Evaluation and Application of a Passive Air Sampler for Atmospheric Volatile Organic Compounds

In this study, we propose a passive air sampler (PAS) that mainly consists of activated charcoal sorbent, a radial diffusive body, and a protective shield for the outdoor monitoring of volatile organic compounds (VOCs). Due to its reusable accessories and small size, this PAS is cost-effective and highly convenient for outdoor deployment. The estimated sampling rates (SRs) for 28 VOCs were very close to the theoretical values over a 6-month field test. We evaluated the performance of the PAS with benzene, toluene, ethylbenzene, and xylenes (BTEX) and determined the method detection limits (MDLs) (0.02–0.04 μg m), recovery values (95.3 ± 9.8–99.8 ± 7.5%), repeatability (3.2 ± 2.2– 4.6 ± 2.9%), and uncertainties (10.1 ± 3.8–12.9 ± 4.5%), which are comparable to those of previously reported samplers and meet the requirements of European Standards (EN) 13528. The deployment times for obtaining reliable time-weighted average (TWA) atmospheric concentrations of BTEX ranged from 2 days to a minimum of 3 months according to corresponding background concentrations in the ambient air. Moreover, the atmospheric concentrations of BTEX measured by the sampler in this study compared favorably with values from both active and passive sampling methods, with R ≥ 0.97 (p < 0.0001). Finally, a 12-month field test was performed after evaluating the sampler, and the results showed apparent seasonal variations of atmospheric VOC concentrations, which were dominated by the sampling location and wind direction.


INTRODUCTION
VOCs have gained attention during recent decades both indoors and outdoors (Murphy et al., 2010;Han et al., 2013;Cheng et al., 2017;Choi et al., 2017).They can easily enter the body through the air and cause health risks in response to long-time exposure (WHO, 2005).Moreover, they are important precursors of urban aerosols and tropospheric ozone (Shin et al., 2013;Pugliese et al., 2014;Han et al., 2017).The Hazardous Air Pollutants (HAPs) released by the United States Environmental Protection Agency (U.S. EPA), as well as the European Environmental Agency (EEA), involved certain VOCs such as BTEX as the extremely hazardous compounds.Benzene has also been appointed as a Group-I human carcinogen by the International Agency for Research on Cancer (IARC) (Demirel et al., 2014).Automobile exhausts, industrial emissions, petroleum refining and storage, surface coatings, solvent use, and natural emissions of animals and plants are the major sources of outdoor VOCs (Klimont et al., 2002;Qiu et al., 2014;Wang, 2014;de Castro et al., 2015;Mo et al., 2016;Zhang et al., 2017).Therefore, it is critical to study and monitor these compounds by developing reliable and effective techniques.
The atmospheric VOCs sampling can generally be divided into active and passive methods.Compared to active sampling, passive sampling has the advantages of low-cost, small size, uncomplicated construction, simple operation, easy transport, and non-power drive, which is thus particularly suitable for the long-term monitoring in remote regions and wilderness areas, simultaneous largescale measurement, and recording of individual exposure (Huang et al., 2018).Since the passive sampling of sulfur dioxide was introduced in 1973 (Palmes and Gunnison, 1973), studies of PAS have been continuously developing, e.g., aspects of sampling approaches, media, and analytes (Mayer et al., 2014).The performance of passive samplers for atmospheric VOCs has been validated in many laboratory and field studies, e.g., low detection limits, long deployment time, multiple sorbents, and optional sampling rates (Batterman et al., 2002;Jia et al., 2007).However, many PASs were designed for indoor sampling and lack accessories for outdoor application.Moreover, users usually directly employed the SRs which were derived from the laboratory to calculate the outdoor concentrations of VOCs which might possibly cause serious biases.In addition, current PASs are still costly particularly for developing countries, e.g., OVM, SKC, and Analyst samplers, which are all imported with high taxes.
In this paper, we present a PAS design for the precise monitoring of TWA atmospheric concentrations of VOCs in the field over a 12-month deployment time.It is costeffective with cheap accessories which can be reused after washing and is highly convenient for outdoor deployment with a small size.The influence of meteorological factors on SRs has been evaluated over a 6-month test.Furthermore, the sampler was compared to both active and passive sampling methods for monitoring atmospheric concentrations of BTEX.Finally, a 12-month field test was performed to measure the TWA concentrations of 28 VOCs under various meteorological conditions.

Sampler Design
This sampler design originates from a study for gaseous mercury (see Fig. 1) (McLagan et al., 2016), and it is used for monitoring atmospheric VOCs in this study.It consists of 530 ± 30 mg activated charcoal sorbent (Sinoreagent, No. 40007561) with a grain size of 36/50 mesh packed in a cartridge made of stainless steel cloth (53 mm height and 5.8 mm diameter), a radial diffusive body (Radiello, No. RAD120) made of microporous polyethylene (1.7 mm thick, 16 mm diameter, and 25 ± 5 µm average porosity), and a protective shield made of polyethylene terephthalate (76 mm height and 72 mm diameter).The addition of the shield protects the sampler from wind and precipitation, which overcomes the shortage of indoor samplers.Before and after deployment, a polyethylene cap is screwed onto the bottom of the shield and sealed with PTFE tape, then the sample is stored below -20°C before analyzing.During deployment, a polypropylene lid fitted with a stainless steel mesh screen replaces the polyethylene cap on the bottom of the shield.

Study Design
To derive SRs of BTEX in the ambient air, 8 samplers were deployed on the roof of the Ningbo Urban Environment Observation and Research Station (NUE ORS) of the Chinese Academy of Sciences (CAS) (29.75°N, 121.89°E), and each 2 samplers were retrieved after 1 week, 2 weeks, 3 weeks, and 1 month, respectively, as a deployment period, totaling a 6-month period in the field (48 samplers in all).Temperature, wind speed, and relative humidity ranged from -0.2-28.5°C,0.1-20.3m s -1 , and 26.9-100.0%,respectively.Simultaneously, two thermal desorption tubes (SUPELCO, No. 20235-U) both packed with the same 530 mg activated charcoal as sampler and sealed with quartz wool at both sides, were connected in series to uptake the analytes from air via sampling pump (No.CYB-301, Nanjing Tigerking Instrument Co., Ltd.), and were deployed about 30 cm from the samplers at the same height.The ambient air was pumped through outside and inside tubes in sequence at a constant flow rate.These tubes were retrieved after 2 weeks and 1 month, respectively, together with the samplers.At the beginning of each monthly deployment period, two blank samplers and tubes were hermetically sealed and stored, waiting to be analyzed together with other samples at the end of the month.

Theory
A PAS quantifies the TWA concentration of the analyte in air according to its amount in the sampler's sorbent over a fixed period of time.Due to the lack of external power, the PAS depends on turbulent diffusion and/or molecular diffusion to sorb the analyte, which is transferred from a sampled medium to a collecting medium by a concentration gradient between the two media (Seethapathy et al., 2008).
SR is a critical factor for quantifying the analyte of a PAS.An ideal SR is a constant at different sampling sites, while in fact, it varies with environmental factors, e.g., temperature, relative humidity, and wind speed (Özden Üzmez et al., 2015).Thus, it needs to be calibrated according to specific conditions to obtain more precise sampling data (Du et al., 2013).
The empirical SR (mL min -1 ) of an individual PAS can be determined via Fick's first law as: where M represents the mass of sorbed analyte (µg), t represents the deployment time (min), and C represents the TWA atmospheric concentration of the analyte over the deployment time (µg mL -1 ).Assuming that the analyte goes through an air-side boundary layer, the pore space of the diffusive body, and the internal airspace of the diffusive body sequentially, the theoretical SR can be estimated from the geometry of the sampler as: where D represents the molecular diffusion coefficient of the analyte in air affected by temperature T and pressure P (cm 2 min -1 ), h represents the height of the diffusive body (4.7 cm), v represents the porosity (0.496 ± 0.001), r a (1.8 cm), r d-out (0.8 cm), r d-in (0.63 cm), and r s (0.29 cm) represent the radii corresponding to the outside of the airside boundary layer, the diffusive body, the internal air space, and the sorbent cylinder, respectively (McLagan et al., 2016).

Analytical Methods
The procedure of sample analysis was as follows: introducing 2 mL of CS 2 (Aladdin, No. C103306) to the exposed cartridge in a glass tube, stirring the tube and letting it stand for 30 min.Injecting 1.0 µL of the CS 2 solution in the gas chromatograph (Agilent, No. 7890B) (250°C injector temperature, 5 mL min -1 nitrogen flow as carrier gas) with flame ionization detector (250°C, 300 mL min -1 air flow, 30 mL min -1 hydrogen flow, 25 mL min -1 nitrogen flow as make-up gas), the analytes were then separated via a GC column (SUPELCO, No. 25358, 105 m × 0.53 mm × 3.0 µm).In this study, the following temperature program was used: held for 10 min at 35°C, then followed by warming up to 200°C at a heating rate of 4°C min -1 , and held for 10 min.
Before use, activated charcoal was exposed to 450°C for 3 h.The values of blank samples showed no significant increase after storage for 1 month and were all below the MDLs.All cartridges and glass tubes were discarded after use.The diffusive bodies needed to be washed if they got dirty from airborne dust, while in this study they were all used only once.The breakthrough of the pumped samples in the test was prevented by connecting two thermal desorption tubes, and the sorption efficiencies were all above 90%.The flow rates of pumps remained at 200 mL min -1 , and were calibrated before and after deployment using a digital liquid flowmeter (JCL-2010, Juchuanghb Instrument Co., Ltd.) with the uncertainty of 3.7 ± 1.6% at 2σ.
The MDLs of BTEX were derived by multiplying three times the standard deviation of seven blank samplers ranging from 0.02 to 0.04 µg m -3 , which were comparable to many previous reported samplers or even lower (see Table S2).The method quantification limits (MQLs) were derived by multiplying ten times the standard deviation of the field blanks ranging from 0.07 to 0.13 µg m -3 .Recoveries were derived by analyzing three standard concentrations, ranging from 95.3 ± 9.8 to 99.8 ± 7.5%.The repeatabilities of the samplers in the field were derived by the mean relative standard deviation (RSD) ranging from 3.2 ± 2.2% to 4.6 ± 2.9%, which were more precise than many previous reported samplers, particularly for outdoor samplers.Considering both of the pumps and samplers, the expanded uncertainties of this passive sampling method under field test for BTEX ranged approximately from 10.1 ± 3.8 to 12.9 ± 4.5% at 2σ, thus meeting the requirements of the European Standards (EN) 13528 (Brown, 2000).In addition, the performance values of 28 VOCs are listed in Table S1.

Sampling Rates
After the 6-month field test, SRs of BTEX in the ambient air were derived from Eq. ( 1).The SRs are equal to the values of the regression line slopes in Fig. 2, i.e., 77.11 ± 6.47, 66.01 ± 5.74, 63.07 ± 6.57, 56.22 ± 7.54, and 60.29 ± 7.47 mL min -1 at 298 K and 1013 hPa, for benzene, toluene, ethylbenzene, m/p-xylene, and o-xylene, respectively.These values are all lower than the SRs measured in exposure chambers by the Radiello sampler as 80, 74, 68, 70, and 65 mL min -1 (Sigma Aldrich).The SRs determined in the laboratory are typically much higher than in the ambient air (Tolnai et al., 2000).This is possibly due to the lower outdoor temperatures, which might cause slower molecular diffusion in air.Another reliable reason is that the protective shield might increase the thickness of the air boundary layer and decrease the uptake rates of BTEX in the sampler (McLagan et al., 2017).Furthermore, wind speed has different influences on VOCs during outdoor deployment (Huang et al., 2018).As presented in Fig. 3, sampling rates of BTEX from PAS in this study were more stable between the wind speed from 1.9 and 4.5 m s -1 and did not have various increasing trends of high wind speed.Compared to Radiello sampler, the addition of the protective shield of the PAS helped to reduce the effect of wind speed on the SRs, particularly for higher wind speeds in the field.
All of the parameters in Eq. ( 2) were precisely measured from the sampler, and r a was derived from a model estimating SRs in a PAS for organic vapors (Zhang and Wania, 2012).The diffusion coefficients of BTEX in air were calculated at 298 K and 1013 hPa (Lugg, 1968).The theoretical SRs of BTEX were estimated as 75.23, 68.53, 60.94, 54.81, and 58.68 mL min -1 , respectively.These values are very close to the empirical SRs.In addition, the SRs of 28 VOCs are listed in Table S1.

Deployment Times
Assuming background BTEX concentrations of 2.04, 2.85, 1.75, 2.28, and 2.08 µg m -3 which were the mean concentrations of BTEX over 6 months, and using the evaluated SRs and MDLs above, the minimal deployment times of BTEX by the sampler are approximately 11, 5, 8, 13, and 14 h, respectively (McLagan et al., 2016).To ensure reliable quantification, it is advisable to take deployments of approximately 35,18,29,41, and 45 h, on the basis of the MQLs.The deployment times are inversely proportional to the BTEX concentrations in the field, which means that it will take much shorter deployment times under higher concentration conditions, e.g., in industrial regions.
The sorption efficiencies of pumped samples obviously indicated that the sorbent in the tubes did not approach its equilibrium uptake capacity after 1 month.Compared to the flow rates of pumps and the evaluated SRs above, the sampler is able to be deployed at least 3 months at the assumed background BTEX concentrations.Similarly, if the BTEX concentrations are lower than the background concentrations, e.g., in remote regions, the corresponding deployment times will be much longer.

Comparison between Active and Passive Samplers
The atmospheric concentrations of BTEX derived by the PAS in this study over a 6-month field test were calculated from Eq. (1) using the evaluated SRs above.Meanwhile, the mean concentrations of BTEX corresponding to the deployment times of the passive sampling were calculated from the proton transfer reaction mass spectrometer (PTR-MS), which was kept monitoring online at NUEORS.In Fig. 4, the regression line slopes of benzene and toluene are both close to 1, with R 2 ranging from 0.97 to 0.98 (p < 0.0001), indicating excellent agreement between the concentrations derived with both the methods.The ethylbenzene and xylenes were not shown for the same molecular weight that could not be distinguished by PTR-MS.
Radiello samplers were deployed at the same time with the samplers in this study over a 3-month period at NUEORS, following the procedure described in 2.2.Derived from the corresponding SRs, the BTEX concentrations of the two samplers are favorably compared, with R 2 ranging from 0.97 to 0.99 (p < 0.0001) (see Fig. 5).In addition, the performances of 16 passive samplers for BTEX from published studies are listed in Table S2.Compared to those commercial samplers, the sampler in this study showed excellent performance in the aspects of MDLs, recoveries and SRs.Particularly, it provided the best values of the repeatability and uncertainty, which are not even reported by many samplers.
Three different brands of activated charcoals with grain sizes of 20/50 mesh were tested by the same sampler in this study over a 3-month period at NUEORS (see Fig. 6).The BTEX concentrations by the three sorbents are almost close to each other, with R 2 above 0.99 (p < 0.0001).It indicates that the sorbent used in this study has a good performance compared to other commercial products.Most of the sorbents used in PASs were activated charcoal.Graphitized Carbon Blacks, Porous polymers, and Carbon molecular sieves were also commonly used (see SI Table 2).More introductions about sorbent materials can be found in the relevant review (Huang et al., 2018).

Field Tests
A field test started from Nov. 2016 and ran until Oct. 2017 at NUEORS.Two samplers were deployed for each month before replacement, and one blank sampler was hermetically sealed and stored at the same time.All samples were analyzed followed by the method in 2.4.Fig. 7 shows the mean monthly concentrations of BTEX derived from PAS in this study at NUEORS over a 12-month period.Obviously, all concentrations of BTEX were much higher during winter and reached peaks in December, then slowly decreasing to the lowest values in summer, finally increasing again in autumn.This result can be explained by meteorological factors and the geographic feature of NUEORS showed in Fig. 8.The NUEORS is surrounded by several industrial regions including auto parts factories, package plants, paper plants, etc.The pollutants from the industrial regions were transported by winds from the north and northwest in winter and descended at NUEORS, then sorbed by the samplers at the roof.In contrast, the winds from the southeast in summer blew clean air from the sea to NUEORS.Apparently, the sampling location and wind direction are the main determinants of BTEX concentrations in ambient air over a long deployment period, which is consistent with the study by Wang et al. (2016).Also, the photochemical reactions contributed to decrease the concentrations of atmospheric VOCs particularly in summer for higher solar radiation.In addition, the mean monthly concentrations of 28 VOCs are listed in Table S3, which show almost identical seasonal variations as BTEX.

CONCLUSIONS
In this study, we present a PAS for assessing atmospheric concentrations of VOCs that is particularly suitable for long-term monitoring in the field.Due to its reusable accessories and small size, this PAS is cost-effective and highly convenient for outdoor deployment.The addition of the shield protects the sampler from wind and precipitation, overcoming the shortcoming of indoor samplers.The sorbent used in this study exhibits high performance compared to other commercial products, and the sampler itself displays a comparable performance to that of previously reported devices for atmospheric BTEX in terms of MQLs, repeatability, recovery, and uncertainties.
The SRs obtained from the 6-month field test are very close to the theoretical values and lower than the corresponding SRs of the Radiello sampler, which is possibly due to the lower uptake rates of the PAS and lower outdoor  The mean monthly concentrations of BTEX derived from the PAS at NUEORS from Nov. 2016 to Oct. 2017 indicate that the sampling location and wind direction are the main determinants of ambient VOC concentrations over a long period.It is anticipated that additional samples taken in the field under various meteorological and geographical conditions will be obtained and analyzed.Finally, it is important to calibrate the SRs for meteorological factors when conducting outdoor sampling, although the biases usually remain small in normal environments.

Fig. 1 .
Fig. 1.Structural schematic of the PAS in this study.The diagram is not to scale.

Fig. 2 .Fig. 3 .
Fig. 2. The uptake volumes of BTEX by the PAS versus the deployment times (p < 0.0001).The uptake volume is the ratio of M and C from Eq. (1).The Xylene includes m/p-xylene and o-xylene.

Fig. 4 .Fig. 5 .
Fig. 4. Comparison of passive and active atmospheric concentrations of Benzene and Toluene derived from PAS in this study and PTR-MS.The red solid line is the 1:1 relationship.(p < 0.0001)

Fig. 6 .
Fig. 6.Comparison of atmospheric concentrations of BTEX derived from PAS in this study using three different sorbents.RAD: activated charcoal of Radiello with a grain size of 35/50 mesh (No.RAD130).SIN: activated charcoal of Sinoreagent with a grain size of 36/50 mesh used in this study (No.40007561).SUP: activated charcoal of SUPELCO with a grain size of 20/40 mesh (No.10275).The Xylene includes m/p-xylene and o-xylene.The red solid line is the 1:1 relationship.(p < 0.0001)

Fig. 8 .
Fig. 8. Location of NUEORS and the surrounding industrial regions.