Cavity ring-down spectroscopy is a relatively new method for making isotopic measurements of carbon dioxide, methane and water vapor at atmospheric concentrations (O'Keefe and Deacon, 1988). Applications for instruments using cavity ring-down spectroscopy include monitoring of greenhouse gas emissions (Chen et al., 2010; Crosson, 2008), monitoring carbon storage and sequestration (Krevor et al., 2010), studying plant respiration (Cassar et al., 2011; Munksgaard et al., 2013), and process monitoring in the automotive and pharmaceutical industries (Gupta et al., 2009). Recent attempts to apply this technique to monitoring of active volcanic centers have been successful (Lucic et al., 2014, 2015; Malowany et al., 2014), but in some instances there have been anomalous responses from the Picarro G1101-i cavity ring-down spectrometers (CRDSs). Volcanoes emit a range of gases whose concentrations can be much higher than their concentrations in the ambient atmosphere. In particular, hydrogen sulfide gas is abundant in certain volcanic centers and can produce interference in the near-infrared spectrum in which the instrument operates. Our goal was to characterize and quantify this interference for future applications of the CRDS in volcanic environments.

Carbon isotopes are powerful tracers of volcanic gases and degassing processes (Gerlach and Taylor, 1990; Taylor, 1986) and are currently analyzed along with a suite of other geochemical tracers to monitor activity at active volcanoes (Carapezza et al., 2004). CRDS has a promising future monitoring activity at volcanic centers and tracking real-time changes in the isotopic composition of volcanic gases. However, interference of H2S with the isotopes of carbon dioxide prevents accurate measurements of the 12CO2 and 13CO2 concentrations, resulting in erroneous δ13C measurements. To use CRDS at volcanic centers, the interference of H2S gas needs to be characterized and removed. This paper reports the results of laboratory tests using carbon dioxide of a known isotopic composition spiked with different amounts of H2S to assess the nature of the H2S interference upon the CRDS. These controlled experiments were designed to qualitatively and quantitatively characterize the interference of H2S from low concentrations (1 ppb) to those observed at volcanic centers (> 10 000 ppb). To use these instruments for in situ measurements, a quick and efficient way of removing H2S from the sample gas prior to analysis is needed. Metals which have a high affinity for acid species, such as copper and zinc, react rapidly with H2S to form metal sulfides. If H2S can be removed from a sample gas without altering the isotopic composition of carbon dioxide, then the successful application of CRDS in H2S-rich environments will only require application of a simple metal scrub prior to analysis.

Interference was first observed with the addition of 20 ppb H2S, causing a change in δ13C of -0.5  ‰ from the 995 ppm CO2 standard (δ13C = -28.5 ‰ ). As H2S concentrations increased, the δ13C decreased proportionally (Fig. 2). A sample without H2S returned a stable δ13C value, but with increasing amounts of H2S the δ13C value started to decrease over the course of a single run. This resulted in an increasingly negative slope in the raw δ13C signal with the addition of greater amounts of H2S.

The decrease in the measured δ13C resulted from changes in the 12CO2 and 13CO2 concentration measurements in the presence of H2S. Figure 3 shows an increase in the 12CO2 concentration and a significant decrease in the 13CO2 concentration measured by the CRDS when comparing samples diluted with variable amounts of H2S to the same diluted samples that had been scrubbed of H2S. The percent change in the 12CO2 and 13CO2 concentrations are represented by Eq. (1), illustrating how the addition of H2S affects the measurements of the carbon isotopes used to calculate the δ13C value: % change in 12CO2 concentration=12CO2 with H2S-12CO2 with copper scrub12CO2 with copper scrub×100,% change in 13CO2 concentration=13CO2 with H2S-13CO2 with copper scrub13CO2 with copper scrub×100. There is an apparent decrease of nearly 50 % in the 13CO2 concentration reported by the CRDS with the addition of 20 000 ppb H2S, whereas the 12CO2 concentration has an apparent increase of only 3.5 % for the same amount of H2S. The end result is that H2S causes a large negative interference on the δ13C value measured by the instrument, predominantly governed by a negative interference with the 13CO2 concentration. This apparent decrease is a result of instrument interference between the H2S molecule and the 13CO2 and 12CO2 molecules in the absorption spectra. The gas samples prepared with the H2S and CO2 mixture had elevated 12CO2 and depleted 13CO2 with respect to the same sample after the H2S had been removed with the copper scrub.

Furthermore, the addition of H2S to the CO2 standard gas to create our gas mixture resulted in a decrease in the true chemical concentration of both 12CO2 and 13CO2. The real changes in CO2 concentration are known by the large difference between the predicted dilution of CO2, and that measured after H2S had been removed. Figure 4 shows this decrease in both the 12CO2 and 13CO2 concentrations when H2S is added compared to the pure CO2 standard gas. Dilution of the standard occurs by addition of 3 mL of H2S via syringe to 1000 mL of the 995 ppm CO2 standard in a gas bag. This should result in a decrease of 12CO2 and 13CO2 concentrations of only 2.9 ppm and 0.032 ppm, respectively. However, the observed decreases in the 12CO2 and 13CO2 concentrations are much greater than the predicted dilution, 45 ppm for 12CO2 and 0.6 ppm for 13CO2. This is on the order of 15 times greater than the predicted dilution. Since the decrease in CO2 concentration cannot be explained by dilution when H2S is added, we propose that there is a compound reaction which consumes CO2 with the addition of H2S to the gas mixture.

Sample analyses with H2S concentrations from 1 to 500 ppb show a linear interference of -1 ‰ for every 23 ppb H2S added (Fig. 5). The interference was successfully eliminated by reacting samples with copper. Figure 6 shows a larger range of samples from 1 ppb to 20 000 ppb. The higher H2S concentrations still show a linear interference, but the interference is smaller at -1 ‰ for every 37 ppb H2S. We believe that this discrepancy is a result of diluting the CO2 standard gas with larger volumes of H2S during sample preparation. The suite of samples from 1 to 500 ppb H2S had larger volumes of diluted H2S (1 ppm) added than the sample suite from 500 to 20 000 ppb. The larger dilutions resulted in lower CO2 concentrations, suggesting that the H2S interference also depends on CO2 concentration. Hence, we ran a further series of experiments to examine this effect.

The set of experiments performed at a range of CO2 concentrations (500 to 3000 ppm CO2) revealed that the H2S interference also depends strongly on the CO2 concentration (Fig. 7). The interference from H2S is much smaller at high CO2 concentrations and is quite large at atmospheric concentrations. For example, an interference of -1 ‰ resulted from the addition of 21 ppb H2S at 500 ppm CO2, whereas at 3000 ppm CO2 an interference of -1 ‰ required the addition of 154 ppb H2S. Thus, the H2S interference is also dependent on the CO2 concentration of the sample. During experiments performed at a fixed H2S concentration, it was found that the H2S interference with δ13C was inversely proportional to the CO2 concentration of the sample. Figure 8 illustrates the variation of the H2S interference at different CO2 concentrations and shows this inverse relationship between CO2 concentration and H2S interference.

Carbon isotopic measurements of CO2 using cavity ring-down spectroscopy have a clear interference in the presence of H2S that is dependent on both the H2S and CO2 concentrations. At lower CO2 concentrations, the H2S interference was more pronounced due to the relatively higher proportions of H2S contained within the sample. This may explain the discrepancy between the slopes of Figs. 5 and 6, where there was more dilution of the CO2 standard gas at lower H2S concentration (Fig. 5) than at higher H2S concentration (Fig. 6), resulting in a larger H2S/CO2 ratio in samples with lower CO2 concentrations.

The H2S interference with the G1101-i CRDS is an inherent property of the spectral lines that are fitted to determine the 12CO2 and 13CO2 concentrations (Fig. 9). The specific spectral lines used in the Picarro G1101-i were chosen to avoid overlapping ambient levels of common gas species encountered in atmospheric air (i.e., H2O, CH4, NH3, etc.). In the case of water vapor for example, where it is not possible to choose CO2 lines that are free from overlap, the system measures and corrects for such species to the extent that they interfere with either the 12CO2 or 13CO2 spectral features. For H2S specifically, the chosen spectral lines avoid the strongly absorbing H2S spectral lines, but there are weaker lines that partially overlap with both spectral features of the CO2 used in the system. At typical ambient levels of H2S for which the spectroscopy of the G1101-i was designed, these weak lines have no measurable effect on the reported CO2 concentrations or carbon isotope ratio. However, at elevated levels, they begin to cause the observed measurement bias.

Since the isotope ratio measurements in CRDS use ratios of the absorption peaks of the two spectral lines of the CO2 isotopologues, it is the relative concentration of H2S to the CO2 that determines the effect of the H2S on the isotope ratio (Fig. 9). The more pronounced effect of H2S on the isotope ratio at lower CO2 concentrations is due to a weak H2S spectral line slightly overlapping the 13CO2 line such that when the ratio of CO2 to H2S concentration is low, the measured 13CO2 line will be more affected by the H2S since it makes up a larger proportion of the overall measured line shape. There is a similar overlapping H2S line near the 12CO2 peak that has a similar (but opposite sign) concentration-dependent effect on the reported 12CO2 concentration as compared to the 13CO2 concentration. The reason for the sign difference of these two H2S concentration-dependent effects is related to how the independent spectroscopic fitting algorithms used for each peak to calculate the isotopologue concentrations interpret the change in line shape imparted by the interfering H2S signal.

In addition to the H2S interference, there was an unanticipated decrease in both the 12CO2 and 13CO2 concentrations with the addition of H2S to the standard gas that could not be accounted for solely by dilution (Fig. 4). We propose that a reaction between CO2 and H2S is occurring to consume CO2 upon combination in the Tedlar^® bags. However, we have not directly measured any products; hence we are uncertain as to what reaction will be consuming these reactants at atmospheric conditions. Isotopic readings of our gas mixture indicate that the effects of any reactions are small compared to the effects of H2S, so the major concern for these and future experiments is the removal of H2S from all samples prior to analysis.

Although all H2S experiments conducted in this study use an older model (G1101-i) of the carbon isotope analyzer from Picarro, all subsequent models have maintained the same spectral lines for CO2, and their H2S performance is presumed to be equivalent. We have verified that no spectroscopic corrections for H2S have been applied to any model of the carbon isotopic analyzer, and as such the copper scrub proposed here is a simple and effective solution to the H2S interference for all current models. The operating lines of the instrument were chosen to minimize strong overlap of spectral lines from ambient levels of small molecules found in ambient air such as ammonia, water vapor, H2S etc., and as such the H2S interference only occurs at concentrations > 1 ppb. Normal atmospheric concentrations are much less than this amount, and no correction for the H2S overlap has previously been warranted. In non-atmospheric conditions, such as those on active volcanoes or sour gas plants, these concentrations are more common and H2S should be considered as an interferant.

Isotopic measurements using this particular implementation of CO2 spectroscopy in cavity ring-down spectrometers have a clear and quantifiable interference resulting from the presence of H2S in excess of a few parts per billion. Laboratory experiments using controlled amounts of H2S mixed with a CO2 gas of known concentration and isotopic composition show that the interference is linear and dependent on both the H2S and CO2 concentrations of the sample. The H2S interference arises as a result of the line choice for this type of spectrometer (Picarro^© G1101-i), which avoids interference with other common atmospheric species such as H2O, CH4, NH3, etc., but it has some small lines remaining in the range of H2S that causes the interference observed at high H2S concentrations. All models of the carbon analyzer from Picarro^© use the same spectral lines and, thus, are susceptible to the same type and magnitude of interference with H2S. The most practical approach to eliminating H2S interference when measuring the δ13C value is the use of a metal scrub, for example copper, to remove all H2S before the sample is run through the CRDS. Removing this interference is an important step to making real-time measurements of δ13C of CO2 with cavity ring-down spectrometers in environments with high sulfur concentrations, such as actively degassing volcanoes. Volcanoes have a range of CO2 concentrations (400–1 000 000 ppm), and the H2S interference is significant in the operational range of the CRDS (0–3000 ppm). Therefore, the most practical approach to eliminating the interference is with a simple scrub for all samples containing H2S in excess of 1 ppb.