Environ. Sci. Technol. 2001, 35, 4857-4867
Sampling Atmospheric Carbonaceous Aerosols Using a Particle Trap Impactor/Denuder Sampler BRIAN T. MADER, RICHARD C. FLAGAN, AND JOHN H. SEINFELD* Departments of Environmental Science and Engineering and of Chemical Engineering, California Institute of Technology, Pasadena, California 91125
A particle trap impactor/denuder system has been developed and tested for the sampling of ambient carbonaceous aerosols. Use of a particle trap impactor allows for a reduction of particle bounce and re-entrainment at high particle loadings, and operation at high volumetric flow rates is achieved without the use of oiled impaction substrates, thus facilitating the chemical and physical analysis of the organic compounds comprising the collected gas (G) and particle (P) phases. Honeycomb denuders have a greater density of channels for a given denuder crosssectional area than parallel plate or annular denuders; for a given sampling flow rate, honeycomb denuders can be fabricated in more compact shapes and will have a greater amount of surface area for the collection of gases. Field testing of the sampler was conducted primarily at night to minimize the evaporation of organic carbon (OC) from collected particles, which can result from the heating of collected particles as ambient temperatures rise during the day. In side-by-side testing with an open-face filter pack sampler, the denuder system was found to minimize positive gas adsorption artifacts caused by the adsorption of gaseous OC to quartz filter fiber (QFF) surfaces. In the denuder sampler, negligible amounts of OC were observed on a QFF placed downstream of a particle-loaded QFF, suggesting that OC detected on the backup QFF in the filter pack sampler resulted primarily from the adsorption of ambient G-phase OC rather than OC evaporated from particles collected on the front filter. Equations are presented for the evaluation of the magnitude of positive and negative sampling artifacts. Analysis of these equations indicates that the mass of OC evaporated from filter-bound particles present downstream of a denuder depends on (i) the volume of OC-free gas passed through the filter, (ii) the P-phase concentration and the P/G partition coefficients (Kp) of the compounds comprising the P-phase OC, (iii) the temperature (T) (values of Kp are inversely proportional to T), and (iv) the mass fraction of carbon in the compounds comprising P-phase OC. For these reasons, the magnitude of evaporative losses of OC in denuder samplers may vary among different sampling events. In addition, a method utilizing gas chromatography/mass spectrometry has been * Corresponding author phone: (626)395-4635; fax: (626)796-2591; e-mail:
[email protected]. 10.1021/es011059o CCC: $20.00 Published on Web 11/09/2001
2001 American Chemical Society
developed for determination of inertial impactor collection efficiency and denuder particle transmission efficiency. Using this method, only a single extraction of the sampler components is necessary, thereby reducing the number of extractions and analyses over conventional approaches by at least a factor of 2.
Introduction Typically 10-50% of the mass concentration (µg/m3) of atmospheric aerosols is comprised of organic carbon (OC) (1). Semivolatile organic compounds (SOCs) are a class of compounds comprising OC having vapor pressures between 10-11 and 10-4 atm and partitioning between gas (G) and particle (P) phases in the atmosphere (2). Aerosol optical properties, atmospheric lifetime, and behavior as cloud condensation nuclei depend on the carbonaceous material in the particles (1). Proper sampling of atmospheric carbonaceous aerosols that contain SOCs partitioned between the G and the P phases has occupied attention in the past (3-9). It is important to collect atmospheric aerosol in such a way as to separate gaseous compounds from the particle phase without biasing the measurement of the compounds’ G/P distributions. G- and P-phase SOCs are typically separated using filter/ sorbent samplers. The adsorption of a gaseous SOC to filter surfaces can cause a positive bias in the measured P-phase concentration (cp, ng/µg) and a negative bias in the measured G-phase concentration (cg, ng/m3) of that compound. These biases result in a positive artifact in the compound’s measured P/G partition coefficient (Kp ) cp/cg). Similarly, when collecting ambient atmospheric aerosols on a filter, the adsorption of gaseous OC can result in a positive artifact in the measured OC concentration (µg of C/m3) (3, 10, 11). When using filter pack samplers, a common method to correct for positive biases involves the use of backup filters in which a second filter (backup filter) is placed downstream of the front filter. Since the backup filter is exposed only to particle-free air, SOCs measured on this filter originate only from the gas phase. To correct for positive artifacts, the mass of the given SOC measured on the backup filter is subtracted from the mass observed on the particle-loaded front filter. This correction assumes that (i) the mass/filter area amounts of each SOC sorbed on the front and backup filters are equal, (ii) the gas adsorption capacity of the front and backup filter are equal, and (iii) the SOC observed on the backup filter did not result from evaporation of particles collected on the front filter. Mader and Pankow (10) have suggested that since the front filter will tend to reach equilibrium with the incoming gaseous SOCs first, such G/filter equilibrium might only be achieved after both filters have reached equilibrium with the gaseous SOCs in the sample air. Thus, if sampling ends before equilibrium is reached on both filters, such backup filter corrections will underestimate the extent of the gas adsorption to the front filter, and the mass amount subtracted from the front filter amount will be too small. Short sampling times exacerbate this problem. Recent laboratory work by Kirchstetter et al. (12) support the assertions of Mader and Pankow (10) and also suggest that the gas adsorption capacity of filters may vary among different lots of filters from the same manufacturer. Since aircraft measurements of OC are conducted at land speeds ≈180-540 km/h; sampling times are often minimized in order to achieve satisfactory spatial resolution between OC measurements. For this reason, measurements of OC VOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4857
FIGURE 1. Important processes occurring in a denuder sampler. made using filter pack samplers present in aircraft are especially prone to gas adsorption artifacts. To allow further reductions in sampling times, methods other than filter pack sampling must be developed. Denuders have been used to first remove gaseous OCs from the sample airstream before collecting the particles on a filter. Denuder samplers have been constructed of parallel plates (6, 13-15), annular tubes (8, 16-18), bundles of capillary tubes (19), and glass honeycomb configurations (7, 20, 21). These samplers operate at flow rates in the range of 2 to 225 L/min. Since some individual organic chemical species found in ambient atmospheric aerosols have atmospheric concentrations in the picogram to nanogram per cubic meter range, flow rates of some of these systems may be insufficient to collect enough mass of particles to measure individual organic species comprising the aerosol using current analytical chemical methods. Moreover, some of these systems are too large and/ or fragile to be placed in aircraft used for air sampling. Particles of different sizes may have different chemical compositions (22); often it is necessary to sample only particles of a certain size range. Plate inertial impactors and cyclones are commonly used as inlets to separate particles as a function of their aerodynamic size (23). Although such inlets can theoretically provide sharp size cutoffs, both types of inlets are subject to particle bounce and re-entrainment at high particle loadings (24). Moreover such impactors only approach their theoretical collection efficiencies when coated with organic greases or oils (25-29); however, such coatings severely confound the chemical analysis of organic aerosols. Particle trap impactors are a type of virtual impactor where particles are impacted into a cavity having a depth to width ratio .1. Such impactors do not require greases or oils. Biswas and Flagan (24) designed a particle trap impactor that provides a sharp size cutoff with a maximum collection efficiency of 90%; moreover, the collection efficiency was constant at high particle loadings. These properties of particle trap impactors make them suitable for sampling organic aerosols. New Features and Advantages of the Particle Trap Impactor/Denuder Sampler. To shorten sampling times and collect a sufficient mass of organic aerosol for measurements of the individual organic compounds, a high-volume particle trap/denuder sampler was designed for use on ground and 4858
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001
aircraft platforms. The denuder portion of the sampler removes gaseous OC that would adsorb to quartz fiber filters, and the particle trap collects aerosol above a given aerodynamic diameter. This sampler has a number of advantages over some previously existing technologies: (i) Reduction of particle bounce and re-entrainment at high particle loadings and operation at high volumetric flow rates is achieved without the use of oiled impaction substrates, thus facilitating the chemical and physical analysis of the organic compounds comprising the collected gas and particle phases. (ii) Honeycomb denuders have a greater density of channels for a given denuder cross-sectional area than parallel plate or annular denuders; for a given sampling flow rate, honeycomb denuders can be fabricated in more compact shapes and will have a greater amount of surface area for the collection of gases. (iii) For a given wall thickness, metallic denuders are mechanically stronger than conventional glass or ceramic denuders. Air samplers with coated metallic denuders are more rugged than conventional denuders. (iv) Operation at airflow rates higher than most conventional impactor (or cyclone)/denuder configurations can be achieved. The flow rates achievable with a particle trap impactor/denuder sampler allow sufficient amounts of picograms per cubic meter levels of specific organic compounds to be collected in a shorter amount of time than with most current impactor/ denuder combinations. An impactor/denuder sampler such as that described here can be used in aircraft-based sampling and also allows for increased temporal resolution when used in ground-based air sampling. In this paper, the design and performance of the sampler is presented, and a framework is developed to identify the important parameters affecting the magnitude of positive (gas adsorption) artifacts and negative (evaporation) artifacts. Particle Trap Impactor/Denuder Design. Particles greater than 10 nm diameter have diffusion coefficients in air at least 100 times smaller than those of gas molecules. Consequently, under laminar flow conditions, gases diffuse to the walls of an open channel much more rapidly than particles. On the basis of the physical-chemical properties of the gaseous compounds and particles of interest, denuder samplers can be designed to separate gases from particles. As illustrated in Figure 1, in an ideally operating system an ambient mixture of semivolatile gases and particles enters
FIGURE 2. Particle trap impactor/denuder sampler. a denuder, the gases diffuse to the walls of the denuder and are removed onto/into coating present on the wall, while particles are transmitted without loss through the denuder and deposited onto a filter downstream of the denuder. In reality, several artifacts can occur that complicate the evaluation of the G/P distribution of SOCs as measured in such a system: (i) small particles can be deposited via diffusion to the walls of the denuder, (ii) SOCs can evaporate from the particles as the gas phase is depleted of SOCs during transport through the denuder (negative artifact), (iii) gaseous SOCs can escape (breakthrough) the denuder and be adsorbed on the front and/or backup filter (positive artifact), and (iv) SOCs can evaporate from filter-bound particles (negative artifacts). Backup filters can be used to adsorb gaseous SOCs that evaporate from collected particles; however, the gas collection efficiency of the backup filter may be less than 100%. With these processes in mind, a particle trap impactor/ denuder sampler was developed consisting of four major sections: (i) a particle trap virtual impactor to remove particles above a given aerodynamic size, (ii) a coated stainless steel honeycomb denuder to remove gaseous OC that would adsorb to quartz fiber filters, (iii) a pair of filters to collect particles and evaporated gases; and (iv) a computerprogrammable volumetric flow controller to enable remote control of the sampling flow rate. The sampler is shown in Figure 2. The particle trap impactor contains a jet inlet to focus the aerosol stream onto the particle trap. Different diameter jets can utilized to provide different particle stream velocities, the purpose of which is to control the desired impactor particle size cutoff. In the experiments to be reported here, the jet diameter was 2.606 cm. The particle trap body consists of a metal disk containing ≈200 cavities, each oriented parallel to the jet. The ratio of the diameter of the disk to the diameter of the jet was 2. The top of each cavity is chamfered at 45° to eliminate horizontal surfaces and minimize particle bounce. The depth of each cavity is 15× its diameter. The ratio of the distance between the top of the particle trap body and the bottom of the jet to the diameter of the jet was 2.0 for the experiments reported here and can be adjusted to optimize the desired particle size cutoff. There is no airflow through the cavities of the particle trap. The denuder consists of a stainless steel honeycomb disk that can be coated with different stationary phases. The stationary phase tested was XAD-4 and was coated onto the denuder by Restek Chromatography Products (Bellefonte, PA). The denuder is designed such that at operating flow rates the residence time of the particles in the denuder is less than 0.2 s to minimize the mass of particle-phase SOC evaporated during transport through the denuder (30-32).
The Gormley-Kennedy equation (33, 34) was used to estimate the residence time and honeycomb cell diameter such that 99.999% of gaseous SOCs would be removed. (The ability of the denuder to remove gaseous SOC is discussed in an upcoming section.) The residence time, at a flow rate of 300 L/min, is short enough to ensure that 1, it is likely the particle will impact upon a surface. Measured values of ηi for the high-volume particle trap impactor used in the current study, the particle trap impactor of Biswas and Flagan (24), and an ungreased plate impactor studied by Pak et al. (29) are shown as a function of xStk in Figure 4. Also shown in this plot are data corresponding to a theoretical plate impactor (i.e., well-greased and perfectly operated). As shown in Figure 4, the high-volume particle trap impactor exhibits a reasonably sharp size cutoff. Moreover, the impactor has a maximum collection efficiency of 100% and a significantly higher collection efficiency for larger particles as compared to ungreased plate impactors. Pak et al. (29) suggest that for the ungreased plate impactor the dramatic decrease in ηi for larger particles was a result of particle bounce. For the sampling of organic aerosols, a particle trap impactor is preferable to an ungreased plate VOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4861
FIGURE 5. Fraction of particles lost to the wall of the sampler and the jet nozzle (ηw) as a function of particle diameter. impactor. As determined using the DOS and DOP calibration aerosols, for operation at a flow rate of 300 L/min, the highvolume particle trap impactor has a measured d50 of 7 µm, where d50 is aerodynamic diameter of particles collected with 50% efficiency. Losses of particles to the walls of the sampler (upstream of the denuder) and the nozzle of the jet were also evaluated. The collection efficiency of the wall and nozzle ηw was defined as
ηw )
Mw Mi + Md + Mf + M w
(3)
where Mw is the mass of aerosol extracted from the walls of the sampler (upstream of the denuder) and the nozzle of the jet. Values of ηw are plotted as a function of particle diameter in Figure 5. Losses are negligible for dp < 5 µm and never exceeded 10% at any size. Note that airflow in these regions is turbulent. (Rejet > 10 000 at 25 °C), where the characteristic length scale for Rejet is the jet diameter. Particle Transmission Efficiency of the Denuder. The denuder transmission efficiency was tested by comparing the amounts of EC measured on the QFFf,a and QFFf,c filters of the denuder and open-face filter pack sampler, respectively (Figure 3). At ambient temperatures, EC is effectively nonvolatile, residing exclusively in the particle phase. Moreover ambient EC is primarily present on particles with aerodynamic diameters between 0.07 and 1.0 µm (39). Using an estimate of the diffusion coefficient of particles of a given diameter and the residence time of particles in the denuder, it is possible to calculate, as a function of particle diameter, the fraction of particles that is lost to the walls by diffusion. In this way, particle transmission efficiency of the denuder can be estimated. Such calculations indicate that, for particles of 0.1 and 1.0 µm, approximately 3 and 0.2%, respectively, of particles entering the denuder will be lost to the denuder walls. If the mass of particles lost to the denuder walls by diffusion significantly affected the measured mass concentration of particles in this size range, the concentration of EC (µg/m3) measured using QFFf,a would be significantly less than that determined using QFFf,c. As shown in Table 2, for all 13 samples in which trains A and C were used, no significant difference existed in the EC concentrations measured using the denuder in train A or the open-face filter pack sampler. Thus, it is likely that an insignificant mass of particles in the size range of 0.07-1.0 µm is lost to the walls of the denuder portion of the sampler. Minimization of Positive Gas Adsorption Artifacts from the Use of the Denuder. Using two denuder samplers in parallel, it is possible to evaluate the reduction in the positive 4862
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001
gas adsorption artifact that results from the use of a denuder upstream of a QFF. The sampler configuration for these sampling events is shown in Figure 3 and described previously. Briefly, train B was configured with a TMF to remove P-phase OC and EC, a denuder, and a filter (QFFb,b) to adsorb G-phase OC not collected by the denuder. As shown in Table 2, seven sampling events were conducted with a QFFb,b present; in all these sampling events an insignificant mass of OC was measured on these filters. All samples had been field blank-corrected, on average the mass of OC present on QFF blanks was ≈3% of the mass of OC observed on QFFf,a and QFFf,c. These results indicate that the denuder effectively removed gaseous OC that otherwise could have adsorbed to QFFf,a causing a positive gas adsorption artifact in the measured P-phase OC concentration. Evaluation of Possible Evaporative Losses of OC on Particles Collected on Filters Downstream from the Denuder. Many compounds that comprise organic aerosols are semivolatile and partition between the G and P phases in the atmosphere (40). At equilibrium a given compound will have a P/G distribution governed by an equilibrium coefficient, Kp:
Kp (m3/µg) )
cp cg
(4)
When G-phase SOCs are removed during flow through a denuder, cg f 0 and SOCs will evaporate from the particle phase in an attempt to reestablish P/G equilibrium. This process can occur either during transport of particles though denuders (31) or when SOC-free gas is passed through a particle-loaded filter (41). As discussed by Liang and Pankow (41) and Rounds et al. (42, 43) for a given filter-bound P-phase SOC, if cg ) 0 (as is the case with a denuder operating at 100% efficiency), the rate of evaporation is such that the initial cg of the SOC (downstream of the filter) can be calculated using the equilibrium P/G partitioning coefficient of the given SOC:
cg )
cp Kp
(5)
The value of cp will be reduced as the given compound evaporates from the particle phase. Liang and Pankow (41) and Rounds et al. (42, 43) show that, for polycyclic aromatic hydrocarbons (PAHs) and n-alkanes, eq 5 is valid for situations where cp is approximately g70% of the initial value; as cp is reduced further, the overall rate of evaporation becomes limited by the rate of intraparticle diffusion; under these circumstances, eq 5 would overestimate the true cg. Using eq 5, it is possible to determine the maximum (equilibrium) possible evaporative loss for the given compound:
cp me ) Vscg ) Vs Kp
(6)
where Vs is the sample volume (m3). Thus, during denuder sampling, the mass of a given compound evaporated from the filter depends on the volume of SOC-free gas passed through the filter, the equilibrium P-phase concentration of the compound, and its Kp value. Liang and Pankow (41) show for environmental tobacco smoke (ETS) particles at 20 °C that 350 times more phenanthrene (log Kp) -3.18) is evaporated from particles collected on a filter than chrysene (log Kp) -0.66). Kp is function of temperature (T) (primarily through the influence of T on the vapor pressure of the compound) (44, 45); therefore, values of me will be T dependent.
TABLE 2. Concentrations (µg of C/m3) of OC and EC As Measured Using the Various Sampler Configurations sample event 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 a
train A (denuder:filter:filter)a filter
OC
error
EC
error
QFFf,a QFFb,a QFFf,a QFFb,a QFFf,a QFFb,a QFFf,a CIGb,a QFFf,a QFFb,a QFFf,a CIGb,a QFFf,a CIGb,a QFFf,a CIGb,a QFFf,a CIGb,a QFFf,a CIGb,a QFFf,a QFFb,a QFFf,a QFFb,a QFFf,a QFFb,a QFFf,a CIGb,a QFFf,a CIGb,a
10.75 0.00 18.66 0.00 8.94 0.00 8.43 51.22 10.83 0.00 8.93 52.54 6.72 39.15 7.29 66.72 7.38 37.54 8.40 57.27 7.73 0.00 3.81 0.00 6.02 0.05 3.49 22.95 7.67 45.74
0.72 0.24 1.25 0.42 0.67 0.28 0.59 2.71 0.77 0.25 0.77 2.90 0.51 2.11 0.62 3.56 0.58 2.06 0.69 3.14 0.65 0.21 0.42 0.20 0.53 0.03 0.41 1.34 0.64 2.50
2.61 0.00 2.15 0.00 2.34 0.00 1.54 nab 2.46 0.00 0.24 na 1.68 na 1.88 na 1.47 na 1.72 na 1.42 0.00 0.43 0.00 1.00 0.00 0.50 na 1.39 na
0.32 0.20 0.44 0.35 0.34 0.23 0.24 na 0.36 0.24 0.31 na 0.27 na 0.33 na 0.28 na 0.33 na 0.33 0.29 0.19 0.25 0.26 0.23 0.23 na 0.32 na
train B (filter:denuder:filter)a filter
OC
QFFb,b 0.00
error
EC
train C (filter:filter)a
error
filter
CIGb,b QFFb,b
QFFf,c 0.20 0.00 0.20 QFFb,c QFFf,c 0.00 0.28 0.00 0.23 QFFb,c QFFf,c 51.45 2.72 na na QFFb,c QFFf,c 0.00 0.25 0.00 0.24 QFFb,c
CIGb,b 52.75 2.95 na
CIGb,b CIGb,b QFFb,b QFFb,b QFFb,b CIGb,b CIGb,b
Values for trains A-C are field blank corrected.
b
EC
field blanks error
20.17 2.59 11.79 1.57 11.28 2.89 10.96 0.95
1.75 0.36 0.81 0.28 0.72 0.30 1.95 0.25
2.17 0.00 2.56 0.00 1.47 0.00 2.32 0.00
0.90 0.35 0.34 0.23 0.23 0.16 0.32 0.24
na
QFFf,c QFFb,c QFFf,c 63.45 3.40 na na QFFb,c QFFf,c 37.73 2.07 na na QFFb,c QFFf,c 46.46 2.55 na na CIGb,c QFFf,c 0.08 0.05 0.00 0.21 QFFb,c QFFf,c 0.00 0.24 0.00 0.25 QFFb,c QFFf,c 0.00 0.23 0.00 0.23 QFFb,c QFFf,c 23.56 1.36 na na CIGb,c QFFf,c 47.11 2.56 na na CIGb,c
CIGb,b 48.14 2.58 na CIGb,b
error
filter
OC
error
EC
error
QFF blnk 0.85
0.24 0.09 0.20
QFF blnk 1.34
0.42 0.05 0.35
QFF blnk 0.89
0.28 0.03 0.23
QFF blnk 0.06 CIG blnk 6.09 QFF blnk 0.20
0.16 0.02 0.16 0.47 na na 0.25 0.02 0.24
QFF blnk CIG blnk QFF blnk CIG blnk QFF blnk CIG blnk QFF blnk CIG blnk QFF blnk CIG blnk QFF blnk
0.20 1.00 0.20 0.46 0.26 0.72 0.21 0.55 0.27 0.81 0.27
0.24 0.00 0.20
QFFb,b 0.00 QFFb,b
OC
na
8.78 1.87 10.51 1.89 9.37 1.57 10.26 76.15 10.05 2.51 5.22 1.33 7.22 1.36 4.86 24.15 8.66 57.04
0.63 0.27 0.81 0.36 0.70 0.30 0.85 4.08 0.82 0.43 0.54 0.32 0.63 0.32 0.52 1.42 0.72 3.10
1.67 0.00 1.94 0.00 1.46 0.00 1.67 na 1.35 0.00 0.45 0.00 1.09 0.00 0.42 na 1.15 na
0.28 0.19 0.35 0.21 0.30 0.21 0.37 na 0.37 0.29 0.20 0.25 0.29 0.23 0.25 na 0.34 na
0.29 13.07 0.23 5.55 0.12 9.37 0.08 6.74 0.07 10.72 0.14
0.03 na 0.02 na 0.21 na 0.06 na 0.23 na 0.07
0.20 na 0.19 na 0.27 na 0.21 na 0.29 na 0.29
QFF blnk 0.10
0.24 0.20 0.25
QFF blnk 0.07
0.23 0.09 0.23
QFF blnk CIG blnk QFF blnk CIG blnk
0.24 0.54 0.26 0.81
0.11 6.00 0.05 10.78
0.08 na 0.05 na
0.24 na 0.26 na
na, not applicable.
Several authors have observed that QFF can adsorb gaseous OC (3, 11, 46-49); indeed a significant mass of gaseous OC was found adsorbed to QFFb,c present in the open-face filter pack sampler (Table 2). Moreover, Eatough et al. (14) observed that QFF can adsorb gaseous OC evaporated from collected particles. In the current study, QFFb,a was placed behind QFFf,a to collect gaseous OC evaporated from collected particles (Figure 3). As shown in Table 2, with the exception of sample event 13, no significant OC was observed on QFFb,a. (Note: All samples had been field blank-corrected, on average the mass of OC present on QFF blanks was ≈3% of the mass of OC observed on QFFf,a.) Since QFF may not collect gaseous OC with 100% efficiency, CIGs were used in eight sampling events as backup filters in trains A and B. (It has been reported that CIFs collect gaseous OC with 80-100% efficiency (14, 37, 50).) As shown in Table 2, a significant amount of OC was found on CIGb,a and CIGb,b; in the eight samples, OC levels were on average a factor of 6 times greater than those observed on blank CIG (OC levels determined on CIGb,a and CIGb,b in Table 2 are field blank corrected). For six of the eight samples there was no significant difference in the amount of OC on each CIG. The precision of the OC analysis has been found to be a function of OC concentration (51) and was determined as follows: 3-5 replicate measurements of OC were made using filter punches having the same OC loading, the standard deviation was calculated among these samples. The precision was determined in this manner over the range of filter OC loadings typically observed in the field. For the levels of OC observed on CIGb,a and CIGb,b, the average precision of the CIG analysis was 6%. In all eight sampling events conducted using CIGs, the levels of OC on QFFf,c > QFFf,a. Should this difference be primarily due to the evaporation of P-phase OC from QFFf,a (negative artifacts), then the OC concentration measured using CIGb,a should be larger than CIGb,b by a value equal to the difference in OC measured using QFFf,c and QFFf,a.
Although no significant difference in the OC levels on CIGb,a and CIGb,b was observed for six of the eight samples, the difference in OC concentration determined using QFFf,c and QFFf,a was within the precision of the OC analysis of the CIGs; it is not possible to determine whether evaporation of P-phase OC was primarily responsible for the lower levels of OC determined using QFFf,a as compared to QFFf,c. Had the level of OC on CIGb,a been significantly greater than CIGb,b, the amount of OC evaporated (OCevp) from particles collected on QFFf,a would have been calculated as follows:
OCevp (µg of C/m3) ) OCCIGb,a - OCCIGb,b
(7)
where OCCIGb,a and OCCIGb,b are the OC concentrations (not blank corrected) determined using filters CIGb,a and CIGb,b, respectively. Considering the precision of the OC analysis of CIGs (i.e., propagating the error of the OC analysis method for CIGs through eq 7) and the levels of OC determined using QFFf,a, it is possible to estimate that for these six sampling events less than 33% of the OC mass collected on QFFf,a was evaporated during sampling. In sample event 10, the level of OC measured using CIGb,a was significantly higher than using CIGb,b. From these data, it would suggest that the true OC concentration would be 19.21 µg of C/m3, which is more than 2.5 times higher than the OC levels observed on the previous and subsequent nights. Moreover, this would correspond to the highest OC level measured in the study and a value 2.5 times higher than the average OC level measured at this site. It seems unlikely that this value is correct. In sample event 7, the level of OC measured using CIGb,a was lower than CIGb,b; this observation is not consistent with the evaporation of OC from collected particles. The relatively high and equal levels of OC found on CIGb,a and CIGb,b suggests that some gaseous OC escapes the XAD4-coated denuders; however, this OC is not adsorbed by QFF. VOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4863
FIGURE 6. Thermograms of filter punches from sample event 11. FID response to evolved carbon (nA) is normalized by the volume of air passed through the given filter (m3) and plotted as a function of time. Also plotted are the analyzer oven temperature and carrier gas chemical composition. This phenomenon has also been observed with XAD-coated glass annular denuders used by Lewtas et al. (37), who suggested that this OC was gaseous volatile organic compounds (VOCs). In the current study, the gaseous OC collection efficiency of the denuder could be estimated as follows:
ηg,oc ≈
OCCIGb,c - OCCIGb,b OCCIGb,c
(8)
Equation 8 is an approximation of ηg,oc since it must be assumed that CIG collect gaseous OC with 100% efficiency and that TMFf,b and QFFf,c have the same gaseous OC collection efficiency. CIGs were used as backup filters in trains B and C during sampling events 10, 14, and 15. The average collection efficiency was ≈20% but varied from nearly 0 to 40%. Although the XAD-4-coated denuders do not remove gaseous OC with 100% efficiency, as shown previously, such denuders did remove gaseous OC that would adsorb to QFF; thus, these denuders did minimize positive gas adsorption artifacts. These observations suggest that only a small fraction (by mass) of the compounds comprising ambient gaseous OC have vapor pressures low enough and/or filter/gas partition coefficients high enough such that they will partition onto QFF. Comparison of OC Measured on Filters Present in Denuder and Open-Faced Filter Samplers. Comparing the results for trains A and C (Table 2), a significantly higher amount of OC was observed on QFFf,c than on QFFf,a. In addition, significant OC was observed on QFFb,c, but no significant OC was observed on QFFb,a. The OC adsorbed to QFFb,c could have originated from either the adsorption of ambient gaseous OC or the gaseous OC evaporated from particles collected on QFFf,c. Previously, it was shown that the denuders mostly removed gaseous OC that would adsorb to QFF; therefore, any OC on QFFb,a would have originated primarily from gaseous OC evaporated from particles collected on QFFf,a. However, insignificant amounts of OC were found on QFFb,a. It is quite unlikely that in these sampling events the evaporation of OC from particles collected on QFFf,c would be enhanced relative to QFFf,a. The OC observed on QFFb,c must have resulted primarily from the adsorption of ambient (native) gaseous OC to the filter surface and not from OC evaporated from particles collected on QFFf,c. Thermograms from the OC/EC analysis of filter punches from QFFf,c, QFFb,c, QFFf,a, and QFFb,a collected during sample event 11 are shown in Figure 6. In this plot, the flame ionization detector (FID) response of the OC/EC analyzer was normalized by the volume of air passed through the 4864
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001
filter during sample collection. This value (normalized FID response in Figure 6) is proportional to the amount of carbon evolving from the filter punch during analysis and can be used to directly compare of the amount of carbon evolved from the various filter samples collected in a given sampling event. Also shown in Figure 6 are the OC/EC analyzer oven temperature and carrier gas composition as a function of time. Primarily OC is evolved from the sample under a He atmosphere, while pyrolyzed OC and EC is evolved under a He:O2 atmosphere. As illustrated in Figure 6, the majority of the difference in the OC levels on filters QFFf,c and QFFf,a is due to differences in the amount of OC evolving during the first temperature step of the OC/EC analysis. As discussed previously, OC observed on QFFb,c was primarily due to the adsorption of ambient (native) gaseous OC to filter surfaces; the majority of OC present on QFFb,c evolved during the first temperature step. QFFb,a was used to collect gaseous OC evaporated from particles collected on QFFf,a; an insignificant amount of OC was observed on QFFb,a during any temperature step. The observation that no measurable OC was found on QFFb,a could be explainable as follows. The mass of OC evaporated from particles collected on QFFf,a was very low and/or such OC is not adsorbed by QFF. Although the gaseous OC collection efficiency of QFF may not be 100%, as shown previously such gaseous OC can be adsorbed by QFF (14, 46). It is possible for these samples that the Kp values of the compounds comprising the majority of P-phase OC and/or the sample volume were such that the mass of OC evaporated from particles collected on QFFf,a was small. It is possible that a significant fraction of some compounds (having low Kp values) may have evaporated from particles collected on QFFf,a but that these compounds may have comprised a small percentage of the total ambient P-phase OC mass. Furthermore, the majority of samples were collected at night under conditions of decreasing temperature; such conditions would minimize evaporation of OC from collected particles. The larger denuder (train D) was used in sampling events 11 and 12. For these sampling events, the OC concentration was 6.92 ((0.70) and 2.71 ((0.45) µg of C/m3, respectively, and the EC concentration was 1.42 ((0.42) and 0.46 ((0.23) µg of C/m3, respectively. The EC concentrations were not significantly different than those observed in trains A or C. The OC levels were 10-30% lower than was measured using train A. No significant OC was observed on backup QFF in train D. These data suggest that some OC was present on particles collected by the impactor. As mentioned previously, EC was expected to be present on particles smaller than 1 µm; the collection efficiency of the impactor for particles of this size was essentially zero. Equations for the Evaluation of the Magnitude of Positive and Negative Sampling Artifacts When Measuring OC. There is currently much debate regarding the importance of positive sampling artifacts (as caused by the adsorption of gaseous OC to filter surfaces) (3, 4, 9, 10, 12, 52, 53) versus negative artifacts (as caused by the evaporation of OC from filter-bound particles) (6, 13, 14, 50, 54) during measurements of ambient P-phase OC. In the following section, a framework is developed to identify the important parameters affecting the magnitude of each type of artifact. As shown by Mader and Pankow (10), at G/filter equilibrium the mass (µg) of a given compound adsorbed from the gas phase to a QFF is proportional to its G-phase concentration, cg (µg/m3):
mfa ) Kp,faceAfcg
(9)
where Kp,face (m3/cm2) is the compound’s filter face area normalized filter/G partition coefficient and Af is the crosssectional area of the filter face (cm2). (Note: For a given
compound, the value of Kp,face can vary significantly among different filter types, T, and relative humidities (RH) (10, 49, 55).) Moreover the minimum air sample volume to reach filter/G equilibrium with a single filter is
Vmin,f (m3) ) Kp,faceAf
OC evaporated from QFFf,a) would be the sum of the individual me,C values for all compounds evaporating from filter-bound particles:
me,OC )
(10)
∑m
e,C,i
) Vs
i
For a given compound, the value of mfa can be rewritten in units of mass of carbon adsorbed by multiplying mfa by the mass fraction of carbon (fc) in the given compound:
mfa,C (µg of C ) ) mfa fc fc )
(11)
Nc × 12 MW
(12)
where Nc is the number of carbon atoms in the compound and MW is the compound’s molecular weight. The mass of OC adsorbed from the gas phase onto a QFF (mfa,OC) (i.e., the mass of OC on QFFb,a, QFFb,b, or QFFb,c) would be the sum of the individual mfa,C values for all the G-phase organic compounds in contact with the filter:
mfa,OC )
∑m
fa,C,i
∑K
) Af
i
p,face,icg,i fc,i
(13)
i
Analysis of eq 13 indicates that the mass of gaseous OC adsorbed to filters will depend on (i) the compounds comprising gaseous OC (Kp,face is compound dependent), (ii) the G-phase concentration of the compounds comprising gaseous OC, (iii) the type of filter used (values of Kp,face vary among filters), (iv) the filter face area, and (v) the T and RH to which the filter is exposed (values of Kp,face are T dependent (55) and for QFF are RH dependent (49)). For a situation where the adsorption of gaseous OC to the filter surface is the major sampling artifact, the measured OC concentration as determined from the analysis of a particle-laden filter would be
mOC + mfa,OC Vs (14)
OCmeas (µg of C/m3) ) OCtrue + OCads )
where mOC is the mass of P-phase OC collected on the filter, OCtrue is the true ambient P-phase OC concentration, and OCads is the measured OC concentration due to the adsorption of gaseous OC to the filter surface. When all gaseous OC compounds achieve equilibrium with the filter, OCads can be calculated as
OCads (µg of C/m3) )
mfa,OC
∑K
p,face,icg,i fc,i
Af )
Vs
i
Vs
(15)
i
cp,i fc,i
(17)
Kp,i
The minimum measured P-phase OC concentration, as determined from the analysis of a particle-laden filter present downstream of a denuder, would be
mOC - me,OC Vs (18)
OCmeas (µg of C/m3) ) OCtrue - OCevp )
and
∑m 3
OCevp (µg of C/m ) )
e,C,i
i
) Vs
∑ i
cp,i fc,i
(19)
Kp,i
Analysis of eq 17 indicates that the mass of OC evaporated from filter-bound particles present downstream of a denuder depends on (i) the volume of SOC-free gas passed through the filter, (ii) the P-phase concentration and Kp values of the compounds comprising the P-phase OC, (iii) the temperature (values of Kp are inversely proportional to T (40)), and (iv) the mass fraction of carbon in the compounds comprising P-phase OC. For these reasons, the magnitude of evaporative losses of OC in denuder samplers may vary among different sampling events; such losses have been observed to vary greatly among locations, time of day, and seasons (5). As discussed previously, Kp values are inversely proportional to T. During outdoor sampling, 20-deg diurnal fluctuations in T are not uncommon; for some SOCs such a change in T would result in approximately a factor 4 change in Kp (56). Therefore, should T increase during a sampling event (i.e., as occurs when sampling begins in the evening or morning and ends in the afternoon), a substantial increase in evaporative losses of OC from filter-bound particles can occur. Conversely, control of filter temperature in air samplers could be used to reduce such evaporative losses of OC.
Acknowledgments This work was supported by Office of Naval Research Grant N00014-96-0119 and National Science Foundation Grant ATM-0001934.
Nomenclature Roman
According to eq 15, the OC concentration measured on QFFb,a, QFFb,b, or QFFb,c will be proportional to the total G-phase OC concentration in contact with the filter. Our observations are in agreement with eq 15, i.e., the OC concentration measured using QFFb,b < QFFb,c as a consequence of a denuder being used in train B to reduce ∑icg,i. Equation 6 described the maximum mass of a given compound that will be evaporated from filter-bound particles present downstream of a denuder, converting to units of micrograms of C:
me,C ) me fc
∑
(16)
When cp,i is approximately g70% of the initial value, the maximum mass of OC evaporated from filter-bound particles present downstream of a denuder (me,OC) (i.e., the mass of
Af
cross-sectional area of filter face (cm2)
Cc
slip correction factor (dimensionless)
cg
G-phase concentration (ng/m3)
CIF
carbon-impregnated cellulose filter
CIG
carbon-impregnated glass fiber filter
cp
P-phase concentration (ng/µg)
DOP
dioctyl phthalate
DOS
bis(ethylhexyl) sebacate
dp
particle diameter (cm)
EC
elemental carbon
fc
mass fraction of carbon in a given compound
VOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4865
G
gas
GC/MS
gas chromatography/mass spectrometry
F
particle density (g/cm3)
(m3/µg)
µ
gas viscosity (g/cm s)
Greek
Kp
particle/gas partition coefficient
Kp,face
filter face area normalized filter/G partition coefficient of an SOC (m3/cm2)
ηg,oc
gaseous OC collection efficiency of denuder (dimensionless)
Md
mass of a given compound extracted from denuder
ηi
particle collection efficiency of impactor (dimensionless)
me
maximum mass of a given compound evaporated from filter
ηw
particle collection efficiency of sampler walls + jet nozzle (dimensionless)
meC
maximum mass of a given compound evaporated from a filter (µg of C)
meOC
maximum mass of OC evaporated from a filter (µg of C)
Mf
mass of a given compound extracted from filter (ng)
mfa
equilibrium mass of a given compound adsorbed from gas phase to a QFF (µg)
mfa,C
equilibrium mass of a given compound adsorbed from gas phase to a QFF (µg of C)
mfa,OC
equilibrium mass of OC adsorbed from gas phase to a QFF (µg of C)
Mi
mass of a given compound extracted from impactor cavity (ng)
Mw
mass of aerosol extracted from interstage walls and nozzle of impactor (ng)
MW
molecular weight of a compound
nA
nanoamps
Nc
number of carbon atoms in a given molecule
OC
organic carbon
OCads
concentration of OC adsorbed from gas phase onto a QFF (µg of C/m3)
OCCIGb,a
concentration of OC determined using filter CIGb,a (µg of C/m3)
OCCIGb,b
concentration of OC determined using filter CIGb,b (µg of C/m3)
OCevp
maximum concentration of OC evaporated from particles collected on filter downstream of denuder (µg of C/m3)
P
particle
QFF
quartz filter fiber
Rejet
Reynolds number of air in jet nozzle (dimensionless)
SOC
semivolatile organic compound
Stk
Stokes number (dimensionless)
T
temperature
TMF
Teflon membrane filter
U
gas velocity (cm/s)
Vmin,f
minimum air sample volume to reach filter/G equilibrium
VOAG
vibrating orifice aerosol generator
Vs
sample volume (m3)
W
jet nozzle diameter (cm)
4866
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001
Literature Cited (1) Seinfeld, J. H.; Pandis, S. N. Atmospheric Chemistry and Physics: From Air Pollution to Climate Change; John Wiley and Sons: New York, 1998; pp 1113, 970, 99. (2) Pankow, J. F. Atmos. Environ. 1994, 28, 185-188. (3) McDow, S. R.; Huntzicker, J. J. Atmos. Environ. 1990, 24 (10), 2563-2571. (4) Turpin, B. J.; Huntzicker, J. J. Atmos. Environ. 1994, 28 (19), 3061-3071. (5) Eatough, D. J. BOSS. The Brigham Young University Organic Sampling System: Determination of particulate carbonaceous material using diffusion denuder sampling technology. In Gas and Particle Phase Measurements of Atmospheric Organic Compounds. Advances in Environmental, Industrial and Process Control Technologies; Lane, D. A., Ed.; Gordon and Breach Science Publishers: Amsterdam, 1999; pp 233-285. (6) Cui, W.; Eatough, D. J.; Eatough, N. L. J. Air Waste Manage. Assoc. 1998, 48, 1024-1037. (7) Koutrakis, P.; Sioutas, C.; Ferguson, S. T.; Wolfson, J. M.; Mulik, J. D.; Burton, R. M. Environ. Sci. Technol. 1993, 27, 2497-2501. (8) Gundel, L. A.; Lee, V. C.; Kariyawasam, R.; Mahanama, R.; Stevens, R. K.; Daisey, J. M. Atmos. Environ. 1995, 29 (14), 17191733. (9) Turpin, B. J.; Saxena, P.; Andrews, E. Atmos. Environ. 2000, 34, 2983-3013. (10) Mader, B. T.; Pankow, J. F. Environ. Sci. Technol. 2001, 35 (17), 3422-3432. (11) Turpin, B. J.; Hering, S. V.; Huntzicker, J. J. Atmos. Environ. 1994, 28, 3061-3071. (12) Kirchstetter, T. W.; Corrigan, C. E.; Novakov, T. Atmos. Environ. 2001, 35, 1663-1671. (13) Eatough, D. J.; Lewis, L.; Hansen, L. D.; Lewic, E. A.; Farber, R. J. Aerosol Sci. Technol. 1989, 10, 438-449. (14) Eatough, D. J.; Wadsworth, A.; Eatough, D. A.; Crawford, J. W.; Hansen, L. D.; Lewis, E. A. Atmos. Environ. 1993, 27A (8), 12131219. (15) Fitz, D. R. Aerosol Sci. Technol. 1990, 12, 142-148. (16) Possanzini, M.; Febo, A.; Liberti, A. Atmos. Environ. 1983, 17 (12), 2605-2610. (17) Lane, D. A.; Johnson, N. D.; Barton, S. C.; Thomas, G. S.; Schroeder, W. H. Environ. Sci. Technol. 1988, 22, 941-947. (18) Coutant, R. W.; Callahan, P. J.; Kuhlman, M. R.; Lewis, R. G. Atmos. Environ. 1989, 23 (10), 2205-2211. (19) Krieger, M. S.; Hites, R. A. Environ. Sci. Technol. 1992, 26, 15511555. (20) Koutrakis, P.; Ferguson, S. T.; Wolfson, J. M. Denuder for Gas Sampling, U.S. Patent. Harvard University, 1994. (21) Sioutas, C.; Wang, P. Y.; Ferguson, S. T.; Koutrakis, P.; Mulik, J. D. Atmos. Environ. 1996, 30 (6), 885-895. (22) Hughes, L. S.; Allen, J. O.; Kleeman, M. J.; Johnson, R. J.; Cass, G. R.; Gross, D. S.; Gard, E. E.; Ga¨lli, M. E.; Morrical, B. D.; Fergenson, D. P.; Dienes, T.; Noble, C. A.; Liu, D.-Y.; Silva, P. J.; Prather, K. A. Environ. Sci. Technol. 1999, 33 (20), 35063515. (23) Marple, V. A.; Rubow, K. L.; Olson, B. A. Inertial, gravitational, centrifugal and thermal collection techniques. In Aerosol Measurement: Principles, Techniques, and Applications; Willeke, K., Baron, P. A., Eds.; Van Nostrand Reinhold: New York, 1993; pp 206-232. (24) Biswas, P.; Flagan, R. C. J. Aerosol Sci. 1988, 19 (1), 113-121. (25) Rao, A. K.; Whitby, K. T. J. Aerosol Sci. 1978, 9, 77-86. (26) Rao, A. K.; Whitby, K. T. J. Aerosol Sci. 1978, 9, 87-100. (27) Lawson, D. R. Atmos. Environ. 1980, 14, 195-199. (28) Turner, J.; Hering, S. V. J. Aerosol Sci. 1987, 18 (2), 215-224.
(29) Pak, S. S.; Liu, B. H.; Rubow, K. L. Aerosol Sci. Technol. 1992, 16, 141-150. (30) Kamens, R. M.; Odum, J. R.; Fan, Z. H. Environ. Sci. Technol. 1995, 29, 43-50. (31) Kamens, R. M.; Coe, D. L. Environ. Sci. Technol. 1997, 31 (6), 1830-1833. (32) Strommen, M. R.; Kamens, R. M. Environ. Sci. Technol. 1999, 33, 1738-1746. (33) Gormley, P.; Kennedy, M. Proc. R. Ir. Acad. 1949, 52A, 163-169. (34) Ferm, M. Atmos. Environ. 1979, 13, 1385-1393. (35) Fuchs, N. A. The Mechanics of Aerosols; Pergamon: Oxford, 1969; p 205. (36) Berglund, R. N.; Liu, B. Y. H. Environ. Sci. Technol. 1973, 7 (2), 147-153. (37) Lewtas, J.; Pang, Y.; Booth, D.; Reimer, S.; Eatough, D. J.; Gundel, L. A. Aerosol Sci. Technol. 2001, 34, 9-22. (38) Birch, M. E.; Cary, R. A. Aerosol Sci. Technol. 1996, 25, 221-241. (39) Venkataraman, C.; Friedlander, S. K. Environ. Sci. Technol. 1994, 24 (4), 563-572. (40) Pankow, J. F. Atmos. Environ. 1994, 28 (2), 189-193. (41) Liang, C.; Pankow, J. F. Environ. Sci. Technol. 1996, 30, 28002805. (42) Rounds, S. A.; Pankow, J. F. Environ. Sci. Technol. 1990, 24, 1378-1386. (43) Rounds, S. A.; Tiffany, B. A.; Pankow, J. F. Environ. Sci. Technol. 1993, 27, 366-377. (44) Yamasaki, H.; Kuwata, K.; Miyamoto, H. Environ. Sci. Technol. 1982, 16 (4), 180-194. (45) Pankow, J. F.; Bidleman, T. F. Atmos. Environ. 1991, 25A (10), 2241-2249.
(46) Cotham, W. E.; Bidleman, T. F. Environ. Sci. Technol. 1992, 26 (3), 468-478. (47) Hart, K. M.; Pankow, J. F. Environ. Sci. Technol. 1994, 28, 655661. (48) Storey, J. M.; Luo, W.; Isabelle, L. M.; Pankow, J. F. Environ. Sci. Technol. 1995, 29, 2420-2428. (49) Mader, B. T.; Pankow, J. F. Atmos. Environ. 2001, 35, 12171223. (50) Tang, H.; Lewis, E. A.; Eatough, D. J.; Burton, R. M.; Farber, R. J. Atmos. Environ. 1994, 28 (5), 939-947. (51) Mader, B. T.; Flagan, R. C.; Seinfeld, J. H. Manuscript in preparation. (52) McDow, S. R. Sampling Artifact Errors. In Gas and Particle Phase Measurements of Atmospheric Organic Compounds; Lane, D. A., Ed.; Gordon and Breach Science Publishers: Amsterdam, 1999; pp 105-126. (53) Kim, B. M.; Cassmassi, J.; Hugo, H.; Zeldin, M. D. Aerosol Sci. Technol. 2001, 34, 35-41. (54) Cui, W.; Machir, J.; Lewis, L.; Eatough, D. J. J. Air Waste Manage. Assoc. 1997, 47, 357-369. (55) Mader, B. T.; Pankow, J. F. Atmos. Environ. 2000, 34, 48794887. (56) Pankow, J. F.; Bidleman, T. F. Atmos. Environ. 1991, 25 (10), 2241-2249.
Received for review June 12, 2001. Revised manuscript received September 10, 2001. Accepted September 18, 2001. ES011059O
VOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4867
