ARTICLE pubs.acs.org/est
Memory Effects on Adsorption Tubes for Mercury Vapor Measurement in Ambient Air: Elucidation, Quantification, and Strategies for Mitigation of Analytical Bias Richard J. C. Brown,*,† Yarshini Kumar,† Andrew S. Brown,† and Ki-Hyun Kim‡ † ‡
Analytical Science Division, National Physical Laboratory, Teddington, TW11 0LW, United Kingdom Atmospheric Environment Laboratory, Department of Environment and Energy, Sejong University, 98 Goon Ja Dong, Gwang Jin Goo, Seoul 143-747, Republic of Korea
bS Supporting Information ABSTRACT: The short- and long-term memory effects associated with measurements of mercury vapor in air using gold-coated silica adsorption tubes have been described. Data are presented to quantify these effects and to determine their dependence on certain relevant measurement parameters, such as number of heating cycles used for each analysis, age of adsorption tube, mass of mercury on adsorption tube, and the length of time between analyses. The results suggest that the long-term memory effect is due to absorption of mercury within the bulk gold in the adsorption tube, which may only be fully liberated by allowing enough time for this mercury to diffuse to the gold surface. The implications of these effects for air quality networks making these measurements routinely has been discussed, and recommendations have been made to ensure any measurement bias is minimized.
’ INTRODUCTION Air quality continues to be a concern because of its potentially adverse effects on human health and environmental sustainability. Of the pollutants present in air, mercury is of particular interest because of its combined qualities of toxicity and potential for bioaccumulation in aquatic and terrestrial biosystems. Moreover, because of the long atmospheric half-life and long-range transport of gaseous mercury species, emissions produced in one country or continent can have a significant impact on the ambient concentrations in a different country or continent. Therefore, pollution of the atmosphere and wider environment by mercury is now acknowledged to be a global problem.1,2 One of the major routes of human exposure to mercury is through inhalation. In order to assess the exposure of the general population to mercury in air, to ensure compliance with national and international legislation to limit mercury in air concentrations, and to further the global research effort, many nations now have air quality monitoring programs in place.3 In most cases, total gaseous mercury (mostly gaseous elemental mercury with varying but very small amounts of reactive gaseous mercury) is measured. It is clearly essential that the concentration data produced should be accurate and moreover traceable to the SI system of units to ensure comparability over time and location.4 r 2011 American Chemical Society
To achieve this goal, many analytical instruments and standardized measurement methodologies have been developed. The majority of measurements of total gaseous mercury in air rely on “trap and desorb” technology.5 Ambient air is pulled through a device capable of trapping any mercury contained; usually, the mechanism of trapping is by surface amalgamation on a high surface area gold structure contained within the “trap” or “adsorption tube”. The adsorption tube can then be heated up to high temperatures at which point the mercury desorbs and may be detected using atomic fluorescence spectroscopy or atomic absorption spectroscopy. The use of an adsorption tube has two major benefits: (1) it allows preconcentration of the sample to improve overall method detection limits in terms of ambient mass concentrations, so ambient background concentrations (on the order of 1 ng m3) are easily measurable, and (2) it allows for the remote sampling of air at multiple locations followed by measurement at on central location. The latter ensures that operating a multisite network is cost-effective as only one analytical instrument Received: April 30, 2011 Accepted: August 15, 2011 Revised: July 13, 2011 Published: August 15, 2011 7812
dx.doi.org/10.1021/es201454u | Environ. Sci. Technol. 2011, 45, 7812–7818
Environmental Science & Technology is required. In addition, measurement results only rely on the calibration of one instrument, so they are likely to show better comparability and less systematic bias. The role of the absorption tube is clearly critical to the accuracy of the measurement results obtained. It has previously been observed that the efficiency of collection and desorption on these tubes contributes to the overall measurement uncertainty.6 This paper demonstrates that there may be drawbacks to the use of remote sampling methodologies rather than installing automatic measuring instruments at each monitoring site as a result of a memory effect which these adsorption tubes can display under some conditions. While there has been some recent measurement science research into determinations of mercury vapor in air,711 very little of this has focused on the performance of the adsorption tube and none on possible memory effects. A couple of studies have identified the presence of a possible memory effect12,13 and proposed that short and long-term effects exist; there has been no work to quantify this effect, its dependence on input variables, or its potential to bias measurement results. This paper now addresses that deficiency.
’ EXPERIMENTAL SECTION The National Physical Laboratory (NPL) currently measures total gaseous mercury at 13 monitoring stations around the UK as part of NPL’s operation of the UK Heavy Metals Monitoring Network on behalf of the UK Government.3 Sampling is performed by pulling ambient air though a mercury adsorption tube containing gold-coated silica (Amasil, PS Analytical, UK) at an approximate rate of 100 mL/min using a pump (NMP 05 S, KNF Neuberger, UK) for sampling periods of one week to one month (depending on the expected ambient concentrations). The pump is calibrated, traceable to national standards, and its flow is measured at the beginning and end of each sample. An in-line particulate filter (0.4 μm GN-4 Metricel, Pall, UK) at the sample line inlet protects the adsorption tube from exposure to particulate matter. PFA tubing is used throughout, and the distance between the sample line inlet and the adsorption tube is kept to a minimum (approximately 5 cm). Typical masses collected on the adsorption tubes range from 0.1 ng to 50 ng. Sampled adsorption tubes are retuned to NPL where analysis takes place using a 10.525 Sir Galahad analyzer with an atomic fluorescence detector (PS Analytical, UK). The instrument is calibrated by use of a gastight syringe, making multiple injections of known amounts of mercury vapor, from a “bell-jar” calibration source8 onto the permanent trap of the analyzer. Prior to a series of measurements, the linearity of the instrument is confirmed using a series of five injections across the measurement range and the recovery of mercury from an adsorption tube dosed with a known mass of mercury is measured. Once linearity is established and the recovery is within allowable method parameters, the mercury content of the adsorption tubes is determined relative to an additional onepoint calibration. This calibration injection is made prior to the measurement of each adsorption tube and tailored to a level close to the mass expected to be present on the tube. Sampled adsorption tubes are placed in the remote port of the instrument and heated to 500 °C, desorbing the mercury onto a permanent trap. Subsequent heating of this trap to a similar temperature then desorbs the mercury onto the detector. The adsorption tubes go through this heating cycle three times. It is assumed that the result of the third analysis of the adsorption tube represents the blank associated with the adsorption tube, such that the analytical intensity
ARTICLE
attributable to the adsorption tube, Atube, is given by: Atube ¼ A1 þ A2 2A3
ð1Þ
where Ai is the analytical intensity associated with the ith sequential heating cycle. Because of the analytical procedure used, it is assumed that the adsorption tube is clean following analysis and ready to be used again for air sampling. As a result of remote sampling and centralized analysis conducted by the air quality network and ongoing research activities into mercury vapor measurement, NPL maintains an ensemble of approximately 100 adsorption tubes, replacing broken and degraded tubes periodically. In this way, the ensemble contains tubes of many different ages. At any one time, about 25 adsorption tubes are out in the field sampling air, 25 sampled adsorption tubes are awaiting analysis, 25 are awaiting dispatch to monitoring sites, and 25 are engaged in research investigations. Given that adsorption tubes are additionally periodically rotated between research and air quality network duties, the period between analysis for any given tube may be several months. If an adsorption tube has been inactive for a prolonged period or its status is unknown, it is good practice to reclean it prior to its next use by running an analysis on the tube. During this process, it has been noticed on occasions that the analytical response from some adsorption tubes is substantially higher than expected. This indicated that there might be a long-term memory effect associated with these adsorption tubes. In order to investigate this effect and its short-tem analogue, a full study was made on a variety of adsorption tubes to elucidate the dependence of the memory effects on relevant parameters such as: time between analyses, age of the adsorption tube, and mass of mercury originally collected on the desorption tube. To this end, synthetic dosing of tubes took place in the laboratory using known quantities of mercury removed from the bell-jar calibration apparatus and injected into a stream of mercury-free nitrogen flowing through the adsorption tube. Each analysis comprised five consecutive heating cycles, six in the case of investigation of the short-term memory effect. To investigate the short-term memory effect, adsorption tubes were dosed with between 5 and 50 ng of mercury and then analyzed using 6 repetitive heating cycles. To investigate the effect of adsorption tube age on the long-term memory effect, adsorption tubes of varying age were dosed with 5 ng of mercury, analyzed immediately, then analyzed for a second time after 1 week. To investigate the effect of the time between analysis on the long-term memory effect, adsorption tubes were dosed with 5 ng of mercury, analyzed immediately, and then analyzed for a second time after x days and for a third time after 2x days. To investigate the effect of originally dosed mass of mercury on the long-term memory effect, adsorption tubes were dosed between 5 and 50 ng of mercury, analyzed immediately, and then analyzed for a second time after 1 week. Measurement points presented are an average of the results from five adsorption tubes unless otherwise stated. The results of these analyses are presented either as analytical intensities or as masses of mercury liberated during analysis calculated using the instrument sensitivity determined by calibration of the instrument using the bell-jar apparatus prior to each heating cycle. Ordinary least-squares fitting, weighting according to the uncertainty in the y-component of the data, has been performed throughout using NPL’s XLGENLINE software.14 7813
dx.doi.org/10.1021/es201454u |Environ. Sci. Technol. 2011, 45, 7812–7818
Environmental Science & Technology
Figure 1. Analytical intensity obtained for each heating cycle, normalized to the fifth and sixth heating cycle, during the initial analysis of adsorption tubes dosed with different masses of mercury (in ng, as indicated in the key). The inset shows the intensity ratio, in percent, of the second to first heating cycles as a function of the mass of mercury initially dosed onto the adsorption tube, with a best-fit line included as a guide to the eye.
A field-emission scanning electron microscope (FE-SEM), Carl-Zeiss-Supra 40, was also used to examine the gold-coated silica material in the adsorption tubes. The SEM operating parameters were as follows: an accelerating voltage of 10 kV, a working distance of 3.7 mm, and a final lens aperture of size 30 mm. Secondary electron images were acquired using an In-Lens detector.
’ RESULTS AND DISCUSSION Short-Term Memory Effect. We indentify the presence of a short-term memory effect as a result of the incomplete desorption of mercury after the first heating cycle of the adsorption tube. This effect and its dependence on the mass of mercury loaded onto the adsorption tubes is shown in Figure 1. It is clear that the vast majority of mercury is liberated from the adsorption tube during the first heating cycle. For tubes with relatively low masses of Hg (for instance 10 ng or less), as might be experienced during ambient air sampling, the intensity is generally only a few percent of the intensity obtained during the first heating cycle. Above 10 ng, the percentage of the analytical intensity observed on the second heating cycle to that on the first rises to (by interpolation of the experimental data) about 5% at 25 ng and 10% at 50 ng. The profile of the repeat analysis from the second heating cycle onward is very similar, regardless of mass loading, showing a further significant decrease on the third heating cycle, and then more minor decreases in the fourth and fifth heating cycles, with almost no change observed between the fifth and sixth heating cycles. Regardless of original mass loading, a relatively similar intensity is recorded from the adsorption tubes after the fifth and sixth heating cycles. For most practical purposes, it is clear that using eq 1 to calculate the total intensity from an adsorption tube is fit for purpose, achieving better than 99.5% recovery of the mercury adsorbed on the tube for a 5 ng loading. Using eq 1 for much higher loading may result in recoveries only in excess of 97%, but this is still fit for purpose given the target combined standard uncertainty for measurements made under the European air quality directive is 25% (relative). It is clear therefore that a shortterm memory effect exists such that it requires several heating cycles to reach a steady state response for the mercury adsorption
ARTICLE
Figure 2. Analytical intensity from an adsorption tube dosed with approximately 5 ng of mercury, which was analyzed immediately (the first analysis, Day 1), and then subsequently after 1 week (the second analysis, Day 8) and then 2 weeks (the third analysis, Day 15), normalized to the minimum response from each analysis. Each analysis comprised 5 heating cycles, as shown. Note that the x-scale is logarithmic.
tube. Hence, a more rigorous but time-consuming analysis procedure to obtain the analytical intensity attributable to the adsorption tube for high accuracy work would be: Atube ¼ A1 þ A2 þ A3 þ A4 4A5
ð2Þ
Furthermore, it is observed that Ai for i g 5 is slightly greater that the intensity observed for an injection of blank gas onto the permanent trap of the instrument. This could be as a result of very small quantities of mercury liberated during analysis in the pipework between the remote trap (where the adsorption tubes are placed) and the permanent trap (where the calibrations injections are made, and where mercury liberated from adsorption tubes in the remote trap are reabsorbed). However, this slightly elevated response is still present even after new pipework is installed, suggesting that the origin of the signal is from the adsorption tube. These baseline levels are discussed later. Long-Term Memory Effect. An extreme example of the longterm memory effect is displayed in Figure 2. The nature of the long-term memory effect is clear from these results. It seems that by the fifth heating cycle on day 1 that a relatively stable baseline has been reached. However, when the adsorption tube is reanalysed after one week, there appears to be a significant quantity of mercury liberated, which then decreases to a baseline value again after five heating cycles. The third analysis after two weeks results in only small quantities of mercury being liberated over and above the baseline response. Hence, we may quantify the long-term memory effect as the quantity of mercury liberated from the adsorption tube, over and above the baseline level established during the first set of heating cycles, as a result of additional sets of heating cycles (the second, third, and subsequent analyses). We will call this quantity the excess mercury. In general, the second analysis seems to liberate an additional quantity of mercury from the adsorption tube, whereas it is much rarer to see significant mercury liberated on the third and subsequent analyses. Three parameters have been investigated with respect to the long-term memory effect of adsorption tubes: (1) age of the adsorption tube (which is taken to be an approximate surrogate for the number of previous heating cycles experienced); (2) the mass of mercury originally dosed onto the tube; and (3) the length of time between the first set of heating cycles and consecutive sets of heating cycles. The results of these investigations are given in Figures 3, 4, and 5. 7814
dx.doi.org/10.1021/es201454u |Environ. Sci. Technol. 2011, 45, 7812–7818
Environmental Science & Technology
Figure 3. Mass of excess mercury obtained from the second analysis of adsorption tubes originally dosed with 5 ng of mercury as a function of adsorption tube age. Each measurement point is the average of five repeats. The error bars represent the standard deviation of each point, and the dotted line represents a weighted least-squares best fit linear relationship.
Figure 4. Mass of excess mercury obtained from the second and third analyses (as indicated on the chart) of adsorption tubes originally dosed with 5 ng of mercury as a function of the time between analyses. Each measurement point is the average of five repeats. The error bars represent the standard deviation of each point. The dotted line represents a weighted least-squares best fit quadratic relationship (up to 20 d) and weighted linear relationship (after 20 d), for the data associated with the second analysis. The dashed line represents a weighted least-squares best-fit linear relationship, for the data associated with the third analysis.
The gradient on the relationship determined in Figure 3 by weighted least-squares regression was +0.002 ( 0.003 ng/yr. Not only is this gradient small compared to the absolute values observed but also, because of the large standard deviations associated with each point, the uncertainty in this gradient is larger than the gradient itself. It is therefore reasonable to state that there is no significant relationship between the excess mercury recovered as a result of the long-term memory effect and the adsorption tube age. The data in Figure 4 shows a clear increase in the mass of excess mercury recovered from the second analysis as the time between analyses increases. The increase slows as the time between analyses increases, resulting in a plateau for times in excess of approximately 20 days, where no increase in excess mercury is observed. For the third analysis, no significant gradient is observed (0.0001 ( 0.0010 ng/d) although, similarly to Figure 3, a small quantity of between 10 and 20 pg continues to be liberated consistently from the adsorption tube. Figure 5 indicates a significant positive relationship between the excess mercury measured following the second analysis and
ARTICLE
Figure 5. Mass of excess mercury obtained from the second and third analyses (as indicated on the chart) after one and two weeks, respectively, as a function of the mass of mercury originally dosed onto the adsorption tube. Each measurement point is the average of five repeats. The error bars represent the standard deviation of each point. The dotted line represents a weighted least-squares best fit linear relationship, for the data associated with the second analysis. The dashed line represents a weighted least-squares best-fit linear relationship, for the data associated with the third analysis.
the original mass for mercury dosed onto the tube: a gradient of +0.0016 ( 0.0004 ng/ng. However, the third analysis shows an insignificant gradient of +0.0002 ( 0.0008 ng/ng. The uncertainty in the relationship determined for the second analysis is sufficiently large so that it is not possible to propose with any confidence whether the relationship shows a plateau, similar to Figure 3, for high mercury masses or whether it continues to increase linearly. Several observations may be made about the long-term memory effect from the results presented. It is worth noting that the data shows a relatively high spread. This is thought to arise mainly from variability in the adsorption tubes themselves, and therefore, to some extent, the manifestation of the long-term memory effect is related to the individual history and characteristics of each adsorption tube. This imposed variability decreases the certainty with which conclusions may be drawn. This notwithstanding, it is possible to determine some general characteristics of the long-term memory effect. First, the excess mercury measured shows no dependence on tube age and history. This implies that there is no dependence on the structure of the gold surface in the adsorption tube, which is thought to change rapidly upon first usage of the tube, owing to the first Hg adsorption and heating cycles, but only to change very slowly after that.15 Second, it seems that the second analysis removes the majority of the excess mercury present on the tubes. The third analysis in most cases shows very little excess mercury above the baseline level. Third, it is apparent that the longer an adsorption tube is left after the first analysis the greater the quantity of excess mercury is liberated during the second analysis. However, there seems to be a maximum to the quantity of excess mercury liberated during the second analysis, presumably equal to the sum total of mercury remaining on the adsorption tube after the first analysis. Finally, the excess mercury recovered shows some proportionality to the mass of mercury initially dosed onto the adsorption tube. It is also instructive to consider the microstructure of the adsorption media within the adsorption tube. This has been done by scanning electron microscopy (an exemplar image is shown in Figure S1, Supporting Information). This has shown the porous nature of the surface silica structure and the relatively low surface coverage of gold on this silica surface. Generally, the 7815
dx.doi.org/10.1021/es201454u |Environ. Sci. Technol. 2011, 45, 7812–7818
Environmental Science & Technology gold is dispersed across the surface as 2050 nm islands, with occasional larger agglomerations on the surface. At a rough approximation, on the basis of the observed surface coverage of gold on each silica sphere and the number of silica spheres in the adsorption tube, the theoretical capacity of each tube is on the order of 10 μg of mercury. From these observations, some theories may be proposed to explain the long- and short-term memory effects. It may be assumed that when mercury is dosed onto the adsorption tube the vast majority is adsorbed onto the gold surface where it forms an amalgam; the affinity of silica for mercury is significantly less than that of gold. We now propose that a small proportion of this mercury in the surface amalgam diffuses into the bulk gold. Previously, it has been found that only reactive gaseous mercury (RGM) species (i.e., oxidized mercury) diffused into bulk gold16,17 and that elemental mercury at room temperature would be expected to remain as an amalgam on the gold surface, in which case the measurement of excess mercury measured as a result of the long-term memory effect would be a possible way of measuring RGM using a standard TGM sampling tube. However, as discussed later, the mercury diffusion into the bulk may not occur during sampling but instead during analysis when temperatures are elevated. It is not unreasonable to suggest that the quantity of mercury diffusing into the bulk under these circumstances might be a relatively constant percentage of the total mercury dosed onto the tube (hence, the results of Figure 5), with variability introduced by subtle changes in individual tube characteristics. During the first analysis, the majority of the mercury is liberated the first time the adsorption tube is heated. However, some mercury will not be completely removed by this first heating cycle. There may be a number of reasons for this. First, just like any multiphase system, there will be a partitioning of mercury in the gas and solid phase, and while the high temperatures which the tube is exposed to will mean that this equilibrium is shifted heavily in favor of mercury in the gas phase, some will still remain amalgamated to the gold. Second, heating is only conducted for a limited time in order to protect the instrument and adsorption tube from damage, and while most mercury will be released very shortly after heating, the mass of mercury liberated will show an exponential decay with time; this function may not fully decay by the time the heating is stopped. Third, the porous structure of the silica in the adsorption tube and the tight packing of the silica spheres may mean that some mercury liberated from the surface of the gold is not carried away efficiently by the gas stream flowing through the tube and reabsorbs elsewhere as the tube cools. These proposals explain the nature of the short-term memory effect, whereby the adsorption tube needs to be heated several times in succession before a baseline response is produced. At this point, the adsorption tube produces no excess mercury over and above its baseline response despite further heating. Furthermore, we now propose that the high temperatures experienced during analysis will also allow greater mobility of the mercury to move within the gold and therefore may encourage some surface adsorbed mercury to move deeper into the bulk of the gold, possibly only a few nanometers into the bulk. Any mercury absorbed into the bulk gold in this manner will most probably not be liberated at any significant rate by continued heating, unlike any remaining surface adsorbed mercury. Just as we propose a partitioning between mercury in the gas phase and surface adsorbed mercury, so there is a partitioning between surface adsorbed and bulk absorbed mercury which is
ARTICLE
known to be significantly in favor of surface adsorbed mercury.16,17 Thus, with the previously surface adsorbed mercury removed by the first analysis, it is proposed that this equilibrium is re-established as a result of mercury absorbed a few nanometers into the bulk gold diffusing toward the surface and becoming surface adsorbed. Given the relative small size of the gold deposits on the silica structure, the diffusion distances can only be a maximum of 10 nm and probably much less; however, as Figure 4 suggests, this process takes several days to complete. Therefore, after a period of time when the second analysis occurs, all of the mercury, which has migrated to the surface of the gold, is now desorbed during heating and measured, resulting in observation of the long-term memory effect. The same partitioning would be expected to happen during the second analysis as well, but by the time of the third analysis, the remaining memory effect will be very small compared to the baseline response and the variability of the tubes and is so not detectable. It is interesting to note that the proposed diffusion rate for mercury in bulk gold at room temperature, as observed in Figure 4, is on the order of many days. However, the long-term memory effect, which is proposed here to occur as a result of mercury becoming absorbed into the bulk, is seen in adsorption tubes that are first analyzed only a few hours after being dosed by only gaseous elemental mercury, when it has previously been observed that usually only oxidized mercury species diffuse into bulk gold.18 Hence, these observations provide additional evidence that the heating of the tubes during analysis may initially act to increase the rate at which mercury is absorbed within the bulk gold, possibly by lowering the activation energy barrier from the surface mercury to bulk mercury partitioning equilibrium. This observation has significant implications for monitoring in the field where non-negligible RGM components in ambient air are to be expected, furthering the possibility of additional mercury diffusion into the bulk gold. Adsorption Tube Baseline. It is notable that the baseline levels from these adsorption tubes correspond to approximately 10 pg above a measurement of mercury-free gas. The origins of these baselines, which show some variation between tubes, are not entirely clear. The baseline shows very little variation with repeated analysis, suggesting perhaps that it is less to do with the continual and consistent release of very low levels of mercury and more to do with the measurement of an interfering substance, most probably submicrometer sized particulate matter. The origin of this particulate matter is either from the carrier gas used for analysis (which although undergoing prefiltration will still contain some particles above 100 nm and many particles below 100 nm in diameter) or more probably from the degradation of material in the adsorption tube during heating. The particulate matter acts to cause a false positive signal from the atomic fluorescence instrument by scattering the incident radiation in the direction of the detector. This “mercury-free” baseline is significantly higher for new adsorption tubes; indeed, the manufacturers recommend repeated heating of new adsorption tubes prior to use to reduce this baseline to normal values. There is also experimental evidence gained during this study that suggests that this baseline might increase again for older tubes, possibly as they start to degrade more rapidly after a very large number of heating cycles. Implications for Ambient Air Monitoring of Memory Effects. Providing the adsorption tube is heated at least three times during a first analysis, it has been shown that the short-term memory effect may be largely eliminated. For normal ambient 7816
dx.doi.org/10.1021/es201454u |Environ. Sci. Technol. 2011, 45, 7812–7818
Environmental Science & Technology conditions where approximately 2 ng of mercury may be collected during a normal sampling period, heating the adsorption tube three times should recover in excess of 99.5% of the mercury that may be liberated during the first analysis cycle. For a similar sample of 2 ng of mercury collected on an adsorption tube, it has been shown that maximum long-term memory effects corresponding to between 5% of this mass (0.1 ng of excess mercury) have been observed in the second analysis. This potential bias is significant when compared to the overall uncertainty of the measurement method (on the order of 15% at the 95% confidence level). However, when the likely operation of an air quality network employing remote sampling and central measurement is considered, a large proportion of this possible measurement bias would be expected to be eliminated. The reason for this is that, providing the adsorption tubes are exposed to a similar quantity of mercury during each sampling period and experience a similar length of time between analyses, the excess mercury, resulting from the long-term memory effect from sampling period 1 and not recovered during the single analysis of this adsorption tube, would instead be liberated during the analysis of the sample collected during sampling period 2 and would be equal to the excess mercury not recovered as a result of the long-term memory effect attributable to sampling period 2. In this respect, the long-term memory effect is probably only relevant if sampling or analysis conditions change significantly for individual tubes from one analysis to the next. (This effect is explained and displayed diagrammatically in Figure S2, Supporting Information). In general, then, the long-term memory effect would be expected to have relatively little impact on air quality networks that rely on remote sampling followed by analysis at a central laboratory. However, it is conceivable that exposure of the same adsorption tube at an industrial site with high ambient concentrations, followed by exposure at a rural background site, might result in biases of up to 15%, comparable with the expanded uncertainty of the entire measurement. Therefore, there are certain recommendations that should be followed to ensure that the long-term memory effect imposes as little measurement bias as possible on air quality measurements: (1) Individual adsorption tubes should be deployed to locations where they are likely to sample similar masses of mercury during successive sampling periods. This may mean ensuring a set of tubes is used only at one particular site or at several sites where concentrations are very similar. Given that TGM concentrations in air show relatively little variation, planning such a strategy should be fairly simple. (2) New adsorption tubes should be conditioned with repeated injections of mercury (of expected ambient sample masses) and subsequent analysis prior to dispatch to monitoring locations (this should be part of the validation procedure for new adsorption tubes in any case). (3) As far as is practicable, the period of time between successive analyses of the same adsorption tube should be kept constant. If large differences in the time between analyses occur, it may be worth redosing and reanalyzing the adsorption tube in question in the laboratory to minimize the possible bias resulting from the long-term memory effect. (4) Account should be taken of the monitoring history of an adsorption tube when ratifying results from air quality measurements, and the uncertainty of the measurement should be increased if necessary to take account of any likely bias as a result of the long-term memory effect. (Attempting correction for the long-term memory effect is not recommended because of the variability observed in the effect.)
ARTICLE
The effect of the issues discussed above on automatic analyses of mercury vapor are less simple to discern. In general, only one heating cycle is performed to desorb each sample, and samples comprise very low mercury masses (often
