Experimental investigation of the scavenging of gaseous mercury by sea salt aerosol
- First Online:
- Cite this article as:
- Malcolm, E.G., Ford, A.C., Redding, T.A. et al. J Atmos Chem (2009) 63: 221. doi:10.1007/s10874-010-9165-y
- 111 Views
Sea salt aerosol may be an important sink for reactive gaseous mercury (RGM) in the marine boundary layer, reducing ambient RGM concentrations and transferring the mercury (Hg) to the oceans and coastal ecosystems. The goal of this study was to determine the affinity of gaseous mercury for sea salt aerosol (SSA) by conducting adsorption experiments with sea salt-coated sampling denuders. In the first set of experiments, ambient outdoor air was passed through denuders coated with either KCl, as in the widely accepted method to sample RGM, or with NaCl, a primary component of sea salt aerosols. On the one sampling day in which RGM was above the MDL, the NaCl coated denuders removed Hg from the ambient air, equivalent to 87% of the RGM in the air (as determined by KCl denuders). For the second set of experiments HgCl2 generated in the laboratory was passed through denuders coated with KCl and either NaCl or sea salt. The NaCl denuders collected an average of 99 ± 16% of the mercury that the KCl denuders collected. Newly coated sea salt denuders collected 88 ± 17% of the amount of mercury that the KCl denuders collected, but interestingly the sea salt denuders capacity decreased with repeated use. These experiments demonstrate that HgCl2, a major component of RGM has a strong affinity for NaCl and sea salt and is therefore likely to be scavenged by SSA. This study adds to the growing evidence that RGM is scavenged by sea salt aerosols and therefore more quickly deposited to the ocean and coastal environment.
KeywordsGas-particle partitioningReactive gaseous mercuryAtmospheric mercuryParticulate mercuryMarine boundary layerSodium chloride
Pioneering mercury research in lacustrine and forested watersheds demonstrated that mercury could be transported through the atmosphere on regional and global scales, affecting even remote locations. These studies led to the realization that an understanding of the atmospheric chemistry of mercury is necessary for predicting its atmospheric transport and deposition. In the last decade, many researchers have shifted attention to coastal and open ocean environments. What they have found is a unique and dynamic environment. The chemical cycling of atmospheric mercury is greatly influenced by the chemistry of the marine boundary layer (MBL), which includes many reactive halogen compounds. The halogen gas bromine, along with other photochemically produced oxidants (Lin and Pehkonen 1999; Seigneur and Lohman 2008; Selin 2009), rapidly convert some of the abundant, long lived elemental gaseous mercury (Hgg0) to oxidized gaseous forms of mercury, collectively known as reactive gaseous mercury (RGM). RGM is then quickly removed from the atmosphere through wet and dry deposition processes. Several recent modeling studies, field measurements, and laboratory experiments have begun to elucidate these processes and their importance for the local and global transport and deposition of mercury (For example see: Sprovieri and Pirrone 2008; Holmes et al. 2009; Engle et al. 2008). There is an additional importance to understanding mercury (Hg) cycling in the MBL because many people throughout the world are exposed to mercury through consumption of marine fish. The full impact of anthropogenic atmospheric emissions on the concentrations of mercury in seafood is just beginning to be understood (Sunderland et al. 2009; Kraepiel et al. 2003). This study adds to the growing evidence that RGM is scavenged by sea salt aerosols and therefore more quickly deposited to the ocean and coastal environment.
The chemical form of mercury affects its transport and deposition, so a complete understanding of atmospheric mercury’s chemical and physical transformations is crucial to predicting the fate of mercury emissions. Three major forms of mercury exist in the atmosphere: elemental gaseous (Hg(g)0), reactive gaseous (RGM or Hg(g)II), and particulate phase (Hg(p)). Hg(g)0 typically comprises over 95% of Hg in the atmosphere and is relatively inert, with a lifetime estimated to be 6 months to 1 year (Lin and Pehkonen 1999). Consequently, Hg(g)0 can be distributed on a global scale. RGM, on the other hand, is water-soluble and reactive with a high dry deposition velocity, causing it to be deposited to ecosystems close to its source (Lin and Pehkonen 1999; Schroeder and Munthe 1998). RGM is operationally defined, based on widely used sampling techniques; however, the separate chemical components of RGM have not been individually measured in the field. They are expected to include Hg(II) gases such as HgCl2, HgBr2 and HgO. Hg(p) is also readily deposited by wet or dry deposition mechanisms, dependent on the size and chemical properties of the aerosol particle.
Open ocean and coastal environments have high concentrations of sea salt aerosols (SSA), which are released from the ocean surface (Lewis and Schwartz 2004). Owing to high humidity in coastal environments and their hygroscopic nature, sea salt particles are typically deliquesced, having an associated aqueous phase that could facilitate mass transfer of RGM. Soluble gases like RGM preferentially partition into the aqueous phase, as evident by the Henry’s Law Coefficient of HgCl2 (3.7 × 10-5 Pa-m3/mol). In contrast, the Henry’s Law Coefficient for the less soluble Hg(g)0 is 729 Pa-m3/mol. Consequently, RGM is expected to partition more rapidly into SSA. Based on the reactive nature of RGM gases, we predict that RGM has a large mass accommodation coefficient for deliquesced or dry sea salt particles, so that a large fraction of the RGM molecules striking a SSA will be captured and dissolve into the SSA. Through this process, we expect that SSA is an important sink for RGM in the marine boundary layer, significantly reducing ambient RGM concentrations. Such scavenging of Hg gases would cause Hg to be rapidly cycled at the atmosphere-ocean interface from gaseous, to aerosol, to oceanic forms. Deposition of aerosol enriched in Hg via this process may constitute a significant global mercury flux to the oceans. This process may also increase local deposition in coastal cities where there are large sources of anthropogenic gaseous mercury and may also enhance deposition to manmade and natural surfaces that are coated with sea spray.
As artifact-free measurements of Hg(p) are difficult to obtain, much of what we know about the gas-particle partitioning of mercury comes from models and a select few field and laboratory studies. Modeling studies by Pirrone et al. (2000) and Hedgecock and Pirrone (2001) suggest that Hg(p) with high chloride content may play a role in oxidizing mercury, and then releasing HgCl2(g) to the atmosphere. A measurement study in Florida (Malcolm et al. 2003) suggested that some form of gaseous mercury may be incorporated into SSA, increasing the transfer of mercury from the atmosphere to ocean. Mercury concentrations in SSA were higher than could be explained by mercury in seawater alone. Calculations based on major cations in the aerosol revealed that Hg(p) concentrations were 3000× greater than expected if seawater was the only source of the Hg to the aerosols. A hypothesized explanation for these observations was that mass transfer of RGM to existing SSA comprised most of the particulate Hg concentration. At sites along the US Gulf Coast, Engle et al. (2008) observed Hg bound to coarse aerosol, and hypothesized that some of this was Hg bound to sea salt aerosol.
Selin et al. (2007) found that their global mercury model could not reproduce the concentration and diurnal behavior of RGM over the Pacific Ocean without including uptake of RGM by SSA. Without this sink, the modeled RGM peak concentrations were too high and did not decrease quickly enough in the afternoon. In a laboratory study of gas-particle partitioning, Rutter and Schauer (2007) found that HgCl2(g) preferentially partitions to dry NaCl particles. Reactive gaseous mercury compounds also preferentially partitioned to particles of sodium nitrate and potassium chloride. Partitioning of reactive mercury shifted to the gas phase for ammonium sulfate and organic aerosols. Holmes et al. (2009) used a chemical box model to investigate the production and destruction of RGM in the MBL. They predicted that under typical MBL conditions about 85% of total HgII is partitioned into SSA at equilibrium. Deposition of these aerosols accounted for 65-90% of the total RGM loss for modeling scenarios over the Pacific and Atlantic Ocean.
The aim of this study was to determine the affinity of gaseous mercury for sea salt aerosol by conducting adsorption experiments with sodium chloride and sea salt-coated sampling surfaces. Adsorption of ambient RGM, HgCl2, and Hg(g)0 were evaluated.
For these experiments, denuders were coated with either sodium chloride or seawater, to simulate SSA. Air containing mercury was passed over the denuders’ coated surfaces and adsorption by the salt was quantified. Experiments were done using both ambient, outdoor air and a N2 carrier gas with HgCl2(g) addition. Several other RGM gases in addition to HgCl2 have been predicted to occur in the atmosphere. HgCl2 is often chosen as a representative species for laboratory experiments (Landis et al., 2002; Rutter and Schauer 2007).
Diffusion denuders are a common sampling device used when reactive gases must be segregated from aerosols during collection. They typically consist of single or concentric tubes, whose inner surfaces are coated with a sorbent or reactive trapping medium for the sampling gas. Large particles are removed upstream of the denuder with cyclones or impaction inlets, and the air stream is passed through the tube at a flow rate sufficient to ensure laminar flow. This prevents small particles from impacting on the coating surface, but allows the gas of interest to diffuse to and be retained by the coating. Thus by varying the denuder coating and geometry, denuders have been designed which selectively collect gases of interest, and minimize collection of particles and unwanted gases (Appel 1993).
Potassium chloride (KCl) coated quartz annular denuders have been shown in the laboratory to be an effective method for selectively collecting RGM without collection of Hg(g)0. Landis et al. (2002) demonstrated that their capture efficiency for HgCl2 was greater than 97%. The denuders chosen for this study are identical to those used in Landis et al. (2002) (manufactured by URG Corporation). They are annular denuders and consist of an outer tube that contains an inner tube, or annulus (Landis et al. 2002). The inside of the outer tube and outside of the annulus are coated with KCl salts, so that when air is passed through the denuder, the RGM diffuses to the KCl coating. The RGM is then thermally desorbed for quantification. For these experiments, denuders were not only coated with the standard KCl solution, but also with sodium chloride (NaCl) solution and sea water.
The NaCl and sea water coatings were chosen to chemically simulate SSA. The chemical composition of newly formed SSA is similar to that of bulk seawater, with the following constituting most of the salinity: chloride (55.04%), sodium (30.61%), sulfate (7.68%), magnesium (3.69%), calcium (1.16%) and potassium (1.10%). Minor constituents, including other ions and organic compounds total less than 1% of seawater’s salinity (Trujillo and Thurman 2008). As SSA age, their composition can change however, such as by reactions with acidic gases which can reduce the chlorine content and increase the sulfur content (Lewis and Schwartz 2004). NaCl was chosen as a coating for this study because it corresponds to the two primary ions of seawater (85%), it has a high melting and boiling point (801°C and 1413°C; CRC 2009) making it suitable for a thermal denuder, and it has been successfully used as a denuder coating for HNO3 sampling (Appel 1993). Seawater was also chosen as a coating since its more complex chemical composition will more closely mimic real SSA.
Before coating, the denuders were rinsed twice with 3.5 M hydrochloric acid solution and three times with deionized water, then dried with activated carbon filtered air. Coating was accomplished by dipping the denuder into an aqueous solution of 2.4 M KCl, 2.4 M NaCl, or concentrated seawater. The inlet was dipped three times in deionized water to help reduce the formation of salt crystals. Care was taken during the whole process to prevent the introduction of liquids into the annulus of the denuder through the vent hole. The seawater was collected from the Atlantic Ocean at approximately 0.6 m from the shoreline in Virginia Beach, VA (36º50’43” N, 75°48′26″). The water was collected in acid cleaned bottles that were rinsed three times with seawater. The salinity of the seawater was measured using a refractometer and then concentrated by gently boiling with a hotplate to reach a concentration of approximately 2.4 M.
After coating, the denuders were dried by blowing mercury free air through them. The inlet was cleaned of any crystals with a Kimwipe wetted with deionized water. Before use, denuders were blanked by heating in a furnace to 525 ° C for 30 min with mercury free air passing through. Denuder blanks were routinely measured for quality control as reported below.
2.2 Ambient experiments
2.3 Manifold experiments
Since ambient RGM concentrations were often below the detection limit during ambient experiments, a glass manifold system was designed so that a controlled, artificial source of RGM could be passed through the denuders. RGM was supplied by a permeation source constructed from Teflon tubing and loaded with mercuric chloride (HgCl2). The manifold system and the denuders were warmed with heat tape, and N2, supplied by a compressed gas cylinder (ultra high purity grade), served as the carrier gas for HgCl2.
Each manifold experiment compared denuders coated with either 1) NaCl vs. KCl or 2) sea salt (SS) vs. KCl. Denuders were attached to ports on the manifold by Teflon-coated glass elutriators. The relative position of the denuders was varied among the experiments. Backup KCl denuders were also placed downstream of the NaCl and SS denuders to capture any RGM not removed by the NaCl. The downstream end of each KCl denuder was connected to a DTM to measure the flow rate through each of the four denuder setups. Both the manifold and denuders were heated to 50°C ± 5oC by heat tape and covered with Reflectix brand insulation to help maintain a uniform temperature.
In order to minimize the residual mercury in the manifold system before beginning an experiment, the permeation source sidearm was removed from the manifold while the entire system was flushed with N2 gas for 30 min. During the flushing time, the heat tape on the manifold was turned on. Immediately following the flushing, the denuders (already wrapped in heat tape) were attached to their designated port and DTM. When all connections were complete, the heat tape on each denuder treatment was turned on. Temperature was monitored on the denuders and the manifold. Once all the denuder treatments and manifold were within the 50°C ± 5°C range, the sidearm containing the permeation source was reattached to the manifold and the N2 gas was turned on. The N2 flow rate for each experiment was between 5 and 12 lpm. The experiments were conducted for one to 6 min then the gas was turned off, and the sidearm was removed from the system. Next, the heat tapes were turned off and the DTM values were recorded for each denuder treatment. The denuders were then ready for analysis.
2.4 Denuder analysis
The Hg was thermally desorbed from the denuders onto gold coated sand traps (Brooks Rand) by heating to 525°C in a clamshell furnace (Landis et al. 2002). The heating time was a minimum of 15 min for KCl denuders and 30 min for NaCl and sea salt (SS) denuders. Two gold traps were used in sequence after the denuder in case of breakthrough of Hg through the first trap. The traps were then analyzed by cold vapor atomic fluorescence (CVAFS) on a Tekran 2600 analyzer with a Brooks Rand Amalgamation Control Module. Calibration was completed each analysis day by injection of Hg0 vapor standards onto gold coated sand traps (Brooks Rand Air Calibration Unit and traps) (Dumarey et al. 1985). The standard curve consisted of a minimum of 4 standards with an r2 value of at least 0.999.
2.5 Quality control and calculations
Clean techniques were employed throughout the study. All sampling components were acid cleaned with 3.5 M HCl and dried in a High Efficiency Particulate Air (HEPA) filtered drying cabinet. In addition, particle-free gloves were worn at all times, and samples and sampling equipment were stored and transported in new zip top bags. Denuders were quantitatively blanked before each experiment to minimize contamination.
Denuder blanks were collected throughout the study for all denuder types. The mean ± standard deviation of the denuder blanks for the ambient study were 0.011 ± 0.012 and 0.013 ± 0.014 ng/denuder for the KCl and NaCl denuders, respectively (n = 23, n = 11). The mean ± standard deviation of the denuder blanks for the manifold study were 0.016 ± 0.015, 0.012 ± 0.014, 0.007 ± 0.005 ng for the KCl, NaCl and SS denuders respectively (n = 47, 10, and 10). An ANOVA test showed no significant difference among all of these blank types (p = 0.210), so the blanks were pooled to give an overall mean ± standard deviation of 0.013 ± 0.013 ng (n = 101). An estimate of the MDL for all experiments was calculated as 0.026 ng, based on the standard deviation of these pooled blanks times the student’s t-value (df = 100). All experimental values were blank corrected by subtracting the mean blank value of 0.013 ng. Since denuder measurements are reported as a mass and not a concentration, mass values were corrected for any differences in flow rate among denuders within each experiment. Uncertainty in denuder measurements was also estimated using the standard deviation of the blanks, 0.013 ng.
The manifold setup was tested and optimized before data collection began. This included two experiments in which four KCl denuders were attached to the ports. This comparison allowed for evaluation of the variability of Hg delivered to the different ports and captured by individual denuders. The mean ± standard deviation of the each of the two tests was 0.229 ± 0.028 and 0.329 ± 0.026 ng. The standard deviation of these tests is roughly double that of the blanks, and may reflect the added variability from the experimental procedures using the manifold.
Mercury collected on the backup KCl denuders during the ambient experiments was similar to the denuder blank values (mean 0.013 ± 0.018, n = 7). Mercury collected on the backup KCl denuders during the manifold experiments was higher, suggesting some breakthrough from the NaCl and SS denuders at these higher concentrations (mean 0.62 ± 0.82, n = 9; one data point missing due to broken denuder). Backup denuders were not used behind the lead KCl denuders, so it is not known if a similar amount of breakthrough would occur for both types of denuders. Despite these higher values the mass on the backup denuders was always less than the mass on the lead denuder, with the exception of one denuder pair on 02/11/09 (see Results and Discussion).
Statistical analysis was done with the SPSS statistical package software.
3 Results and discussion
3.1 Ambient air experiments
In an effort to collect a greater mass of Hg than the MDL, the sampling time for subsequent experiments was increased to 24 h. Two NaCl denuders and two KCl denuders were co-located and sampled simultaneously. On 6/16/06 Hg mass from these four lead denuders were at or below the MDL (≤0.026 ng). Heavy rain during the second half of sampling is likely to have contributed to the low concentrations of RGM, which is easily scavenged by precipitation and wetted surfaces (Malcolm and Keeler 2002). On 6/28/06, the KCl denuders captured 0.049 and 0.043 ng Hg and the NaCl denuders captured 0.040 and 0.041 ng Hg (Fig. 1). On average the NaCl denuders collected 87% of the mercury that the KCl denuders collected for this day.
Although concentrations during all these experiments were below or close to MDL levels, the experiment on 6/28/06 suggests that NaCl did remove Hg from the ambient air. This implies that some RGM compounds have an affinity for sodium chloride, a hypothesis that was further tested by the manifold experiments. The unlikely possibility that the Hg collected by the NaCl was elemental gaseous mercury is discussed below.
3.2 Manifold experiments
Summary of the results from all ambient outdoor and manifold laboratory experiments
NaCl or SS denuders used?
NaCl or SS denuder mean Hg collected (ng)
KCl denuder mean Hg collected (ng)
% Hg collected on NaCl or SS relative to KCla
87 ± 26
88 ± 4
85 ± 5
119 ± 7
102 ± 8
94 ± 2
81 ± 4
57 ± 3
49 ± 3
56 ± 4
89 ± 6
63 ± 4
In order to compare the NaCl and SS experiments, the ratio of Hg collected on each NaCl to its paired KCl denuder and each SS to its paired KCl denuder was calculated for all experiments with results above the MDL (ambient and manifold; n = 10 for NaCl/KCl, n = 6 for SS/KCl; excluding low efficacy SS experiments noted above). A t-test showed no significant difference between the means of these two groups (p = 0.420), suggesting similar HgCl2 collection for NaCl and fresh SS. Given that there was no significant difference in means between the NaCl and SS data, they were pooled into one data set for further analysis. A paired t-test showed no significant difference between the mass of Hg collected on the KCl versus the corresponding NaCl or SS denuder (p = 0.173, n = 16), supporting the hypothesis that NaCl and SS have a high affinity for HgCl2, similar to that of KCl.
3.3 Does gaseous elemental mercury adsorb to NaCl?
Ambient elemental gaseous mercury (Hg0) was not measured at our sampling location, but is expected to be about 1.5-2.5 ng/m3 based on data from similar locations (Lamborg et al. 2002; Schroeder and Munthe 1998). Therefore the Hg0(g) concentration should have been approximately 1000× greater than that of RGM at our outdoor site. If Hg0 did adsorb to the NaCl, we would have expected to measure much higher concentrations on the NaCl denuders, yet on 4 days, the Hg collected on the NaCl denuder was below the MDL. For example, on 6/16/06, the Hg collected by the two KCl and two NaCl denuders were all at or below the detection limit of 0.026 ng, or 1.2 pg/m3. This represents only about 0.1% of the Hg0(g) that was available for collection to the NaCl denuder. KCl denuders are selective for RGM, and should not collect Hg0. The close match between the Hg collected on the KCl and NaCl denuders in the ambient experiment on 6/28/05 and in all of the HgCl2 manifold experiments thus further supports the conclusion that the form of Hg collected in the ambient experiments was RGM. Therefore we can deduce that Hg0(g) is not adsorbing to the NaCl denuders in our experiments.
3.4 Implications for mercury deposition and transport
The lifetime of SSA can range up to a few days and is a function of wind speed and particle size (Sander and Crutzen 1996). Sea salt penetration onto land is greatest in the first few kilometers but can exceed 50 km inland (Gustafsson and Franzen 2000). In some cases, sea salt particles may be transported to the free troposphere as well (Patterson et al. 1980). Aerosols may continue to scavenge Hg gases throughout their lifetime. Consequently, the effects of Hg scavenging by sea salt may be globally important, particularly to marine and coastal ecosystems. Models by Selin et al. (2007) and Holmes et al. (2009) demonstrated that a fast sink, like scavenging by SSA was required to reproduce RGM concentrations and diurnal patterns in the marine boundary layer. Holmes et al. (2009) further hypothesized that the uptake of RGM by SSA would be limited by mass transfer.
SSA can also serve as cloud condensation nuclei and be scavenged by wet deposition processes. Scavenging of RGM by SSA may thus indirectly increase wet deposition of mercury. In a recent modeling simulation, this process increased Hg wet deposition along the Northeastern U.S. coast (Selin and Jacob 2008). In the coastal environment, sea spray and dry deposition of SSA will coat natural and manmade surfaces. Salts deposited to leaf surfaces have been shown to increase surface wetness, even at moderate relative humidity (Burkhardt and Eiden 1994). These sea salt coatings may enhance dry deposition of RGM to surfaces coated with dry salts or with aqueous films and dew.
3.5 Future studies
Direct measurements of mercury on SSA are difficult, due to the potential positive (Lynam and Keeler 2002) and negative sampling artifacts (Lynam and Keeler 2005; Malcolm and Keeler 2007). Consequently, many studies have focused on only fine fraction (<2.5 μm) aerosols. These studies underestimate the total concentration and dry deposition of Hg(p). In a study on the Atlantic coast of Florida for example, during periods of onshore airflow, the fine fraction Hg(p) (<2.5 μm) only accounted for 25% of the Hg(p) on aerosols less than 10 μm (Malcolm et al. 2003). The larger particles will deposit more quickly, while smaller particles can be transported further inland. Future studies should evaluate the role of SSA in all size fractions on local and regional deposition.
This is the first study to experimentally investigate the gas-particle partitioning of RGM to sea salt that has used ambient air or sea salt generated from seawater. These experiments demonstrate that HgCl2, a major component of RGM has a strong affinity for NaCl and fresh sea salt and is likely to be scavenged by SSA. This adds to the evidence that RGM is scavenged by sea salt aerosols, and thereby could be more quickly dry deposited.
Funding was gratefully supplied by the Jeffress Memorial Trust and Virginia Wesleyan College. We thank Dr. Thomas Atkinson and the Florida Department of Environmental Protection for the loan of the sampling manifold and Dr. Gerald Keeler at the University of Michigan Air Quality Laboratory for additional equipment. We are grateful to Dr. Margaret Reese of Virginia Wesleyan College and Dr. Terry Glover of Bloomfield College for statistical assistance and Dr. Mary Lynam of Marygrove College for advice on methods. We thank the many students at Virginia Wesleyan College who have assisted with this research, especially Ivy Ozmon and A.J. Ward. We appreciate the advice of two anonymous reviewers for suggestions that improved the manuscript.