performance of the 13× higher nonlinearity CSP compound in a similar OPO setup [6], the authors highlighted the advantage of the lower nonlinearity BGS stemming from of its 10× higher surface damage resistance compared with CSP.
The first BGSe nanosecond OPO ever reported was pumped by a Q-switched Ho:YAG laser at 2.09 μm and used a y-cut crystal (θ=40.8°, j=0°) for type-I (o-ee) phase-matching in the xz plane [26]. However the tunable 3-5 μm idler wavelength did not exceed the ∼5 μm limit of oxide-based ferroelectrics and the downconversion idler pulse energy (<1.4 mJ at 20 mJ pump) was quite modest despite the sample length used (dimension 6×6×30 mm 3 ). With a similarly cut but shorter BGSe sample (dimension 10.33×11.95×14.57 mm 3 ), a 1064-nm pumped BGSe OPO was reported in [27], demonstrating a considerably improved output performance in terms of both idler MIR tunability (2.7-17 μm) and output pulse energy (∼3×). Using a wide pump beam diameter (2w=5 mm) to accommodate for the available high-energy from the Nd:YAG laser (up to E p =250 mJ for 8 ns pulse at f=10 Hz), 3.7 mJ at 7.2 μm from 63 mJ of pump pulse energy without any significant saturation in the observed 40% quantum conversion efficiency was obtained. This limitation was not due to the BGSe damage threshold but to that of the Au-coated OPO end mirror. As an important finding, the comparison of the nonlinear down-conversion efficiency of two samples, one phase-matched for type-I (o-ee) phase-matching in the xz dielectric plane ), allowed to assess that the nonlinear tensor elements d 16 and d 23 must have the same sign [27].
In this article, we report on a BaGa 4 Se 7 nanosecond OPO pumped for the first time by a Q-switched Nd:YLiF 4 (Nd:YLF) at 1053 nm, the shortest OPO pump wavelength used for this crystal. Contrary to [27], the pump laser has not only a slightly shorter wavelength but also much lower pulse energy (maximum energy 1.3 mJ). Low threshold of oscillation (E p,th =0.25 mJ) was achieved by reducing the pump beam diameter to 2w∼1 mm, thus the reported performance corresponds to the operation of the OPO far below parametric gain saturation (and damage threshold). The MIR idler wave tunability extends from 2.6 μm to 10.4 μm, greater than for a mercury thiogallate (HGS) OPO pumped by the same Nd:YLF laser and employing an identical crystal length [14]. Despite the higher χ (2) nonlinearity of HGS (4× that of BGSe), the idler output performance of our present BGSe OPO is 4 to 5 times higher than that of the HGS OPO, delivering a maximum pulse energy of 45 μJ at ∼3.3 μm and 14 μJ at ∼8 μm, whereas the HGS OPO yielded less than 8 μJ over its whole tuning range (4.2-10.8 μm). We report on surface damage threshold measurements of BGSe at 1053 nm, confirming the high damage threshold (>2 J/cm 2 for 10 ns pulse and f=100 Hz repetition rate) reported in [23] for similar experimental conditions. Additionally, with our phase-matched OPO data, we could infer the accuracy of three recently published dispersion and thermo-optic dispersion relations of BaGa 4 Se 7 .

Experimental nanosecond OPO set-up
The BaGa 4 Se 7 optical element (dimensions 3×7×11.8 mm 3 , see in figure 1) used in this experiment was grown at High Technologies Laboratory, Kuban State University (Russia) using the Bridgeman-Stockbarger technique [20,21]. It was cut for type-I (o-ee) phase-matching (θ=45°, j=0°) with the pump laser polarized ordinarily along the 3 mm side (direction of the y-dielectric axis) and the signal and idler waves polarized within the xz plane (top or bottom facet of the shown element in figure 1). The wider transverse aperture of 7 mm is designed to allow an extended angular tuning of the phase-matching. The crystal is maintained at a temperature of T=40°C(±0.1°C) by means of a thermo-electric Peltier element. The crystal was single-layer antireflection (AR) coated at ∼1.5 μm, resulting in a transmission greater than 85% over the bandwidth 1-2 μm as shown from FTIR spectrum (figure 2). The transmission in the mid-IR idler range (5-11 μm) is above 72%, slightly improved with respect to the uncoated sample transmission.
The experimental layout of the BGSe OPO is sketched in figure 3. The pump source is a diode-pumped Q-switched Nd:YLF laser (Laser-compact Group model TECH-1053-N). The pump wavelength is  λ p =1.053 μm and the pulse duration is τ=16 ns, with a pulse repetition rate set to f=100 Hz. The specified bandwidth of the laser is ∼1 cm −1 (30 GHz), within the pump acceptance bandwidth of the type-I phasematching process. The maximum laser pulse energy is 1.3 mJ and the measured beam quality factors are M x 2 =M y 2 =1.3. Due to the linear OPO cavity, the Q-switched Nd:YLF laser is protected from feedback radiation by a Faraday isolator (FI). A half-wave plate (HWP) is used to rotate the laser polarization vertically with respect to the xz plane of BGSe so as to satisfy the o-ee phase-matching requirement. Given the low pump pulse energy, in order to reach threshold of parametric oscillation, the pump beam diameter is reduced to a measured spot diameter of 2w x =0.84 mm and 2w y =0.84 mm by a f=300 mm BK7 lens.
The pump pulses are directed to the linear OPO resonator by a 45°incidence ZnSe dichroic bending mirror (BM) which is HR-coated in a narrow 1.0-1.1 μm band (average reflectivity 97% at normal incidence) and highly transmitting (T average =90%) from 1.15 μm to approximately 10 μm. Its use at 45°incidence results in further pump transmission leakage from the cavity. Several combinations of interference filters (F) were used to isolate as much as possible the idler wave from the pump and signal waves residual leakage from BM mirror.
The OPO cavity, configured as a double-pass pump, signal-resonant oscillator (SRO), is formed by the input-output ZnSe coupler (IOC) and a gold (Au) mirror M2 and the cavity length was kept as short as possible (L cav 18 mm) to increase the number of signal-wave roundtrips within the cavity so as to reduce the pump threshold. The IOC has a long radius of curvature R=−1000 mm so as to improve transverse signal beam confinement. Its transmission at λ p =1.053 μm is 93%. The IOC coupler was dielectrically coated so as to be reflecting for part of the entire signal wave band (1.17λ s 2.155 μm): Its reflectivity ranges from an average R s ≈90% from 1.28 μm to 1.8 μm and gradually decreases to R s ≈60% at 1.18 μm and 2.1 μm. Its average transmission is T∼75% in the idler range between 2.28 μm and 11 μm. Due to the narrow HR band of the bending mirror, most of the signal wave exiting IOC could leak through BM, and to separate the mid-IR idler from the leaking signal wave-especially for short idler wavelengths below λ i ≈2.6 μm-two dielectric mirror filters had to be used to minimize signal component contamination of the measured idler pulse energy. However these two high-pass filters could efficiently block only the signal wavelengths below ∼1.683 μm, so that in the idler range from ∼2.2 μm (near the 2λ p degeneracy point) to 2.6 μm the idler pulse energy could not be properly measured due to strong signal contamination (thus only idler data from 2.6 μm have been plotted in figure 4). The idler energy was measured with a pyroelectric detector (OPHIR model VEGA PE10-C). To determine idler wavelength in spectral range from 2.6 μm to 10.4 μm the signal wavelength was measured by a commercial wavelength meter (HighFinesse models LSA L IR) and the idler wavelength was retrieved from energy conservation ( p s i The BGSe sample was hold inside a copper mount that is temperature-regulated at T=40°C by means of a thermo-electric Peltier element. For wavelength tuning purpose, the BGSe mount is positioned on a motorized high-precision rotation mount (STANDA model  for the calibration of the internal θ phasematching angle (deduced from the external incidence angle using Snell's law). The rotation stage itself is mounted on a motorized X-translation stage (STANDA model 8MT173-20), which automatically adjust the lateral position of the BGSe crystal respective to the cavity axis as the crystal is rotated. A PC controller monitored the θ-position and X-lateral position of the crystal. To accommodate rotational tuning of the phase-matching (the BGSe cut angle θ=45°corresponds to an idler wavelength of λ i ≈8 μm), the cavity length was increased by linear translation of mirror M2 and IOC.  Figure 4 shows the BGSe OPO idler pulse energy at λ i =8.07 μm (corresponding to sample at normal incidence and a short cavity length L cav =18 mm) as the pump pulse energy is increased.

Experimental results: conversion efficiency and mid-infrared tuning
At this idler wavelength, λ s =1.211 μm with a IOC reflectivity R s =70%. The threshold pump energy is extrapolated to be E p,th =250 μJ. This threshold is about 25× lower than that reported in the previous BGSe-OPO employing a 5× larger pump beam diameter [27]. The intensity threshold of a singly-resonant, critically phase-matched, nanosecond OPO can be expressed as [28,29], is the mode-coupling coefficient (g 0.465 s  in our case assuming equal confocal parameter for the resonating signal and the pump); l 11.8 mm c = is the crystal length; L l c eff < is an effective parametric gain length [29] taking into account the walkoff angle of the extraordinary signal wave ( 2.62 mrad 1.7 s r = = ); L cav is the cavity length; 0.05 cm a » / is the signal absorption coefficient; 16 t = ns is the pump pulse width. The coefficient g is the ratio of the forward and backward pump amplitude ( 1 g~in our double-pump pass SRO) and ) is an overall reflectivity coefficient for a round-trip cavity ( for an AR-coated sample, and R R 0.9 in out =~is the IOC coupler reflectivity for the signal wave ) for θ=45°, this rough and indirect estimation pledges for both nonlinear tensor elements to have the same sign, given that d 10 23 » pm/V, as already pointed out in [27] from the comparison of the performance of type-I versus type-II BGSe OPOs. Assuming different signs for d 16 and d 23 would result in a much lower d eff outside the boundaries reported above. A direct measurement of the relative sign and magnitude of d 16 remains thus necessary.
Turning back to figure 4, the maximum idler energy at 8.07 μm (∼14 μJ) was limited by the available Nd:YLF pulse energy. The data could be excellently fitted with a straight line, showing no sign of saturation as expected since the BGSe OPO is pumped even farther from parametric gain saturation as done in [27] that used up to 63 . BaGa 4 Se 7 nanosecond OPO measured idler energy at ∼8 μm versus pump pulse energy measured in front of the OPO cavity (after BM). The idler energy was the net energy measured just after the BM mirror (hence corrected for transmission loss due to BM and the two additional filters, but not for the transmission of IOC). mJ of Nd:YAG pumping with a much larger beam diameter (2w=5 mm) to avoid damage. Comparing with the performance of a ns OPO pumped with the same Nd:YLF laser and employing a similar length mercury thiogallate (HGS cut for type-II eoe at θ=47°; j=0°with d d eff 36 @ ), and which yielded 2 μJ of idler at λ i =7.5 μm (normal incidence) [14], the performance of this BGSe OPO is about 7× superior despite the 4× superior value of HGS effective nonlinear coefficient. Because the OPO mirror cavity reflectivities were almost comparable, this difference in performance may be mainly attributed to the superior optical quality of BaGa 4 Se 7 with respect to HgGa 2 S 4 whose growth in large size, good quality is extremely difficult. Furthermore with its larger energy bandgap (E g =2.79 eV instead of 2.64 eV for BGSe [6]) one would have expected a better performance for the HGS OPO. The measured s=2.8% energy slope efficiency in figure 4 corresponds to q s 21% i p l l = = ( ) / quantum conversion efficiency, far from pump saturation and back-conversion. Figure 5 displays the idler energy measured after IOC mirror as a function of wavelength. For reasons explained in the previous section, we did not plot the data below λ i =2.6 μm down to 2.2 μm because of imperfect separation of the idler energy from the leaking signal energy. The MIR coverage (2.2-10.5 μm) obtained with a single BGSe crystal exceeds the one (5.6-10.8 μm) obtained from the previously quoted HGS OPO, that needed two samples (θ=47°and at θ=60°) to cover that idler range [14]. This broad coverage with a single sample highlights the broader phase-matching capability of the monoclinic BaGa 4 S 7 compared with the uniaxial defect chalcopyrite HgGa 2 S 4 . The general decreasing trend of the OPO performance in figure 5 relates to the i 1 ldependence of the parametric gain. The OPO could deliver a maximum of 50 μJ of idler energy at ∼3.3 μm (λ s =1.55 μm). The peak around 3.3 μm coincides with an OPO configuration where the IOC reflectivity is R s 95% between 1.4 μm and 1.62 μm. The dip at 4.2 μm and those in the 5.5-6 μm range are probably due to atmospheric absorbing molecular species (such as water). The slight enhancement bump centered at 8.1 μm corresponds to the usually observed enhancement of such nanosecond OPOs when the crystal is phase-matched at normal incidence with respect to the linear cavity axis [25,27], resulting in a subcavity etalon effect originating from the plane parallel crystal facets that are uncoated for the idler wave. Compared to [27], the magnitude of the bump is weaker due to the 10× reduced diameter or the pump beam.
As to the plateau behavior between 9.2 and 10.4 μm, it corresponds to signal wavelengths for with the reflectivity of the IOC coupler drops severely between R=20% and 50%. The much weaker signal wave resonance probably prevents a genuine parametric oscillation, giving rise rather to an amplified parametric generation.
The phase-matching angles associated with the MIR tuning curve shown in figure 5 are plotted in figure 6.
As seen the OPO idler wave could be tuned from nearly the degeneracy point (λ s =λ i =2λ p =2.106 μm) to 10.5 μm. Actually the near-IR wavelength meter could measure directly the lower wavelength part of the idler branch (plotted with filled circle symbol) while the MIR part was deduced by measuring the signal wave (filled square symbol) and making use of pump photon energy conservation. For obvious reasons due to the limited pump pulse power of the Nd:YLF laser (E p <1 mJ at the OPO) combined with decreasing parametric gain (resulting in higher threshold), we could not achieve the largest ever reported MIR tuning in [27]-up to 17 μm -obtained with a similar BGSe sample but a high-power Nd:YAG pump laser (E p >100 mJ). Very recently new dispersion formulae (Sellmeier equations) for BGSe have been proposed by three different groups [31][32][33], and their accuracy has been tested with our experimental phase-matching data taken at relatively low pump power, i.e. without any self-heating bias (due to residual absorption) introduced by highpower phase-matched data. In [32] thermo-optic Sellmeier equations for the three principal dielectric axis of BGSe were constructed for the first time, using direct index of refraction measurements of prisms maintained at some fixed temperatures spanning 25-150°C and employing the minimum deviation angle technique to derive the index of refraction. Mercury, Helium and Cadmium lamp lines between 0.546 μm and 2.325 μm were used for the temperature-dependent index measurement at a relative accuracy level of a few parts in 10 −4 . The authors worked out temperature-dependent Sellmeier equations which yield the (blue) dash-dot curve labelled 'Zhai et al (T=40°C)'. Large deviation from the measured experimental data can be observed not only near the degeneracy turning point but also at long-wave idler above 4 μm. Setting the temperature to T=25°C would lead to an even larger discrepancy of Zhai's equations for long wave MIR. Such large discrepancies find their origins in the customary inability of direct index measurements-even at the 10 −4 relative accuracy-to reflect experimental phase-matched data at less than a few degrees. Another origin of the inaccuracy of the Sellmeier data proposed in [32] is the limited MIR wavelength range of their index measurements (the longest wavelength being 2.325 μm).
The second set of Sellmeier equations (at room temperature), obtained by refinement of the two-pole Sellmeier equations in [21], that we used to model our experimental phase-matched data are taken from [31]. The authors used phase-matched difference-frequency generation (DFG) data extending down to 11 μm to propose refined dispersion relations for BGSe. The theoretical curve based on their Sellmeier equations is plotted with dot (green) line in figure 6. While these dispersion relations account fairly well our experimental data near the degeneracy turning point, they fail to account for experimental idler wavelength data above ∼4 μm again, despite the fact that the DFG data used in [31] extended to 11 μm. The same discrepancy for idler wavelengths >4 μm was also noted in the Nd:YAG laser pumped BGSe OPO in [27].
Finally the most accurate Sellmeier equations that reproduce fairly well our entire OPO phase-matched dataset is the ones worked out by Kato et al at room temperature [33]. Kato et al worked out their dispersion relations from second-harmonic generation (SHG) of the signal/idler components of a 1.064 μm pumped AgGaS 2 OPO, supplemented with data from phase-matched sum frequency generation (SFG) of a CO 2 laser (10.59 μm) and the signal/idler wave from the AGS OPO. The calculated curve based on their refined two-pole Sellmeier equations is plotted with solid(black)-line in figure 6. We infer the relatively fair accuracy of their dispersion relations to the fact that they were based on a rather dense wavelength set spanning the whole 1-11 μm range.
The pulse duration of the pump, signal, and idler waves were also investigated. Figure 7 shows the temporal traces of the incoming pump pulse, and output signal and idler pulses. The pump and signal pulses were detected with a fast InGaAs photodetector (Thorlabs Inc. model FGA-04) while the MIR idler pulse was monitored with a fast Mercury Cadmium Telluride (HgCdTe) photodetector having a response time less than 1.5 ns (VIGO Systems model PVM-10.6). All temporal signals were displayed on a 2.5 GHz bandwidth oscilloscope (LeCroy Wavesurfer 62XS). Unlike in the high-depletion Nd:YAG pumped OPO described in [27] for which the signal build-up time was shorter, in our low power experiment this build-up time is slightly longer. Also the idler FWHM pulse width was found to be shorter than the pump pulse width contrary to [27], in agreement with the linear regime of conversion efficiency shown in figure 4. Figure 8 represents the 2D (a) contour plot and 3D profile (b) of the mid IR idler beam profile at 7.6 μm, measured with a commercial infrared Spiricon PYROCAM-III camera equipped with LiTaO 3 pyroelectric detector (active area, 12.4×12.4 mm; element size, 0.1×0.1 mm). The Si lens with focal length 100 mm was used for the measurement of beam widths as function of propagation distance. The M 2 measurement was carried out at the following conditions: E pump =670 μJ, f=1000 Hz. As can be noticed the idler beam profile is far from a Gaussian pattern as often expected for nanosecond OPOs, exhibiting in particular a pedestal in the vertical direction. A software automatically evaluates the beam diameter at 1/e 2 intensity level. For the plot in figure 9, the distance between the lens and the camera was varied from 13.5 cm to 36.5 cm.

The beam profile and M 2 measurements
The plot in figure 9 displays the experimental values of square of the beam width D g in terms of Gaussian approximation, both along the horizontal (black squares) and vertical (red bold circles) transverse directions. The fitting of the squared widths with a quadratic polynomial was made according to international standard ISO 11146-1-2005. The resulting M 2 factors obtained from the fits are M 2 x =24 and M 2 y =20 for the idler beam at λ=7.6 μm. Compared to the previous high-energy, high-efficiency BGSe OPO experiments reported in [27],  x =40 and M 2 y =60 were then reported, due to backconversion distortions and the large Fresnel number cavity.

Damage threshold measurements
Damage threshold of MIR nonlinear materials are key parameters for high-power laser applications. Owing to its large bandgap, BaGa 4 Se 7 displays ∼4× higher surface damage threshold than the ternary chalcopyrite such as AgGaS 2 or AgGaSe 2 [22]. Damage thresholds for low repetition rate pulse are usually expressed in terms of energy fluence F (in unit of J/cm 2 ) while in high repetition rate regime or continuous-wave, they can be expressed in terms of intensity (in unit of W/cm 2 ). The first damage thresholds measurements on BaGa 4 Se 7 , performed with 1.064 μm pulses (τ=5 ns, f=1 Hz) focused to 2w=0.4 mm beam diameter, have been previously reported and yielded F th =2.8 J cm −2 (corresponding to an intensity I th =557 MW cm −2 ) [22]. But since damage may also depend on crystal quality it is worth measuring it with respect to the higher quality sample of the same origin used in [27], for which a twice lower surface damage threshold of F th =1.4 J cm −2 (τ=14 ns, f=100 Hz, 2w=5 mm) has been estimated. This fluence translates to I th =100 MW cm −2 which is the highest damage threshold value among non-oxide MIR nonlinear compounds, except for LiGaS 2 and BaGa 4 S 2 (BGS) having even larger bandgap than BGSe. The twice smaller damage threshold measured by the authors of [27] was attributed to the 10× larger beam diameter they used compared to the 0.5 mm diameter used in [22]. Indeed sampling larger surface addresses more local surface defects or impurities.
In order to characterize the damage threshold of our sample, several thin (∼1-2 mm) plates cut with the same orientation and from the same ingot as the present OPO sample were prepared and polished. Some of them were AR coated as the OPO sample, but these AR-coated test plates definitely displayed twice lower damage threshold than the uncoated ones, meaning that the damage observed relates rather to the dielectric ARC layer. In the following we report thus only on damage of uncoated samples. In order to increase the energy fluence of the Nd:YLF laser (with maximum pulse energy 1.3 mJ in a M 2 <1.3 beam quality factor and a pulse duration τ=16 ns) the laser beam was focused down to a diameter 2w=200 μm and the surface of each sample was mapped every mm 2 and scanned with the focused beam. Each mapped area was exposed during 5 min and the sample surface was examined after each exposure with a microscope in order to identify eventual dark spots. The pulse energy fluence is then incremented in steps of 0.05 J/cm 2 and the process repeated until trace of dark (burned) spots appear under microscope examination. The damage fluence threshold is defined as the fluence just preceding the observation of these surface dark spots. We studied the damage formation at 3 repetition rate values of the Q-switched laser: f=100, 150 and 200 Hz. Figure 10(a) shows the picture of one of the test plates after irradiation, displaying a constellation of faint and much darker and larger spots, the latter ones appearing beyond the defined threshold fluence (just after observation of the faint spots). As the laser pulse energy is increased, the larger surface spots develop as craters inside the material. On much thinner samples-that had to be fixed on a metal holder with a ring aperturethese craters may even evolve into a crack, fracturing the sample. Such a crack is highlighted by the white vertical arrow in figure 10(b). The final results of the damage threshold measurements are summarized in table 1, each row representing a fixed repetition rate of the Nd:YLF laser. The damage threshold F th and I th refer to an average irradiated area πw 2 while the last column is the peak on-axis intensity threshold defined as 2×I th . From the table it can be seen that the onset of damage threshold diminishes as the pulse repetition rate is increased. The decrease is relatively modest (∼1%) from f=100 to 150 Hz so that the damage threshold can be considered as constant up to 150 Hz, but an abrupt decreasing trend is noted at f=200 Hz (∼11%). Extrapolation of our data to f=1 Hz as in [22] is not straightforward, but our value F th ∼2 J/cm 2 is rather consistent with the value 2.8 J/cm 2 given in this reference at f=1 Hz, τ=5 ns and λ=1.064 μm. These values, related to small beam diameters, are among the highest for a mid-IR nonlinear crystal and confirm the beam diameter-dependent threshold value hinted in [27].

Concluding remarks
We have demonstrated a low threshold BaGa 4 Se 7 nanosecond optical parametric oscillator pumped for the first time by a Q-witched Nd:YLiF 4 low power laser at 1.053 μm. With less than 1 mJ pump pulse energy, the OPO threshold was as low as 250 μJ at an idler wavelength of 8.1 μm. The comparison of this experimental threshold with theory allowed us to estimate a nonlinear effective coefficient (for type-I (o-ee) down-conversion in the xz plane) exceeding 12 pm V −1 . Such a large value seems to indicate that the relative sign of the d 23 and d 16 tensor elements of BaGa 4 Se 7 should be the same. A maximum idler energy up to 45 μJ at 3.3 μm and 14 μJ at 8.1 μm has been achieved. A wide mid-IR idler wave coverage of 2.6-10.5 μm has been obtained using a single type-I cut crystal, confirming the promising status of this newly discovered biaxial monoclinic chalcogenide compounds for efficient mid-IR nonlinear down-conversion over its full transparency range (0.47-18 μm). Our downconversion phase-matching angle measurements were further compared with recently constructed linear dispersion relations for this material, with a good agreement found with the Sellmeier equations of Kato et al [33]. Finally our damage threshold measurements (∼2 J/cm 2 at 100 Hz repetition rate) are found in fair agreement with a previous measurement performed at 1.064 μm by another group [23] at similar beam diameter conditions. Figure 10. (a) Uncoated BaGa 4 Se 7 plate showing the formation of surface dark damage spots of increasing size as the Nd:YLF laser pulse fluence is increased above the damage fluence threshold; (b) Thinner sample as seen through the circular hole of its metallic holder: As the fluence is further increased, craters are formed at the dark spots locations (red arrows) and may even induce a crack trail (white vertical arrow).