Quantitative passive soil vapor sampling for VOCs-part 2 : laboratory experiments †

Geosyntec Consultants, Inc., 130 Research Canada. E-mail: tmcalary@geosyntec.com; 0861 University of Waterloo, Waterloo, Ontario, Fondazione Salvatore Maugeri, Padova, Ita Craneld University, Craneld, UK Columbia Analytical Services, Simi Valley, United States Environmental Protection Age Eurons/Air Toxics, Inc. (formerly Air Toxic Arizona State University, Tempe, AZ, USA † Electronic supplementary informa 10.1039/c3em00128h Cite this: Environ. Sci.: Processes Impacts, 2014, 16, 491


Introduction
Volatile organic compounds (VOCs) including chlorinated solvents and petroleum hydrocarbons are common conta-minants in soil and groundwater.Subsurface vapor intrusion to indoor air is a pathway of concern in human health risk assessments, which creates a need for effective monitoring technologies for VOC soil vapor concentrations.7][8] As a result, many regulatory guidance documents caution that passive soil vapor sampling is not quantitative and should only be used as a screening tool. 9,10This paper describes controlled laboratory experiments designed to demonstrate the performance of passive samplers under conditions that would be typical for soil vapor monitoring.This research team has related articles on mathematical modeling 11 and eld experiments. 12assive samplers are dened here as devices that contain a sorbent medium and uptake VOC vapors passively by diffusion or permeation.Concentrations are calculated using eqn (1): The mass (M) of VOC sorbed and sample duration (t) are measured typically with accuracy of 10% or better, so the key factor controlling the accuracy of passive diffusive sampler concentrations is the uptake rate (UR) of the sampler.
Uptake rates for quantitative passive samplers can be obtained in three main ways: (1) supplied by the vendor based on controlled exposure chamber tests, (2) interpolated from the uptake rates of similar compounds based on ratios of diffusion or permeation coefficients or (3) eld-veried via side-by-side contemporaneous duplicate samples collected using a conventional sampling method ("eld-veried uptake rates").The passive diffusive samplers included in this study all have experimentally measured vendor-supplied uptake rates, which distinguishes these devices from qualitative or semi-quantitative passive samplers (e.g., Petrex tubes™, 13,14 EMFLUX Cartridges™, 7 Beacon BeSure Passive Soil Gas Technology™, 15 and Gore™ Modules (formerly known as the Gore-Sorber®), 8 and similar devices that are not specically designed to control the uptake rate).Where vendor-supplied uptake rates were not available for some of the compounds in this study, they were interpolated from values for compounds of similar structure and mass (and collectively referred to hereaer simply as "uptake rates").No adjustments were made for temperature or pressure because all tests were performed at 24 C and atmospheric pressure.
Several of the samplers in this study are available in more than one variety with different uptake rates.High uptake rates allow lower concentrations to be measured in a shorter period of time.However, if the uptake rate is too high relative to the face velocity of air near the sampler, then the sampler may cause a localized reduction in the vapor concentrations, and an associated low bias in the measured concentration, which is referred to as the "starvation effect".For soil gas sampling, there is a greater risk of starvation compared to indoor or outdoor air sampling because soil gas ow rates are typically very low or negligible, and replenishment of vapors to the vicinity of the passive sampler occurs primarily by diffusion. 11his testing program focused on different compounds, concentrations and samplers (uptake rates, sorbent and extraction method).Test protocols for evaluating occupational indoor air quality monitors 16 were considered, but not employed because they address variables such as temperature, humidity and sampling duration, but in the subsurface, the humidity is almost always high, the temperature is insulated to some extent and long sample durations were not needed to quantify the concentration range of this study for most of the compounds tested for most of the samplers.Most of the tests were conducted using a steady gas velocity of 5 cm min À1 (owrate of 100 mL min À1 in a 5 cm diameter cylinder) through the exposure chamber to minimize the starvation effect in order to focus on the performance of the passive samplers for different compounds and different concentrations in a high humidity environment.Water can be adsorbed by carbon-based sorbents and this can cause poor retention of weakly sorbed analytes or interference during analysis, so the high humidity typical of soil gas was considered likely to pose challenges for some samplers.The gas velocity of 5 cm min À1 was very low compared to typical indoor air velocities (600 to 3000 cm min À1 is a common range of air ow velocities for testing passive samplers designed for indoor air quality monitoring 17 ), in keeping with the intent of assessing the performance of the passive samplers under conditions approximating soil vapor sampling.A series of samples collected under stagnant conditions was also included.

Experimental design and sampling methods
A concentration range of 1 to 100 parts per million by volume (ppm v ) was tested to evaluate the performance of the samplers over a sufficiently wide range to assess whether their response is linear with concentration.Ten VOCs were included in the supply gas mixture, spanning a range of chemical families (chlorinated ethenes, ethanes and methanes, aliphatic hydrocarbon and aromatic hydrocarbon) and properties (vapor pressure, solubility and solid phase partitioning) to represent the range of VOCs typically encountered in assessing contaminated land (Table 1).Two standard J-size cylinders were custom-lled with these compounds at concentrations of 10 and 100 ppm v in N 2 .These were prepared by Air Liquide America Specialty Gases LLC of Santa Fe Springs, CA. Naphthalene (NAPH) and 1,2,4trimethylbenzene (124TMB) have much lower vapor pressures than the other compounds, and to avoid potential condensation issues, NAPH was added at a concentration of about 1 ppm v in the 10 ppm v supply gas and neither compound was included in the 100 ppm v supply gas mixture.For the test at 1 ppm v concentrations, the 10 ppm v supply gas was diluted 10 : 1 with ultra pure nitrogen using a mass ow controller to deliver 9 mL min À1 of the supply gas and a needle-valve to deliver about 90 mL min À1 of nitrogen (veried periodically with a soapbubble owmeter).For the 10 and 100 ppm v tests, the supply gasses were delivered undiluted at a ow rate of about 100 mL min À1 , controlled using a mass ow controller and veried using a soap-bubble ow meter.
The following samplers were used in this study SKC Ultra™. 17This is a badge sampler, which operates by diffusion through a 2 cm diameter plastic cap with about 300 holes ($0.5 mm each) and is available with various adsorbent media to suit different target analytes.The Ultra has relatively high uptake rates ($10 mL min À1 ) because it was designed to provide good sensitivity over an 8 h occupational sample period with analysis by solvent extraction and gas chromatography/ mass spectrometry (GC/MS).It is also available with a lowuptake rate cap having only 12 holes, which reduces the uptake rates by a factor of about 25 to avoid saturating the sorbent when using the sampler in high concentration environments or for long sample durations.There is also a variety (the Ultra II™) designed for use with thermal desorption GC/MS for increased sensitivity.The Ultra with activated carbon and solvent extraction analysis was used for the 10 and 100 ppm v tests and the Ultra II with Carbograph 5 and thermal desorption analysis was used for the 1 ppm v tests to minimize the risk of non-detect results.
Radiello®. 18This sampler operates by diffusion through a porous plastic cylinder housing with a large (23 cm 2 ) crosssectional area that results in relatively high uptake rates.The Radiello is available with two different housings: the yellow body has uptake rates of $20 to 30 mL min À1 and was specically designed for use with analysis by thermal desorption.The white body has uptake rates that are a factor of about 2.5 times higher than the yellow body, and was designed for use with analysis by solvent extraction.The yellow body was used with solvent extraction in this study to reduce the risk of a low bias via starvation and avoid saturation of the adsorbent.The uptake rates were assumed to be the same as those for the thermal sorbent, which is reasonable if both sorbents act as a zero sink throughout the sample duration (this is a fundamental assumption for all sorptive passive samplers).The sample duration was only 30 minutes in this study, so the assumption that the graphitized carbon sorbent acts as a zero sink is considered reasonable.
3M OVM 3500™. 19This is a badge sampler which operates predominantly by diffusion through a microporous polypropylene sheet and adsorption onto an activated carbon sheet of similar size for analysis by solvent extraction GC/MS.The OVM has relatively high uptake rates ($20 to 50 mL min À1 ) because of the relatively large ($10 cm 2 ) cross sectional area.No low-uptake or thermal desorption varieties are available.
Waterloo Membrane Sampler™. 20,21In the WMS sampler uptake occurs via permeation through a membrane of polydimethylsiloxane and VOCs are retained by an adsorbent within a glass vial.The membrane has low permeability for water vapor (water may compete for adsorptive sites or interfere with the analysis) and prevents advective uptake by turbulence (which can cause a positive bias for other passive samplers in high velocity environments).The WMS sampler is available in either a 1.8 mL vial (WMS™) with an exposed membrane surface of about 0.24 cm 2 or a 0.8 mL vial with a smaller membrane area (0.079 cm 2 ) and proportionately lower-uptake rates (WMS-LU™).Analysis by thermal desorption is also an option for improved sensitivity or shorter sample durations, if needed, but was not required in this study.
Passive ATD tube samplers. 22,23Passive ATD tube samplersconsist of a standard Automatic Thermal Desorption (ATD) tube, which facilitates sample preparation because it can be placed directly on a GC/MS thermal desorption auto-carousel unit for analysis via EPA Method TO-17. 25Chemical desorption is also possible, but less common.The ATD tube sampler is normally used with a dust screen cap that has an opening larger than the tube itself ($4.5 mm I.D.), but can be tted with a cap (specially designed for this study) that has a $1.4 mm I.D. opening that reduces the uptake rates by a factor of about 10.
The laboratory apparatus consisted of a 1 m long Â 5 cm diameter glass cylinder with three side ports (inuent at the bottom, effluent at the top and a sampling port in the middle).A schematic diagram of the apparatus is shown in Fig. 1.The interior surface of the glass cylinder was passivated using a silanization process.The outer wall of the cylinder was wrapped with 1.6 cm diameter Tygon tubing, which was used to circulate water for temperature control.The cylinder and tubing were placed inside a 10 cm diameter clear acetate tube for structural support and mounted to a frame for stability.Two PVC and stainless steel gate valves were secured to the top of the acetate pipe by friction with Teon™ tape acting as a seal.The gate valves formed an air-lock, to allow samplers to enter and exit the chamber with minimal disruption to the concentrations inside.A supply of gas containing known concentrations of selected VOCs was humidied and fed through the apparatus.When deployed in the exposure chamber, the badge samplers (3M and SKC) had their face vertical, the WMS and ATD samplers faced down and the Radiello was aligned near vertical.
Stainless steel and nylon tubing were used to deliver the supply gas to the exposure chamber, with compression ttings used at all connections.All ttings were leak-tested by connecting the apparatus to a 100 mL min À1 ow of pure helium and monitoring all the ttings with a helium meter.

View Article Online
Adjustments were made as necessary until there were no measurable helium leaks in the regions immediately outside of the ttings.Three identical humidication vessels were used (one for each concentration) and the water in each vessel was spiked with a mixture containing each of the 10 neat liquid VOCs mixed in proportions such that aer dissolving into the water in the humidication vessel, the water would be approximately in equilibrium with the supply gas according to Henry's Law (Table 2).Each humidication vessel contained about 1 L of distilled, deionized water and a Teon-coated magnetic stir bar.The stir bars operated continuously and the supply gas was delivered to the bottom of the humidication vessel through 1/4-inch glass tubing with a porous ceramic cup at the bottom to generate a large number of small gas bubbles.This apparatus consistently delivered steady source vapor concentrations with a relative humidity of about 80%.
All three supply-gas systems were set up simultaneously (Fig. 1 shows only one for simplicity) and allowed to run continuously for a week at about 100 mL min À1 and monitored periodically with a MiniRae 1000 photoionization detector (PID) and sampled using an active (pumped) sorbent tube lled with Anasorb 747 and analyzed by solvent extraction GC/MS to document the attainment of stable conditions prior to the experiments.The temperature and relative humidity were monitored using a Madgetech RHTemp101A datalogger.
Testing was performed starting with the concentrations at 1 ppm v , followed by 10 ppm v and 100 ppm v to reduce potential effects of carryover from one test to the next.At least 60 h were allowed for the chamber to equilibrate with each new concentration.At a ow rate of 100 mL min À1 , more than 180 times the volume of the test chamber passed through the chamber prior to sampling.The sample port at the mid-point of the test chamber was periodically monitored during the stabilization period using the PID to assess the stability of total ionizable vapor concentrations inside the test chamber and verication testing using pumped sorbent tubes (50 mL min À1 for 20 min with Anasorb 747) and solvent extraction GC/MS analysis until concentrations stabilized.NAPH was slower to equilibrate than the other compounds, presumably because of its tendency to adsorb even to relatively inert surfaces.
For the 1 ppm v test, three replicates of each of the ve passive samplers and the 1 L Summa canister samples were collected over 30 minutes in random order (denoted using lower case a, b and c in Table ESI 1A-C †).For the 10 ppm v and 100 ppm v tests, additional Summa canister samples were collected at the beginning and end for a total of ve active samples (denoted a through e).For the 1 and 10 ppm v tests, samples were deployed with no delay between them.PID measurements made aer the 10 ppm v tests indicated that some of the samplers may have sufficient uptake to inuence the concentrations inside the chamber (e.g., 10% lower PID readings aer the sample period compared to before for the samplers with higher uptake rates), so a 5 minute interval was allowed for re-equilibration between samples during the 100 ppm v tests.The effect of this change is discussed further in the results section.
Analyses were performed by the laboratories considered by the study team to be most familiar with the respective samplers.

Low uptake rate sampler tests
Additional tests were performed using available low uptake rate varieties of the passive samplers.Two tests were performed at the midpoint concentration (10 ppm v ) with the supply gas ow velocity held at 5 cm min À1 (100 mL min À1 ) for the rst test to maintain consistency with the rest of the experiments.The second was performed with the supply gas shut off to assess the performance of the samplers in a setting with no net gas ow ("stagnant" conditions), which is a worst-case condition for low biases attributable to the starvation effect.No attempt was made to assess whether thermal convection may have contributed to advection within the column, but the temperature was held as constant as possible, so thermal convection was likely negligible.The SKC low-uptake sampler had no detectable concentrations for either of the rst two tests, so a third test was performed at 100 ppm v under stagnant conditions (only the SKC and ATD tube samplers were used in this test).The low-uptake varieties of passive samplers used for these tests were: Radielloyellow body with charcoal.SCK Ultra -12-hole cap with charcoal.WMS-LU -0.8 mL vial with Anasorb 747.ATD tubelow-uptake cap with Tenax TA.No low-uptake version of the 3M OVM 3500 is available, so it was not included in this set of tests.
Inter-laboratory testing was performed to ensure each analytical laboratory could adequately analyze samplers.Each analytical  View Article Online laboratory adhered to its own QA/QC program (method blanks, surrogate analysis, internal standard analysis, laboratory duplicate analysis, etc.).No signicant QA/QC issues were identied.

Results
The concentrations measured using each of the passive samplers and the Summa canisters are presented in Fig. 2(a-c 3M OVM 3500: not detected in 0.1 ppm v samples.SKC Ultra: not detected in the 1 ppm v samples.WMS: low bias of about 8Â in the 0.1 ppm v samples and 3Â in 1 ppm v samples. 2) MEK -Radiello: low bias by a factor of about 2 to 3.
ATD tube: not detected in the 1 and 10 ppm v samples.3M OVM 3500: low bias by a factor of about 3 to 5. SKC Ultra: high bias with thermal desorption @ 1 ppm v and low bias via solvent extraction at 10 and 100 ppm v .
WMS: not detected in the 1 ppm v samples, low bias by 2Â in 10 ppm v samples.
Naphthalene and 1,2,4-trimethylbenzene were the two compounds with the highest and second highest Koc values (Table 1), and MEK was the compound with the highest solubility.Less soluble and less sorptive compounds yielded better agreement between the passive samplers and Summa canisters.
The accuracy of the passive samplers is summarized in Table 3, which shows the relative concentration (C/C 0 ), where C is the average passive sampler concentration and C 0 is the average Summa canister concentration for each compound, sampler and concentration.Overall, the C/C 0 values were within the range of 0.5 to 1.67 (corresponding to an RPD of AE50% between the passive and active samplers) in 83% (110 of 133) of sampler/ compound pairs with detectable results.The C/C 0 values were generally higher for the 100 ppm v tests, which might be attributable to the fact that the chamber was allowed to reequilibrate for 5 minutes between samples.The compounds that showed the poorest comparison between the passive and active samplers were MEK and naphthalene.These compounds were specically included in this research because they were expected to be challenging compounds for passive samplers.

View Article Online
Note that for the 1 ppm v test, the SKC Ultra sampler was used with Carbograph 5 as the sorbent for better sensitivity and the result showed a high bias for MEK, which demonstrates the importance of sorbent selection.The precision of the passive samplers is summarized in Table 4, which shows the relative standard deviation (RSD, the standard deviation divided by the mean) for all the compound and sampler combinations.The RSD values for the passive samplers were comparable or better than the values for the Summa canister samples.In most cases, the RSD values were less than 15%, which is consistent with passive sampling protocol requirements for occupational monitoring, 26 especially  at the 10 and 100 ppm v levels where the mass was more readily resolved against reporting limits.A linear regression analysis was performed to calculate the slope, intercept and correlation coefficient (R 2 ) of the relation between the relative concentration (C/C 0 ) and absolute concentration in the chamber.An ideal correlation would have all C/C 0 values equal to 1.0, which would result in a regression with a slope of zero, an intercept of 1.0 and a correlation coef-cient (R 2 ) of 100%.Table 5 provides the regression parameters calculated.The intercepts were slightly lower than 1 (0.7 mean for 50 observations), which is attributable to the change in procedure for the 100 ppm v tests where 5 minutes was allowed between samplers for re-equilibration of the chamber concentrations, which resulted in slightly higher concentrations for the 100 ppm v test.Otherwise, the slopes were near zero for all but 124TMB and NAPH in the WMS and 3M OVM 3500 samplers.The R 2 values were above 80% for all but: MEK and NHEX for the WMS.124TMB for the ATD.12DCA, 124TMB and NAPH for the Radiello.
BENZ for the 3M OVM 3500 and.most of the compounds with the SKC Ultra.This demonstrates that different compounds pose challenges for each of the samplers, which is an area for further research.
The results for the low-uptake rate samplers are provided in Table 6.The Radiello sampler (yellow body), WMS-LU (0.8 mL vial) and the ATD tube sampler with the low-uptake rate cap (Markes International, Wales) showed average results within a factor of 0.72, 1.08 and 0.72, respectively of the Summa canister results in the 10 ppm v test at a ow rate of 100 mL min À1 , which shows the low uptake rate samplers have a comparable accuracy to the regular uptake rate samplers.Under no-ow conditions, the passive samplers showed average C/C 0 values of 0.47, 0.73 and 0.1, respectively, which were lower (by a factor of 0.65, 0.68 and 0.71, respectively) compared to the samples collected with 100 mL min À1 ow in the chamber.The low bias under no-ow conditions was similar for all three samplers even though they have considerably different uptake rates (about 25 mL min À1 for the Radiello, about 0.5 mL min À1 for the WMS-LU and about 0.05 mL min À1 for the ATD tube).The low-uptake rate Radiello also showed a low bias of 100Â for 124TMB, and a low bias of 5Â for tetrachloroethene (PCE) under no ow conditions, which are the compounds with the highest organic carbon partitioning coefficient (Koc) values and lowest free air diffusion coefficients (excepting NAPH which was not detected by the Radiello).The ATD tube sampler showed a high bias of 2Â for BENZ and 9Â for NAPH and a low bias of about 10Â for 1,1,1-trichloroethane (111TCA), carbon tetrachloride (CTET) and 124TMB.The SKC/Charcoal sampler with the low-uptake rate cap showed detectable concentrations for only 3 compounds in the 100 ppm v stagnant test, but the concentrations were quantied within a factor of 2 for all three.The WMS-LU sampler showed concentrations within 2Â for all compounds under both owing and stagnant conditions.

Conclusions
The results of this study indicate that passive samplers can provide vapor concentration measurements in settings similar to those expected to be encountered in passive soil vapor sampling and therefore may be a practical alternative for monitoring soil vapor concentrations for many of the volatile organic compounds of interest for human health risk assessment.Most of the concentrations measured with the passive samplers were within a factor of 2 or less of the concentrations measured with Summa canister/EPA Method TO-15 and the precision of the passive samplers was as good or better than the Summa canisters.This is encouraging considering that the passive samplers and analytical methods are all different and the samples were analyzed in different laboratories, and none of the vendor-supplied uptake rates were derived specically for short (30 minute) duration, high (80%) humidity, and low (5 cm min À1 ) face velocity settings.Low-uptake rate varieties of four of the samplers yielded similar accuracy to the regular uptake rate samplers, which is encouraging because low uptake rate samplers are expected to minimize the starvation effect in applications of passive soil vapor sampling. 12Highly soluble compounds (like MEK) or highly sorptive compounds (like NAPH) appear to be more challenging to quantify accurately than other compounds.The laboratory testing apparatus cannot simulate eld sampling of soil vapor exactly, so further in situ testing is needed.Field conditions could involve a broader range of chemicals, concentrations, sample durations and sampler design modications (sorbents, uptake rates).Until more is known about these variables, it is prudent to perform intermethod comparisons as a quality assurance procedure (e.g., collect adjacent samples for analysis by conventional methods in a certain percentage of locations to enable calculation of site-specic or eld-veried uptake rates).
Fondazione Salvatore Maugeri in Padova, Italy analyzed the Radiello samplers via solvent extraction GC/MS.The University of Waterloo analyzed the WMS samplers via solvent extraction GC/MS.AirZone One Ltd of Mississauga, Ontario analyzed the
), for the 1, 10 and 100 ppm v tests, respectively.The concentrations were calculated by dividing the mass of each compound adsorbed by each sampler (as determined by the analytical laboratory) by the product of the uptake rate and sample duration (30 min).The passive sampler concentrations were divided by the average of the concentrations measured with the Summa canister samples and EPA Method TO-15 analysis and presented as normalized (C/Co) concentrations.Tables ESI † 1A-C in the ESI present the uptake rates, individual concentrations measurements, the mean, standard deviation and the relative standard deviation (RSD, the standard deviation divided by the mean) for the three replicates for each sampler at each concentration level.Most of the samplers provided concentrations within a relative percent difference (RPD) of AE50% of the Summa canister values, with the following exceptions: 1) Naphthalene -Radiello: not detected.

Table 2
Volumes of pure compounds added to humidification vessel for 100 ppm v test

Table 3
Average concentrations measured with passive samplers divided by average concentrations measured with Summa canisters (C/C 0 ) a a NAnot available for SKC because two different sorbents were used.NDnot detected.NTnot tested.This journal is © The Royal Society of Chemistry 2014 Environmental Science: Processes & Impacts Paper Open Access Article.Published on 24 January 2014.Downloaded on 29/05/2017 15:45:44.This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

Table 4
Relative standard deviation (RSD) of concentrations measured in test chamber a a NDnot detected.NTnot tested.

Table 5
Linear regression parameters for normalized (C/C 0 ) concentration data for 1, 10 and 100 ppm v tests at 5 cm min À1 face velocity and 30 min sample duration a a * -not considered representative because of apparent laboratory blank contamination in 1 ppm v samples.This journal is © The Royal Society of Chemistry 2014 Environ.Sci.: Processes Impacts, 2014, 16, 491-500 | 497 Paper Environmental Science: Processes & Impacts

Table 6
Low-uptake rate sampler results (in mg m À3 ) for three tests: 10 ppm v with 100 mL min À1 flow; 10 ppm v stagnant, and 100 ppm v stagnant