Passive air sampling of gaseous elemental mercury: a critical review

Introduction Conclusions References

and thus GOM is expected to contribute a relatively minor proportion to the overall uncertainty of a GEM measurement (Pandey et al., 2011). For the sake of simplicity, we will therefore refer to the sorbate as GEM in this review. Starting with a discussion of the basic elements of passive air sampling, we summarise the rationale for passive air sampling of GEM. Features and specific requirements of effective PASs for GEM 5 are discussed vis-a-vis existing designs. We conclude by outlining future perspectives in GEM passive air sampling research. This review differs from the recent review by Huang et al., which had a much wider scope and discussed man-made and naturally occurring (e.g. moss and lichen) PASs for Hg, as well as wet and dry Hg deposition samplers (Huang et al., 2014). Furthermore, Huang et al. (2014) reviewed samplers 10 for GEM, GOM and PBM; whereas we focus this review on PASs for GEM only.

Basic elements of passive air sampling
Before discussing PASs for GEM specifically, there are underlying principles and important terminology in the chemistry and physics of passive sampling, regardless of the chemical of interest, that require introduction. In passive air sampling, the amount 15 of a target compound taken up in a sorbent exposed to the atmosphere for a given period of time is quantified. Without electricity to pump air across the sorbent, PASs must rely on turbulent and/or molecular diffusion to deliver a target compound to the sorbent. To interpret the sorbed amount as a concentration of the target compound in air, a sampling rate (SR) specific to a particular sampler design and target compound 20 must be known. SRs can be estimated theoretically, but are usually determined by calibration using simultaneous measurements with reliable active sampling techniques (Gustin et al., 2011;Huang et al., 2012;Lyman et al., 2010;Mattoli et al., 2007;Skov et al., 2007;Zhang and Wong, 2007). Since considerable variability in sampling rates can result from turbulent airflow, PAS designs often include a diffusive barrier to assure Introduction The SR L 3 T −1 for a diffusive PAS, which quantifies the volume of air that is effectively stripped of target compound per unit of time, using Fick's first law, is governed by: where D is the molecular diffusion coefficient of the target compound in air L 2 T −1 . 5 L is the diffusive path length [L], which, in the case of a PAS with a porous diffusive barrier, needs to account for the tortuous path taken by a molecule diffusing through the barrier. A is the surface area of the collection surface L 2 ; m is the mass of sorbed compound [M]; t is the deployment time [T]; C is the concentration of the compound in air M L −3 . Once a SR for a particular compound and PAS has been determined, 10 a concentration can be calculated from the mass of sorbed analyte by rearranging Eq.
(1). The uptake rate (UR) M T −1 in a diffusive sampler, which is the change in the mass of sorbed target compound over time, can also be calculated using Fick's first law: 15 where dC dL is concentration gradient of the target compound in air across L. Note that whereas SR is independent of the analyte concentration C, UR is not.
During passive sampling, it is normally assumed that the SR is independent of the amount of sorbed analyte m. If the concentration C in air is constant, this implies that m is a linear function of deployment time t (Fig. 1). As long as this assumption is valid, 20 a sampler is said to operate in the linear uptake phase (Bohlin et al., 2007;Hayward et al., 2010). If deployment is too long, the SR will decline, either because the number of available sorption sites declines as the sorbent becomes saturated or because an equilibrium distribution between atmospheric gas phase and sorbent material is approached. During this curvilinear uptake phase, thermodynamic factors in addition to Introduction surface area that could unintentionally sorb the target substance. The most prominent radial type sampler is the commercially produced Radiello ® , which is the reference sampler housing for volatile organic compounds in Europe (Król et al., 2010). There are also PAS designs without a porous diffusion barrier that do not seek to explicitly standardize the diffusion distance; the turbulent flow of air is allowed to 5 pass freely over the sorbent (Shoeib and Harner, 2002;Xiao et al., 2007). Because SRs for such samplers are generally quite high, such designs are popular in PASs for semivolatile organic chemicals, whose air concentrations are so low that high SRs are often required to sorb quantifiable amounts (Huang et al., 2012). The SR for such samplers is more susceptible to the influence of wind speed and even wind direction, causing substantially reduced accuracy and precision (Huang et al., 2012;May et al., 2011).

The rationale for a passive air sampler for gaseous elemental mercury
Mercury has been regulated by national agencies, international agreements and most recently on a global scale through the United Nations Environmental Program's Mina- 15 mata Convention (Driscoll et al., 2013;Pandey et al., 2011;Renner, 1999;Schroeder and Munthe, 1998;Selin, 2009;UNEP, 2013). Under the Minamata Convention Hg mining is to be phased out, air emissions are to be controlled and artisanal/small-scale gold mining is to be regulated (UNEP, 2013). Scientific evaluation of the effectiveness of control measures on mercury releases is stipulated under the convention and will 20 require long term monitoring at numerous sites around the world (UNEP, 2013).
While active air sampling works effectively when measuring GEM concentrations at sites with access to reliable electricity (Munthe et al., 2001;Schroeder et al., 1995), PASs provide an economical, zero-energy alternative for a number of specific sampling needs. The cost of passive sampling allows for a much larger number of samplers and Introduction efficacy of Hg regulations, as well as a basis from which to assess potential impacts of global climate change on long time scales. The accessibility of PASs is also critical for Hg monitoring and research in developing countries where the cost of active samplers might otherwise prohibit it. The potential for improved monitoring of Hg at remote sites, where the maintenance of active samplers would be difficult, is also intriguing since 5 certain remote populations may be at elevated risk to Hg exposure. This is evident in aboriginal communities in Canada's North for example, where consumption of traditional foods with elevated concentrations of MeHg, such as fish and marine mammals, places these communities at greater risk than non-aboriginal communities not subsisting on such diets (Govt. of Canada, 2012;Tian et al., 2011;Wheatley and Wheatley, 10 2000). This also extends to include Hg exposure (both MeHg in food and GEM in air) in communities in close proximity to small scale artisanal gold mining activities (Barbosa et al., 2003;Sousa and Veiga, 2009). Current atmospheric Hg models have a large and often unknown degree of uncertainty, due in part to a lack of rigorous evaluation against real data (AMAP/UNEP, 2013; 15 Keeler et al., 2009). Finer spatial monitoring could improve the accuracy of the modelling and future predictability of atmospheric Hg dispersion, transformation and phase exchange processes that are fundamental to understanding the spatial and temporal variability of Hg in the global atmosphere (Keeler et al., 2009;Lin et al., 2006). The increased spatial resolution of data that is possible with PASs could also facilitate Hg 20 source characterisation. Mapping with high spatial density sampler deployments would expedite the identification of both known and unknown Hg sources, as well as populations at greater exposure risk.
The small size and weight of certain PAS designs also facilitates their use in recording GEM concentrations in homes and workplaces or even in monitoring personal ex- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | characterisation would likely benefit from personal Hg sampling within the general community across a wide range of Hg exposure scenarios. In summary, there are different rationales for passive sampling of GEM. The most compelling rationales for a GEM PAS include: A. Measuring ambient GEM levels close to the global average over extended periods

Low-level detection
The most basic requirement is that a GEM PAS sorbs a sufficient amount of GEM for accurate and precise quantification. According to Eq. (1), the amount of GEM sorbed in a PAS increases linearly with the air concentration C, the SR, and the deployment time t. There is an upper limit to the SR, because of the requirement for keeping it 15 tightly controlled (see Sect. 4.2). The length of the deployment period is limited both by the desired temporal resolution and the equilibrium uptake capacity, i.e. the need to remain in the linear uptake phase ( Fig. 1). At a site with an average GEM concentration of 2 ng m −3 a sampler with a low SR of 0.001 m 3 day −1 (typical of a tube type sampler) deployed for 1 day, 1 week, Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | C) will not require the low MDLs needed for a PAS that aims to measure background atmospheric GEM concentrations (rationale A).

Tightly controlled sampling rates
The European Union has stipulated that a maximum uncertainty of 50 % for measurement of GEM concentration is acceptable (European Commission, 2005). However, 5 such a general standard of uncertainty is unlikely to be appropriate for all monitoring needs. Ambient background concentrations of GEM typically vary by less than 0.5 ng m −3 (approximately < 25 % of the average concentration) (Gustin et al., 2011;Kellerhals et al., 2003;Kock et al., 2005;Lyman and Gustin, 2008). If the sampler application requires the ability to discriminate such variability (rationale A), Gustin et al. (2011) estimated that PASs need to be able to resolve GEM concentration differences of 0.1 ng m −3 . This implies that very high precision and accuracy of the measurement is essential. In particular, meteorological factors that may vary between sampling sites (such as temperature, relative humidity, and wind speed) cannot be allowed to have a strong effect on the SR. Otherwise, it will be impossible to attribute small differ- 15 ences in the amounts sorbed in different samplers to differences in air concentrations. Such high precision and accuracy in the SR of a PAS can probably only be achieved by tightly controlling the diffusive distance through the use of a diffusive barrier. Because tightly controlling the SR inevitably means a lower SR, such PASs require more sensitive analytical techniques and/or longer deployment times (see Sect. 4.1). GEM 20 concentrations in and around point sources are not only higher than at background sites, but they can vary by multiple orders of magnitude (Carpi and Chen, 2001;Liu et al., 2009;Malm et al., 1995). If the primary mandate of a PAS is the characterisation of such concentration variability close to sources (rationale B and C) the accuracy and precision of the PAS's SR need not be as high as for PASs recording background 25 levels, where variability is low. Precision and accuracy, of course, not only depend on a tightly controlled SR, but also on the precision and accuracy of the analytical method for determining Hg concentrations in the sorbent. 34615 ACPD 15,2015 Passive air sampling of gaseous elemental mercury: a critical review

High and stable equilibrium uptake capacity
An important feature of PASs is their ability to provide time-averaged concentrations over extended time periods. In many deployment scenarios, especially in remote regions and in developing countries, travel to and from the sampling sites is associated with considerable time, effort, and cost, which limits the number of site visits. In such 5 cases, deployment times that extend over months or even years are desirable. Long deployments demand sorbents that not only have a very high uptake capacity, but that can retain this capacity for a long time. While it is possible to envisage scenarios involving deployments of a week or less (especially for rationales B and C), a low capacity sampler will inevitably be limited in its applicability, especially if its capacity may be 10 subject to variability (e.g. as a result of meteorological factors or ageing processes). A high capacity sampler can still serve in short deployments as long as it sorbs reliably quantifiable amounts.

Simple and inexpensive sampler and analysis
A good design should be simple to produce and be made from inexpensive materi- 15 als. The samplers should be easily and quickly deployed and retrieved. The analysis should be straightforward. Ideally, exposed samplers should be easily stored (even for extended time periods) and transported.

Method validation
To find wide acceptance, a PAS's performance must also be well validated against 20 other broadly accepted methods, usually active air monitoring systems with satisfactory precision and accuracy (Ebinghaus et al., 1999;Munthe et al., 2001;Schroeder et al., 1995).

34616
ACPD 15,2015 Passive air sampling of gaseous elemental mercury: a critical review

Existing passive air samplers for gaseous elemental mercury
Passive mercury sampling in air has been documented in the literature for over 35 years (McCammon and Woodfin, 1977). Over the years, 10 different PASs for GEM have been proposed (Fig. 2); their main characteristics are summarised in Table 1 (2001) involves a permeable tissue, it is placed directly on top of the sorbent and not in the entrance to the diffusion tube and thus does not serve to reduce air turbulence in the diffusion tube. 15 The designs by Skov et al. (2007;Fig. 2I) and Gustin et al. (2011;Fig. 2G) are radial diffusion type samplers using either a commercial Radiello ® diffusive barrier (Skov et al., 2007;Gustin et al., 2011) or one made from expanded polytetrafluoroethylene (PTFE) (Gustin et al., 2011;Peterson et al., 2012). The sampler design by Huang et al. (2012;Fig. 2E), which has been adapted from a sampler for semivolatile organic compounds 20 and involves the sorbent placed in the space between two bowls, makes no attempt to have the SR limited by a molecular diffusion step.
In two of the samplers, diffusion through a barrier (rather than through air) is meant to be the rate-limiting step to GEM uptake in the sorbent. In the PAS developed by the 3M Company and tested by McCammon and Woodfin (1977; Fig. 2A), GEM diffuses ACPD 15,2015 Passive air sampling of gaseous elemental mercury: a critical review  Fig. 2D), it is assumed that diffusion through a low density polyethylene (LDPE) lay-flat tube membrane is the rate-limiting step, i.e., is slower than diffusion through the air boundary layer surrounding the membrane and slower than oxidation within the liquid sorbent. No experiments were conducted to validate this. 5 6 Do existing samplers meet the identified requirements?

Do existing gaseous elemental mercury passive air samplers sorb amounts above detection limits?
While the lowest GEM concentration that can be determined with a PAS depends on its SR and the deployment time, the MDL of the analytical technique is an important 10 consideration. Most of the existing PASs for GEM rely on spectroscopic techniques for analysis. Four studies relied on thermal desorption followed by Cold Vapour Atomic Fluorescence Spectrometry (CVAFS) to quantify the amount of GEM sorbed on a goldbased sorbent (Brown et al., 2012;Gustin et al., 2011;Huang et al., 2012;Skov et al., 2007), or a silver wire (Gustin et al., 2011). Four studies used Atomic Absorption Spec-15 troscopy (AAS): Mniszek (2001) first dissolved Hg from a hopcalite sorbent using nitric acid prior to analysis by AAS; Zhang et al. (2012) analysed the mercury sorbed to carbon using an automated analyzer that relies on thermal decomposition, amalgamation followed by AAS; Brumbaugh et al. (2000)  was 2 ng m −3 over a four week deployment, a value slightly above global atmospheric averages, and thus has limited applicability in monitoring background concentrations. PASs relying on non-spectroscopic techniques are clearly unsuitable for monitoring GEM at ambient levels. While the method used in the analysis of the sampler tested by McCammon and Woodfin (1977) is not revealed, it has a very high MDL ( Table 1) that renders it impractical for deployment in all but the most contaminated environments. The sampler design by Mattoli et al. (2007) also has a high MDL (Table 1), due to the use of electronic resistivity for analysis. Nonetheless, PASs that effectively quantify higher concentrations are not necessarily rendered obsolete by designs that can adequately monitor GEM at lower concentrations. The equilibrium uptake capacity of samplers targeting trace concentrations of GEM may be too low for use in high concentration environments, potentially leading to an underestimation of actual concentrations.

Do existing gaseous elemental mercury passive air samplers have tightly
controlled sampling rates? 20 The SRs of existing PASs for GEM range over 4 orders of magnitude ( Several studies have conducted controlled chamber experiments to test to what extent the SR of a PAS is influenced by meteorological conditions and other factors. Changing temperature has the two-fold effect of shifting partition equilibria for GEM between the sorbent and air, and altering the rate of GEM diffusion, both potentially affecting SRs. Four studies have looked at the impact of temperature on uptake. Both the diffusion through the barrier film in the 3M-sampler (McCammon and Woodfin, 1977), and that through the LDPE membrane of the sampler by Brumbaugh et al. (2000) were found to be dependent on temperature. McCammon and Woodfin (1977) (Levy et al., 2009). Also in samplers where the rate-limiting step is diffusion through air, temperature has the potential to affect the SR through its influence on the diffusion coefficient of Hg in air (Eq. 2). However, the effect is quite small and neither Skov et al. (2007) nor Guo et al. (2014) testing at 6, 27, 36.5 and −10, 0, 15, 25, 35 • C, respectively, observed 15 a temperature dependence of the SR. The effect of variable wind speeds on SRs is likely the principal concern in outdoor passive air sampling. Diffusive samplers assume the presence of a laminar layer of stagnant air surrounding the sorbent, with the rate limiting step for analyte uptake in most samplers being diffusion through this layer (Shoeib and Harner, 2002). Wind 20 has the effect of increasing SRs by decreasing the thickness of the stagnant air layer, essentially reducing the diffusive path length and in turn leading to increased and/or more variable SRs (Bartkow et al., 2005;Moeckel et al., 2009;Pennequin-Cardinal et al., 2005;Plaisance et al., 2002;Zhang et al., 2012). Four studies exposed their samplers to controlled GEM concentration at different wind speeds, again with diver- exposure chamber (range of wind speeds tested was not given), presumably because of the relatively small and long diffusion tube. On the other hand, Skov et al. (2007) measuring at wind speeds of 0.002, 2.75, 5.15, 7.1 m s −1 , noted an almost doubling of the SR between wind-still conditions and the highest wind speed (Fig. 3a). The dependence of SR on wind speed was described as non-linear, with greater effects observed 5 at low wind speeds (Fig. 3a) (Skov et al., 2007). The wind speed effect on GEM uptake in this sampler is similar to what has been observed for the uptake of benzene also using a Radiello ® -based diffusive barrier (Pennequin-Cardinal et al., 2005;Plaisance et al., 2002). However, Skov et al. (2007) did not test wind speeds between 0.002 and 2.75 m s −1 (region of greatest effect on SR), and adding a linear trendline improved the relationship from the power trendline that was originally fitted to the data (Fig. 3a). Guo et al. (2014) tested their sampler at wind speeds of 0.5, 1, 2, 3, 4, and 5 m s −1 and observed a linear relationship between SR and wind speed, which led to a more than 6-fold increase in SR between 0.5 and 5 m s −1 (Fig. 3b). Even the application of a wind speed dependent SR model could not entirely account for the effect of wind speed on  2007) did not observe a significant effect of relative humidity in laboratory testing of synthetic air (presumably low relative humidity) and high relative humidity air (approximately 90 % relative humidity) using a gold sorbent. Similarly, Guo at al. (2014) observed no depen-25 dence of GEM uptake to their activated carbon-based PAS across a relative humidity range from 25 to 90 %. If uptake in a PAS has been measured simultaneously with active monitoring system measurements (Gustin et al., 2011;Huang et al., 2012;Nishikawa et al., 1999;Zhang ACPD 15, 34605-34645, 2015 Passive air sampling of gaseous elemental mercury: a critical review  , 2012), it is possible to test whether the SR of a PAS is constant under field conditions. We can plot the uptake rate, UR (amount of GEM sorbed per day), observed in the field-deployed samplers against the GEM concentration during the deployment period ( Fig. 4) (Gustin et al., 2011;Huang et al., 2012;Nishikawa et al., 1999). Ideally, such a plot should go through the origin (suggesting no uptake if no GEM is present) and be linear suggesting that ( data not available in manuscript) also showed no relationship. The scatter of the data for the activated carbon based PAS by Zhang et al. (2012; blue diamond data series Fig. 4d) was less, especially in the initial field testing. However, active measurements only covered ∼ 11 % of the sampling period, which could bias the comparison if GEM levels fluctuated strongly. Subsequent testing of the PAS by Guo et al. (2014;green 15 triangle and red square data series Fig. 3d) revealed a significant effect of wind speed on SR, which weakened the linear relationship between UR and actively measured GEM concentration. The data in Table 1  Mniszek (2001) field tested the SR of their samplers outdoors. However, data was not reported in the format shown in Fig. 4. The precision of the PAS by Brown et al. (2012) compared to actively measured values were low (although active measurements were only determined on a weekly basis and may not give an accurate estimation of mean concentrations during 25 the deployment period). Overall uncertainty of these results was determined to be ∼ 30 % (Brown et al., 2012). Part of this uncertainty was attributed to the low SR of the design, which despite the length of deployments resulted in a very low amount of GEM being sorbed by the PAS (Brown et al., 2012). Brumbaugh et al. (2000) did not ACPD 15,2015 Passive air sampling of gaseous elemental mercury: a critical review take any actively measured GEM samples along with the passive measurements, and hence no assessment of calibrated SR accuracy could be made (Brumbaugh et al., 2000). Mniszek (2001) assessed the effectiveness of their axial badge PAS in high concentration workplace scenarios in the Polish chlor-alkali industry. On average the relative percent difference of passive measurements from active measurements was 5 high at 173 ± 177 % (Mniszek, 2001). However, the active measurements were taken by pumping air across the same hopcalite sorbent used in the PAS and it was itself still being developed. Prokopowicz and Mniszek (2005) then tested the same passive and active sampling setups in seven hospitals and one residence across Poland that also had elevated GEM. Here, results were much closer with the relative percent difference 10 of passive measurements from active measurements being 16.2±15.1 % (Prokopowicz and Mniszek, 2005). The higher limits of detection for the sampling method used meant the PAS could not be used for deployments at background concentrations of GEM.

Do the sorbents used in existing samplers have a sufficiently high and stable uptake capacity?
Among the ten samplers reviewed here, all but one have relied on metals as sorbents: either gold, silver, or hopcalite, an alloy of manganese and copper. Gold-based sorbents are especially popular, having found use in eight of the samplers. This preference for gold is due to its high affinity to Hg (Greaves et al., 1997;Gustin et al., 2011;Rex et al., 2006), and presumably also due to its common and successful use in ac-20 tive air monitoring of GEM. Gold has been deployed as a film (Mattoli et al., 2007), foil (McCammon and Woodfin, 1977), and solid cylinder (Skov et al., 2007), but it is most often used as a thin coating on a carrier material, partly to increase the surface area to volume ratio of the sorbent and partly to reduce the amount of gold required and therefore the cost. Examples are sputter-coated quartz plates (Gustin et al., 2011;25 Peterson et al., 2012), quartz fibre filters (Huang et al., 2012;May et al., 2011), goldcoated silica (Brown et al., 2012), and chromosorb (Nishikawa et al., 1999). The sampler design by Brumbaugh et al. (2000)  ride/nitric acid solution, which is filled and sealed into a LDPE membrane pouch. One PAS utilised silver in the form of a solid wire (Gustin et al., 2011), while another used a sulphur-impregnated activated carbon sorbent (Zhang et al., 2012). Activated carbon has a large surface area to volume ratio, which results in a high number of sites available for interaction with GEM (Karatza et al., 1996(Karatza et al., , 2000Lee et al., 2006;Vidic 5 et al., 1998). Sulphur-impregnated activated carbon in particular may have considerable potential for uptake of GEM due to the affinity of Hg to sulphur (Suresh Kumar Reddy et al., 2013;Vidic et al., 1998;Vidic and McLaughlin, 1996;Vidic and Siler, 2001;Zhang et al., 2012). The uptake capacity for GEM has not been directly measured for any of these sor-10 bents. One way to indirectly determine the uptake capacity of a PAS is to establish an uptake curve for a period longer than the linear uptake period. Somewhat surprisingly, studies to determine uptake curves have rarely been performed and most studies on PASs for GEM have relied on a single (possibly replicated) length of deployment. Specifically, what has been rare in the literature are uptake experiments involving the 15 simultaneous deployment of multiple samplers (including sample replication) for variable time periods and the plotting of an uptake curve displaying adsorbed amounts vs. deployment time. Mniszek (2001) did expose eight samplers simultaneously for periods ranging from 1 to 8 h, yielding highly linear uptake curves (Fig. 5a), however the short overall deployment period prevents the derivation of information on the PAS's uptake 20 capacity. Nishikawa et al. (1999) conducted a rudimentary uptake experiment, which involved exposures of 11 and 126 days for two different lengths of their diffusor tube (3 and 30 mm) and uptake curves appear to be linear over such a long time (Fig. 5b). Brumbaugh et al. (2000) sampled over 4, 8 and 12 weeks outdoors. While the data (Fig. 5c) do suggest reasonably linear uptake, GEM concentrations during the deployment were not recorded. Brumbaugh et al. (2000) also investigated whether the SR depends on the length of PAS exposure by comparing day-long with week-long deployments. SRs after one week were slightly, but consistently lower by 16.5 % than after 24 h, suggesting that the sampler's linear uptake phase is shorter than a week. We suspect that some of the scatter in Fig. 4 is due to a low, irreproducible, and/or unstable uptake capacity of the employed metal sorbents, in addition to poorly constrained SRs. In particular, passivation, memory effects, and physical degradation affect the performance of gold-based sorbents over extended deployments. Passivation occurs when GEM binding sites on a sorbent become obscured by interfering compounds or 5 when reactions between atmospheric constituents and sorbed Hg strip some of the analyte from the sorbent over time (Gustin et al., 2011;Huang et al., 2012;Mattoli et al., 2007). While silver and gold sorbents can both undergo passivation, complexes between atmospheric interferents and silver are typically more stable than those with gold (Gustin et al., 2011). Elevated levels of chlorinated and sulphonated compounds in the atmosphere affect noble metal PAS uptake the most (Brown et al., 2012;Mattoli et al., 2007;McCammon and Woodfin, 1977;Luo et al., 2010). Furthermore, passivation of gold tends to increase with increasing deployment time and Hg concentration (Gustin et al., 2011). Passivation may thus reduce the maximum possible deployment period of a PAS and, if interferents are not completely removed in analysis, the sor-15 bent's long-term regeneration capacity is reduced.

ACPD
Conceivably, passivation could also occur with other sorbents. The relatively high number of active sites on activated carbon should reduce the effect of passivation, but research to date has been inconclusive. Diamantopoulou et al. (2010) reported that while HCl, SO 2 , and O 2 enhanced GEM uptake to virgin activated carbons, CO 2 20 inhibited it. These experiments were conducted at elevated concentrations and temperatures relevant to Hg sequestration from flue stack emissions (Diamantopoulou et al., 2010). While the concentrations of potentially interfering compounds in ambient air will be much lower, this is nonetheless a concern that may require experimentation at interferent concentrations, temperatures and humidities relevant to actual deployments 25 of PASs.
Memory effects, or the potential to retain residual Hg after thermal desorption, may also compromise sampler accuracy (Brown et al., 2011(Brown et al., , 2012Luo et al., 2010). Memory effects have been estimated to contribute between 0.1 and 5 % error to individual  (Brown et al., 2011;Luo et al., 2010). Brown et al. (2011) observed both short-term and long-term memory effects in the analysis of gold sorbent sampling tubes deployed for 1 to 4 weeks in an active sampler. Short-term memory effects occur when Hg is insufficiently flushed from the sorbent during desorption (Brown et al., 2011) and can be addressed through the use of properly tuned desorption cycles that 5 maximize the recovery of sorbed Hg by ensuring flow rate, desorption temperature and duration are sufficient to flush all Hg from the sorbent (Brown et al., 2011;Luo et al., 2010;Skov et al., 2007). Long-term memory effects occur when Hg, especially in its oxidized forms (Brown et al., 2011;Morris et al., 2002;Nowakowski et al., 1997), diffuses from surface sorption sites into the bulk material (Brown et al., 2011;Dumarey 10 et al., 1985;Luo et al., 2010). Brown et al. (2011) proposed that the diffusion of Hg into the bulk material may actually be facilitated by thermal desorption and not occur during sampling. They suggest that two analytical desorption cycles be applied with a delay of approximately 20 days between cycles to allow Hg in the bulk material to migrate back to the surface, and hence, allow the determination of all sampled Hg (Brown et al., 15 2011). Alternatively, managing sorbents through deployment only in environments of consistent GEM concentrations (not using the same sorbent at a point source site then at a background site), over consistent deployment times is likely to reduce or remove the long-term memory effect altogether (Brown et al., 2011). This is due to a consistent amount of Hg migrating into and out of the bulk sorbent during desorption and deploy-20 ment, respectively, when GEM concentrations during deployments are similar (Brown et al., 2011). Physical degradation of the sorbent may interfere with the accuracy of measurements, particularly in the case of sorbent material reuse. With time and repeated use, sorbent coatings of noble metals onto inert substrates will slowly lose their structural 25 integrity and partially disintegrate, resulting in the addition of particulate matter to the analysis stream (Brown et al., 2011). While the exact nature of this particle interference is unknown, it has been suggested that there may be some interaction between the desorbed GEM and the sorbent particles in the gas stream that may interfere with  (Brown et al., 2011). In order to minimize these effects, careful accounting of age, condition and changes in desorption and analysis characteristics should help to identify problematic samplers.
6.4 Are existing gaseous elemental mercury passive air samplers inexpensive and simple? 5 Noble metal, especially gold-based, sorbents are relatively expensive. This is certainly the case when these materials are used in solid forms as in the PAS by Skov et al. (2007). Activated carbon is a low-cost material, particularly when considering the small amount of material required per sampler (Zhang et al., 2012 used approx. 1 g of carbon per sampler). However, analytical methods for activated carbon are destruc-10 tive by nature (Zhang et al., 2012), while analysis of noble metals also regenerates the sorbent allowing re-deployment. This potentially mitigates elevated material costs for noble metals. Sorbent analysis may be a considerable expense as well. The costs associated with sampler housings (and external shields) must also be considered, and while commercially available diffusive barriers such as the Radiello ® tested by Skov 15 et al. (2007) and Gustin et al. (2011) reduce quality control issues they are also likely to increase the cost of the PAS. Overall, the cost of existing designs is presumably quite low, possibly with the exception of the sampler by Skov et al. (2007). It also appears that manufacture, deployment, retrieval and storage of existing PAS is generally simple, although handling of the liquid filled tube of the sampler by Brumbaugh et al. (2000) is 20 more challenging.

General reflection on existing gaseous elemental mercury passive air samplers
At the present time, existing GEM PASs do not appear to have the precision and accuracy that is required to distinguish between the relative small concentration differences 25 that can be expected between background sites. In particular, samplers relying on gold- based sorbents struggle with low, variable, and/or deteriorating uptake capacity and SR inconsistencies, especially over longer deployments, due to issues such as passivation and memory effects. While activated carbon overcomes some of the problems of goldbased sorbent, the SR of the carbon-based PAS by Zhang et al. (2012) is influenced by wind. While few samplers have succeeded in completely eliminating the effect of wind 5 on uptake kinetics, the SR of samplers incorporating a diffusive barrier and/or a long and narrow diffusion tube can be expected to be less susceptible to wind. The high sensitivity of the analytical techniques typically used for Hg quantification may allow for sampler designs that sacrifice a relatively high SR in order to reduce the effect of wind. One of the major failings of research to date has been the limited use of uptake 10 experiments in GEM PAS testing involving the simultaneous deployment of multiple samplers for variable time periods. Uptake curves enable the effective deployment period and the uptake capacity of the sampler to be determined as well as the SR if active measurements are also taken. with sulphur than for a commercially available sulphur-impregnated product has previously been reported (Vidic et al., 1998;Vidic and McLaughlin, 1996). Although noble metals perform well during the amalgamation of GEM in active air monitoring systems and in other Hg-related analytical instruments, their use in PASs so far has not been convincing, largely because of sorbent passivation. It is possi-5 ble that the large surface area to volume ratio created by nano-structured gold and silver sorbents (James et al., 2012;Liu et al., 2008) could mitigate the effects of passivation due to the sheer abundance of available binding sites. Hence nano-structured sorbents may have more enhanced performance in GEM PASs than noble metal sorbents in more traditional forms. Liu et al. (2008) coated chabazite (aluminum-silicate) nano-particles of high polarity with silver, yielding a nano-sorbent with a high reactivity with Hg and thus high uptake. Similarly, James et al. (2012) successfully measured the uptake of GEM onto commercially available 4-tert-butylthiophenol functionalized 2-5 nm gold nanoparticles using shifts in the characteristic localized surface plasmon resonance of the gold nanoparticles caused by sorbed Hg. The regenerability of these 15 sorbents was excellent; Liu et al. (2008) reported 96.9-102.9 % GEM recovery over 15 sorption and desorption cycles under the same GEM concentrations and James et al. (2012) determined no reduction in signal even after 30 regenerations. It is possible however, that more binding sites may simply equate to more sites for passivation and further testing is required to confirm the relevance of nano-structured sorbents 20 for GEM PASs. Other considerations that need to be addressed in the application of nano-technology to passive Hg sampling are the high costs of commercially available nano-structures and potential inconsistencies in the reproducibility of nano-structure synthesis and substrate attachment (Fraunhofer et al., 2004). 25 Active air sampling techniques for GEM share many characteristics with PAS methods. Active samplers utilize gold cartridges, which collect ambient air at a given flow rate over the course of deployment and are analyzed in situ by CVAFS (Brown et al., 2011 The work by Brown et al. (2011) on the issue of memory effects in gold sorbents is applicable to PASs for GEM due to the widespread use of similar sorbents. Furthermore, handling and analysis techniques used in active samplers can, in many cases, be directly imported to research using a single PAS or even a network of the devices, assisting in the minimization of systematic errors. Among the many recommendations put forward, one of the most crucial is network planning. That involves the development of PAS histories for repetitively deployed PASs, the long-term goal being the eventual establishment of PAS cohorts that can be used interchangeably at sites with similar expected concentrations (Brown et al., 2011). In doing so, variations in PAS measurements as a result of memory effects are minimized due to the near constant carry-over 10 of Hg diffusing into the bulk gold between samples (Brown et al., 2011).

Establishing testing protocols
The testing of existing PAS designs has varied considerably from one device to the next, making PAS comparison problematic. Although opinions on exactly how to test the various factors that influence PAS measurements may differ, establishing a list of 15 factors that require testing would likely facilitate sampler comparisons, and in turn enhance future research in the field. Ideally, SRs should be estimated theoretically from the sampler design as well as being determined by calibration using a recognized active air monitoring system under both controlled, stable conditions and environmentally relevant conditions. As suggested by Gustin et al. (2011) the comparison of theoretical 20 and measured SRs would immediately identify PAS inefficiencies. However, theoretical or modelled assessments of SRs must have low uncertainties for this to be the case. Indeed, not accounting for reduced SRs caused by the "tortuous" path taken by molecules diffusing through the porous barrier may have been a substantial contributor to the discrepancies between theoretical and calibrated SRs in certain PAS designs 25 (e.g. Brown et al., 2012;Gustin et al., 2011;Skov et al., 2007). The influence of meteorological factors should be isolated and examined individually. Uptake curves should be computed for all PASs as they not only allow the determination of effective deploy-34630 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ment time ranges, but also the potential to ascertain uptake capacities of the sampler sorbents. Furthermore, if the PAS is designed for outdoor deployment, thorough testing across the full range of the conditions a PAS may experience should be undertaken along with simultaneous measurements from an active air monitoring system to assess calibrated SRs and the sampler's effectiveness in realistic settings. Finally, it is impor-5 tant for each specific sampler to attempt to determine the exact species of Hg being sampled (GEM, TGM or even TGM + PBM). This may be tested in the laboratory or by sampling alongside an automated Hg speciation sampler at sites where GOM and/or PBM are known to be elevated in proportion to GEM, such as coastal polar locations. 10 This critical review contains no new data. All data from tables and figures were obtained from data in existing manuscripts. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | mercury in Canada: results from the Canadian Atmospheric Mercury Measurement Network (CAMNet), Atmos. Environ., 37, 1003-1011, 2003.: Comparison of long-term trends and seasonal variations of atmospheric mercury concentrations at the two European coastal monitoring stations Mace Head, Ireland, and Zingst, Germany, Atmos. Environ., 39, ACPD 15,2015 Passive air sampling of gaseous elemental mercury: a critical review Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Vidic, R. D., Chang, M.-T., and Thurnau, R. C.: Kinetics of vapor-phase mercury uptake by virgin and sulfur-impregnated activated carbons, J. Air Waste Manage., 48, 247-255, 1998. Wheatley, B. and Wheatley, M. A.: Methylmercury and the health of indigenous peoples: a risk management challenge for physical and social sciences and for public health policy, Sci. Total Environ., 259, 23-29, 2000.