Compact highly sensitive multi-species airborne mid-IR spectrometer

Abstract

We report on the development and airborne field deployment of a mid-IR laser-based spectrometer. The instrument was configured for the simultaneous in situ detection of formaldehyde (CH₂O) and ethane (C₂H₆). Numerous mechanical, optical, electronic, and software improvements over a previous instrument design resulted in reliable highly sensitive airborne operation with long stability times yielding 90 % airborne measurement coverage during the recent air quality study over the Colorado Front Range, FRAPPÉ 2014. Airborne detection sensitivities of ~15 pptv (C₂H₆) and ~40 pptv (CH₂O) were generally obtained for 1 s of averaging for simultaneous detection.

1 Introduction

Sensitive and selective laser-based trace gas absorption spectroscopy has become a widely applied method for research and industrial applications [1, 2]. Various suitable optical techniques and approaches have been developed and incrementally improved over the last decade [3]. Instruments targeted for use in atmospheric research applications are challenged by the need for very high sensitivity, precision, and accuracy, while being subjected to unforgiving environmental operating conditions present on most airborne platforms and at remote ground sites [4–7]. This paper describes numerous advances over a predecessor airborne spectrometer [8] called difference frequency generation absorption spectrometer (DFGAS), which is based on a tunable difference frequency generation laser source and a modified long path length Herriott optical absorption cell. Like many other laser spectrometers, the instrument described here (CAMS, compact atmospheric multi-species spectrometer) utilizes selective optical gas absorption of strong rotational–vibrational absorption lines in the mid-infrared spectral region coupled with long optical paths and the ability to detect small changes in laser radiation to obtain low detection limits in the parts-per-trillion (pptv) range.

To maximize the signal-to-noise ratio (SNR) and precision, this requires careful engineering and design of mechanical, electronic, and optical components, which we will sequentially discuss. Specifically, we provide details in this paper regarding the: (1) laser source; (2) new optical sampling/absorption cell; (3) electronics; (4) environmental temperature and pressure control of the optical system; (5) data acquisition and processing software; and (6) gas flow and calibration systems. A photo-rendered CAD model of the CAMS spectrometer module discussed here is shown in its final configuration in Fig. 1.

Fig. 1

Photo-rendered CAD image of CAMS optical core showing the compact integration of optical, electronic, and gas flow components with a total length (69 cm), width (39 cm), and height (18.5 cm), and total weight of 22 kg. 1 Fiber amplifier tray, 2 seed diode lasers and electronics, 3 DFG module, detectors, and optics detailed in Fig. 4, 4 multi-pass cell, 5 gas inlet and outlet

2 Instrument design and setup

2.1 DFG laser source: lasers, fiber amplifier, PPLN stage

A quality tunable laser source is essential for obtaining high precision measurements, regardless of the spectroscopic technique. Important laser source properties include high beam quality and stability (near-Gaussian spatial mode, low divergence, and large beam waist lead to minimal magnification when coupling the laser beam into a multi-pass absorption cell), sufficient power, narrow linewidth, high side mode suppression, linear tuning, and associated long-term reliability [6]. Such properties, respectively, allow minimization of noise due to scattered photons and in absence of other sources of technical noise lead to an effective increase in signal above the electronic instrument noise floor, selectivity, and long-term spectroscopic stability. Laser sources based on difference frequency generation (DFG) using telecom grade-based distributed feedback(DFB) diode and fiber laser sources offer these characteristics and offer unique multi-wavelength emission capabilities [9].

The CAMS’ tunable laser source is based upon difference frequency generation using quasi-phase matching in congruently grown periodically poled lithium niobate (PPLN), as shown in Fig. 2. As seed sources, two pairs of DFB diode and fiber laser sources were fusion spliced and amplified in Yb³⁺ and Er³⁺/Yb³⁺ doped fibers [10, 11]. For the detection of ethane (C₂H₆) at 3.35 µm (2,986.7 cm⁻¹), a 1,063.5 nm DFB laser was paired with a 1,558 nm DFB laser. Similarly, for the detection of formaldehyde (CH₂O) at 3.53 µm (2,831.6 cm⁻¹), a 1,083 nm DFB fiber laser and a 1,562 nm DFB laser was applied. The ~1 and ~1.5 µm lasers were amplified in Yb³⁺ and Er³⁺/Yb³⁺ doped dual-clad optical fibers, respectively, using counter propagating multi-mode pump lasers operating at 976 nm. The individual output powers for each wavelength depends mostly upon the wavelength-dependent fiber gain, fiber length, and seed powers. Adjusting the seed powers allows flexible adjustment of the respective power products governing the idler powers generated at 3.35 and 3.53 µm and can mitigate the wavelength-dependent conversion efficiency (typically, 200–300 µW W⁻² cm⁻¹ in the 3–4 µm wavelength range) to generate equivalent idler powers. Figure 3 shows the amplified output powers for the ~1 and 1.5 µm wavelength channels for various relative seed power levels versus launched pump power levels. The measured output power levels include insertion losses introduced by splices, fused fiber components, and output isolators. The amplifiers were operated in saturation with combined seed power levels of over 20 mW for each channel. The splice connecting the single-mode seed fiber and the dual-cladding fiber was optimized to couple into the doped cores by observing the transmitted power as well as monitoring the loss coupled into the pump claddings. A counter propagating pump configuration was chosen for better suppression of forward- and backward-propagating amplified spontaneous emission (ASE) and stable amplification with lower seed powers. In the case of the Yb fiber amplifier, this was particularly important due to the lower gain at 1,083 nm, which competes with ASE at shorter wavelengths subjected to higher gain. In addition, a bandpass pump laser protection filter was used for the Yb fiber channel pump laser. Polarization of the pump and signal lasers were adjusted by polarization controllers prior to the optical amplification. Linear polarization states were maintained at the output of the fiber amplifier chain for optimal DFG. After passing through fiber optic amplification, both pump and signal beams are combined and launched from a single-mode fiber into a 50 mm long by 1 mm thick anti-reflection (AR)-coated (for pump, signal, and idler wavelengths) PPLN crystal. Both generated beams are collimated using a single CaF₂ lens. An AR-coated Ge window removes the unconverted signal and pump wavelengths. The maximum mid-IR power generated with the fiber amplifier combination above was ~200 µW per wavelength channel.

Fig. 2

Difference frequency generation-based tunable mid-IR laser source. L _Sx DFG signal lasers, L _Px DFG pump lasers, P _FA 976 nm multi-mode fiber amplifier pump lasers, Mon _x monitoring photodiodes, WDM wavelength division multiplexer, TFB tapered fiber bundle signal and FA pump combiner, PPLN periodically poled lithium niobate (nominal grating period = 30.1 µm)

Fig. 3

Optical fiber amplifier performance for the Er/Yb (signal) and Yb (pump) wavelengths. Shown are the effective slope efficiencies (S.E.), which include insertion losses and unabsorbed pump power

One unique advantage of fiber optically pumped, single-pass DFG is the high, near-Gaussian beam quality. The DFG source’s relatively large beam waist of ~100 µm only requires a single f = 50 mm focal length lens for capture and imaging into a modified 89.6 m Herriott multi-pass absorption cell (magnification ~10), which provides very good positional and optical background stability. In contrast, direct emitters such as lead–salt and quantum cascade lasers, with their much smaller beam waists, require significantly higher magnifications, which in turn places higher demands on precise adjustment of the first collection optic and its positional stability. Both of these attributes result in enhanced stability of DFG-based systems.

As shown in Fig. 4, the two beams are directed through a Brewster-type beam splitter (BTBS) [12, 13]. This beam splitter is designed to separate the beams into three parts of different optical intensity while suppressing noise originating from residual reflections from plane parallel surfaces that create optical interference fringes. One part (~1 %) is passed through the BTBS and directed through gas reference cells filled with pure ethane and formaldehyde at low pressures before being detected by the reference (REF, a photovoltaic HgCdTe, PV MCT) detector. The REF signal is used to lock the center of the absorption features to the center of the wavelength scans. About 30 % of each wavelength-modulated beam is reflected off the front surface of the BTBS and directed onto a second PV MCT detector, which is designated as the amplitude modulation detector (AMD) in Fig. 4. This detector is used to capture the modulated optical background laser structure, which may in part originate from nonlinear residual amplitude modulation (RAM) of the signal lasers when they are scanned and modulated. This background structure is further changed by a host of additional convolved optical structures from the entire dispersive optical (fiber) train leading to random and fringe-like wavelength-dependent intensity changes of as much as 10⁻⁴ to 10⁻³. Although the noise appears as a fringe-like structure similar to an optical interference between two parallel surfaces, it does not behave as such over time. The noise continuously changes its structure from periodic to non-periodic shapes and shows respective amplitude variations. Weibring et al. [14] discuss in greater detail the temporal behavior of such noise, which we refer to here simply as background optical noise structure.

Fig. 4

Optical setup. M Mirror, BTBS Brewster-type beam splitter, Gx gas reference cell, OAP 60 deg off-axis parabolic mirror, CD cell detector, AMD amplitude modulation detector, REF reference detector MPC multi-pass cell

The third beam emanating from the BTBS is directed into a modified Herriott multi-pass absorption cell (see Sect. 2.2). After traversing an 89.6-m effective optical pathlength, the output beams emerge geometrically separated by ~5 mm from the input beams and are directed onto a third PV MCT detector, which we designate as the cell detector (CD) in Fig. 4. The background optical noise structure, which is common mode to CD and AMD chains, is captured and removed in real time to resolve absorption sensitivities down to 10⁻⁷. Weibring et al. [8, 14] provide greater details regarding all the requirements necessary for matching this common-mode noise in the AMD and CD detection chains by 1/1,000 and will not be further discussed here with the exception of minimization of scattered light. As will be discussed in the next section, the success of removing such large background optical noise structure critically depends upon reducing added non-common-mode noise generated from cell scattered light. As will be shown, the multi-pass cell employed in CAMS is designed to minimize this effect.

2.2 Multi-pass cell (MPC)

For maximum performance, an optical gas absorption cell should provide a long effective absorption path length, offer a small sampling volume to facilitate a fast gas exchange and response time, and minimize effects of optical scatter and noise superimposed onto the absorption signal. The CAMS instrument described here employs a (non-astigmatic) Herriott cell to achieve these attributes. We provide in this section scattered light simulations to show that detected scattered light emanating from the cell mirrors is significantly reduced relative to an astigmatic design. This is particularly important in the present instrument, where high-precision matching of the temporal signal profiles between the CD and AMD optical arms is required. Such reduced scattered light, which to our knowledge has not been the focus of other publications discussing alternative cell designs [15], may also yield performance improvements in other laser spectrometers. This will be further discussed shortly. In addition, the patented MP cell [16] offers a number of other optical and gas sampling advantages. Its design offers the ability to readily change optical path lengths by a change of mirror separation, a significantly reduced sampling volume compared to traditional Herriott cells, and a mostly all-glass internal surface to minimize chemical adsorption effects. The reduced volume is accomplished by reducing the sampling zone to the annulus of two concentric tubes. Its design includes two dielectrically enhanced metal-coated (R ~ 99.6 % between 2.8 and 3.8 µm) spherical mirrors (f = 1500 mm), which are mounted to a center glass-covered carbon fiber tube. Two circular end plates hold the internal assembly and attach to an outer glass shell. The mid-IR beams are coupled into the cell through a 12.7-mm-diameter AR-coated CaF₂ window mounted in the end plate and a slotted hole machined on the edge of the input spherical mirror. The cell was configured for an effective optical path length of 89.6 m and transmitted ~55 % of the coupled light. The use of a stepping mirror assembly re-launches the recirculating beam to a different radius with respect to the center longitudinal axis of the cell prior to exiting the cell. This effectively doubles the optical path length of a traditional Herriott cell configuration. In addition, the input and output beams are separated by ~5 mm and opposing angular direction (~2.5°), which facilitates a good optical separation close to the entry aperture of the multi-pass cell.

Most notably, as discussed previously, the design offers superior optical performance resulting in significantly reduced optical scattering. Scattering is minimized by two different features of the CAMS cell: (1) This cell employs spherical mirrors made from fused silica substrates that are highly polished and coated with dielectric-enhanced metal coatings, and this reduces scattering compared to diamond-turned metal surfaces where micro-grooves cannot be entirely eliminated; and (2) the unique geometric configuration reduces the number of scattered photons incident on the CD detector. The reduction in scattered photons reaching the CD detector directly determines the degree of effective subtraction of the DFG background optical noise structure measured by the AMD when zero air (air scrubbed of formaldehyde and ethane) is directed into the sampling cell. Compared to our spectrometers utilizing a Herriott cell employing astigmatic mirrors with reflective silver coatings, we were able to obtain significant improvement in the ease of matching between the AMD and CD detectors when zero air is sampled. The ease of matching is measured by the ability to obtain identical baseline shapes represented by the time (wavelength dependent) signal amplitude and signifies the lack of optical noise emanating from the multi-pass cell reaching the CD. This also resulted in an increased stability time, thus increasing our ambient measurement duty cycle from ~60 to over 90 % during the most recent airborne science mission.

Figure 5 shows optical simulations (FRED, Photon Engineering, LLC) of the astigmatic-type Herriott cell and the new CAMS non-astigmatic Herriott. A beam propagation factor of M ² = 1.24 was measured for the fiber-pumped DFG source described in this paper and the simulations performed with M ² = 1. Figure 5 shows a forward and backward ray tracing for a 100-m astigmatic Herriott cell (ROC_x = 675.9 mm, ROC_y = 740 mm, Mirror separation = 549.21 mm, 182 reflections, ROC = Radius Of Curvature) and the new 89.6-m non-astigmatic Herriott cell (ROC = 3,000 mm, Mirror separation = 546.21 mm, 164 reflections). The backward ray trace provides a first-level scatter path from the mirror to the detector. The detector power was set equal to A*L*sin²(θ), where A is the area of the detector (π × 0.5²), L is the mirror radiance (unity), and θ is the emission angle (1.5°). The modeled power is then P = 5.38 × 10⁻⁴ W. The total power incident on the mirror is numerically the total stray light power on the detector for a unit radiance mirror. The detector was modeled as a 41 × 41 grid plane on a 0.5 × 0.5 mm semi-diameter detector (1 × 1 mm square diameter), and 50 rays were launched randomly into a 1.5 semi-angle (3° full angle) from each grid point. The ray interactions with the back mirror surface were accumulated with a script to calculate the cumulative irradiance over the number of passes through the cell. The integrated power on the back mirror of the astigmatic cell is 20.5 and 2.1 mW for the non-astigmatic Herriott cell. This tenfold reduction in scattering is further reduced by the use of highly polished optical substrates (not simulated here) to yield significant increase in performance as will be presented below.

Fig. 5

Optical scattering simulation. a and b show, respectively, the modeled scattering and beam spot pattern for an astigmatic Herriott cell, while c and d depict these for the non-astigmatic Herriott cell used in CAMS. For all cases, the irradiance was scaled to log (base 10) with the floor set -3 decades below the peak

2.3 Detection electronics

Instrument control, electronic detection, and processing often determine the non-optically induced noise floor. Thus, our instrument was designed to minimize electronic-induced signal degradation effects through electronic noise filtering, low-noise laser drive and control circuitry, low-noise detector amplification design, and opto-isolated housekeeping control. Figure 6 shows the electronic signal processing chain starting with a reverse-biased 1-mm² 4-stage TEC-cooled (−80 °C) photovoltaic mercury cadmium telluride (PVMCT) detector connected to the input of an ultra low-noise and low-distortion transimpedance amplifier (TIA) with an effective gain of 44 K. The TIA output is then split into a AC and DC channel, namely a high-pass (HP) filter with 5 × gain and a low-pass filter, respectively. The DC low-pass filter configures the instrument for direct optical absorption, in which the appropriate seed laser is simply current tuned across the absorption line of formaldehyde or ethane by a triangle waveform. A rapidly synced optical chopper to the scans enables the measurement of the total optical power and transmitted intensities for a direct computation of the calibration concentrations using the Beers-Lambert law. This mode was used daily in determining absolute calibration mixing ratios for formaldehyde and ethane compressed gas standards that were added to the instrument inlet from first principles knowing sampling temperature, pressure, line strength, and absorption path length. Gas standards of CH₂O and C₂H₆ were sequentially sampled with only the corresponding signal and pump laser pair enabled.

Fig. 6

Electronic signal processing. TIA transimpedance amplifier, PV MCT photovoltaic HgCdTe detector, SW electronic switch, AD analog to digital converter, TEC CTRL thermoelectric cooler control

2.4 Software processing and lock-in demodulation and phase-sensitive detection

Measurements requiring absorption sensitivities approaching 10⁻⁷ also require sophisticated software to control, acquire, and process the instrument data. Our instrument utilizes fast and dedicated FPGA-based platforms along with optimized processing algorithms.

For the DFGAS instrument [8], a phase-locked loop (PLL) lock-in detection technique was applied. For CAMS, this approach was replaced by a coherent scan and sample scheme, including sample acquisition locked to the waveform generation. This allowed suppression of asynchronous noise and post-processing storage of raw data.

Second harmonic detection of the wavelength-modulated absorption signals were obtained as follows [17]. The respective signal seed lasers for the C₂H₆ and CH₂O channels were current tuned with a 25-Hz triangular waveform. Superimposed, sinusoidal waveforms at 40 and 50 kHz were applied resembling a traditional wavelength modulation spectroscopy approach. The signals were then acquired by a 16-bit, 800kS/s data acquisition card and processed by employing computer demodulation software, as will be discussed. An electronic noise floor corresponding to an effective detectable absorbance of 1 × 10⁻⁷ (1 Hz) was measured after the system was calibrated. The electronic noise floor includes all noise sources from the entire system (hardware, software, fitting algorithms) when the system is operated without the lasers turned on. After acquisition, signals were averaged and then transformed by software processing into frequency space, while the sinusoidal waveform provided clean FFT planes for bandpass filtering around their respective harmonic signals in the frequency plane [18]. The filtered signals were then transformed back to the time domain and fitted using a singular value decomposition (SVD) routine to calibrated absorption signal signatures obtained during calibration routines as verified by direct absorption (further described in the next section).

The fundamental modulation frequencies are utilized to apply a computerized balanced detection of the CD/AMD signals using their power spectral density (PSD) in the scan endpoints. The residual CD/AMD difference, i.e., instrument background, is recorded during a “zero” measurement with clean (void of C₂H₆ and CH₂O) air in the absorption cell and subsequently subtracted from the recorded measurements. Automatic algorithms record new backgrounds and perform a “zero” measurement when the instrument precision is degraded below a user-determined criteria as discussed further below.

2.5 Spectrometer enclosure and gas flow system

Often, noise introduced through mechanical and optical effects by way of ambient variability of temperature, pressure, vibration and shock limits the instrument performance. Weibring et al. [8] discuss the requirements for precise temperature and pressure control of the entire optical system for high performance.

To accomplish this, the entire optical compartment (78 cm × 44 cm × 21 cm) of the CAMS instrument was well insulated and actively temperature and pressure controlled. An external air-conditioning unit continuously circulated temperature-stabilized air (±0.1 °C max) throughout the optical enclosure. A miniature 400 W air conditioner compressor unit (Termotek, AG, Germany) was mounted to the vessel. Air was circulated with fans alongside the bottom of the spectrometer baseplate and returned over the top, then passed through a heat exchanger and recirculated. Generated heat was channeled to heat sinks exposed to the airflow, while most optical and mechanical parts were shielded or insulated to minimize temperature gradients.

Maximum ambient cabin temperatures approaching 37 °C were measured during some research flights conducted at low altitude and elevated ambient summer temperatures. Even for these extreme ambient conditions, the air-conditioning system’s cooling capacity never exceeded 75 %. The enclosure temperature was maintained to a constant temperature of 23 °C. An additional heating element mounted close to the heat exchanger provided additional flexibility in maintaining this temperature set point for much colder ambient conditions. The optical bench was mounted inside a pressure vessel and attached to the enclosure through vibration isolators. The pressure in the enclosure was stabilized to a pressure of ~80 kPa (±0.01 kPa) using a pump, and pressure controller (PC) together with a controlled flow (~100 std cm³/min) of dry cylinder make up zero air, as shown in Fig. 7. The leak rate of the vessel was measured to be less than 20 std cm³/min. The enclosure pressure was chosen to be approximately mid-range between the highest and lowest anticipated cabin pressure to minimize the pressure differential across the enclosure vessel. In addition to a safety pressure relief valve on the pressure controller, the enclosure was automatically vented to atmospheric pressure by a 3-way solenoid valve when system power was shut off. Vacuum feed-throughs for all the electrical and flow plumbing components provided the necessary pressure seal to maintain constant enclosure pressure under all cabin pressure scenarios. This pressure-stabilized enclosure, which pressure stabilizes all the optical components without flexing these components, is a significant improvement of CAMS over its predecessor DFGAS, which did not have this option. Cabin pressure changes by as small as 1.3 kPa in 1 min often caused DFGAS background changes that degraded its instrument performance. Although the exact cause of the pressure-induced degradation is still under investigation, it is clear that this feature in CAMS adds stability.

Fig. 7

Gas flow system. PC pressure controller, FM flow meter, FC flow controller

The gas flow system illustrated in Fig. 7 provides an effective continuous ambient gas sampling, zeroing, or calibration mode operation. Ambient air is sampled through a heated (35 °C) electropolished 12.7 mm OD (all diameter dimensions hereafter refer to the outside diameter) stainless steel tube (inlet in Fig. 7) that has a restriction on the outlet to provide a small RAM-air pressure boost for sampling at high altitudes and high speeds (~8 kPa at ~200 m/s above ~10 km). Large particles greater than ~1-µm (estimate) are excluded by subsampling at right angles through ~30 cm of heated 12.7 mm OD electropolished stainless steel tubing, which in turn is connected to several feet of 12.7 mm OD heated PFA Teflon line. The exact length of line depends upon the aircraft platform. The sampled air is directed through an MKS T3B pressure-controlling throttle valve (labelled PC(T3B) in Fig. 7). This valve, which has a low pressure drop of ~1 Torr at flow rates of ~10-slm (where standard conditions refer to 1 atmosphere pressure and 273 K), in conjunction with the RAM-air pressure boost is necessary for sampling on the NASA DC-8 at high altitudes and high speeds. The sampled air at the output of the pressure-controlling valve is directed into the optical absorption cell, which is enclosed in the pressure- and temperature-controlled vessel. The sampling pressure inside the MPC (volume = 1.5 L) was maintained at 6.6 kPa (±0.01 kPa). The spectrometer base platform consisting of laser sources, optical components, detectors, detection and control electronics, and a signal housekeeping FPGA-based computer is mounted on vibration mounts attached to the bottom of the enclosure. A convoluted flexible Teflon sampling line connects the MPC cell through sealed ports of the pressurized enclosure and minimizes mechanical coupling to the spectrometer. After passing through the MPC cell (~3.9 slm, 1/e-folding sampling time ~1.5–1.8-s), the sample gas is directed through a flow meter and is drawn outboard via a scroll pump (Anest Iwata ISP 90, 108 L/min).

As shown in Fig. 7, calibration standards in zero air and/or zero air without standards can be added near the inlet entrance. This is accomplished by overflowing the inlet with an excess (~2 slm) of the desired sample by switching in the flow from a 3-way solenoid valve following the zero air flow controller (FC) in Fig. 7. Zero air is provided employing a scrubber pair (heated Pd/Al₂O₃ catalyst beads from Aadco corporation followed by an unheated SnO₂ scrubber). The beads are heated to 380 °C. Compressed cabin air is used to feed the scrubber at a pressure of ~90 kPa and the combination of the pressure controller upstream of the scrubber, and the flow controllers downstream ensure constant residence time under all aircraft cabin pressures. The scrubber pair was found to be necessary under highly varying ambient water vapor conditions, which on occasion revealed CH₂O breakthrough, as evidenced by persistent negative concentration values at high altitudes. This was obviated by the use of the low-temperature SnO₂ scrubber at the output of the heated Pd/Al₂O₃ catalyst beads. Complete ethane removal was accomplished by the heated catalyst beads. For an abundance of additional caution, a secondary line connected to a cabin-mounted zero air cylinder is used to periodically compare against the scrubber pair to ensure complete removal of CH₂O and C₂H₆ under a range of ambient concentrations. A manual 3-way valve dictated which zero air source is employed. For calibration, the output of two high-concentration gas cylinders (ppm range) is added to a dilution air flow (dil air FC), which in turn is added to the zero air flow. The desired concentrations can be selected by appropriate flow controller settings. When not in use, the standards are drawn away by the vacuum pump by switching in the 3-valve adjacent to the suck-back flow controller. The suck-back flow controller is set to draw back ~1.5–2 times the combined flow of the standards and the dilution flow so that any outgassing in the standard lines are drawn away from the inlet. This provides fast removal of the calibration gases.

Employing the inlet system of Fig. 7, we calibrate and chemically zero nearly the entire system inlet, which in the case of CH₂O is extremely important to eliminate inlet line/cell outgassing effects after sampling large plumes from fires and other sources. In addition to laboratory tests with the entire inlet, calibration standards were typically added before and after each research flight and during maintenance days using various gas–phase calibration standards. Occasionally, during periods at high altitudes while transiting from and to scientifically important locations on research flights, calibration gases were also sampled. This verified accurate spectroscopic performance and chemical sampling accuracy during actual flight sampling conditions. As discussed previously, the absolute mixing ratio of the calibration standards passing through the entire system was frequently determined by direct absorption. Background zero air measurements were acquired more frequently between ambient measurement cycles. For measurements during the 2012 DC³ (deep convective clouds and chemistry) study, and the 2013 SEAC⁴RS (studies of emissions and atmospheric composition, clouds and climate coupling by regional surveys) study, this was accomplished once every minute to effectively capture changes in optical background structure. However, during the 2014 FRAPPÉ (Front Range Air Pollution and Photochemistry Éxperiment), the CAMS system was considerably more stable and background zero air measurements were typically acquired once every 5–10 min.

Ideally, the frequency of the zeroing cycle is determined by the instrument stability employing an Allan-Werle variance plot [19]. In most cases, such a plot shows an instrument improvement in SNR with averaging time until the technical-induced noise grows faster than improvements through longer averaging. Such variance measurements are, however, unpractical during airborne operation, as they require sampling a stable calibration or zero air sample gas concentration at the expense of valuable ambient acquisition. During the 2014 FRAPPÉ study, we adopted an alternative tool, the fit precision, to determine dynamically the optimal instrument time between zero air background acquisitions. As discussed by Wert et al. [20], this precision is the product of two terms: the first involving the squared deviations between the ambient spectrum and the fit spectrum summed over each of the wavelength channels and a term corresponding to the absolute uncertainty of the model basis vectors (a measure of the linear independence of the basis vectors) determined from the variance–covariance matrix. Wert et al. [20] found that although the fit precision estimate positively correlated with replicate precisions during constant air mass legs, it yielded results that were overly optimistic because it did not encompass all the factors involved in the measurement cycle. Nevertheless, Wert et al. [20] showed that fit precisions could be used as a useful real-time diagnostic of precision trends largely independent of ambient concentrations. We employed this approach during the 2014 FRAPPÉ study by monitoring when ambient fit precision estimates started to degrade beyond 5 % of the last 10 s of measured averaged concentration or when the ambient sampling interval exceeded 600 s. Either condition triggered an automated background acquisition lasting 15 s in duration. Before and after each ambient-background measurement state switch, we instituted a 10-s delay (~5 1/e-folding times) before data were used in the analysis. An additional delay period of 6 s was employed for state switches involving calibration-background changes. In addition, the operator could override zeroing at any time to either force a zero or avoid a zero cycle for times of high scientific interest. Adopting this approach, we extended our ambient data coverage during the 2014 FRAPPÉ study to over 90 % compared to the earlier CAMS deployments (60 %) without loss of measurement precision.

3 Development history and airborne results

The fundamental instrument described here was first developed and flown during the DC³ field campaign aboard the G-V airplane operated by NCAR, followed by the SEAC⁴RS field campaign on the NASA DC-8, and lastly the FRAPPÉ field campaign aboard the NCAR C-130 airplane. As discussed, the fundamental CAMS design is based on its predecessor DFGAS spectrometer [8]. However, CAMS was completely re-engineered using a number of advanced components and approaches as described above. For this new instrument, a number of additional advances were made after each field campaign. For the first two campaigns, the instrument was configured for the detection of formaldehyde only. An actively pressurized and temperature-controlled enclosure was added for the SEAC⁴RS campaign. Relatively stable cabin pressure and temperature aboard the G-V aircraft provided sufficient stability during the DC³ campaign but was inadequate for operation on the NASA DC-8 and NCAR C-130 aircraft where respective changes of several hundred Torr and over 20 °C were experienced.

For the 2014 FRAPPÉ campaign, the fiber optic amplification was optimized for multi-wavelength amplification at 1 and 1.5 µm. Signal acquisition and processing were further expanded and optimized for simultaneous detection of C₂H₆ and CH₂O. Figure 8 shows the single-channel performance during the 2013 campaign during a high-altitude flight leg. Assuming a minimum amount of air mass variation, we can estimate a conservative detection sensitivity of 17 pptv (average 1 s replicate standard deviation) for five consecutive 60-s acquisition cycles. This performance corresponds to a detectable absorption of 5 × 10⁻⁷ or 5.6 × 10⁻¹¹ cm⁻¹ for CH₂O. For the two-channel operation, the detection sensitivities were slightly reduced due to the split of optical power between each wavelength channel. For the FRAPPÉ mission, the electronic signal gain was reduced by 4× (gain changed from 20× to 5× in the HP section) to accommodate possible higher total optical powers limited by the dynamic range of the data acquisition interface. Typical 1-s precisions of 15–20 and 40–50 pptv for C₂H₆ and CH₂O when zeroing after ~5 min of sampling, respectively, were achieved, whereas single-channel operation yielded ~10 and ~20 pptv 1-s precisions for C₂H₆ and CH₂O, respectively.

Fig. 8

Single-channel detection performance at high altitude and stable air mass. The means and standard deviations are shown for each 60-s ambient measurement cycle

Figure 9 shows 1 h of boundary layer flight measurements of CH₂O and C₂H₆. Represented by the plot gaps, the instrument was only zeroed seven times (40 s each) providing a very high measurement duty cycle of 92 %. The entire research flight is shown in Fig. 10 and color coded for concentrations. For clarity, the C₂H₆ colored concentration span is set from 1 to 10 ppbv, which omits the display of higher peaks measured up to 20 ppbv during this research flight. For reference, the gray-highlighted area shows the data plotted in Fig. 9. During the measurements shown in Fig. 10, the aircraft was maintaining 1,000 ft above ground level when flying the ‘lawn mower’ pattern across the front range and passing over several oil and gas operation sites in the Denver Front Range. Several narrow and wider peaks of high C₂H₆ concentrations were measured, as seen in Fig. 9. Narrow peaks (as short as a few seconds) as well as the absence of enhancements in formaldehyde indicate fresh direct emissions, while wider peaks with simultaneous formaldehyde enhancements indicate photochemically processed plumes or flaring, which would yield such enhancements. While C₂H₆ is directly emitted from well and processing sites, significant formaldehyde enhancements are photochemically produced primarily by other VOCs (volatile organic compounds) [21].

Fig. 9

Measurement of CH₂O and C₂H₆ during FRAPPE research flight #2 (07/27/2014). See text for discussion

Fig. 10

Flight track and color coded measured concentration of CH₂O and C₂H₆ during research flight #2 (07/27/2014)

To further illustrate the significant improvements of the instruments spectroscopic performance and robustness, Fig. 11 shows the measured concentrations of C₂H₆ and CH₂O for a different research flight during which the aircraft performed a series of violent maneuvers including pitch, roll, and yaw. Such maneuvers were carried out for the calibration of instruments that measure the wind speed and direction and were first developed by D. Lenschow [22]. Significant acceleration and forces are presented for such maneuvers and can reveal if a scientific instrument measurement performance is affected. These tests were performed at higher and lower altitudes. Figure 11 shows the maneuvers at an altitude of ~5.9 km and a location that minimizes ambient influences from urban pollution. This in turn allows a direct comparison of replicate measurements during the three maneuvers without the simultaneous influence of air mass changes. The instrument was operated with fixed zeroing every 60 s and sampling intervals and allowed us to observe any optical (sampling) as well as gas system (zero) effects during the maneuvers. One can see that no significant changes of the measured standard deviations are present during these maneuvers, which were performed during zero and ambient sampling intervals and confirm that the entire spectrometer system is not susceptible to such aircraft perturbations. This is a significant improvement from our previous DFGAS spectrometer, which showed a measurable susceptibility to such perturbations.

Fig. 11

Performance during Lenschow Maneuvers (see text for discussion). In the flight sections, we provide the average and SD for aircraft pitch, roll, and yaw maneuvers (degrees) along with the corresponding retrieved CH₂O and C₂H₆ concentrations (pptv). Section A (constant pitch and roll, varying yaw), pitch = 2.3 ± 0.2, roll = −26.3 ± 0.1, yaw = 166 ± 107, CH₂O = 305 ± 43, C₂H₆ = 642 ± 27; Section B (constant pitch, varying roll and yaw), pitch = 2.0 ± 0.2, roll = 22.6 ± 12.4, yaw = 166 ± 107, CH₂O = 201 ± 50, C2H6 = 667 ± 34; Section C (constant pitch, roll and yaw), pitch = 1.3 ± 0.3, roll = 1.1 ± 0.8°, yaw = 40.1 ± 0.4, CH₂O = 227 ± 51, C₂H₆ = 654 ± 20; Section D (varying pitch, constant roll and yaw), pitch = 1.3 ± 2.8, roll = 0.4 ± 0.6, yaw = 34.6 ± 1.2, CH₂O = 170 ± 58, C₂H₆ = 635 ± 28

Extensive post-mission laboratory tests of the CH₂O and C₂H₆ channels were carried out to detect potential spectroscopic interferences. In the case of C₂H₆, various gases were tested, including: methanol, n-butane, i-butane, propane, acetaldehyde, acetone, acetonitrile, methyl ethyl ketone, benzene, toluene, and trimethylbenzene. Methanol was found to be the only major interference, yielding an interference of +3.4 % (i.e., 1 ppbv of methanol produces a C₂H₆ artifact of +34 pptv). In the CH₂O channel, methanol was also found to interfere, yielding an artifact of 2.6 %. Presently, both are removed using aircraft methanol measurements simultaneous with our measurements employing the above factors. In the future, this interference will be removed from the C₂H₆ and CH₂O absorption features by appropriate spectral fitting routines.

The improved CAMS spectrometer has accumulated over 300 flight hours during the DC³, SEAC⁴RS, and FRAPPÉ campaigns with negligible down time, operating on three unique airborne platforms.

4 Summary

We presented the design and performance of an improved DFG-based mid-IR spectrometer suitable for the simultaneous sensitive, selective, and robust detection of C₂H₆ and CH₂O on airborne platforms. The systems’ tunable laser source design allows flexible wavelength configurations in the 3–4 µm wavelength range by using different standard near-IR diode laser seed sources. The optical amplifier, DFG module, and optical signal chain including the multi-pass absorption cell supports operation in such wavelength range without any exchange or adjustment of optical or electronic components. Desired operating wavelengths are determined by the choice of seed (signal and pump) laser wavelengths, which can be easily exchanged and fusion spliced to the system without the need for any further realignment. Due to the very stable operation, zero gas consumption is very low and could conveniently be supplied from a gas cylinder instead of catalytic converters.

Very high single- and dual-channel detection sensitivities and high duty cycles (>90 %) were demonstrated during three field campaigns. Various electronic, mechanical, and opto-mechanical improvements have made this instrument rugged to sustain a wide range of unforgiving airborne perturbations without loss of spectroscopic performance. Single-channel (~5 × 10⁻⁷) and dual-channel (~1.5 × 10⁻⁶) absorption sensitivities were obtained during flight. The performance improvements of this spectrometer to previous designs [8] have been obtained through careful optimization of all system components. In particular, the minimization of optical scattering emanating from the multi-pass cell, along with low-noise fiber optic amplification, stable gas flow and calibration routines, a temperature- and pressure-controlled spectrometer enclosure, shock and vibration damping, low-noise electronics and power supplies, and effective data acquisition, and processing algorithms have all contributed to the performance improvements reflected in the higher sensitivity (in airborne environments) and much longer stability times (~10×).

The presented results suggest a high maturity of DFG-based laser spectrometers applied to airborne research. With the current research experience and understanding, future DFG-based systems can now be designed and integrated into more cost effective, smaller, and less complex architectures while offering similar or better high-sensitivity measurements for a wide range of trace gas species on a single optical platform.