1. Introduction

The mid-infrared spectral region of 6 – 9 μm contains multiple molecular fingerprints of various chemical compounds, such as greenhouse gases, air pollutants, explosives [1] or disease markers in exhaled human breath [2]. Nowadays, spectroscopic systems and trace-gas detection schemes based on optical frequency comb (OFC) technology provide high speed, excellent accuracy and quantum-noise limited sensitivity [3,4]. There are two main approaches to generate broadband, widely tunable and powerful comb radiation in the MIR, namely: optical parametric oscillators (OPOs) and DFG. In general, OPOs offer high output power with broad spectral coverage [5,6], but they require a free-space resonator, which free spectral range (FSR) needs to be locked to the FSR of the pumping laser. This makes OPO sources bulky, vulnerable to any external disturbances and very impractical in field applications. On the other hand, the DFG approach is much more flexible – it does not require a cavity, therefore it can be implemented almost fully using optical fibers, which ensures alignment-free operation and reduced complexity [7]. The DFG process can be also quite efficient – recently, DFGs with optical to optical efficiency at the level of 16% were reported [8,9]. Additionally, a difference frequency signal generated between two trains of laser pulses sharing the same carrier-envelope phase will have a passively cancelled carrier-envelope offset frequency (f_CEO) [10], which might be advantageous in some particular applications in cavity-enhanced spectroscopic techniques [11].

The MIR spectral band spanning from approximately 6 to 10 μm can be obtained in the DFG process via mixing of 1.55 μm with 1.8 – 2 μm radiation, both achievable from fiber-based sources (based on Erbium- and Thulium-doped gain media). In order to avoid the necessity of synchronizing two separate pulsed laser sources, most of the MIR DFG systems are seeded with one master source [12–20]. The signal for the DFG process is then provided via nonlinear conversion of the 1.55 μm pump to longer wavelengths in highly nonlinear fibers (HNLFs), e.g. by Raman-induced soliton self-frequency shift (SSFS). As a result, a MIR idler in the range from 6 to even 18 μm can be achieved, depending on exact input wavelengths and nonlinear crystal parameters [12]. For this purpose, several crystals might be used, e.g. silver selenogallate (AgGaSe₂) [12,13], silver thiogallate (AgGaS₂) [14], gallium selenide (GaSe) [15–17], cadmium silicon phosphide (CdSiP₂) [18], orientation-patterned gallium arsenide (OP-GaAs) [19], or gallium phosphide (OP-GaP) [20]. Recently, OP-GaAs crystals have gained much attention as efficient nonlinear media for OPOs and DFGs [19,21]. However, O. Heckl et al. have revealed, that three-photon absorption in OP-GaAs strongly limits the output power achievable from OPO sources [22]. This can be avoided by using OP-GaP crystal [6,23]. Hence, the crystal can be also used in the DFG setups pumped at 1 µm [24,25]. To date, there was only one literature report on OP-GaP-based DFG – K. Lee et al. demonstrated generation of a widely tunable idler in the 6 – 11 μm range with output power up to 60 mW [20]. However, the setup contained a free-space delay line and both beams interacting in the DFG were combined using free-space optics. In the literature, several long-wavelength DFGs (ranging up to 18 μm) have been reported with the use of different crystals other than OP-GaP [12–19]. However, all of the reported setups utilized free-space optics for spatial and temporal overlapping of the pump and signal pulses. This may imply the necessity of day-to-day alignment of the free-space optics part for achieving satisfactory performance.

Here, we report a DFG source based on an OP-GaP crystal with exceptional simplicity, in which the pumping laser system is entirely fiberized (including the delay line and beam combining) and built only of single-mode, polarization maintaining (PM) fibers and components. In our design, both pump and signal waves are combined into one common single-mode fiber and focused in the crystal by an achromatic optical system. This makes the DFG system monolithic, robust and alignment-free. As a result of the DFG process, we have obtained a MIR idler with average power of 7.4 mW at 7500 nm central wavelength, with >400 nm of full-width at half-maximum (FWHM) bandwidth, at 48 MHz repetition frequency. By tuning of the signal wavelength and crystal QPM period, other idler wavelengths from the range of 6000 to 8800 nm are easily achievable. Moreover we have experimentally confirmed the high coherence of the signal pulses, which makes the source valuable for cavity-assisted spectroscopic techniques. Applicability of the system to the absorption spectroscopy was proved by the measurement of the methane absorption spectrum in the 7.5 – 8.3 µm spectral range.

2. Experimental setup and results

The experimental setup of the MIR DFG source is shown in Fig. 1. The system is seeded by an Er-doped fiber oscillator (OSC) mode-locked with a graphene saturable absorber, similar to that presented previously [26]. The pulses from the seed are evenly split into two branches. One part is amplified in an Erbium-doped fiber amplifier (EDFA-1) to approx. 120 mW and coupled into a PM highly nonlinear fiber (HNLF). The HNLF (developed in the Laboratory of Optical Fiber Technology, Maria Curie-Sklodowska University, Lublin, Poland) shifts the pump spectrum towards longer wavelengths via Raman-induced soliton self-frequency shift (SSFS) [27]. Afterwards, the converted pulses are amplified in an Tm-doped fiber amplifier (TDFA) to approx. 150 – 235 mW (depending on the wavelength of the input signal), forming a signal for the DFG process. The second part of the seed is also amplified (EDFA-2) to approx. 175 mW of average power. Afterwards, both arms are combined with a 1550/2000 nm wavelength division multiplexer (WDM) into a common single-mode fiber. The temporal overlap between the 1.56 μm pump and the wavelength-shifted signal pulses is ensured by a manually controlled fiber-coupled variable delay line (VDL) with 500 ps optical delay range. The pump and signal beams are co-focused on the nonlinear crystal by means of a connectorized, achromatic, reflective collimator integrated with an air-spaced achromatic doublet with a focal length of 75 mm (COL). Such a configuration minimizes the focal length shift between the pump and signal beams in the nonlinear crystal plane to ~70 µm, while the diameters of both beams at the focal plane are close to 25 µm. As nonlinear medium for the DFG process, a 3 mm thick OP-GaP crystals are used with six QPM periods: 51, 55, 58, 60, 61.1, and 62 μm. The crystals with 1x5 mm² clear aperture are anti-reflection (AR) coated for 1.5 – 2.1 μm (input) and 6 – 10 μm (output face). The idler beam is collimated with a 75 mm barium fluoride (BaF₂) lens. It is separated from the unabsorbed pump and signal light using an commercially available long-pass filter (F) with a 4.5 µm cut-on wavelength.

 

Fig. 1 Experimental setup of the DFG source. OSC: graphene-based mode-locked oscillator, EDFA: Er-doped fiber amplifier, TDFA: Tm-doped fiber amplifier, HNLF: highly nonlinear fiber, DCF: dispersion compensating fiber, VDL: variable delay line, WDM: wavelength division multiplexer, COL: collimator, F: long-pass filter.

Download Full Size | PDF

The performance of the pump arm is summarized in Fig. 2. All parameters were measured directly in front of the nonlinear crystal. The optical spectrum is centered at around 1570 nm and has a FWHM of 87.9 nm. The pulse duration was 76 fs assuming a sech² shape. The pump pulses were amplified up to 175 mW of average power, which corresponds to a pulse energy of ~3.6 nJ.

 

Fig. 2 Optical spectrum (a) and autocorrelation (b) of the 1570 nm pump pulse.

Download Full Size | PDF

The signal pulses generated in the HNLF could be easily shifted even up to 2 μm. Wavelength tuning could be obtained just by changing the pump power of the EDFA-1. In order to check whether the shifted solitons can be useful for measurement applications, their coherence was investigated in an unequal-path Michelson interferometer [28]. The measured spectral interference signals of shifted solitons are depicted in Fig. 3 (left panel). The visibility of the modulation on this signal directly corresponds to the degree of coherence [28,29]. Hence, it was calculated as defined V(λ) = [I_max(λ)-I_min(λ)]/[I_max(λ) + I_min(λ)], where I_max and I_min are the maximum and minimum intensities in the interferogram, respectively [29]. Here, the shifted solitons are highly coherent as evidences by their visibility of ~0.85−0.9.

 

Fig. 3 Interference patterns of consecutive soliton pulses including: the calculated fringe visibility function (red line), the upper (blue line) and lower (orange line) envelopes of the interferograms, used for obtaining I_max and I_min, respectively (left panel). Optical spectra (middle panel) and autocorrelation traces (right panel) of the amplified solitons used in the DFG experiment.

Download Full Size | PDF

The optical spectra and autocorrelation traces of the amplified shifted solitons are depicted in Fig. 3. Because the amplifier operates in nonlinear regime the output spectra are significantly broadened and distorted. Depending on the soliton central wavelength, the average output power varies between 152 mW (for 1965 nm) and 235 mW (for 1953 nm). The differences are caused by the need to set the right pump power level to obtain the best pulse quality at the fixed length of the fiber between the TDFA and the COL. The uneven gain curve of the used Tm-doped fiber in broad spectral range is another factor influencing the average output power. The duration of the signal pulse varies between 105 and 65 fs.

It is worth noting that the length of the optical fibers in both branches was precisely optimized in order to achieve the shortest output pulses at both wavelengths. In order to optimize the pulse duration and compensate the anomalous dispersion of single-mode fibers in both branches, additional dispersion compensating fibers (DCF1 and DCF2) were used prior to the amplification stages. The DCF1 used for the 1.55 μm pump was a 1 m-long segment of commercially-available fiber (OFS PM-DCF-1550) with group velocity dispersion (GVD) of 0.127 ps²/m at 1560 nm wavelength. In the signal branch (DCF2) we have used a self-made PM microstructured fiber with 0.075 ps²/m GVD at 1930 nm.

Figure 4 shows the generated MIR idler spectra for different signal wavelengths (tuned from 1953 to 1965 nm), obtained with six different crystal QPM periods, measured with a Fourier-transform spectrometer (Thermo Scientific Nicolet iS50, FTS). It can be seen, that the presented DFG source covers almost the entire band from 6000 to 9000 nm even without the necessity of signal tuning, only by changing the crystal period. The achievable idler tuning range is currently limited by the QPM periods in the available crystals. The FWHM of the tunable idler varies from 460 to 900 nm depending on the chosen period and signal wavelength. In our experimental setup, the distance between the OP-GaP crystals and the FTS was approx. 2.2 m, therefore the spectra measured in the 6-7.5 µm are strongly affected by water absorption lines.

 

Fig. 4 Normalized MIR spectra obtained with different QPM periods and different signal wavelengths: 1953, 1956, 1961 and 1965 nm.

Download Full Size | PDF

The maximum achieved output power of the DFG was 7.4 mW, recorded at 7500 nm central wavelength of the idler, crystal temperature of 30°C, 1953 nm centered soliton and 60 μm QPM period. This power level corresponds to approx. 7% photon conversion efficiency taking into account the pump and signal photons. The average idler power higher than 1 mW is maintained over the entire tuning range at any QPM period what is depicted in Fig. 5(a).

 

Fig. 5 Average idler output power vs. QPM period measured at T = 30°C (a), average output power of a free-running system measured over a 60 minutes period (b)

Download Full Size | PDF

It should be emphasized that despite the lack of any active control of the pump and signal pulses overlap (we do not use any real time control of the delay line), the fluctuations of the output power over time are at a low level. During a 60 minute-long test in a laboratory environment, the systematic output power drop form 6.5 mW to 6.2 mW was observed and is presented in Fig. 5(b). The output power of the system can be fully stabilized by adding an active control loop, which will be the subject of our future works.

It is worth noting that the crystal temperature is not a critical parameter for the DFG output power which is directly related to the broadband character of the pump and seed optical spectrum. We have investigated the performance of the system (set at ~8000 nm central wavelength) as a function of the crystal temperature. The output power drops by approx. 19% (from 6.2 to 5.1 mW) while heating the crystal from 30 to 70°C. The temperature dependence change of the phase-matching period results in the red-shift of the idler spectrum as can be seen in Fig. 6(a). Additionally, the shape and the FWHM of the idler spectrum evolves to the optical spectrum measured for the crystal with 62 µm phase-matching period, which is presented in Fig. 4 (upper panel). Nevertheless, those results show that keeping the crystal at nearly room temperature is beneficial and provides the best DFG performance. Therefore, holding the crystal in a temperature-stabilized oven is unnecessary, which is an important feature from the viewpoint of potential field applications.

 

Fig. 6 Optical spectra of the idler recorded for different crystal temperatures from 30 to 70°C (a). The RF spectrum of the idler signal (b).

Download Full Size | PDF

The generated idler was also detected with a 1 GHz MIR photodetector (VIGO Systems) coupled with a radio frequency (RF) spectrum analyzer. The generated pulses were repeated with a frequency of ~48 MHz with signal to noise (S/N) ratio better than 60 dB (measured with filter resolution bandwidth of 80 Hz), as shown in Fig. 7. Broad and flat spectrum of harmonics presented in the inset of Fig. 6 is free of any parasitic modulations which confirms stable pulse operation of the entire system.

 

Fig. 7 Absorption spectrum in transmission of 0.23% of methane in a cell at atmospheric pressure and 0.75% of atmospheric water in the free space path, compared with the model spectra (methane – red, water – blue; inverted for clarity).

Download Full Size | PDF

Finally, the DFG source is suitable for absorption spectroscopy. It was successfully used in combination with a commercial FTS to measure the absorption spectrum of the ν₄ band of methane. The sample was contained in a 10 cm-long cell and diluted in N₂ at atmospheric pressure. Figure 7 shows the transmission spectrum (black) plotted together with the model spectra, calculated using the line parameters from the HITRAN database [30], of 0.23% of CH₄ (red) and 0.75% of H₂O (blue, both inverted for clarity). The water is atmospheric, present in the free-space path of 2.2 m separating the DFG output and the FTIR. The agreement between the measurement and simulation is good, and the relative noise on the baseline reaches 3 × 10⁻³ for 12 minutes averaging time.

Summarizing, we have demonstrated an entirely fiberized DFG source delivering a milliwatt-level idler tunable in the range of 6000 – 9000 nm at 48 MHz repetition rate. The system is designed using only polarization maintaining fibers and components, ensuring robustness and invulnerability to external disturbances. It is seeded by a self-starting master oscillator operating at 1.56 μm wavelength, based on a graphene saturable absorber. The maximum achieved power of the idler was 7.4 mW at 7500 nm central wavelength Broad spectral coverage and smooth envelope makes the presented DFG an excellent source for laser spectroscopy in the MIR. We have confirmed its applicability in this field by measurement of methane absorption spectrum in the 7.5 – 8.3 µm spectral range. What is an important intrinsic feature of DFG systems, the carrier-envelope offset in the generated MIR comb is passively canceled. Therefore, the absolute stabilization of the generated MIR comb-like spectrum requires only stabilization of the repetition frequency and linking the seed laser to a reference in the 1550 nm band. We believe that the unprecedented simplicity and compactness paves the way for applications of such DFG sources in field-deployable optical frequency comb spectroscopy systems.