1 Introduction

Tunable, high-average-power, picosecond sources are of significant interest for a variety of applications, including pump–probe spectroscopy [1] and upconversion imaging [2]. Such sources in the mid-infrared (mid-IR) are of great demand for minimally invasive surgery [3] and spectroscopy in the molecular fingerprint region [4]. Access to the mid-IR region with spectral coverage beyond 4 µm is best achieved using nonlinear frequency conversion sources, in particular optical parametric oscillators (OPOs), driven by commercially available pump lasers in combination with suitable nonlinear crystals [5]. The advent of fiber laser technology offering simple, robust and reliable architecture in a compact footprint has enabled various scientific and technological applications [6], including pumping of OPOs [7, 8]. Fiber lasers have also been used to pump crystalline bulk solid-state mid-IR lasers. For example, Tm-fiber lasers have been successfully deployed to pump Ho-based solid-state lasers and Cr:ZnSe lasers [9]. However, to date, it has remained challenging to exploit Tm-based fiber lasers for pumping frequency conversion sources, particularly in the continuous-wave (CW) and picosecond time-scales, due to the lack of readily available commercial sources with suitable characteristics. On the other hand, Yb-fiber laser technology has been one of the most successful platforms for nonlinear frequency conversion sources. Pumped by Yb-fiber lasers at 1064 nm, OPOs based on MgO:PPLN have been extensively demonstrated in various time-scales from CW to ultrafast picosecond and femtosecond domain [10,11,12], providing wide and continuous tuning in the near-to-mid-IR up to ~ 4 µm, limited by long-wavelength transparency cut-off of oxide-based nonlinear materials due to multiphonon absorption. On the other hand, alternative nonlinear crystals such as ZnGeP2 and orientation-patterned GaAs (OP-GaAs) with extended mid-IR transparency have been developed to extend the spectral range of frequency conversion sources beyond ~ 4 µm [13]. However, such crystals require pumping above ~ 2 µm, which still remains challenging, mainly due to the lack of widespread availability of viable CW and mode-locked Tm/Ho-based lasers offering the required spatial and spectral quality together with high power, thus motivating the development of alternative pump sources near 2 µm [14]. These limitations have primarily driven the interest in novel nonlinear materials that can use the well-established 1-µm pump laser technology to generate mid-IR radiation beyond ~ 4 µm. Recent efforts in this direction have led to the development of a new nonlinear material, cadmium silicon phosphide, CdSiP2 (CSP) [15], a chalcopyrite semiconductor with transparency range extending from ~ 1 to 7 µm. CSP is a negative uniaxial birefringent crystal possessing a high nonlinearity (deff ~ 84.5 pm/V) and moderate thermal conductivity (13.6 W/m K). It can be grown in large apertures and long interaction lengths up to 20 mm, using horizontal gradient freeze technique. Importantly, the large bandgap of CSP allows pumping near 1 µm, while enabling parametric generation in the 6–7 µm spectral region in mid-IR under type-I (e → oo) noncritical phase-matching (NCPM), thus avoiding the detrimental effects of spatial walk-off.

Several ultrafast femtosecond and picosecond frequency conversion sources based on CSP have been previously reported [16,17,18,19,20,21]. In the picosecond time-scale, these include OPOs and optical parametric generators (OPGs) pumped by high-pulse-energy mode-locked and amplified solid-state lasers at 1064 nm [21,22,23,24,25,26,27,28]. The first high-energy picosecond OPO, based on a 9.5-mm-long CSP crystal, was non-collinearly pumped by a Nd:YAG master-oscillator power-amplifier (MOPA) system, providing 12.6 ps pulses at 25 Hz repetition rate at 1064 nm, generating 0.56 mJ (14 mW) of output energy (average power) at 6.4 µm [22]. Later, we demonstrated a picosecond OPO based on 12-mm-long CSP crystal, pumped by a Nd:YAG MOPA providing 8.6 ps pulses at 1064 nm and 20 Hz repetition rate, generating tunable mid-IR radiation in the 6.1–6.6 µm with up to 1.5 mJ (14 mW) of output [23]. A picosecond OPG using a flashlamp-pumped Nd:YAG amplifier providing 20 ps pulses at 1064 nm and 5 Hz repetition rate was also demonstrated, using a 12-mm-long CSP crystal, and generating tunable mid-IR radiation in the 6153–6731 nm range with up to 33 µJ (0.2 mW) of output at 6234 nm [24]. Recently, the OPO in Ref. [22] was upgraded to generate higher pulse energies of up to 1.1 mJ (25.5 mW) in the 6.5–7 µm range [25]. In addition, a number of OPOs and OPGs based on CSP have been demonstrated by pumping at kHz repetition rates, resulting in microjoule pulse energies in the mid-IR. These include an OPG based on 8-mm-long CSP crystal pumped by a mode-locked Nd:YVO4 laser at 1064 nm, providing 8.6 ps pulses at 100 kHz, generating 1.54 µJ (154 mW) at 6204 nm [26]. In another experiment, a passively Q-switched Nd:YAG MOPA at 1064 nm providing 400 ps pulses at 1–10 kHz was used to pump an OPO based on 7-mm-long CSP crystal to generate 4.3 µJ (4.3 mW) of idler energy at 6150 nm [27]. Finally, a diode-pumped Nd:YAG MOPA providing 500 ps pulses at 1–10 kHz was deployed to pump a seeded OPG generating 8.5 µJ (8.5 mW) of idler output at 6100 nm [28].

All of the above picosecond frequency conversion sources based on CSP developed to date have been driven by Nd-based bulk solid-state laser/amplifier systems providing high pump pulse energies at low (Hz–kHz) repetition rates. On the other hand, the development of picosecond OPOs pumped by mode-locked laser oscillators of low pulse energy at high repetition rate has been challenging due to the low peak intensities available to drive the nonlinear gain, combined with the increasingly stringent demands on the nonlinear material quality with regard to low transmission loss over long interaction lengths. While high-repetition-rate picosecond OPOs have been extensively demonstrated using the well-established nonlinear material, MgO:PPLN, providing multi-Watt average powers in the 1.5–4 µm spectral range [29], the development of such sources based on CSP has to date remained challenging due to the lack of availability of crystals of sufficiently high optical quality and low transmission loss over relatively long interactions lengths (10–20 mm) .

We recently reported the first high-repetition-rate picosecond OPO based on CSP, which was synchronously pumped by an Yb-fiber laser at 1064 nm operating at ~ 80 MHz repetition rate [30]. The successful demonstration of this OPO was enabled by major improvements in the optical quality of CSP crystal over long interaction lengths. Here we show that such improvements in CSP crystal quality and reductions in transmission loss further enable the deployment of output coupling in the OPO cavity, allowing the extraction of significant amounts of near-IR signal power in addition to the mid-IR idler power. We also report on the characterization of the CSP crystal with regard to transmission loss and thermal lensing, as well as detailed OPO performance including temperature tuning, power across the tuning range, power scaling, spectral as well as power stability, beam quality and temporal characteristics of the signal pulses. The OPO, which is configured as a singly-resonant oscillator (SRO) with output coupling for the resonant signal, provides tunable near-IR signal radiation across 1284–1264 nm together with mid-IR idler output across 6205–6724 nm, at high average powers at ~ 80 MHz pulse repetition rate. It generates up to 44 mW of signal power at 1284.2 nm and as much as 95 mW of idler power at 6205 nm, with more than 50 mW of average power over > 55% of the mid-IR tuning range. The OPO exhibits excellent spatial quality in both output beams with high passive long-term power stability better than 2.4% rms for the idler and 1.9% rms for the signal over > 17 h.

2 Experimental setup

The schematic of the high-repetition-rate picosecond OPO based on CSP is shown in Fig. 1. The pump source is a commercial Yb-fiber laser (Fianium, FP1060-20) providing up to 20 W of average power in 20 ps pulses at 79.5 MHz repetition rate. The laser operates at a central wavelength of 1064 nm with a double-peak spectrum and a full-width half-maximum (FWHM) spectral bandwidth of ~ 1.4 nm. A combination of a half-wave plate and a polarizing beam-splitter is used to systematically control the average pump power. A second half-wave plate is used to yield an extraordinary (e) pump polarization in the CSP crystal. The pump beam is focused to a waist radius of wo ~ 50 µm at the center of the crystal with a lens (L1) of focal length, f ~ 125 mm, corresponding to a focusing parameter of ξ ~ 0.4. The CSP crystal used as the OPO gain medium is 16.3-mm-long, with an aperture of 4 × 6 mm2, and cut at θ = 90° (ϕ = 45°) for type-I (e → oo) interaction under NCPM. The crystal is housed in an oven that can be adjusted from room temperature to 200 °C with stability of ± 0.1 °C, and its end-faces are antireflection (AR)-coated (R < 0.5%) over 1035–1300 nm for the pump and signal, with high transmission (R < 5%) over 4000–8000 nm for the idler. Using the relevant Sellmeier equations [31], we calculated the pump spectral acceptance bandwidth for the crystal under NCPM at room temperature to be ~ 1.4 nm, which very closely matches the FWHM pump spectral bandwidth, thus allowing efficient parametric conversion. The OPO is configured in a standing-wave X-cavity with two plano-concave mirrors (r = 150 mm), M1,2, and a plane mirror, M3. In order to partially extract the signal from the cavity, we deploy a plane output coupler (OC) as a fourth cavity mirror. All mirrors are highly transmitting for the pump (T > 97%) at 1064 nm and idler wavelengths (T > 98%) over 5500–7500 nm, while highly reflecting (R > 99%) for the signal over 1200–1400 nm, thus ensuring SRO configuration. The OC is partially transmitting (T ~ 5%) over the signal wavelength range. In order to separate the idler from the pump, we use a dichroic mirror, M4, with high transmission for the idler (T > 95%) and high reflection for the pump (R > 99%). The total round-trip optical length of the OPO cavity is ~ 3.77 m, ensuring synchronization with the pump laser repetition rate at 79.5 MHz.

Fig. 1
figure 1

Schematic of the experimental setup for high-repetition-rate, picosecond, mid-IR OPO based on CSP. FI Faraday isolator, λ/2 half-wave plate, PBS polarizing beam-splitter, L lens, M mirrors, OC output coupler

Full size image

3 Results and discussion

3.1 Crystal transmission

In order to assess the material quality, we initially characterized the CSP crystal for transmission at the pump wavelength for extraordinary polarization, as required for type-I (e → oo) parametric generation, with the pump beam focused to waist radius of wo ~ 40 µm at the center of the crystal. The result is shown in Fig. 2. As the pump power at the input to the crystal was gradually increased, the transmitted power was observed to increase linearly up to 2.5 W. A linear fit to this measurement resulted in a transmission of ~ 84% for the 16.3-mm-long CSP crystal at 1064 nm, corresponding to a linear absorption coefficient of α ~ 0.11 cm−1 for extraordinary polarization. This value is close to the absorption coefficient of CSP at 1064 nm, specified to be 0.12 and 0.16 cm−1 for ordinary and extraordinary polarization, respectively. Further increasing the pump power beyond 2.5 W resulted in a significant degradation in the transmitted beam quality, together with a drop in the transmission. Hence, we limited the maximum pump power to 2.2 W, corresponding to a pulse energy of ~ 28 nJ and peak intensity of ~ 27 MW/cm2 in our experiments. Under this condition, we did not observe any significant intensity-dependent drop in the transmission of the CSP crystal, due to the relatively low pump intensity used in this experiment. Moreover, we monitored the transmitted pump beam through the CSP crystal as the pump power was increased gradually from the minimum to 2.5 W. The spatial beam profiles of the transmitted pump at a distance of ~ 30 cm from the center of the CSP crystal, at two different pump powers of 0.1 and 2.5 W, are shown in Fig. 3a, b, respectively, clearly indicating thermal lensing at the high power, with the effective beam diameter increasing from 5.1 to 6.6 mm.

Fig. 2
figure 2

Transmission measurement of the 16.3-mm-long CSP crystal at 1064 nm for extraordinary polarization

Full size image
Fig. 3
figure 3

Spatial beam profile of the transmitted pump beam through the 16.3-mm-long CSP crystal at an input pump power of a 0.1 W, and b 2.5 W

Full size image

3.2 Temperature tuning

As an initial step in characterizing the OPO, we performed temperature-tuning measurements by varying the temperature of the CSP crystal. Given the 6-mm thickness of the crystal, each time the temperature was varied, sufficient time was spent to reach a steady-state value before the output wavelength was recorded. The crystal temperature was then measured on the top surface using a thermocouple. The generated signal and corresponding idler wavelength as function of crystal temperature are shown in Fig. 4. Here, the filled and hollow circles correspond to the measured data, while the solid and dashed lines represent the theoretical calculations using three recent Sellmeier equations and thermo-optic coefficients for the CSP crystal [31,32,33]. The signal wavelength was measured using a near-IR spectrum analyzer, while the idler wavelength was calculated from energy conservation. The experimental data can be seen to closely match the theoretical calculations from Ref. [31] at low temperatures, while a deviation is observed at higher temperatures. This deviation could be attributed to the thermal gradient in the relatively thick CSP crystal. As the crystal temperature is increased from 39 to 134 °C, the signal wavelength decreases from 1284.2 to 1264.9 nm, and the corresponding idler increases from 6205 to 6699 nm. The signal and idler wavelengths tune linearly with temperature at a rate of − 0.2 and + 5.2 nm/°C, respectively. The OPO is tunable over ~ 20 nm in the signal wavelength range, corresponding to 519 nm in the idler, indicating that a small change in the signal wavelength results in a significantly large change in the idler wavelength.

Fig. 4
figure 4

Experimentally measured signal and idler wavelengths as function of the temperature of the CSP crystal in the OPO, together with the theoretical calculations using the recent Sellmeier equations and thermo-optic coefficients

Full size image

3.3 Power across tuning range

Following the wavelength tuning measurements, we performed output power characterization of the OPO by simultaneously recording the average signal and idler output power across the tuning range by varying the temperature of the CSP crystal. The output-coupled signal and single-pass idler average power across the OPO tuning range are shown in Fig. 4. For a fixed pump power of 2.2 W, we were able to generate average signal powers varying from 44 mW at 1284.2 nm to 4 mW at 1264 nm, as shown in Fig. 5a. The simultaneously extracted idler power varies from as much as 95 mW at 6205 nm down to 14 mW at 6724 nm, with > 50 mW over > 55% of the mid-IR idler tuning range, as presented in Fig. 5b. It is to be noted that the data presented here are not corrected for the AR-coating loss from the optics or bulk transmission losses of the crystal, and represent the actual values recorded on the power meter. Further, we investigated the possibility of cavity length tuning of the picosecond CSP OPO at a fixed temperature. However, due to the large group delay dispersion (9251 ± 90 fs2 over 1264–1284 nm) and the narrow gain bandwidth associated with the long crystal, we were only able to achieve signal wavelength tuning of ~ 1 nm, corresponding to an idler wavelength tuning of ~ 26 nm, by varying the cavity length over ~ 400 µm.

Fig. 5
figure 5

Variation of the a signal, and b idler average power across the tuning range of the output-coupled CSP picosecond OPO

Full size image

3.4 Power scaling

We performed average power scaling of the high-repetition-rate picosecond OPO for both signal and idler, with the results shown in Fig. 6a. At a fixed crystal temperature of 39 °C, and for a maximum input pump power of 2.2 W, were able to generate up to 44 mW of near-IR signal average power at 1284 nm together with 95 mW of mid-IR idler average power at 6205 nm. These correspond to a power conversion efficiency of 2% and a photon conversion efficiency of 2.4% for the signal, and a power conversion efficiency of 4.3% and a photon conversion efficiency of 25% for the idler. Linear fits to the measured power scaling data result in slope efficiencies of 2.3% and 5.3% for the signal and idler, respectively. The average pump threshold for the output-coupled OPO was ~ 700 mW and a pump depletion of ~ 57% was recorded at 2.2 W of input pump power. Also shown in the inset of Fig. 6a is the signal spectrum recorded at the maximum power, exhibiting a double-peak structure, which we attribute to the double-peak pump spectrum as well as self-phase-modulation. We further performed power scaling of the mid-IR idler output, with the OPO in pure SRO configuration, by replacing the 5% OC by a plane high reflector (R > 99%) for the signal over 1200–1400 nm. The result is shown in Fig. 6b, where it can be seen that we were able to generate up to 105 mW of idler average power at 6205 nm, with a slope efficiency of 5%. This corresponds to an idler power conversion efficiency of ~ 5% and a photon conversion efficiency of ~ 28%. The average pump power threshold for the OPO in this case was ~ 300 mW, while a maximum pump depletion of ~ 54% was recorded at the highest input pump power of 2.2 W. As evident, no saturation in the idler power was observed at this pumping level even at higher operating temperature of the CSP crystal, indicating the possibility for further mid-IR power scaling, provided the quality of the crystal in terms of transmission is improved. It is to be noted that the presented data do not account for the losses due the AR-coating and the bulk transmission losses of the crystal.

Fig. 6
figure 6

Power scaling of the CSP picosecond OPO a using a 5% signal OC, and b in pure SRO configuration using high reflector

Full size image

3.5 Power stability and beam profile

The simultaneously measured long-term power stability of the near-IR signal and mid-IR idler output from the output-coupled CSP picosecond OPO are presented in Fig. 7. For a fixed input average pump power of 2.2 W and a constant temperature of ~ 39 °C, the extracted signal power at 1284.2 nm exhibits a good passive stability better than 1.9% over > 17 h at an average power of 40 mW, as shown in Fig. 7a. The corresponding idler at 6205 nm is also characterized by a good passive power stability better than 2.4% rms over the same period at an average power of 90 mW, as shown in Fig. 7b.

Fig. 7
figure 7

Long-term power stability of the a signal, and b idler from the output-coupled CSP picosecond OPO

Full size image

The spatial quality of the near-IR signal and the mid-IR idler output from the output-coupled picosecond CSP OPO were also investigated at a crystal temperature of ~ 39 °C, while generating maximum signal and idler average power. The far-field, two-dimensional spatial profiles of the signal at 1284.2 nm and idler at 6205 nm were measured at distance ~ 0.5 m from the cavity, and are shown in Fig. 8a, b, respectively, indicating that the beams exhibit excellent spatial quality with single-peak Gaussian intensity distribution in TEM00 mode profile in both cases.

Fig. 8
figure 8

Spatial beam profile of a signal, and b idler generated from the CSP picosecond OPO

Full size image

3.6 Spectral stability and temporal characterization

Further, we investigated the spectral stability of the signal beam extracted from the output-coupled CSP OPO at a crystal temperature of ~ 39 °C. Using a near-IR spectrum analyzer, we recorded the signal spectra over a period of 15 min, sampled at an interval of 10 s. The result is shown in Fig. 9a, indicating good passive spectral stability. The central wavelength of the signal was recorded to be 1283.6 nm, with a FWHM spectral bandwidth of ~ 1.8 nm, as shown in Fig. 9b. The standard deviation of the signal central wavelength is estimated to be ~ 47 pm.

Fig. 9
figure 9

a Spectral stability of the signal beam extracted from the CSP picosecond OPO, and b variation of central wavelength and FWHM spectral bandwidth of the signal over the measurement period

Full size image

We also performed temporal characterization of the signal pulses extracted from the output-coupled picosecond CSP OPO. We simultaneously recorded the pump and signal pulse trains from the CSP OPO at a crystal temperature of 39 °C, using Si and InGaAs fast photodetectors, respectively, with the results shown in Fig. 10a. As evident, the signal pulses are separated by ~ 12.5 ns, indicating synchronization with the pump laser repetition rate. The signal pulse train also exhibits a noticeable pulse-to-pulse instability, also evident from the long-term power stability measurement in Fig. 7a, which can be improved by active control of the OPO cavity length. Finally, we performed temporal characterization of the extracted signal pulses from the OPO using an intensity autocorrelator, at a CSP crystal temperature of 39 °C, with the result shown in Fig. 10b. The measurement resulted in a FWHM pulse width of 27 ps, corresponding to a temporal duration of ~ 19 ps, assuming Gaussian pulse shape. The corresponding signal spectrum centered at 1284 nm, with a FWHM bandwidth of Δλ ~ 1.9 nm, is shown in the inset of Fig. 10b. These measurements result in a time–bandwidth product of ΔτΔυ ~ 6.5, as compared to ΔτΔυ ~ 7.4 for the pump. Further improvement in the time–bandwidth product can be expected by using intracavity dispersion compensation for temporal and spectral control, as well as optimization of signal output coupling.

Fig. 10
figure 10

a Pump and signal pulse train from the CSP picosecond OPO. b Typical intensity autocorrelation measurement of the signal pulses from the CSP picosecond OPO; inset: corresponding signal spectrum centered at 1284 with Δλ ~ 1.9 nm

Full size image

4 Conclusions

In conclusion, we have demonstrated a high-repetition-rate picosecond OPO based on CSP, synchronously pumped by an Yb-fiber laser at 1064 nm. By deploying an SRO configuration with output coupling for the resonant signal wave, the OPO simultaneously delivers tunable near-IR signal and mid-IR idler radiation at practical average power levels. The OPO output is tunable in near-IR signal across 1264–1284 nm, with corresponding mid-IR idler tuning over 519 nm across 6205–6724 nm. Operating at ~ 80 MHz pulse repetition rate, the OPO provides practical near-IR signal average powers of up to 44 mW at a photon conversion efficiency of 2.4%, which can be further improved by optimization of output coupling. In the mid-IR, it produces idler average powers of as much as 95 mW at a photon conversion efficiency of 28%, with > 50 mW over 6205–6494 nm, corresponding to > 55% of the idler tuning range. We have studied the pump transmission characteristics of the 16.3-mm-long CSP crystal, resulting in ~ 84% transmission for extraordinary polarization, with no intensity-dependent variation in the transmission observed at the average pump power levels used for the OPO. We have also investigated the temperature-tuning properties of the CSP OPO and compared the results with the theoretical calculations using the most recent Sellmeier and thermo-optic coefficients for the material. Under free-running conditions, the idler and signal power from the CSP OPO exhibits passive power stability better than 2.4% rms and 1.9% rms over > 17 h when operating at 6205 and 1284.2 nm, respectively, in good spatial beam quality. The OPO output signal pulses have a Gaussian temporal duration of 19 ps, measured at 1284 nm, with a time-bandwidth product of ΔτΔυ ~ 6.5. Improvements in the time-bandwidth product and cavity delay tuning performance are expected by implementing intracavity dispersion compensation. Due to the significant degradation in the pump beam quality together with the drop in transmission through the CSP crystal, the maximum usable average pump power was limited to 2.2 W in the present experiments. As such, with improvements in the material quality, optimization of the crystal length and signal output coupling, further signal and idler power scaling will be attainable at increased pump powers. Our results demonstrate the potential of Yb-fiber-pumped high-repetition-rate picosecond OPOs based on CSP as viable sources of tunable near-IR and mid-IR radiation, offering practical average powers, high output stability, and excellent spatial beam quality, in compact and rugged table-top design, which will benefit may applications.