2.1 Segmented flow coil equilibrator

The design of our SFCE is shown in Fig. 1. The SFCE is coupled to PTR-MS for measurement of methanol, acetone (2-propanone), acetaldehyde (ethanal), dimethyl sulfide (DMS), isoprene (2-methyl-1,3-butadiene), benzene, and toluene (methyl benzene). These gases cover a large range of solubilities (see Sect. 4.2.1), demonstrating the versatility of the SFCE. The main advantage of this equilibrator lies in its design. Briefly, the segmented flow allows for a large surface area for gas exchange, ample equilibration time, and thus a high degree of equilibration. The simple headspace and water separation system allows for rapid drainage of the sampled water as well as separation of the headspace from water without spray or droplet formation. This enables a fast response time. Due to the ease of changing the water sample intake and low water flow, the equilibrator can conveniently be used for both continuous underway and discrete measurements. The equilibrator is entirely made up of commercially available polytetrafluoroethylene (PTFE) tubing and fittings, which should minimise adsorptive loss and make the equilibrator relatively inexpensive and easy to replicate. The constant flow of water and smooth surfaces should also reduce biofouling and facilitate occasional cleaning.

Figure 1(a) Schematic of the segmented flow coil equilibrator coupled to PTR-MS. (b) Schematic of the jar that was used during the field testing presented here for air–water separation. All other aspects of the SFCE were the same for the two designs.

Download

In our setup, the SFCE takes approximately equal, continuous flows of high-purity zero air (100 cm³n min⁻¹, where the letter n indicates normalised mass flow at 0 ^∘C, 1 atm; controlled by a Bronkhorst mass-flow controller) and unfiltered seawater (100±5 cm³ min⁻¹, controlled by a peristaltic pump, Watson Marlow 120 S/DV with 8 cm long Pumpsil platinum-cured silicone tubing with 4.4 mm inner diameter). We used either ultra-low-VOC zero air (Praxair) scrubbed by a hydrocarbon trap or British Technical Council approved (BTCA) zero air (BOC) oxidised by a custom-made platinum catalyst (heated to 450 ^∘C) as the zero air carrier gas for the SFCE. Complete oxidation of VOCs in the custom-made PT catalyst has been demonstrated previously for both dry air and air that is fully saturated with water at 20 ^∘C (Yang and Fleming, 2019).

The seawater is pumped either from the ship's underway water system or, in the case of discrete measurements, via 900 cm³ glass sample bottles (Sect. 5.1). The carrier gas and water meet in a PTFE tee piece (4 mm inner diameter), which naturally leads to the formation of distinct segments of zero air and water. The segments travel through a coiled, 10 m long PTFE tube (outer diameter 6.35 mm, wall thickness 1.19 mm). Each segment of carrier gas or water is approximately 1.5 cm long, giving an approximate total surface of exchange of 82 cm² in the coil. The coil is immersed in a water bath kept at a constant temperature of 20 ^∘C. The residence time in the 10 m tube is approximately 0.6 min. Laboratory measurements indicate that regardless of the initial water temperature (0–25 ^∘C), the water exiting the equilibrator has a temperature of 20±1 ^∘C. Keeping the temperature essentially constant has the benefit of (i) simplifying calibrations and calculations of aqueous concentration and, (ii) in the case of cold high-latitude seawater samples, increasing the VOC signal in the headspace as warming to 20 ^∘C reduces the gas solubility. A rapid biological response to this warming is not expected in the segmented flow coil due to the very short residence time of 0.6 min.

In the initial design, after equilibration in the coiled segmented flow tube, the equilibrated air–water mixture is separated in a 200 cm³ PTFE jar (Savillex). Here, the sample gas travels towards the PTR-MS and the seawater drains away rapidly via a U-shaped drain (Fig. 1b). The U-shaped drain prevents intrusion of lab air and prevents the sample air from escaping via the water drain. We estimate a response time of about 2 min with the PTFE jar as the air–water separator. This is due to a combination of its sizable internal volume and the production of sea spray inside the jar due to falling droplets. The latter buffer the headspace to step changes in seawater concentration in the case of highly soluble gases. After the Arctic field deployment, the PTFE jar was found to slightly outgas some VOCs (see Sect. 4.1 for further information) and was replaced with a PTFE tee piece (Swagelok, outer diameter 12.7 mm, wall thickness 1.6 mm). This modification improved the system response time to less than 1 min by greatly reducing the volume of the air–water separator and allowing for a smooth separation of the equilibrated air–water mixture without droplet formation (see Sect. 4.3). Thus, the entire SFCE consists of readily available PTFE tubing and fittings routinely used for air sampling. See Fig. 1 for these two designs.

On the top end of the air–water separator, the humid headspace (100 cm³n min⁻¹) is diluted with dry zero air (20 cm³n min⁻¹, same as the carrier gas, controlled by another Bronkhorst mass-flow controller). This prevents condensation in the ∼2 m PTFE tubing between the equilibrator and the heated (80 ^∘C) inlet of the PTR-MS (Fig. 1). The SFCE system is operated at a slight over-pressure (approx. 0.024 mbar above atmospheric pressure) in order to reduce the likelihood of lab air contamination (e.g. due to leaks). A vent is installed upstream of the PTR-MS to avoid pressurising the detector. The vent flow is typically ∼20 cm³ min⁻¹ – the residual between the carrier gas flow (100 cm³n min⁻¹), the dilution flow (20 cm³n min⁻¹), and the PTR-MS intake flow (∼100 cm³ min⁻¹).

The entire SFCE system fits on a bench space of about 40 by 40 cm (see the Supplement). Importantly, the SFCE is designed such that the failure of an individual component does not result in a catastrophic over- or under-pressurisation of the system. For example, if the carrier gas is stopped (e.g. gas supply runs out), the PTR-MS simply measures lab air via the vent and the water is drained from the SFCE as usual. If the water flow from the underway sampling stops, the peristaltic pump will simply pump lab air into the equilibrator. These unexpected failures can be easily identified as lab air typically has much higher concentrations of VOCs than equilibrator headspace in a marine environment. If the PTR-MS fails, the headspace gas simply exits via the vent and/or the top of the U-shaped drain.

Due to the smooth surfaces and constant and complete water renewal, the equilibrator should not be very prone to biofouling. The lack of a membrane for gas exchange means that the degree of equilibration should not vary significantly even if there is some biofouling. To clean the SFCE if necessary, the seawater intake and the water drain pipe are connected to a 10 % HCl solution for 10 min. During this procedure, the PTR-MS is disconnected to sample lab air and the gas flow is stopped. A flow of HCl thus covers all the parts of the equilibrator that are normally exposed to seawater. To resume measurement of ambient seawater, the flow of HCl is stopped and the carrier gas flow is started to drain the HCl safely into the recirculated solution. The equilibrator is typically rinsed with seawater before resuming measurement.

2.2 PTR-MS operation

Equilibrator headspace mixing ratios were initially computed using compound-specific rate constants of the reaction between the VOC in question and the hydronium ions in the drift tube of the mass spectrometer (Yang et al., 2013a; Zhao and Zhang, 2004). Pre- and post-cruise dynamic gas-phase calibrations using a gravimetrically prepared VOC gas standard (Apel-Riemer Environmental Inc., Miami, Florida, USA; nominal volume mixing ratio of 500 ppbv for acetaldehyde, methanol, acetone, isoprene DMS, benzene, toluene) and two Bronkhorst mass-flow controllers agree within 15 % of the computed mixing ratios for all VOCs except isoprene. Isoprene was found to fragment significantly: 17 % of the isoprene molecules are found at the primary ion (m∕z 69) and 30 % and 53 % were found at the fragment ion m∕z 41 and 39, respectively. This is in general agreement with Schwarz et al. (2009). This fragmentation ratio increases with increasing drift tube voltage (see the Supplement). It is possible that some of the mass 79 measured here contains a contribution from fragmenting toluene. However, because the gas standard contains both compounds, it is not straightforward to evaluate the magnitude of this interference.

The PTR-MS measurements can be affected by humidity. Our use of a dilution flow lowers the humidity in the sample gas by 20 % and thereby reduces the measurement sensitivity to humidity. To check for the effect of humidity on the PTR-MS measurement, gas calibrations were carried out at different humidities using three Bronkhorst mass-flow controllers. To produce carrier gas air at different humidities, a flow of moist air saturated in humidity at 20 ^∘C is generated by passing zero air through a wetted SFCE and diluted by varying amounts of dry zero air directly from a gas cylinder. This mixture is scrubbed by the PT catalyst (which does not appear to change the humidity levels) and then added to the flow of VOC gas standard.

The signal of most VOCs monitored is independent of the sample humidity. However, isoprene, benzene, and toluene show a weak humidity dependence in their gas-phase calibrations. Changing the humidity in sample air from completely dry to nearly saturated in humidity at 20 ^∘C, the abundance of the isoprene primary ion increases by 33 % (see the Supplement). This is because the hydronium water clusters do not cause isoprene fragmentation upon ionisation (Schwarz et al., 2009). The opposite is observed with benzene and toluene: primary ion abundance was found to decrease by 12 % and 18 %, respectively, over same humidity range. This is because hydronium water clusters have a lower ionisation energy, thus ionising benzene and toluene less effectively (de Gouw and Warneke, 2007; Warneke et al., 2001); see the Supplement. The humidity-dependant slopes from the gas-phase calibrations were used to correct the measured equilibrator headspace mixing ratios. For a more detailed discussion on the settings of the PTR-MS during deployment, the computation of VOC mixing ratios in the PTR-MS, and the effects of humidity on the signal amplitude and background, please see the Supplement.

4.1 Estimation of backgrounds for seawater VOC measurements

The ideal background, or blank, for seawater VOC measurements would be VOC-free seawater. However, we have been unable to generate or obtain seawater that is free of methanol, acetone, or acetaldehyde due to the high solubility and ubiquity of these gases. Additionally, it is debatable whether any natural seawater may be free of methanol, acetone, or acetaldehyde as there is a lack of knowledge about the cycling of these compounds. The choice of background is most important for these soluble OVOCs as the ratio of background to signal can be quite high and the background can be variable. For example, for acetone the average signal-to-background ratio during the 3-week Arctic field campaign (Sect. 5) was 1.62 with a background standard deviation of 26 %. Below, we discuss three different approaches to estimating the background of seawater VOC measurements.

First, direct measurement of zero air (i.e. bypassing the SFCE) was used to track any drift in the internal PTR-MS background. This simple method of deriving a blank was also used by Yang et al. (2013b). This approach might not provide the most representative of backgrounds for all seawater VOCs because (i) zero air has a much lower humidity than the equilibrator headspace, and humidity could affect the backgrounds of some of the measured compounds (de Gouw and Warneke, 2007) (see the Supplement); and (ii) it does not account for any possible contamination within the equilibrator.

Second, at every sampling station, bottom water (i.e. the deepest water collected by the rosette, which was between 290 and 1700 m, well below the mixed layer) was measured. For some VOCs that are thought to be only produced in the surface ocean and rapidly consumed at depth (such as DMS), it might be expected that their concentrations in deep water would be close to zero. However, there is insufficient field data to know whether this is the case for all the VOCs monitored here. Measurements of methanol and acetone in the north Atlantic show that their concentrations do decrease below the mixed layer (Williams et al., 2004) but do not necessarily go to zero. Similarly, depth profile measurements showed acetone concentrations near the detection limit (0.3 nmol⁻¹ dm³) at 200 m (Beale et al., 2013), while methanol and acetaldehyde concentrations at depth did not decrease as rapidly. We note that for these measurements a flow of dry nitrogen was used as a background, which may be an underestimation of the true system blank (see the Supplement). The chief advantage of using the bottom water measurement as the background is that these samples after equilibration have the same properties (humidity, temperature, exposure to the equilibrator, and collection protocol) as the surface water samples.

The final blank we determined was the “wet equilibrator” blank. This consisted of stopping the water flow into the equilibrator and purging the wet equilibrator (that had been coated with bottom seawater) with zero air for 20 min. During this blank measurement, humidity in the headspace remained constant as small water droplets remained inside the coil and were not substantially dried by the zero air. During the Arctic cruise, the wet equilibrator blank consistently resulted in the lowest reading on the PTR-MS for all VOCs except methanol and acetone (as a result of a contamination, which is discussed below). Thus, in practice the wet equilibrator blank seems to be the best surrogate for a “true” water blank for almost all VOCs measured here. During the wet equilibrator blank, the bottom of the PTFE jar contains approximately 5 mL of residual seawater that is not readily flushed out; in the case of the tee, there is essentially no residual seawater in the air–water separator. During normal seawater measurement, the residence time of zero air and seawater in the equilibrator is approximately 0.6 min. The residence time of zero air during a wet equilibrator blank measurement is about 1.2 min.

The 200 cm³ headspace jar used for separating the headspace from the seawater after equilibration is made from PTFE, which should be inert. However, we found that the empty equilibrator blanks of methanol and acetone were about 0.2 ppbv higher than their deepwater blanks during a cruise. The most plausible explanation for this seems to be outgassing or permeation of methanol and acetone through the walls of the PTFE jar itself, which is suppressed during the water measurement. During the 3-week field deployment, we observed a strong correlation between zero air and bottom water measurements (R²=0.92 methanol, R²=0.69 acetone), suggesting the concentrations of these VOCs at depth are either uniform or very small. Because of the contamination described above, we report seawater acetone and methanol concentrations from this cruise using bottom water as the background; these concentrations should thus be viewed as possible lower-limit estimates. After the cruise, we replaced the PTFE jar with a PTFE tee fitting (Swagelok) and this contamination greatly decreased.

We find that ultrapure Milli-Q water or bottom seawater water is typically free of the less soluble compounds such as DMS, toluene, benzene, and isoprene. This is confirmed by good agreement between the wet equilibrator blanks and the Milli-Q and bottom seawater measurements. This agreement also suggests that our system is not affecting the measured concentrations of these compounds through cell rupturing. The concentrations of methanol, acetone, and acetaldehyde measured in the Milli-Q water during the Arctic cruise were much higher than those in seawater and were highly variable (see the Supplement). Similarly, we found that tap water or bottled drinking water is typically not free of methanol, acetone, and acetaldehyde, likely due to slow leakage of these compounds from the pipes, tubes, and/or containers.

4.2 Estimation of equilibration efficiency

4.2.1 Experimental setup

As a brief recap, for gases that appear to fully equilibrate in the SFCE, seawater concentrations (C_w) are computed from the equilibrator headspace mixing ratio (C_a) using the dimensionless air-over-water Henry solubility constant (H) (Sander, 2015). Headspace equilibrator mixing ratios are converted from ppbv to nmol dm⁻³ using the ideal gas law, and the dilution of equilibrator headspace is accounted for by multiplying measured equilibrator mixing ratios by 1.2 (Sect. 3, Eq. 2).

Where possible, values for Henry solubility recommended by Burkholder et al. (2015) were used for this calculation as those were deemed most reliable. These values represent freshwater solubilities and are converted to seawater solubilities by accounting for salting-out effects (Johnson, 2010). Values of the dimensionless Henry solubility (water over gas) in freshwater and seawater as well as the references for the solubility are displayed in Table 1.

Table 1Dimensionless Henry solubility values (water over gas) in freshwater and seawater used to compute dissolved gas concentrations.

Download Print Version | Download XLSX

Two methods are used to assess the equilibration efficiency of the SFCE: evasion and invasion. In evasion experiments, liquid standards of methanol, acetone, and acetaldehyde were prepared by serial dilution of the pure solvent in the same batch of Milli-Q water. Aliquots of pure, undiluted methanol (Uvasol for spectroscopy) and acetone (high-performance liquid chromatography standard) were dispensed using volumetric pipettes. A 1 cm³ volumetric flask was used to aliquot pure acetaldehyde ($\ge =\mathrm{99.5}$ %, ACS Reagent). Subsequent dilutions utilised a volumetric pipette and volumetric flask to prepare liquid standards ranging from 3 to 30 nmol⁻¹ dm³ for acetone and acetaldehyde and 30 to 300 nmol⁻¹ dm³ for methanol. Liquid standards of isoprene and DMS were prepared gravimetrically airtight each day. A syringe pump (New Era Pump Systems) was used to dynamically dilute DMS and isoprene standards in a flow of Milli-Q water. This yielded DMS standards of up to 7 nmol⁻¹ dm³ and isoprene standards of up to 2 nmol⁻¹ dm³. For this calibration, the flow rate of Milli-Q water is measured at the drain. In evasion calibrations, a solubility-dependent fraction of dissolved VOCs is transferred into the gas phase during the equilibration process. Thus, the final dissolved concentration will be somewhat lower than the initial concentrations. To account for the removal of the fractions of these gases from the seawater during equilibration, a purging factor (PF) based on mass conservation is applied. The PF is the ratio between the dissolved gas concentration before and after complete equilibration in the coil. The derivation of this compound-specific purging factor is presented in the Supplement. At equal airflow and water flow rates, it simplifies to

$$\begin{array}{}\text{(4)}& \mathrm{PF}={\displaystyle \frac{C_{\mathrm{w}}\left(\text{before equilibration}\right)}{C_{\mathrm{w}}\left(\text{after equilibration}\right)}}={\displaystyle \frac{\mathrm{1}}{H}}+\mathrm{1}.\end{array}$$

The precision of the purging factor depends on the precision of the solubility measurement. Since solubilities are reported in this paper to two significant figures, the purging factor is reported here to two significant figures as well. For freshwater, computed purging factors assuming full equilibration and equal zero airflow and water flows are 1.00 for methanol, 1.00 for acetone, 1.00 for acetaldehyde, 1.06 for DMS, 1.18 for benzene, 1.21 for toluene, and 2.57 for isoprene. The same computation in seawater gives the following purging factors: 1.00 for methanol, 1.00 for acetone, 1.00 for acetaldehyde, 1.08 for DMS, 1.22 for benzene, 1.26 for toluene, and 2.96 for isoprene. We see that PF varies from being insignificant (with a value of 1) for highly soluble VOCs to quite large (≫1) for the sparingly soluble gases. To compute the expected headspace mixing ratio during the evasion calibrations assuming full equilibration, the known waterside concentrations are divided by the purging factor. Then the equilibration efficiency is calculated as the measured headspace mixing ratio divided by the expected headspace mixing ratio assuming full equilibration.

We also tested the absorption of gaseous VOCs into the water phase (i.e. invasion). This is especially useful for gases such as benzene and toluene, as we were unable to use liquid standards for these compounds due to their toxicity. During invasion experiments, a flow of VOC gas standard was diluted to varying degrees with VOC-free zero air using mass-flow controllers. This diluted VOC gas standard was then equilibrated with essentially VOC-free Milli-Q water. The assumption of no VOCs in the initial water is reasonable as we used relatively high carrier gas VOC mixing ratios (up to 50 ppbv) and the Milli-Q water is essentially free of DMS, benzene, toluene, and isoprene (Sect. 4.1). The headspace equilibrator mixing ratio is measured and compared to the expected mixing ratio at full equilibration. Calculation of the expected mixing ratios at full equilibration during invasion experiments is presented in the Supplement. For invasion experiments, the equilibration efficiency is calculated as the observed change in mixing ratio over the expected change in mixing ratio.

Ideally, we want to maintain a stable equilibration efficiency of 100 %. This would maximise the signal-to-noise ratio and minimise the measurement uncertainty. This may also reduce the need for frequent calibrations. Robust SFCE calibrations were not performed during the field testing presented here due to logistical constraints. Post-cruise calibrations were carried out on an approximately weekly basis over several weeks, intended to be representative of the duration of the cruise. These calibrations were used to assess the equilibration efficiency of SFCE and uncertainties therein.

4.2.2 Equilibration efficiency of DMS and isoprene

Prior experimentations with a similar setup suggest that the 10 m segmented flow tube presented here is at least a factor of 2 longer than required for full equilibration of DMS (Blomquist et al., 2017). Hence, we expect the soluble VOCs (methanol, acetone, acetaldehyde, DMS) to fully equilibrate due to their higher solubility (Liss and Slater, 1974). Figure 2 shows calibration curves for DMS and isoprene using liquid standards (i.e. evasion) over several weeks. The calibration curve for DMS suggests full equilibration (Fig. 2a); a ∼5 % underestimation of DMS in the mean is within the uncertainty of the solubility. The DMS calibration curves show very little noise and low weekly variability (±4 % SD), suggesting that the SFCE-PTR-MS setup is very stable. The calibration curve for isoprene suggests 62 % equilibration efficiency (Fig. 2b). A greater variability on a weekly basis (±14 % SD) is observed in the isoprene calibration curves, likely due to incomplete (and hence less consistent) equilibration.

Figure 2Evasion calibration curves for DMS (a) and isoprene (b). The average slope of the experimental calibration curve was found to be 1.77 ppbv nmol 1 dm³ ±4 % and 9.12 ppbv nmol 1 dm³ ±14 % for DMS and isoprene, respectively; errors represent the standard deviation over a 3-week period. The full equilibration slope was computed to be 1.87 ppbv (nmol dm⁻³)⁻¹ and 14.69 ppbv (nmol 1 dm⁻³)⁻¹ for DMS and isoprene, respectively (using Burkholder et al., 2015, for DMS solubility and Karl et al., 2003, solubility with Leng et al., 2013, temperature dependence). This suggests approximately 100 % and 62 % equilibration efficiency for DMS and isoprene, respectively. Error bars are too small to display, but the noise associated with the measurement was found to be 0.0069 and 0.00058 ppbv for DMS and isoprene, respectively. This was calculated as the SD of 10 consecutive water blank measurements.

Download

Results from the invasion experiments are displayed in Fig. 3 and confirmed that the equilibrator fully equilibrates for DMS, as the measured and expected gas-phase mixing ratios of DMS match. The equilibration efficiencies of the less soluble gases benzene and toluene were found to be 94±1 % and 95±2 %, respectively. The 5 % difference is within the uncertainty of the solubility of these compounds, and hence for the computation of seawater concentrations we assume that these compounds fully equilibrate. This invasion experiment was also performed for the highly soluble OVOCs (methanol, acetone, and acetaldehyde). These gases were found to be entirely absorbed into the water phase, essentially leading to noise in the measurements of headspace mixing ratios.

Figure 3Invasion calibration curves for benzene (a), toluene (b), DMS (c), and isoprene (d) for which a known amount of standard gas is added to the zero air carrier gas while measuring VOC-free Milli-Q water. Error bars were too small to display, but the noise associated with the measurement was found to be 0.0069 and 0.00058 ppbv for DMS and isoprene, respectively, and 0.043 and 0.042 ppbv for benzene and toluene, respectively. This was calculated as the SD of 10 consecutive water blank measurements. A 1:1 line is included to illustrate the role of the water phase in absorbing these compounds.

Download

The equilibration efficiency of isoprene (the least soluble compound that we measure by far) of 69 % from invasion is similar to that determined in the evasion experiments (62 %) if we use the isoprene solubility from Karl et al. (2003) and the temperature dependence from Leng et al. (2013). We note that there is a large range in the values for isoprene solubility in the literature. Using the solubility values from Yaws and Yang (1992), Leng et al. (2013), or Mochalski et al. (2011) would result in a large and unexpected discrepancy in the equilibration efficiency of isoprene between the evasion and invasion experiments.

Figure 4Evasion calibrations of OVOCs. Displayed are the average experimentally determined slopes of 14 calibration curves of methanol (a) and acetone (b) and 11 calibration curves of acetaldehyde (c). These calibrations suggest possibly lower solubility of these compounds compared to literature values. Shaded area indicates 1σ standard deviation of the variance in the slope during this 3-month period. Average experimentally determined calibration slopes for methanol, acetone, and acetaldehyde were 0.00786±0.00115, 0.0469±0.0145, and 0.0743±0.0190 ppbv nmol⁻¹ dm³. Plotted along this are the predicted slopes using all experimentally determined solubilities as listed in Sander (2015). The recommended solubility by Burkholder et al. (2015) is plotted as a solid thick line in dark blue. The key to the figure is listed in a table in the Supplement, listing the in-figure reference followed by the dimensionless water-over-air Henry solubility in Milli-Q water at 20 ^∘C and the predicted slope using the listed experimentally determined solubility. For full references of the cited solubilities, please refer to Sander (2015).

Download

4.2.3 Equilibration efficiency of OVOCs

Both theoretical considerations (e.g. Liss and Slater, 1974) and experiments with a varying airflow : water flow ratio (Sect. 4.25) indicate that the OVOCs should fully equilibrate within the SFCE. The average slope of 11 calibration curves for acetaldehyde and 14 calibration curves for methanol and acetone over a 3-month period are shown in Fig. 4. Results are compared to the expected mixing ratio computed using every experimentally determined solubility listed in the compilation by Sander (2015). The measurements are also compared to the solubility recommended by Burkholder et al. (2015), which was chosen as a critical synthesis of published solubilities.

The experimentally determined calibration slopes for OVOCs are linear (typical R² above 0.95). However, they are on average about 1.5 times higher than the solubilities recommended by Burkholder et al. (2015). Nevertheless, these experimental mean slopes are within the range of published solubility values. The relative standard deviation associated with the OVOC calibration curves (∼25 %) is much larger than that in the DMS calibration curves (4 %), with the latter an indication for the stability of the PTR-MS–equilibrator system. On a weekly basis, the calibration curves of individual OVOCs correlate with each other, and these OVOCs were diluted together from pure reagents. This suggests that most of the observed variability in OVOC calibration from week to week might be due to errors or contamination in the serial dilution procedure. In order to ensure consistency with previous equilibrator setups (Kameyama et al., 2010), in this paper we report our dissolved gas concentrations using the recommended solubilities from Burkholder et al. (2015). Using the mean of our experimental dissolved gas calibrations would decrease the computed OVOC seawater concentrations by approximately 50 %.

Figure 5Relative signal as a function of water flow into the equilibrator. Error bars represent random error propagation with the initial error determined from the standard deviation of 10 consecutive 6 min blank measurements.

Download

4.2.4 Measurement sensitivity toward airflow : water flow ratio

Air and water at equal flow rates of 100 cm³ at 20 ^∘C were chosen to allow for a sufficiently long equilibration time, large surface area for exchange, and high signal while satisfying the airflow requirements of the PTR-MS. They were also chosen such that the stripping of the soluble compounds from the water phase during equilibration would be small (i.e. purging factor near 1). Additionally, the use of equal flows of air and water simplifies the calculation of dissolved gas concentrations. The water flow was not routinely monitored during the Arctic deployment and decreased by up to 20 % due to ageing of the peristaltic pump tubing. This could influence our measurement through at least (i) the equilibration time and hence the efficiency in the coil as well as (ii) the purging factor. To investigate the influence of these competing factors on the signal, an experiment was performed after the cruise measuring the same solution of liquid standard at different water flows into the equilibrator while keeping the airflow constant (Fig. 5).

The signals of acetone, acetaldehyde, and DMS were found to be independent of the water flow into the equilibrator. These results provide strong experimental evidence that (i) VOCs with solubilities greater than or similar to DMS do equilibrate in the coil, and (ii) the gas flow does not remove a large fraction of these gases from the water phase during the equilibration process (i.e. purging). In contrast, the signal of isoprene was found to decline with decreasing water flow. As the water flow is decreased during this experiment, the purging factor increased at a comparable rate to the decrease in the isoprene headspace mixing ratios. This suggests that the change in purging factor is largely responsible for the change in the isoprene signal (Fig. 5). Consequently, compared to the soluble VOCs, for isoprene there is an additional uncertainty of ∼20 % that is due to the variable water flow during the cruise (see Sect. 4.4).

Figure 6Instrument response to step changes in Milli-Q water OVOC concentration (step size: 20 nmol dm⁻³ acetone, 20 nmol dm⁻³ acetaldehyde, 200 nmol dm⁻³ methanol).

Download

4.3 Measurement response time

A series of liquid standards containing 20 nmol dm⁻³ acetone, 20 nmol dm⁻³ acetaldehyde, and 200 nmol dm⁻³ methanol (Fig. 6) were analysed in order to determine the response and delay time of the equilibrator and to test for any possible memory effect due to wall adsorption and desorption effects. Discrete samples were swapped over rapidly and the water flow into the equilibrator was stopped briefly to avoid interference with the measurement.

The residence time (0.6 min) in the equilibrator segmented flow tube was calculated from the flow of air and water into the equilibrator and the volume of the segmented flow tube. The response time (e-folding time) of the equilibrator response to the step change was estimated using the 8 s PTR-MS measurements to be 35, 33, and 33 s for methanol, acetone, and acetaldehyde, respectively. Thus, the response time appears to be independent of the solubility and comparable to the residence time in the SFCE. The rapid rate of increase and decrease in OVOC concentrations during this experiment also suggests that there was little “carry over” or memory effect. Whilst the response time of the SFCE is less than 1 min, measured equilibrator headspace mixing ratios are typically averaged over 6 min for both underway measurements and discrete measurements to reduce random noise and improve the precision of the measurement.

4.4 Measurement precision and limit of detection

The analytical precision and the limit of detection (LOD) of this system are partly dictated by the noise of the PTR-MS measurement. This in turn depends on the dwell time of the detector at a given mass and thus the time the data are averaged over. Additionally, the measurement precision strongly depends on the gas solubility. For isoprene, the analytical precision additionally depends on the variability in water flow. We compute the analytical precision as the standard deviation (1σ) of 10 consecutive 6 min wet equilibrator blank measurements, which is then converted to a dissolved gas concentration using Eq. (3) for isoprene and Eq. (2) for the other VOCs. The precision is therefore averaged over 70 measurement cycles with a dwell time at each mass of 500 ms, giving an effective dwell time of 35 s. The LOD is defined as 3 σ. The resulting measurement noise and limit of detection for each compound are displayed in Table 2 for 6 min averaged data. These values should approximately halve if the data are averaged over 30 min intervals instead.

Table 2Analytical precision and limit of detection of the seawater VOC measurements.

Download Print Version | Download XLSX

In the case of typically 900 mL discrete samples (see Sect. 5), the measurement time is limited by the water flow rate. A larger water volume may improve the analytical precision by allowing for a longer sampling and averaging time. This is especially relevant for the most soluble compounds, including methanol and acetone.