1. Introduction

Mid-infrared (MIR) coherent sources with emission in 3∼5 µm spectral region have found particularly useful applications in recent years, which include spectroscopy and imaging [1–4], gas or remote sensing [5,6], material processing [7], biomedicine [8,9], strong field science [10], etc. Although materials with rare-earth dopants are receiving more and more attentions in enabling gain in this spectral region [11], the approaches based on nonlinear parametric conversion still hold their distinguished advantages, especially in wavelength tunability and reachable power level, at least to present. Among them, the most widely investigated ones are optical parametric oscillator (OPO), optical parametric amplifier (OPA), and difference frequency generation (DFG). As for them, the most frequently used nonlinear crystal is periodically poled lithium niobite (PPLN), which can be pumped by commonly available laser sources, including Yb-doped fiber laser [12] and Nd: YAG solid state laser systems [13].

In fact, PPLN-based OPO has become a standard approach in efficient generation of watt-level average power in the MIR region, which can be in continuous wave (CW) mode [14,15] or in pulsed mode covering duration from tens of nanoseconds down to sub-100 fs [12,16–19]. Although such an approach is highly successful in conversion efficiency, it requires a resonant cavity and meanwhile the cavity-length related synchronous pumping for a stable pulsed operation. This can make the system complex and circumstance-sensitive. That is why such systems can usually be developed and used in laboratory but can rarely be applied in outdoor fields.

Comparatively, two-branch DFG/OPA systems have shown better robustness and seen some promising outdoor applications recently [1]. This is mainly benefitted from that such systems only include single-pass processes, not requiring resonant cavities. However, they are typically inefficient in parametric conversion; the achieved average power of idler light is typically in the milliwatt or even microwatt levels for the DFG [20–24]. Recent years, the power levels from such DFG sources have been enhanced to watt level but some sophisticated designs had to be taken due to the low damaging threshold of the general PPLN crystal. This can be done through optimized optics together with high power pumping [25] or through time synchronization if a pulsed mode is acceptable [26–28]. Although the OPA can reach much higher power levels, it is in fact not a process that directly enables MIR generation but typically requires some proper seeding source.

As an alternative to the mentioned sophisticated designs, P. Belden et al showed that the average power up to ∼1 W was possible when a CW diode laser was used as the signal source [29]. This is significant for the purposes of system-robustness and practical applications. Such a system is not only free from time synchronization, eliminating the related time jitter, but can also reduce the system complexity and cost. However, it is still unknown if such a system can enable higher power levels; presently the power level is even less than the reported all CW power [25].

In this paper, we demonstrated that as high as ∼3.82 W MIR idler power can be achieved from a CW/pulsed hybrid system. Three main points differ our work from [25]. The first is that we adopt a one-to-many scheme: the signal source has only one longitudinal mode; whereas, the pump source has many longitudinal modes as for the signal and pump branches. Such a scheme can alleviate the quasi-phase-matching (QPM) condition significantly. The second is that a high power CW master oscillator power-amplifier (MOPA) fiber system is used as the signal source, which provides more signal power in the parametric conversion process. The third is the use of a looser focusing configuration, enabling the crystal to tolerate more launched power. The last is that our system can deliver a much shorter idler pulse duration, i.e., ∼18.68 ps (∼2.5 ns in [25]).

2. Experimental configuration

The overall experimental configuration is shown in Fig. 1 schematically. It includes two all fiber MOPA branches, as the pump and signal sources, respectively, and a free-space part for the parametric conversion.

 

Fig. 1. Schematic experimental configuration.

Download Full Size | PDF

The signal branch employs a CW single-frequency distributed feedback laser diode (SF DFB-LD) with emission around 1550 nm as the seeding source. Considering the low emission power of the seeding source, the 1^st stage of amplification utilizes a single mode Er-doped fiber amplifier (EDFA). It comprises a piece of ∼2 m long polarization-maintaining Er-doped fiber (PM-EDF, EDF50-PM EC, OFS Fitel, LLC.) that has a peak absorption at 1530 nm of ∼48.44 dB/m, mode field diameter (MFD) at 1550 nm of ∼5.4 µm, and cladding diameter of ∼124.7 µm. The PM-EDF is core-pumped by a single-mode LD at 975 nm via a PM filter-type wavelength division multiplexer (PM-FWDM). Two stages of cladding-pumped EDFAs comprising the same type of PM Er-Yb co-doped double-clad fiber (PM-EYDF, PM-EYDF-12/130-HE, Nufern) are then followed to further boost the signal power. This PM-EYDF has a peak cladding absorption at 915 nm of ∼3.46 dB/m, peak core absorption near 1530 nm of ∼81 dB/m, core diameter of ∼12 µm, and cladding diameter of ∼130 µm, respectively. The used lengths of the PM-EYDF in the two EDFAs are ∼2 m and ∼2.3 m, successively. Correspondingly, they are pumped by one and two fiber-coupled multi-mode LDs near 910 nm with specified maximum output power of ∼12 W, respectively, via a (2 + 1) × 1 pump/signal tapered fiber bundle (TFB). It should be mentioned that different gain saturations of the two amplifiers have been taken into account [30].

Before feeding the pre-amplified signal light into the final stage of EDFA, a taper-splicing-based mode-field adapter (MFA) is used to smoothly transit the signal light between fibers with considerably different MFDs. The final EDFA employs a piece of ∼3.6 m long polarization-maintaining large-mode-area EYDF (PLMA-EYDF-25P/300 HE, Coherent, Inc.) as the gain fiber and five LDs near 910 nm with specified maximum output power of ∼30 W as the pump sources. The PLMA-EYDF has a peak cladding absorption at 915 nm of ∼5.9 dB/m, core numerical aperture (NA) of 0.09, core diameter of ∼26 µm, and clad diameter of ∼299 µm, respectively. To combine the pump and pre-amplified signal light together, a (6 + 1) × 1 pump/signal tapered fiber bundle (TFB) is used.

Besides the active fibers and pumping components, the signal MOPA includes a PM coupler after the single-mode EDFA and two tap couplers after the two cladding-pumped EDFAs for monitoring light in the forward and backward directions, as well as a PM circulator (PM-CIR) and three PM isolators (PM-ISO) for blocking any backward reflected light. Another capability with each PM-CIR or PM-ISO is that it can only let some particularly polarized light pass through and eliminate light with any other polarization directions. This is due to that each of them contains two polarizers. Thus, these devices can ensure that all the light can only propagate along the slow axis of each piece of the used PM fibers. In fact, the emission from the seeding SF DFB-LD is randomly polarized. The followed PM-ISO ensures that the light becomes linearly polarized and propagates along the slow axis of the subsequent fiber. It should also be noted that, after each cladding-pumped amplifier, a cladding power stripper (CPS) is used to dissipate the residual pump light, avoiding any deleterious effects to the following PM-ISO.

The pump branch closely resembles the signal branch in structure except the different seeding sources and gain fibers. As illustrated in Fig. 1, the seeding source is a mode-locked fiber laser that comprises a saturable absorber mirror (SAM, SAM-1064-18-5ps-0, BATOP GmbH) and a fiber Bragg grating (FBG) partially reflective near 1064 nm to constitute a linear cavity. The gain is enabled by a piece of single-mode PM Yb-doped fiber (PM-YDF, PM-YSF-HI-HP, Coherent, Inc.) with a length of ∼1 m, which is directly butt-coupled to the SAM by using a bare fiber terminator (BFT1, Thorlabs, Inc.). The PM-YDF has a peak absorption at 915 nm of 82.7 dB/m and an MFD at 1060 nm of ∼7.8 µm. The 975-nm pump light is also coupled into the gain fiber via a PM-FWDM, but prior to that a pump protector (PP) is included for preventing any backward ∼1064 nm light from damaging the 975-nm LD.

The following two stages of Yb-doped fiber amplifiers (YDFAs) comprise the same type of PM Yb-doped double clad fiber (PM-YDF, Coherent, Inc.) with lengths of ∼3 m and ∼3.2 m, respectively, to enable appropriate gains, and, correspondingly, one and two 910-nm LDs as the pump sources. A piece of ∼2.9 m long PLMA Yb-doped fiber (PLMA-YDF, Yb1200-30/250 DC-PM, nLIGHT Corp., Finland) is employed in the final YDFA, which is pumped by using four 30-W 910-nm LDs. The PLMA-YDF exhibits a cladding absorption at 920 nm of ∼3.4 dB/m, core NA of ∼0.064, core diameter of ∼30 µm, and cladding diameter of ∼250 µm, respectively.

Both the signal and pump MOPA branches terminate at quartz-rod endcaps (ECs) via fused splicing. Each EC is cleaved with 8°-angle facet and then antireflection (AR)-coated at a specified wavelength region, which is used for beam expansion while preserving minimal beam distortion through eliminating the backward Fresnel reflection. The two delivered beams are then collimated by using the same type of achromatic lens (AL) that has a focal length of 19 mm and diameter of 12.7 mm. To enable a nearly constant focal length, the using of AL is especially critical for either collimating or converging the pump beam considering the pump pulse’s broadband spectrum. A following free-space high power isolator (HP-ISO) is used to direct any backward reflected light out of each beam path. It should be noted that pre-rotations and alignments have been done to enable polarization matching between the output fiber’s slow axis and the input polarizing beam splitter (PBS) of the HP-ISO. The light polarization is further rotated by 45° through the HP-ISO. An achromatic half-wave plates (HWP) is eventually used to align the polarization to match the dipole moment of the followed nonlinear crystal. A highly reflective (HR) dielectric coating mirror (UM10-Y1 HP, Thorlabs, Inc.) is used for deflecting the pump beam and a longpass dichroic mirror (DM, DMLP1180, Thorlabs, Inc.) is used to spatially combine the pump and signal beams together.

The two polarization-aligned and spatially combined beams are finally focused into the nonlinear crystal through another AL that has a focal length of 100 mm and diameter of 25.4 mm. The crystal is a type of commercially available 5% MgO-doped PPLN (MgO: PPLN, MOPO3-1.0-50, Covesion Ltd., UK), which is designed for 1064 nm pumping. Geometrically, it has a thickness, width, and length of 1, 10, and 50 mm, respectively. To adapt for different QPM requirements, it comprises 7 different periodically poled gratings with periods of 28.5, 29.0, 29.5, 30.0, 30.5, 31.0, and 31.7 µm, respectively. For minimizing the input/output reflections, the crystal’s two end facets were AR-coated. As specified, the coating reflection is < 1.5% at 1064 nm (Pump), < 1% within the range of 1400 -1800nm (Signal), and ∼ 3%-2% within the range of 2900-4300 nm (Idler), respectively.

Figure 2 schematically shows the parametric conversion process. As seen in Fig. 2(a), the single spectral component from the SF signal system interacts with each of the multiple components across the emitted spectrum from the pump system. This can result in the generation of the corresponding idler spectral components if only the QPM is satisfied. This is also why we name such an interaction as the one-to-many scheme.

 

Fig. 2. Schematic diagram of the parametric conversion process represented (a) spectrally and (b) temporally.

Download Full Size | PDF

The temporal interaction characteristics, seen in Fig. 2(b), clearly indicate that only a small portion of the CW signal light can interact with the pump pulses and then parametric-converted into idler pulses. In fact, the photon number in the output idler is much greater than the photon number in the input signal pulse, i.e., the signal light during the timeslot corresponding to the pump pulse duration as enclosed by the dashed rectangles in Fig. 2(b). To attain such a case, the signal must have been further amplified in the MgO: PPLN crystal. Thus, besides the initial DFG stage, the parametric process also includes a further optical parametric amplification (OPA) stage. Both the DFG and OPA stage occur in the same piece of MgO: PPLN crystal. Thus, calling such a hybrid process as a single DFG or OPA is not proper, strictly to say. That is why here we use the term “parametric conversion,” mainly in in purpose of avoiding any possible confusion. Some related discussions can be found in [27].

3. Results and discussions

3.1 Seeding and pre-amplifying characteristics of the signal branch

All the spectral profiles from the pump and signal systems were measured by using the same optical spectrum analyzer (OSA, AQ6374, Yokogawa Test and Measurement Corp. Japan) and setting the wavelength resolution at 0.05 nm. Figure 3 plots the emitted spectral profiles from the seeding SF DFB-LD and the pre-amplifiers of the signal branch. It is noted that the peak emission spectrum of the SF DFB-LD is slightly dependent on the drive current when the operating temperature is fixed at ∼25 °C. Through this paper, the applied drive current is fixed at ∼50 mA, enabling ∼8.46 mW output power. Figure 3(a) shows then the overall spectral profile, on which the modulating structure indicates that the LD operates in a sing longitudinal mode, i.e., SF. Such a modulating structure results from the weak interference of the LD’s amplified spontaneous emission (ASE) when it propagates forth and back within the ultrashort cavity of the LD. The inset of Fig. 3(a) plots the details of the spectrum around the peak wavelength of ∼1548.64 nm, exhibiting a 3-dB bandwidth of ∼0.06 nm. As noted, neighboring to such a main spectral peak, there can be seen a secondary much lowered peak, which is a different longitudinal mode. However, considering that the peak intensity of this mode is more than 40 dB lower than that of the main mode, its induced effect can be neglected reasonably from a practical point of view.

 

Fig. 3. Spectral characteristics from (a) the seeding source [Inset: spectrum around the peak wavelength] and (b) pre-amplifiers of the signal MOPA.

Download Full Size | PDF

The emission spectra from the followed three pre-amplifiers are plotted in Fig. 3(b), which were recorded when the launched pump powers for the 1^st through 3^rd amplifiers were ∼200 mW (975 nm), ∼1.87 W (910 nm), and ∼5.70 W (910 nm), respectively. As seen, there is no notable change with the spectral peak parts, but the in-band ASE becomes more and more stronger from the 1^st through 3^rd amplifiers. Still, the signal to noise ratio (SNR) between the emission peak and ASE can reach ∼49.22 dB even at the 3^rd pre-amplifier. As also noted, the 3-dB bandwidth (∼0.04 nm) of the spectrum from the 3^rd pre-amplifier becomes even narrower than that from the seeding source (∼0.06 nm), probably due to the gain-profile-related filtering effect. The 1^st core-pumped amplifier can boost the seeding power to ∼35.5 mW, which is further improved to ∼0.81 W through the followed two cladding-pumped pre-amplifiers.

3.2 Seeding and pre-amplifying characteristics of the pump branch

Figure 4 shows the output characteristics of the passively mode-locked seeding source of the Yb MOPA. Stable mode-locking self-starts when the pump power increases to ∼180 mW. A typical output spectrum at ∼180-mW pumping is plotted in Fig. 4(a), giving an SNR between the peak and ASE power levels of ∼ 55.33 dB. The inset of Fig. 4(a) plots the detailed spectral profile around the peak wavelength. As seen, the peak wavelength is ∼1063.39 nm with a 3-dB bandwidth of ∼0.24 nm. Figure 3(b) plots the corresponding autocorrelation trace (ACT) by using a commercial autocorrelator (PulseCheck, APE, Berlin, Germany). The ACT has a full width at half maximum (FWHM) of ∼11.81 ps, resulting in a pulse duration of ∼8.35 ps with Gaussian-fitting. Figure 4(c) shows a captured radio frequency (RF) trace ranging from 0 to 10 GHz, measured by using an RF signal analyzer (MXA Signal Analyzer N9020 B, Keysight) with a resolution bandwidth (RBW) setting at 100 kHz. The inset of Fig. 4(c) shows the details of the RF spectrum around the fundamental repetition rate of ∼68.336 MHz, which gives an SNR of ∼96 dB with the RBW setting at 1 Hz. The measured average power is ∼6 mW. The single pulse energy can be calculated as ∼87.8 pJ.

 

Fig. 4. Spectral [(a) and the inset], temporal [(b)], and RF [(c) and the inset] characteristics of the passively mode-locked seeding YDF laser.

Download Full Size | PDF

Figure 5 plots the spectra emitted from the 1^st (red) and 2^nd (blue) stage of YDFA, respectively. As seen, the ASE is negligible with the 1^st YDFA, leading to an SNR as high as ∼51.00 dB. However, it increases considerably through the 2^nd YDFA, of which the SNR decreases to ∼48.42 dB. The inset of Fig. 5 further plots the profile details around the spectral peaks. In comparison to the seeding source, the spectrum only experiences a slight broadening through the 1^st YDFA, i.e., from ∼0.24 nm to ∼0.28 nm in 3-dB bandwidth. After the 2^nd YDFA, however, the spectral broadening becomes much more significant; the measured 3-dB bandwidth reaches ∼1.09 nm. Meanwhile, a type of oscillatory structure can be observed, which is due to self-phase modulation (SPM) effect [31]. According to the peak number M, the induced maximum nonlinear phase shifts ${\phi _{max}}$ through the 1^st and 2^nd YDFA are $1.5\; \pi $ and $4.5\; \pi $ rad, respectively, according to the relation [31]

 

Fig. 5. Spectral characteristics of the 1^st (red) and 2^nd (blue) pre-amplifiers. Inset: Profile details at the spectral peak parts.

Download Full Size | PDF

It should be noted that there already is a certain amount of SPM-induced spectral broadening with the spectrum from the seeding source. In fact, we indeed observed the abrupt spectral broadening when the YDF laser was switched from CW to mode locking as the pump power increasing. It should be the dissipative regime enabled by the FBG’s filtering effect that stabilizes the mode-locking state [32]. The structure-less spectral profile is only in that the induced ${\phi _{max}} < \pi $.

3.3 Final amplifiers and the enabled MIR idler generation

Figure 6 depicts the output characteristics of the final amplifiers in both the signal and pump branches. The used signal power for parametric conversion is ∼11.45 W from the final stage of EDFA. The corresponding output spectrum is plotted in Fig. 6(a). As seen, although the ASE becomes significant and can account for up to ∼20.08 dB relative to the background floor across the gain bandwidth, the peak to ASE ratio still reaches ∼34.51 dB. This is acceptable for usual applications. The measured 3-dB bandwidth is ∼0.04 nm, which can be seen as approximately equal to that of the preceding pre-amplifier considering the limited resolution of the used OSA (setting at 0.05 nm) and other possible fluctuations in measurement. Thus, it can be deduced that the signal light is still confined to the single longitudinal mode but has much higher power through amplification. It can also be noticed the secondary longitudinal mode in Fig. 3 disappears into the spectral pedestal now. As expected, it should play no role in the idler generation.

 

Fig. 6. Emission characteristics of the final amplifiers: Spectral profiles of the (a) signal and (b) pump light (Inset: the corresponding ACT with Gaussian fitting); Beam intensity profiles along horizontal (X) and vertical (Y) directions measured at the foci: (c) signal beam; (d) pump beam.

Download Full Size | PDF

Figure 6(b) plots the output spectrum delivered from the final stage of YDFA when the output average power is ∼41.73 W. It can be seen that the spectral profile becomes much more complex than that from the preceding pre-amplifiers. Although it becomes difficult to determine the 3-dB bandwidth due to appeared spike-like peaks on the profile, the central part of spectrum broadens considerably due to the SPM effect increasingly accumulated through the final YDFA.

The inset of Fig. 6(b) plots the measured ACT of the pulses delivered from the final YDFA. The Gaussian fitting gives a pulse width of ∼19.12 ps. However, it can also be noted that the ACT shape significantly deviates from the standa Gaussian profile, showing some weak pedestal. This indicates that the pulses have experienced some distortions through the three stages of YDFA. The probable mechanisms relating to these distortions should include nonlinear effect, temporal noise accumulation and amplification, tilted gain at the off-peak spectral region, etc. Somehow, calculated by using the roughly Gaussian-fitted pulse width, the finally amplified peak power of the pump pulses is ∼31.9 kW. Another feature with the pulses delivered from the final YDFA is that their pulse width reaches ∼2.3 times that of the seeding pulses. Two main aspects might be responsible for so considerable pulse broadening. One is that all the active fibers and the matched passive fibers used in the pre- and final YDFAs have large normal dispersions. The second is that the SPM effect aroused during nonlinear amplification in the YDF and the further propagation in the following passive fiber can also induce continuously increased positive chirp.

Beam sizes of the pump and signal light within the crystal cannot be measured directly, but some measurements at their waists in air can be performed. These were done by slowly moving a scanning-slit optical beam profiler (BP209-IR2/M, Thorlabs, Inc.) around the waist of a beam and finding the minimum size. For each beam, the transverse intensity distributions in both horizontal (X) and vertical (Y) directions measured the its waist were plotted, as seen in Figs. 6(c) and 6(d). Evaluated at a clip level of $1/{e^2}$ (∼13.53%), the X and Y waist widths, noting as ${w_X}$ and ${w_Y}$, are ∼152.1 and ∼147.7 µm for the signal beam and ∼116.4 and ∼117.6 µm for the signal beam, respectively. As seen, both the signal and pump beams have very tiny asymmetries; they are ∼1.5% and 0.5%, respectively, if calculated by using the general formula

In average, the waist diameters are roughly 150 and 117 µm for the pump and signal beam, respectively, and the corresponding Rayleigh lengths are ∼10.1 and ∼11.4 mm. Both the sizes of the beam waists are expected to encounter no noticeable changes when the MgO: PPLN crystal is placed. This is due to the fact that both the pump and signal beam’s Rayleigh lengths are more than twice the length of the used crystal. Thus, their beam convergences should be very weak when incident onto the input facet of the crystal. Comparing to the free-space optics used in [25–27] and [29], our focusing is much looser, which can potentially improve our MgO: PPLN crystal’s power tolerance.

It should also be noted that the absolute coordinates of X and Y in Fig. 6 do not mean the overlapping characteristics of the pump and signal beams. We performed their measurements separately. Thus, such absolute coordinates meant nothing about the two beam waists’ relative transversal positions. In fact, before the two beams were focused into the MgO: PPLN crystal, fine alignments had been done to ensure their transversal overlapping at both nearby and far positions from the deflection mirror. After focusing, the two beams could well overlap transversally. The longitudinal positions of the pump and signal waists were also both near the middle of the crystal but exhibited <1 mm apart (the pump waist is a little in front of the signal waist). Based on the present configuration, the longitudinal overlapping of the two beams’ waists cannot be realized. Further optimizations and investigations might be some of our future considerations.

Considering the specified QPM conditions of the as-used MgO: PPLN crystal and the achieved peak wavelengths in Figs. 6(a) and 6(b), we chose the poling period of ∼30 µm and maintained the operation temperature at ∼162.10 °C in purpose to realize an idler wavelength near ∼3.4 µm. After some fine alignments on the free space optics, the finally realized idler power reached up to ∼3.82 W, when the output spectrum peaked at ∼3400.23 nm and spanned from ∼3370.61 nm to ∼3413.88 nm, as plotted in Fig. 7(c). The idler spectrum was measured by using a Fourier transform OSA (OSA207C, Thorlabs, Inc.).

 

Fig. 7. MIR idler light generated from the MgO: PPLN crystal: (a) Spectral profile and (b) ACT with Gaussian fitting of the idler light; zoomed comparison between the (c) idler and (d) pump spectral profiles; (e) parametric conversion setup in operation.

Download Full Size | PDF

Figure 7(b) plots the ACT of the idler pulse. The corresponding pulse width based on Gaussian-fitting is ∼18.68 ps. The calculated pulse energy and peak power are ∼55.9 nJ and ∼2.99 kW, respectively. Comparing to the launched pump pulse, the idler pulse is slightly shorter in duration, and can be better fitted to the standard Gaussian profile. These should be due to that the parametric conversion efficiency depends directly on the pump power’s temporal distribution. The central part of the pump pulse has higher peak power than its leading and trailing edges. Thus, only the central part can be efficiently parametric converted into idler light, whereas the conversion efficiency would drop increasingly toward the edges. The noise-like pedestal almost disappears with the ACT of idler pulse. This is again due to that the power level of the pump pulse’s pedestal is much lowered than that of its central part, leading to no efficient generation of idler light.

For more clearly comparing the idler and pump spectral profiles, we further plotted their zoomed figures, as seen in Figs. 7(c) and 7(d). It can be seen that the spectral profile of the idler light roughly resembles the peak part of the pump light’s spectral profile. Although the pump spectrum has significant pedestal, it is expected to have too low peak power for parametric conversion. That is why the idler spectrum only resembles to the peak part of the pump spectrum. Another factor contributing to such a similarity is that the signal light from the final EDFA is still mainly confined to a single longitudinal mode, as seen in Fig. 6(a). When the parametric conversion occurs, it is much like that the single longitudinal mode of the signal light sweeps across the multiple longitudinal modes of the pump light. Thus, the produced idler light should be able to preserve the main features of the pump light, such as the coherence, the profile, etc.

Figure 5(e) shows a captured photograph when the parametric conversion system was in operation. The green light and red light result from the second harmonic generation (SHG) of the pump light at ∼1063.3926 nm and the signal light at ∼1548.6389 nm, respectively. However, both the SHG intensities were quite low in fact, since their QPM conditions were not satisfied.

Figure 7 plots evolution characteristics of the measured idler power against the incident signal and pump power, respectively. The idler power in Fig. 7(a) was measured by varying the incident signal power but fixing the pump power at ∼23.1 W. It can be seen that the overall evolution shows a roll-over effect and there exists a best signal power value where the converted idler power reaches the highest. According to our measurement, the best signal power should be close to ∼11.45 W. The pump-dependent evolution of the produced idler power is plotted in Fig. 7(b) by fixing the signal power at such a close-to-best value. As seen, the initial evolution is roughly linear, but starting from ∼22.37 W the idler power increases more and more slowly. Correspondingly, it is also at such a pump power level that the highest conversion efficiency of ∼11.89 mW/W² can be obtained.

The highest measured idler power was ∼3.82 W, which was achieved by increasing the pump power to ∼41.73 W and fixing the signal power at ∼11.45 W. In fact, if only in the purpose of achieving the highest conversion efficiency, we do not have to launch the pump and signal power at so high levels. However, in this work since we mainly intended to obtain the MIR idler power as high as possible, we thus tried to launch the pump power as high as possible and the signal power at the aforementioned highest point, i.e., ∼11.45 W. It should also be mentioned that the ∼3.82 W idler power was the highest power level that we could operate the system stably and safely. In fact, we had observed that the idler power became unstable considerably and then the PPLN crystal was damaged when the output idler power was over ∼4 W. After that, the idler power slowly dropped to only several tens of milliwatts and could no longer rise again no matter how launched power was changed. These results indicate that we probably have achieved what quite close to the highest possible power level based on our current configuration.

The turning point shown in Fig. 8(b), as well the followed slowing down of the increasing rate, should mainly attribute to the change of the spectral profile of the final YDFA with the corresponding output power. Figure 9(a) plots a spectral profile when the output power is ∼19.37 W, and the inset shows its details around the peak wavelength. Such a power level is just below the turning point. Comparing the Figs. 9(a), 6(b), and7(d), it can be noted that the spectral profile changes mainly include spectral broadening, rising of short-wavelength components, and more significant spectral pedestal, when the tuning occurs. These changes can make the operation deviate from the best QPM condition and thus lower the parametric conversion efficiency. In fact, the output power from the final YDFA can reach up to ∼52.30 W. At such a highest power level, the corresponding spectral profile becomes even more complex and some Raman-like components start to rise, as seen in Fig. 9(b). It was at such a power level that our as-used MgO: PPLN crystal became damaged.

 

Fig. 8. Emission idler power against the incident (a) signal and (b) pump power.

Download Full Size | PDF

 

Fig. 9. Emitted spectral profiles from the final stage of YDFA when the output average powers are (a) ∼19.37 W (Inset: the details of the peak part) and (b) ∼52.30 W, respectively.

Download Full Size | PDF

Another fact is that we did not observe significant change of the turning point when the launched signal power varied. This can be seen as another evidence that the turning point in Fig. 8(b) mainly relates to the features of the launched pump pulse.

However, it should be noted that the observed deviation from best QPM condition due to the pump spectral profile variation does not mean the system’s high wavelength sensitivity. The pump power change results in too considerable overall spectral profile variation. Considering that the conversion efficiency might be not so uniform across the spectral bandwidth, the overall accumulative conversion efficiency is possible to see some changes. But, if we can maintain the pump power constant, equivalently the pump spectral profile unchanged, and meanwhile vary the signal wavelength slightly, there should be little change on the conversion efficiency and generated idler power, considering the as-used wide spectral bandwidth from the final YDFA.

In fact, neither our pump nor signal system exhibits wavelength tunability while maintaining a constant output power. Even that, the wavelength sensitivity in QPM can be equivalently checked by using the temperature acceptance bandwidth. Through finely varying the applied temperature to the oven enclosing the MgO: PPLN crystal, it was found that no notable idler power fluctuation could be observed when the temperature varied within several degrees Celsius. This is remarkable in comparison to a system that both the pump and signal sources have narrow linewidths. Such a case can be found in one of our previous works where significant idler power variations could be induced even with only 0.1 °C temperature change [24]. Such a wavelength insensitivity benefited from the as-used one-to-many scheme. The feasibility of such a scheme relies on the crystal’s wide pump acceptance bandwidth [33,34]. It can be expected that there should be no change on the conversion efficiency if the pump system can be further optimized, which include making the spectral profile uniform across the crystal acceptance bandwidth, eliminating the spectral components not belong to the pump pulse, etc.

We also noted that, for any focusing conditions, all above features presented similarly. However, the focusing condition could affect the overall conversion efficiency and the finally generated idler power. This satisfied some general rules as expected, such as that the tighter the focusing, the higher the conversion efficiency, but the lower the tolerated power of the crystal. Relating to Fig. 8, it was noted that, the tighter the focusing, the lower the highest point of the signal power and also the lower the tuning point of the pump power. A looser focusing is expected to be helpful in further improving the generated idler power, which will be one of our future considerations.

4. Conclusions

In conclusion, we have demonstrated a parametric conversion source based a simplified one-to-many scheme, which can generate as high as ∼3.82 W average power at an idler wavelength of 3.4 µm, representing the highest idler power ever achieved with CW signal feeding. The proposed scheme can not only simplify the system design significantly, but also potentially improve the system’s robustness and circumstance-resistance through leaving out the complex time synchronization and alleviating the wavelength sensitivity in QPM, respectively. Hence, it can be a practical option for parametric conversion source capable of watts of MIR power generation.