Detection of Sulfur Dioxide by Broadband Cavity Enhanced Absorption Spectroscopy (BBCEAS)

Sulfur dioxide (SO2) is an important precursor for formation of atmospheric sulfate aerosol and acid rain. We present an instrument using Broad Band Cavity Enhanced Absorption Spectroscopy (BBCEAS) for the measurement of SO2 with a minimum limit of detection of 0.6 ppbv using the spectral range 305.5 – 312 nm and an averaging time of 60 seconds. The instrument consists of high reflectivity mirrors (0.9984 at 310 nm) and a deep UV light source. The effective absorption path length of the instrument is 610 m in a 0.957 m base length. Published reference absorption cross-sections were used to fit and retrieve the SO2 concentrations and were compared to a diluted standard for SO2. The comparison was well correlated, R2 = 0.9985 with a correlation slope of 1.01.

and Multi-Axis Differential Optical Absorption spectroscopy (MAX-DOAS) (Cheng et al., 2019). The LP and MAX-DOAS techniques are not in situ measurements but leverage the spectroscopic signature of SO 2 for quantification. The most applied technique is UV fluorescence with several different manufacturers selling instruments for monitoring. One such instrument, the 43i-Trace Level Enhanced from Thermo Electron Corp. (Franklin, MA, USA) has a detection limit of 0.208 ppbv for a 10 second average but can be as low as 0.05 ppbv for a 300 second average with a 1% or 0.2 ppbv precision. UV fluorescence uses pulsed UV light to excite the SO 2 molecules which then relax to re-emit light at a longer wavelength. The 43i instrument 30 includes a hydrocarbon kicker to remove most interfering hydrocarbons that also fluoresce when excited with UV light. Known interfering species for fluoresence technique include NO, m-xylene, and H 2 O.
Spectroscopic measurement of SO 2 in the UV region is based on its highly structured absorption at wavelengths lower than 320 nm. The structured absorption allows for independent quantification of SO 2 from other gases that absorb in the same wavelength window including NO 2 (Vandaele et al., 1998), SO 2 (Bogumil et al., 2003), BrO (Wilmouth et al., 1999), OClO (Bogumil et al., 2003), and many organic molecules with broad absorptions in the UV, acetone being just one example 50 (Gierczak et al., 1998) (Figure 1).

Experimental
The SO 2 cavity instrument consists of the optical cavity, mounted in a 3D-printed cage assembly sitting on top of an instrument control box. Figure 2a shows a schematic of the instrument including standard dilution and supply as well as gas control valves.   (10 nm FWHM, 310 nm, Edmund Optics). The fiber is then directed to the slit of a grating spectrometer. For this work, two different spectrometers were tested, an Avantes 2048-SPU2 grating spectrometer with a range of 260 -820 nm and a slit width of 10 µm, and an Andor DU440-BV Spectrograph with a SR-303i CCD camera cooled by a Peltier cooler to -20°C with a 65 1200 grooves/mm grating. For the Andor spectrometer the slit was closed until a resolution of 0.26 nm FWHM was achieved, which provided a sharp well-defined line function. The Avantes spectrometer was not ideal in terms of light efficiency being a non-cooled detector with increasingly large read-out noise past one second of integration time and a slit width much too small for the wide range of the detector. The optics are all mounted in an optical cage system constructed of carbon fiber tubes with the braces for the tubes made of 3-D printed parts consisting of either Poly-lactic Acid (PLA) or Acrylic styrene-acrylonitrile 70 (ASA), printed on an Ender3 (Creality) printer. The material of choice was changed for various structural elements depending on the structural and design requirement. ASA parts were vapor smoothed to help provide an air tight seal where needed.
Eventually the 3D printed parts were only used for structural support with stainless-steel tubes inserted into the mirror mounts which sealed via an O-ring to the cavity mirrors. The Teflon tube was held in place between the two stainless steel tubes on each end using a bored-through pipe connection. The reflectivity of the optical cavity was measured using the differential Rayleigh 75 scattering of He and N 2 gas according to the following equation (Thalman and Volkamer, 2010): Where d 0 is the cavity length (95.7 ±0.1 cm), Ray is the extinction due to Rayleigh scattering of the respective gasses (Thalman et al., , 2017 and I is the spectral intensity in the respective gas (N 2 or He). The measured reflectivity was found to be 99.85% and the measured reflectivity, effective pathlength and example spectra for N 2 and He are given in Figure   80 3.
The measured concentrations were retrieved by non-linear least squares fitting of the cavity extinction as given by Fiedler et al. (2003) and Washenfelder et al. (2008): where the (λ) is the wavelength resolved extinction, R(λ) is the mirror reflectivity, d 0 is the cavity base length, I 0 (λ) is the 85 reference spectrum, and I(λ) is the measurement spectrum. Previous works with much higher reflectivity mirrors included a term to account for the Rayleigh scattering of the reference (Washenfelder et al., 2008), this term is omitted for this level of mirror reflectivity ( 99.8%) and only necessary if the Rayleigh scattering contributes significantly to the extinction in the cavity (usually at reflectivities > 99.9%).
The concentrations of the trace gases of interest was retrieved by non-linear least square fitting in IGOR (Wavemetrics) 90 by minimizing the error of the following equation with a 3rd degree polynomial enabling a Differential Optical Absorption Spectroscopy retrieval (Platt and Stutz, 2008): Where σ(λ) is the standard absorption cross section for the given gas and [SO 2 ] is the retrieved concentration of SO 2 (Rufus et al., 2003;Bogumil et al., 2003). The absorption cross-sections were convolved to the instrument slit function using the 95 convolution function in QDOAS (Dankaert et al., 2017). Only SO 2 absorption was retrieved as the absorption cross sections of other gasses are either too small or not in large enough concentrations relative to the sensitivity of the instrument to be fitted. Cross-sections of SO 2 and other possible absorbers are show in Figure 1. Because of the fitted polynomial, the retrieval is only sensitive to the structured (differential) cross-section and is insensitive to broad changes in the light source shape, aerosol scatter (if no filter was used) and other broad-band absorbers (many organic compounds that interfere with fluorescence 100 measurements). Fitting was carried out from 305.5 -312 nm with a 3rd order polynomial and the retrieved concentration was converted to mixing ratio using the measured temperature and pressure.

Comparison to SO 2 standard
Initial testing of the BBCEAS instrument was carried out in comparison to an SO 2 standard cylinder (Airgas, 10.14 ppm SO 2 in N 2 ) diluted using a dilution calibrator (Environics, model 6103). The diluted standard was supplied to a manifold and sampled 105 into the BBCEAS through a 47mm PTFE particle filter (Pall) using a pump at 0.8 lpm. The diluted standard was also sampled by a Thermo 43c SO 2 monitor to observe the response of the calibrator.

Noise Evaluation
To evaluate the limit of detection, N 2 was continuously flowed through the cavity for several hours with a 1 sec integration time. Spectra were then averaged to a range of total integration times and evaluated with the Beer-Lambert Law (Absorption 110 = ln(I 0 /I)) to assess the root mean square noise (RMS). Pure photon counting noise followed the relationship RMS = 1/ √ N, where N is the number of photons collected.

Comparison to SO 2 standard
The BBCEAS followed the response of the SO 2 concentration delivered by the dilution calibrator linearly. Figure 4 shows 115 the measured SO 2 concentrations with time. Figure 5 shows the fit examples of various concentration levels of SO 2 showing unstructured residuals. The correlation of the standard dilution from the calibrator with the BBCEAS retrieved concentrations gave a slope of1.01, an offset of -0.85 ppbv, and an R 2 value of 0.9985 ( Figure 6). The absence of any structure in the residuals suggests no systematic error in the fitting routine. The fit residual was improved by 20% at higher concentrations by the use of the Rufus et al. Rufus et al. (2003) rather that the Bogumil et al. Bogumil et al. (2003) cross-section. The minimum fit residual 120 for the 1 minute average is 3 x 10 -8 cm -1 . The variability of the retrieved concentration at each concentration level indicated a limit of detection of 3.6 ppbv (3-σ) for a 1-minute acquisition.

Noise Evaluation
Signal to noise evaluation was carried out on spectra of N 2 both with the Andor and Avantes spectrometers. The spectra from the Andor spectrometer were taken from the start and finish of the SO 2 comparison tests, allowing for a maximum sampling

Improvements to Limit of Detection
Due to the broadened nature of the absorption lines of SO 2 (~1 nm FWHM), the instrument resolution of the Andor spectrograph (0.26 nm) was unnecessarily narrow. In initial testing the instrument slit was narrowed to reduce the light coming into the spectrograph so that the integration of a single scan was 3 s with 80% saturation on the chip. The slit was able to be opened wider with the 80% chip saturation occurring at the minimum integration time of 0.1 s. This provides a 30x improvement in the 140 signal. Given that the signal-to-noise improvement in the noise evaluation follows the theoretical relationship for photon counting, the RMS noise would improve to 4 x 10 -4 for a 1 minute integration time, yielding a 3-σ limit of detection of 0.6 ppbv in 1 min. This improvement in signal to noise allows for trace level detection of SO 2 in the presence of other structured absorbers.
The demonstration of workable BBCEAS measurements further in to the UV spectral range with lower reflectivity mirrors allows for measurement of an increasingly large number of molecules by BBCEAS in the UV and visible light ranges. Further 145 development of higher-powered UV LEDs provides enough light to access detection limit ranges of atmospheric importance (Washenfelder et al., 2016). Future development of the BBCEAS instrument could be made to lower the power requirements to a level low enough to allow the instrument to be mounted on a mobile platform such as an Unmanned Aerial System (UAS) for SO 2 source identification for larger emitters.