*4.2. Sending Unit*

Figure 4 shows a more detailed sketch of the optical layout inside the sending unit. The cw-QCL (Alpes Laser, 40 mW) with emission wavelength around 4.85 μm is housed together with a collimator in a sealed laser mount. The laser is fixed on a home-made watercooled heat sink. A second cw-QCL (Hamamatsu L12004-2190H-C, 20 mW, Herrsching, Germany) with an emission wavelength centered around 4.56 μm is packaged in a sealed laser mount on a water-cooled heat sink (Hamamatsu A11709-02). An aspheric ZnSe lens (Hamamatsu, A11331-02) inside a precision zoom unit (Thorlabs, SM1ZM) can also be translated perpendicular to the laser beam path (*z* direction) with an *xy* translation mount (Thorlabs, ST1XY-D). The QCLs are thermoelectrically cooled with Peltier elements. To attain stable performance, the heat sinks are connected in series and water cooled with a chiller (Thermo Fisher Scientific, 200 W, Waltham, MA, USA). Two free-space adjustable narrowband optical isolators (Thorlabs, I4500W4 and I4730W5) are used to shield the lasers from back reflections, thus reducing intensity noise and mode hopping. The free beams from both lasers are collinearly combined on a narrowband filter (C1, Spectrogon, NB5040-155 nm, Täby, Sweden) that is used as a long-pass beam combiner transmitting the 4.85 μm beam and reflecting the 4.56 μm beam. In comparison with a 50:50 beam splitter, this filter achieves better overall optical efficiency of 75% for each beam. However, the filter may cause ghosting and beam distortion due to etaloning and limited surface optical quality, respectively.

**Figure 4.** Optical layout of the sending unit.

For the water transitions, two polarization-maintaining (PM) fiber-coupled DFB (distributed-feedback) lasers with emission wavelengths of 1.392 and 1.469 μm (NEL, 20 mW) are clamped into butterfly laser mounts (Arroyo, 203) and thermoelectrically cooled with Peltier elements. For a compact arrangement, they are stacked onto a platform mounted above other optical elements. To prevent damage of long sections of optical fibers, they are coiled into storage reels (Thorlabs, FSR1) separately fixed on the cover. With two 1 × 2 PM fiberoptic couplers (Thorlabs, PN1310R2A1 and PN1480R2A1), the two beams at 1.39 and 1.47 μm are separately split into two PM fractions of 90% and 10% each. To monitor the mode quality and to obtain frequency markers to convert scan time to relative wavelength, the 10% beams are combined using a single-mode (SM) fiber wavelength-division multiplexer (Thorlabs, WD202C-APC, WP9850A) and together propagate through a SM fiber-coupled interferometer (Micron Optics, FFP-I, 0.8 GHz). Its output is collimated and free-space recorded by an amplified InGaAs photodetector (Thorlabs, 0.8 mm2 active area, max. 11 MHz bandwidth). In front of the detector, two narrow bandpass filters (Thorlabs, FB1400-12 as N1, FB1480-12 as N2 in Figure 4) are placed in a dual-position filter holder slider with resonant piezoelectric motors (Thorlabs, ELL6K) such that either 1.39 or 1.47 μm can be selected for etalon analysis to determine the wavelength tuning characteristics. The motor of the filter slider is remotely controlled during the measurement.

The 90% beams are combined using another PM wavelength-division multiplexer (Thorlabs, WP9850A). The coupled beams with the two NIR wavelengths are collimated with an aspherical lens (Thorlabs, C280TMD-C) in a fiber-launch system (Thorlabs, KT110/M). They are subsequently combined with the two collinear propagating MIR beams by a second bandpass filter (Spectrogon, BP-4700-600 nm) operated as combiner C2, with most of the MIR beams being transmitted and the NIR beams reflected off the front surface. For wavelength and tuning characterization, a fraction of the incoming free-space MIR beams is reflected towards a solid germanium etalon and a reference absorption cell, respectively. The detectors record each laser beam through respective narrow bandpass filters (Laser Components, SNB-4860-001793 as M1 and Spectrogon, NB4560-135 nm as M2 in Figure 4) installed in another dual-position filter slider equipped with a resonant piezoelectric motor (Thorlabs, ELL6K). The filtered beams are split into two parts: One passes through the low-pressure reference cell to determine the spectral position of the narrow CO absorption line during a wavelength scan whose peak is identified with the absolute wavelength listed in the HITEMP database. The other part is guided to the 7.62-cm-long solid Ge etalon (free spectral range (FSR) 0.016 cm<sup>−</sup>1). The known fringe spacing enables conversion of tuning time into relative laser wavelength. Information from both channels can be combined to convert scan time into absolute wavelength.

All collinearly combined NIR and MIR beams are then steered to a parabolic mirror (Thorlabs) and are focused into a SM MIR InF3 patch fiber (Thorlabs, P3-32F-FC-1). Due to its small numerical aperture of 0.26, a six-fold reflective beam telescope (Thorlabs, BE06R/M) reduced the beam diameter to approximately 100 μm. This telescope significantly reduced the coupling losses in the fiber. By precisely adjusting the positions of the collimators in front of the PM fiber outputs of the NIR lasers (and the QCL at 4.56 μm), the difference in beam divergence of the four collimated beams can be slightly minimized to achieve better focus quality and reduce coupling losses.

All four lasers are driven by individual controllers (Arroyo, 6300 series) including temperature stabilization and injection current control. The optical setup is installed on a water-cooled aluminum optical board (450 × 600 mm2, Thorlabs), which rested on four aluminum profiles with passive vibration isolators (Figure 5). Below this plate, two fans provided air circulation within the unit. The remaining space accommodated controllers and power supplies for lasers and detectors.

**Figure 5.** Side view of the sending unit mounted on an optical breadboard with the laser controllers below.
