Octave-spanning mid-infrared femtosecond OPA in a ZnGeP2 pumped by a 2.4 μm Cr:ZnSe chirped-pulse amplifier

We report on the highly efficient, octave-spanning mid-infrared (mid-IR) optical parametric amplification (OPA) in a ZnGeP2 (ZGP) crystal, pumped by a 1 kHz, 2.4 μm, 250 fs Cr:ZnSe chirped-pulse amplifier. The full spectral coverage of 3−10 μm with the amplified signal and idler beams is demonstrated. The signal beam in the range of ∼3− 5 μm is produced by either white light generation (WLG) in YAG or optical parametric generation (OPG) in ZGP using the common 2.4 μm pump laser. We demonstrate the pump to signal and idler combined conversion efficiency of 23% and the pulse energy of up to 130 μJ with ∼2 μJ OPG seeding, while we obtain the efficiency of 10% and the pulse energy of 55 μJ with ∼0.2 μJ WLG seeding. The OPA output energy is limited by the available pump pulse energy (0.55 mJ at ZGP crystal) and therefore further energy scaling is feasible with multi-stage OPA and higher pump pulse energy. The autocorrelation measurements based on random quasi-phase matching show that the signal pulse durations are ∼318 fs and ∼330 fs with WLG and OPG seeding, respectively. In addition, we show the spectrally filtered 30 μJ OPA output at 4.15 μm suitable for seeding a Fe:ZnSe amplifier. Our ultrabroadband femtosecond mid-IR source is attractive for various applications, such as strong-field interactions, dielectric laser electron acceleration, molecular spectroscopy, and medical surgery. © 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement


Introduction
Femtosecond laser sources with an ultrabroadband spectral span in the mid-wave infrared (MWIR, 3-8 µm) and long-wave infrared (LWIR, 8-15 µm) portions of the mid-infrared (mid-IR) spectrum have interesting applications, ranging from molecular fingerprinting and studies of ultrafast dynamics [1,2] to strong-field laser sciences [3]. Besides the recently developed ultrafast CO 2 lasers at ∼10 µm and Fe:ZnSe solid-state lasers at ∼4 µm, parametric down-conversions, such as optical parametric amplification (OPA), optical parametric chirped-pulse amplification (OPCPA) and difference-frequency generation (DFG), are the commonly used method to generate high-energy ultrafast mid-IR laser pulses.
Ultrafast OPA with non-oxide nonlinear crystals, such as ZnGeP 2 (ZGP), CdSiP 2 , and GaSe, pumped at ∼2 µm wavelength, is found to be a promising way of accessing the full mid-IR range up to ∼15 µm [4][5][6][7] due to their broad transparency range and favorable phase-matching conditions. Particularly, thanks to its high effective nonlinearity (77 pm/V) and high damage threshold, ZGP has been the most widely used for high-energy mid-IR OPA [4] and OPCPA systems [5]. Mid-IR ZGP OPA with a greater-than-octave-spanning spectral bandwidth was demonstrated using a ∼30 fs, 2.1 µm OPCPA pump source [6], but the conversion efficiency was as low as 4% due to a relatively thin (0.5 mm) ZGP crystal. Picosecond ∼2.05 µm Ho-doped lasers are also suitable for pumping ZGP OPA/OPCPA [5,8], but relatively narrow mid-IR bandwidths have been obtained from those OPA/OPCPA systems. For achieving the broadest mid-IR phase matching bandwidth available from ZGP OPA/OPCPA, a slightly longer pump wavelength at ∼2.5 µm is strongly suggested [4].
Transition-metal-doped II-VI materials, such as Cr:ZnSe, Cr:ZnS, and Fe:ZnSe, are found to be a suitable choice for direct lasing in the mid-IR with broad wavelength tunability and ultrashort pulse generation. These materials combine superb ultrafast laser capabilities with high nonlinearity of wide-bandgap zinc-blende semiconductors. For example, Cr:ZnSe and Cr:ZnS [9][10][11][12][13][14][15][16][17] support all standard regimes of femtosecond lasers and amplifiers at ∼2.4 µm and also enable efficient frequency conversion of ultrashort pulses via χ (2) and χ (3) nonlinearity. They are emerging as excellent mid-IR lasing materials due to the improved crystalline growth and doping process [9][10][11]. In this regard, recently available mJ-level, femtosecond, ∼2.4 µm Cr:ZnSe laser amplifiers [17,18] are an interesting source for pumping ZGP OPA as well as directly driving strong-field interactions [19]. Cr:ZnSe laser amplifiers can possibly establish themselves as an important laser source in the mid-IR range like CPA Ti:sapphire lasers in the near-IR range. For these reasons, robust and relatively compact, femtosecond mid-IR Cr:ZnSe lasers can get increasingly popular in ultrafast mid-IR research.
In this paper, we demonstrate a highly efficient, femtosecond ZGP OPA system fully covering the mid-IR range of 3-10 µm using a 2.4 µm Cr:ZnSe pump laser. We first describe the details of the pump laser and then compare the OPA phase matching conditions in ZGP between 2.4 and 2.05 µm pumping. In the experimental sections, we present a systematic study performed using two different methods for signal generation in the 3-5 µm range: white light generation (WLG) in YAG [19] and optical parametric generation (OPG) in ZGP [4]. We obtain the output energy of up to ∼130 µJ and a conversion efficiency of ∼23% from the OPG seeding. More detailed characterization of pulse energy, signal pulse duration, and beam profiles will be presented. Additionally, we demonstrate the spectrally filtered OPA at 4.15 µm, suitable for seeding a Fe:ZnSe amplifier. Our demonstration offers an energy scalable and technically simple platform of ultrafast laser sources with the possible coverage of the entire mid-IR region for investigating spectroscopic chemical detections and phase-sensitive strong-field interactions [1,6].

Cr:ZnSe pump laser at 2.4 µm
The Cr:ZnSe laser for pumping our ZGP OPA is a femtosecond, mid-IR chirped-pulse regenerative amplifier (CLPF-CPA, IPG Photonics) operating at a 1 kHz repetition rate, pumped by a Qswitched Ho:YLF laser. The Cr:ZnSe regenerative amplifier [18] is seeded by a Kerr-lensmode-locked femtosecond Cr:ZnS laser oscillator. The Cr:ZnS oscillator and Ho:YLF pump are optically pumped by radiations of an Er-doped fiber and Tm-doped fiber lasers, respectively. The diagram of the Cr:ZnSe CPA laser is shown in Fig. 1(a). The main system, as indicated by the red arrows of the 2.4 µm beam, consists of two units: grating-based stretcher/compressor unit and amplifier unit. Stretched pulses are amplified in the Cr:ZnSe regenerative amplifier and then re-compressed in the compressor, which outputs amplified femtosecond pulses. The free-space optical paths are isolated from the environment. This system is cooled by water and purged by industrial nitrogen gas. The output spectrum has a center wavelength of 2.40 µm with the peak at 2.45 µm, as shown in Fig. 1(b). The spectral bandwidth is broader than 100 nm in full-width at half maximum (FWHM) supporting a transform-limited pulse duration of <200 fs. The measurement was carried out after ∼1 hour of warm-up of the system. The oscillations in the spectroscopic signal are mostly attributed to the water vapor absorptions. The power stability measured at 1.23 W (or 1.23 mJ in energy) is better than 5% rms over 10 minutes, as shown in Fig. 1(c). This initial transient appears to be related to the humidity fluctuation of the regenerative amplifier cavity during the warm-up time. In the latter 5-minute range of the same graph, the shot-to-shot stability is as good as 2.5% rms, which can be further improved by reducing the timing jitter of the Ho:YLF pump pulses. The average output power and the shot-to-shot stability do not degrade over an entire day (∼8 working hours). The inset is the near-field output beam profile measured after the compressor. The beam profile is near Gaussian for both axes with some ellipticity of ∼0.6. In Fig. 1(d), the pulse duration of ∼250 fs was measured using intensity autocorrelation (AC) of output pulses.

Phase matching in ZGP for mid-IR OPA
We calculate the phase matching related parameters, such as angular tuning curve, effective nonlinear coefficient (d eff ), group velocity mismatch (GVM), and pulse splitting length, in ZGP pumped at 2.40 µm wavelength for broad mid-IR OPA, using the SNLO software [20] and Sellmeier equation [21]. These parameters are compared with the cases of 2.05 µm pump wavelength that are available with near-IR-pumped OPA/OPCPA systems [22] or Ho-doped lasers [23]. The angular tuning curves for type-I phase matching (e pump → o signal + o idler ) corresponding to the broadest spectral range of the signal and idler beams are shown in Fig. 2(a). With the pump wavelength at 2.40 µm and θ=49.0°, ZGP gives perfect phase matching at around 3.2, 4.5, 5.6, and 9.5 µm while the good phase matching over the full band of 3 − 10 µm can be achieved within ∼1°of angle at 49°. The additional band of pump spectrum along with the beam divergence helps to widen the amplified bandwidth and fill up the spectral gaps. In contrast, the 2.05 µm pump requires ∼4°of angle tuning of ZGP to achieve the phase matching over 3 − 10 µm though it allows excellent phase matching for the range of 7 − 12 µm at θ = ∼50.5°. Figure 2(b) shows the d eff value for two pump wavelengths. Changes in d eff vary in the range of 3 − 12 µm by 1% for 2.40 µm and 5% for 2.05 µm, allowing for efficient amplification across the broad spectral range. Although the difference is not significant, the d eff value at the phase matched angle for the 2.40 µm pump is higher in the 3 − 9 µm range. The GVM curves between pump and signal as well as pump and idler are shown in Fig. 2  than with 2.05 µm pump. The interaction length over which parametric amplification occurs is expressed as the pulse splitting length [24,25], as shown in Fig. 2(d). The temporal walk-off between two pulses becomes the same as the pump pulse duration after the propagation over the pulse splitting length. Using the calculated GVM, the pulse splitting length has been derived with a pulse duration of 250 fs from the Cr:ZnSe CPA laser. The relatively long pulse splitting length allows for the reduction of the required pump intensity to achieve high gain in ZGP crystals, enabling the use of longer crystals for amplification. According to Fig. 2(d), with 2.40 µm pump the pulse splitting does not occur in up to ∼13 mm of ZGP crystal over the entire 3 − 10 µm range, which is ∼2x longer than that with the 2.05 µm pump. Based on this calculation and the commercial availability, a 5 mm-long ZGP crystal (BAE systems) is chosen for the experiment of broadband OPA free of pulse splitting although a ZGP crystal as long as ∼13 mm could be used to increase the gain or lower the pump peak power. All the parameters plotted in Figs. 2(a)-(d) support that the 2.40 µm pump is more advantageous than 2.05 µm pump for broadband OPA in the range of 3 − 10 µm.

OPA setup
The experimental layout of the mid-IR ZGP OPA system is composed of the signal generation stage, the pump arm, and a single ZGP OPA stage, as illustrated in Fig. 3(a). The broadband mid-IR seed beam in the range of ∼3 − 5 µm as OPA signal is generated in the χ (2) or χ (3) nonlinear crystal (NL in Fig. 3(a)) pumped by the common 2.4 µm Cr:ZnSe laser. In our experiment we use either WLG (χ (3) ) based on laser filamentation in a YAG crystal or OPG (χ (2) ) in a ZGP crystal. The 2.4 µm pump beam is split into the seed generation stage (reflection) and the OPA arm (transmission) using an uncoated ZnSe wedge. The ∼100 µJ portion out of the 1.2 mJ, 2.4 µm pump is reflected off the wedge and used for the signal generation.

Signal generation using WLG
The supercontinuum generation (SCG) process from a single laser filament in a solid is widely used for WLG covering multi-octave spectral bandwidth. The WLG seeding has also commonly been used for near-IR-pumped OPA systems. The OPA signal bandwidth of ∼3 − 5 µm is relatively close to the pump wavelength of 2.4 µm and can be reached via WLG. In particular, since the SCG process can maintain the temporal coherence of the pump pulse, i.e., the fixed phase relationship between pump and signal, the carrier-envelope phase (CEP) of idler pulse is passively stabilized in the OPA process due to the cancellation of the common phase offset of pump and signal pulses, which is very useful for driving phase-sensitive strong-field interactions in the few-cycle regime.
In our experiment, we utilize a 6 mm-long YAG crystal as filamentation medium for WLG. YAG is found to be suitable for mid-IR laser filamentation [26] due to the broad transmission range in the mid-IR, high damage threshold and moderately small, negative group-delay dispersion value of -149 fs 2 /mm at ∼2.4 µm suitable for soliton-like pulse propagation without significant pulse broadening. The optical layout of WLG is shown in the signal generation arm in Fig. 3(a). The polarization of the laser beam reflected off the wedge plate is changed to vertical using a half waveplate (WP) in order to later satisfy type-I phase matching at the OPA stage. The pulse energy and beam size are adjusted using a neutral density (ND) filter and an iris, respectively, so that we obtain the single-filament WLG. The input beam diameter of ∼6 mm is found to be optimal when ∼15 µJ pump beam is focused onto YAG using an f =10 cm CaF 2 lens (L1). The position of the YAG crystal relative to the beam focus is further optimized such that we maximize the mid-IR extension to ∼4.5 µm while maintaining the single filament. The optimal crystal position is found to be ∼6 mm behind the focus, measured from the input surface of YAG crystal. The produced white-light beam is collimated with an f =5 cm CaF 2 lens (L2) and loosely focused at the OPA stage with ∼3 mm in diameter.
The WLG spectrum is recorded using a fiber-coupled near-IR spectrometer (NIR-256-2.5, Ocean Optics Inc.) for the wavelength range of 1.0-2.5 µm and a scanning-grating monochromator (MicroHR, Horiba JY Inc.) with a liquid-nitrogen-cooled HgCdTe (MCT) detector for the mid-IR wavelength (2 − 12 µm). Figure 3(b) shows the measured spectrum that spans more than octave bandwidth in logarithmic scale from the near-IR to mid-IR, where the spectra from the two spectrometers are stitched at ∼2.2 µm. The mid-IR spectral edge reaches ∼4.5 µm although the CO 2 absorption at ∼4.3 µm hinders the clear cutoff wavelength. For the purpose of OPA seeding, the residual 2.4 µm pump beam is removed by a long-pass filter (LPF) with the cut-on wavelength at 3.0 µm (EOC Inc.) and the pulse energy in the signal band (3.0 − 4.5 µm) is ∼0.2 µJ.

Signal generation using OPG
The ultrabroadband phase matching and high d eff value, as discussed in section 3.1, make the ZGP crystal very attractive for not only OPA but also OPG as another method of OPA signal (seed) generation when pumped at 2.4 µm of wavelength. For OPG we use a 5 mm-long, anti-reflection (AR)-coated ZGP crystal (θ = 49°, BAE systems) in a collinear type-I phase matching geometry. As shown in Fig. 3(a), we use the same focusing and recollimating optics as the WLG seeding setup for compatibility. The pump polarization is, however, unchanged unlike WLG as the OPG signal beam will have the orthogonal polarization. The pulse energy and beam size are adjusted using the same ND filter and iris to maximize OPG while avoiding any optical damage. With the input beam diameter of ∼2 mm the pulse energy of ∼25 µJ is used for pumping ZGP. The pump beam is focused into the ZGP crystal using the same f =10 cm CaF 2 lens and the ZGP crystal is located ∼2 mm behind the focal position. The produced OPG signal and idler beams are collimated with the same f =5 cm CaF 2 lens and loosely focused at the OPA stage. The residual 2.4 µm pump beam is removed by the 3.0 µm LPF and the OPG spectrum is recorded using the same mid-IR spectrometer. We use two additional LPFs with the cut-on wavelengths at 4.5 and 7.3 µm (EOC Inc.), respectively, to analyze the signal and idler spectra by removing high-order diffractions. Figure 3(c) shows the obtained OPG spectrum, which is the combination of three measurements with three LPFs including the 3.0 µm LPF. The spectrum covers from 3.0 to 8.5 µm in the linear scale. The pulse energy in the signal range is ∼2.3 µJ. The OPG itself has a conversion efficiency into the signal band of ∼9% from the pump energy, which is found to be significantly more efficient than WLG. It should be noted that we have a random phase relationship between the pump and signal pulses in the OPG process. A minor contribution of OPG idler band to the OPA seeding may exist, but the OPA process is still dominated by the OPG signal band because the OPG idler beam is estimated to have nearly an-order-of-magnitude lower fluence than that of the signal beam at the OPA crystal due to the lower pulse energy and larger beam size.

ZGP OPA stage
The signal beam from either WLG or OPG is amplified in an AR-coated 5-mm-long ZGP crystal (θ = 49°, BAE) which is identical to the ZGP crystal for OPG. As shown in the pump arm of Fig. 3(a), the pump beam is delivered to the ZGP crystal after a delay stage and a telescope with two CaF 2 lenses with f =15 cm (L3) and -10 cm (L4), respectively. The telescope down-collimates the pump beam and allows the beam size adjustment at the ZGP crystal for maximizing the gain without inducing the optical damage and excessive superfluorescence (SF) which is essentially OPG competing with OPA in the presence of seeding. The vertically polarized pump and horizontally polarized signal beams are spatially combined at the ZGP crystal for the collinear type-I OPA with a 300 µm-thick Brewster-angle Si wafer that works as a polarizing beam splitter (PBS) to transmit the pump beam while reflecting the signal beam. The pump beam with 0.55 mJ of energy is available for actually pumping the ZGP crystal due to the reflection loss at the metallic mirrors and transmission loss at the wedge plate, the lenses, and the silicon plate. Even though the OPA is basically achieved in a collinear geometry, we can induce ∼2°angle between the pump and signal beams when we need to spatially separate the signal and idler beams.

Optimization of OPA
We first carried out the WLG-seeded OPA experiment and systematically optimized the OPA conditions. The optimization started with the pump beam because the pump beam intensity is the most critical to determine the OPA gain while not damaging the crystal. The pump energy was fixed at ∼0.4 mJ and the pump beam size was adjusted using the telescope in Fig. 3(a), such that we start observing noticeable SF (>10 µJ) without the seed (signal) beam. This condition indicates the OPA gain must be high enough to be measured once the seed beam is spatially and temporally overlapped with the pump beam. Then we aligned the seed beam to the pump beam while letting the seed beam size (>3 mm in dia.) overfill the pump beam size to suppress SF generated from the unseeded part of the pump beam. Finally, the temporal overlap between the two beams was ensured using the delay stage. Once we observed the gain, we fine-tuned the spatial and temporal overlaps again. Figure 4(a) shows the typical amplified spectra from WLG-seeded OPA. The blue solid line represents the amplified spectrum of the signal, the green dashed line, the input WLG seed, and the red solid line, the SF without seeding. The WLG spectrum is plotted in linear scale unlike Fig. 3(b). It should be noted that SF in the presence of a seed is much weaker than SF without a seed because the seed takes over the gain from the noise. The signal gain was as high as ∼4 × 10 2 . To find out the optimal pump beam size for OPA, we scanned the position of f =-10 cm lens in the telescope (L4 in Fig. 2(a)). Figure 3(c) indicates that the beam diameter of 2.0 − 2.2 mm at the ZGP crystal was optimal for obtaining high conversion efficiency while maintaining the minimal SF. The conversion efficiency itself is nearly flat at the beam diameter of 1.5 − 2.2 mm. The pump intensity at 2.0 mm of beam diameter is estimated to be ∼35 GW/cm 2 , which takes into account the ∼30% of pulse energy in the temporal pedestal ( Fig. 1(d)). The conversion efficiency started dropping as the beam diameter gets smaller than ∼1.5 mm. An excessive pump intensity not only induces the strong SF but also can eventually cause a permanent surface/bulk damage.
We characterized the optical spectrum and beam profile of the WLG-seeded OPA output using the mid-IR spectrometer and a mid-IR beam profiler (WinCamD, DataRay), respectively, as shown by Fig. 4(c). For this measurement the pump pulse energy was increased to 0.55 mJ and the signal and idler combined, output pulse energy was 55 µJ. Despite the low seed energy of ∼0.2 µJ in the signal band, the amplified spectrum fully covers the wavelength from 3 to 10 µm, confirming that the ∼2.4 µm pump wavelength is crucial to obtain the octave-spanning phase matching band of ZGP in the mid-IR along with high gain, as discussed in section 3.1. The long-wavelength tail reaches up to ∼12 µm, limited by the absorption band of ZGP, but the spectral contents beyond 10 µm was negligible with the unfavorable phase matching. As a result, we were able to generate more-than-octave-spanning continuum-like spectrum in the mid-IR from the single OPA source. The spectral peak at 2.4 µm is the residual pump beam after the 3 µm LPF. The wavelength range of 3.0 − 4.5 µm is considered as signal band while the 4.5 − 10 µm range, as idler band. Although ∼4.8 µm is the nominal degenerate point of the center wavelength of pump beam, it is not sharply defined due to the broad spectral pump band of ∼2.25 − 2.52 µm ( Fig. 1(b)). The idler itself has an octave-spanning spectrum in full width at 10% maximum and can be compressed to ∼1.5 optical cycles in duration with the passive CEP stability. The absorption line of atmospheric CO 2 at ∼4.3 µm is clearly observed along with some other fine structures. A vacuum, noble-gas purging, or dry-air purging laser enclosure could be used to remove these spectral features and to further increase the output pulse energy. The near-field beam measurements of both signal and idler (the inset of Fig. 4(c)) show decent beam profiles with a single peak and the idler beam size is much larger than the signal beam size, as expected from wavelength-dependent diffraction. Here, a 3.5 − 4.5 µm bandpass filter (BPF) was used for the signal beam and a 4.5 µm LPF, for the idler beam.
The same optimization process was applied to the OPG-seeded OPA. Due to much higher seed energy of ∼2.3 µJ, the OPA output pulse energy (signal + idler) was enhanced to 130 µJ at 0.55 mJ pumping. The corresponding optical spectrum and the spatial beam profile are shown in Fig. 4(d). While the signal and idler combined spectrum covers the same range from 3 to 10 µm as the WLG-seeded case, the spectral contents are more evenly distributed due to the flatter seed spectrum from OPG ( Fig. 3(d)). The near-field beam profiles were also found to be more uniform than the WLG-seeded OPA beams. Detailed comparison of performance between two OPA configurations will be discussed in next section.

Comparison of WLG-seeded and OPG-seeded OPA
We performed systematic studies of WLG-seeded OPA (WLG/OPA) and OPG-seeded OPA (OPG/OPA) using the same experimental setup of Fig. 3(a). The measurements of pulse energy and conversion efficiency vs. pump energy are shown in Figs. 5(a) and (b), respectively, where the 2.4 µm pump beam diameter at the ZGP crystal was set to ∼2.0 mm. The slope of both OPA outputs vs. pump energy in Fig. 5(a) is almost linear, indicating that saturation or back-conversion is not noticeably high. The maximum output energy (signal + idler) was as high as 55 µJ from WLG/OPA and 130 µJ from OPG/OPA. The ∼10 times higher seed energy from OPG is the obvious reason why we obtained much higher output energy from OPG/OPA. The corresponding conversion efficiencies were up to 10% and 23% for WLG/OPA and OPG/OPA, respectively. To the best of our knowledge, the conversion efficiency of 23% is the highest among the octave-spanning or greater-than-octave-spanning OPA systems in the mid-IR. Considering the ∼80% transmittance of the 3 µm LPF filter used for the energy measurement, we even estimate ∼29% of conversion efficiency directly from the OPA. The pulse energy stability is important for reliable applications of OPA output pulses. Figure 5(c) shows the shot-to-shot energy stability of signal and idler combined pulses, measured with the 3 µm LPF to block unconverted pump energy. Here, we set the output pulse energy to 40 µJ and 80 µJ from the WLG/OPA and OPG/OPA, respectively, and measured the shot-to-shot rms stability over 12,000 shots. With the 2.4 µm pump laser stability of ∼4.6% rms, we observed some degradation to ∼6.6% rms with WLG/OPA and ∼8.2% rms with OPG/OPA. In fact, since the energy stability of the Cr:ZnSe pump laser can be enhanced to ∼2.5% rms (the latter half of Fig. 1(c)) with re-optimization along the sufficient warm-up time, the OPA stability can be enhanced to ∼3 − 5% level as well. Regarding the long-term energy stability, no noticeable degradation was observed over many hours. The OPG/OPA was less stable than the WLG/OPA because OPG is based on a spontaneous amplification progress starting from a noise, but it can be further improved if we use a longer ZGP crystal for OPG to saturate the gain.
For the temporal characterization of amplified signal pulses, we used a home-built intensity autocorrelator with a 2-mm-thick polycrystalline ZnSe window as a second-harmonic generation (SHG) and sum-frequency generation (SFG) medium based on random quasi-phase matching (RQPM) [27] that supports ultrabroad phase matching in the mid-IR. The SHG and SFG of the signal pulse in the range of 1.5-2.4 µm were detected using the near-IR spectrometer. Figure 6(a) illustrates the setup of the autocorrelator for the ultrabroadband mid-IR pulses. A 300-µm-thick Si plate was used as a BS at 45°angle of incidence. With a delay stage and a metallic off-axis parabolic mirror (OAP), two beams are temporally overlapped at and focused into the ZnSe window for SFG in a noncollinear geometry. We tuned the input pulse energy such that it is just high enough for SHG from each arm and then maximized the SFG (AC signal) with the delay scan while cutting out the background SHG from each arm using a slit. Figures 6(b) and (c) show the measured AC trace of the pulses from WLG/OPA and OPG/OPA, exhibiting an AC width of 466 and 450 fs in FWHM, respectively. The corresponding pulse duration is 330 and 318 fs, respectively, with the Gaussian assumption of temporal profile. Both cases show the minor pulse broadening from the pump pulse duration (∼250 fs) indicating the minimal GVM between pump and signal (∼10 fs/mm from Fig. 2(c)).
Given the broad spectral bandwidth of the signal beam (∼1000 nm in FWHM), the signal pulses can be compressed to the few-optical-cycle range (<30 fs) for WLG/OPA. The idler pulse duration was not measured in this work, but it is also expected to be close to the pump duration with the minimal GVM (<10 fs/mm over 5 − 10 µm from Fig. 2(c)) for both WLG/OPA and OPG/OPA. The idler pulses from WLG/OPA can be compressed to the sub-two-cycle range given the octave-spanning spectral bandwidth. While the compressibility of OPG/OPA pulses needs to be further investigated, the femtosecond pulse duration of signal beam indicates that the spectral phase is at least partially coherent over the ultrabroad spectral band.

Spectrally filtered OPA
The mid-IR OPA output can be re-optimized for relatively narrowband operation with spectral tunability for specific applications. In this work, we attempted to maximize the OPG/OPA spectrum at 4.15 µm to seed a Fe:ZnSe amplifier operating at the 3.5 − 4.5 µm range [28]. For the spectral filtering we used a BPF at 4.15 µm in ∼200 nm bandwidth with 70% of transmittance after the OPG stage and then re-optimized the spatial and temporal overlaps between the pump and seed beams. The OPA output was filtered again with a 4.2 µm BPF in ∼420 nm bandwidth to remove any residual SF outside the desired spectral band. The resultant spectrum is shown in Fig. 7, where the pulse energy of ∼30 µJ was obtained (0.55 mJ of pump) in the spectral bandwidth of 200 nm. The inset of Fig. 7 confirms a good near-field output beam profile. While the conversion efficiency can be further increased with a longer ZGP crystal, the pulse energy with the spectrum in Fig. 7 is high enough to seed a ultrafast Fe:ZnSe amplifier for multi-mJ ∼4 µm pulse generation [28].

Conclusions
We successfully developed a highly-efficient, robust mid-IR femtosecond OPA system with ZGP that fully covers the wavelength range from 3 to 10 µm, pumped by a 2.4 µm Cr:ZnSe laser operating at a 1 kHz repetition rate. A systematic and comparative study of WLG-seeded OPA and OPG-seeded OPA was presented. The maximum output energy of 130 µJ and the conversion efficiency of 23% were obtained from the OPG-seeded OPA scheme thanks to the high seed energy (∼2.3 µJ), while the moderate pulse energy (55 µJ) and efficiency (10%) were obtained with the WLG-seeded OPA. To the best of our knowledge, our OPG-seeded OPA is the most efficient mid-IR OPA source generating a greater-than-octave-spanning spectrum. The energy scalability is guaranteed with additional OPA stages given that high-power mid-IR pump lasers and high-quality large-area ZGP crystals are available. In addition, we demonstrated relatively narrowband OPA at ∼4.15 µm for seeding a Fe:ZnSe amplifier.
At the demonstrated energy level, ultrashort mid-IR laser pulses are already attractive for applications such as spectroscopic chemical detections, high-order harmonic generation in solids, laser filamentation in solids, and dielectric laser acceleration in nano-structures.