Research Gas/Solid Partitioning of Semivolatile Organic Compounds (SOCs) to Air Filters. 3. An Analysis of Gas Adsorption Artifacts in Measurements of Atmospheric SOCs and Organic Carbon (OC) When Using Teflon Membrane Filters and Quartz Fiber Filters BRIAN T. MADER† AND JAMES F. PANKOW* Department of Environmental Science and Engineering, Oregon Graduate Institute, P.O. Box 91000, Portland, Oregon 97291-1000
Adsorption of gaseous semivolatile organic compounds (SOCs) onto the filter(s) of a filter/sorbent sampler is a potential source of measurement error when determining specific SOCs as well as organic carbon (OC) levels in the atmosphere. This work examines partitioning to both Teflon membrane filters (TMFs) and quartz fiber filters (QFFs) for purposes of predicting the magnitude of the compound-dependent gas adsorption artifact as a function of various sampling parameters. The examination is based on values of Kp,face (m3 cm-2), the gas/filter partition coefficient expressed as [ng sorbed per cm2 of filter face]/[ng per m3 in the gas phase]. Values of Kp,face were calculated based on literature values of the gas/solid partition coefficient Kp,s [ng sorbed per m2 of filter]/[ng per m3 in gas phase] for the adsorption of various polycyclic aromatic hydrocarbons (PAHs), polychlorinated dibenzodioxins (PCDDs), and polychlorinated dibenzofurans (PCDFs) to TMFs, and for the adsorption of PAHs to QFFs. At relative humidity (RH) values below ≈50%, the Kp,face values for PAHs are lower on TMFs than on ambient-backup QFFs. The gas adsorption artifact will therefore be lower for PAHs with TMFs than with QFFs under these conditions. In the past, corrections for the gas/filter adsorption artifact have been made by using a backup filter, and subtracting the mass amount of each compound found on the backup filter from the total (particle phase + sorbed on filter) amount found on the front filter. This procedure assumes that the ng cm-2 amounts of each SOC sorbed on the front and backup filters are equal. That assumption will only be valid after both filters have reached equilibrium with each of the gaseous SOCs in the incoming sample air. The front filter will reach equilibrium first. The minimum air sample volume Vmin,f+b required to reach gas/filter sorption equilibrium with a pair of filters is 2Kp,faceAfilter where Afilter (cm2) is the per-filter face area. Kp,face values, and therefore Vmin,f+b values, depend on the compound, relative humidity (RH), temperature, and filter type. Compound-dependent Vmin,f+b values are 3422
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001
presented for PAHs and PCDD/Fs on both TMFs and QFFs. Compound-dependent equations which give the magnitude of the filter adsorption artifact are presented for a range of different sampling arrangements and circumstances. The equations are not intended for use in actually correcting field data because of uncertainties in actual field values of relevant parameters such as the compound-dependent Kp,face and gas/particle Kp values, and because of the fact that the equations assume ideal stepfunction chromatographic movement of gas-phase compounds through the adsorbing filter. Rather, the main utility of the equations is as guidance tools in designing field sampling efforts that utilize filter/sorbent samplers and in evaluating prior work. The results indicate that some backup-filter-based corrections described in the literature were carried out using sample volumes that were too small to allow proper correction for the gas adsorption artifact for some specific SOCs of interest. Similar conclusions are reached regarding artifacts associated with the measurement of gaseous and particulate OC.
Introduction Quartz fiber filters (QFFs) and Teflon membrane filters (TMFs) are frequently employed in filter/sorbent samplers used to measure the atmospheric gas/particle (G/P) distributions of semivolatile organic compounds (SOCs) (1-7) as well as the concentrations of G- and P-phase organic carbon (8, 9). Unfortunately, the adsorption of gaseous SOCs to filter surfaces can cause positive biases in measured P-phase concentrations (cp, ng µg-1), negative biases in measured G-phase concentrations (cg, ng m-3), and positive biases in measured values of the partition coefficient Kp (m3 µg-1) (3, 10, 11) where
Kp ((ng µg-1)/(ng m-3)) ) cp/cg (m3 µg-1)
(1)
As in G/P partitioning, gas/filter partitioning can be characterized using a Kp coefficient. Normalizing a gas/filter Kp value by the specific surface area af (m2 g-1) of the filter yields (12)
Kp,s (ng m-2/ng m-3) )
Kp(m3 µg-1) 10-6 (g µg-1) af (m2 g-1)
)
106 Kp/af (m3 m-2) (2) where the m2 in the units for Kp,s and in the units for af refer to actual m2 of adsorbing surface area of the filter. This type of normalization is relevant when the gas/filter partitioning occurs by adsorption. When comparing different filters of the same nominal size (e.g., 8 × 10 in.) in terms of their gas-adsorption artifact potentials, another useful partition coefficient is obtained by normalizing Kp values by the ratio of the face area of the * Corresponding author phone: (503)748-1080; fax: (503)748-1556; e-mail:
[email protected]. † Current address: Department of Chemical Engineering, California Institute of Technology, Pasadena, CA 91125. 10.1021/es0015951 CCC: $20.00
2001 American Chemical Society Published on Web 08/08/2001
filter Af (cm2) to the mass of the filter Mf (µg) according to
Kp,face (ng cm-2/ng m-3) )
Kp (m3 µg-1) 2
property
)
Af (cm )/Mf (µg) 3
-2
KpMf/Af (m cm ) (3) The cm2 in the units for Kp,face and Af refers to the cm2 of filter face area and not the actual surface area of the whole filter matrix. The relationship between Kp,face and Kp,s is 3
-2
TABLE 1. Physical Properties of QFFs and TMFs
6
Kp,face (m cm ) ) (afMf/10 Af) Kp,s
(4)
The filter surface area factor Sf (m2 cm-2) is defined here as
Sf (m2 cm-2) ) afMf/106Af
(5)
Kp,face (m3 cm-2) ) SfKp,s
(6)
so that
The m2 in the units for Sf refer to the actual m2 of adsorbing surface area of the filter; the cm2 in the units refer to the cm2 of face area. Relevant values of af, Mf, Af, and Sf are summarized in Table 1. The relative values of Sf indicate that a QFF of a given face area has over three times more sorbing surface area than a TMF of the same face area. Storey et al. (12) and Luo and Pankow (13) have measured Kp,s values for polycyclic aromatic hydrocarbons (PAHs) adsorbing to QFFs. Mader and Pankow (14, 15) have reported Kp,s values for various PAHs, polychlorinated dibenzodioxins (PCDDs), and polychlorinated dibenzofurans (PCDFs) adsorbing to TMFs and QFFs. This paper (1) utilizes those data to compare the affinities of TMFs and QFFs for these compounds and (2) develops an approach that allows estimates of the effect of gas/filter adsorption on measured values of cg, cp, and Kp, all as a function of filter type, compound, relative humidity (RH), total suspended particulate (TSP) material level, and sampling volume.
Adsorption of PAHs on “Ambient-Backup” QFFs and TMFs Air sampling with filter/sorbent samplers is sometimes carried out using both a front and a backup filter, with the intent of using analyses of the backup filter as a means of correcting for the adsorption of gaseous SOCs to the front filter (3, 8, 9). The mass amount of a given compound that will adsorb to a front filter at equilibrium (given a large enough sample volume) is determined by its value of Kp,face for adsorption to the filter, the Af of the filter, and the ambient value of cg of the compound. Predictions of the adsorbed mass amounts can therefore be attempted using either Kp,face data obtained in the laboratory for clean filters or Kp,face data inferred from measurements of SOCs found adsorbed to backup filters of the type of interest. Since backup filters become exposed to the ambient conditions experienced by the front filters, including all of the ambient background G-phase compounds, it seems possible that Kp,s values inferred from backup filters may be more indicative of Kp,s values for front filters than values measured in the laboratory with clean filters of the same type. This seems true for QFFs inasmuch as Mader and Pankow (15) report that at RH ≈ 40%, Kp,s values for PAHs for “ambient-backup QFFs” obtained in suburban Portland, OR were found to be three times smaller than Kp,s values for clean QFFs. This may be due to occupation of some of the more active sites on the QFF surface by compounds found in the ambient air. Ambient backup QFFs from locations other than Portland may well exhibit behavior similar to what Mader and Pankow (15)
af Af Mf Af/Mf Sf
QFFs
TMFs
dimensions (cm × cm) 20.3 × 25.4 20.3 × 25.4 specific surface area (m2 g-1) 1.65 0.21 filter face area (cm2) 516 516 filter mass (µg) 4.02 × 106 8.55 × 106 2 -1 -4 face area/mass (cm µg ) 1.3 × 10 6.0 ×10-5 filter surface area factor (m2 cm-2) 0.013 0.0035
observed. For TMFs, there does not appear to be a difference in the sorption properties of ambient-backup and clean filters (14). The Kp,s values of Luo and Pankow (13) and Mader and Pankow (14, 15) can be used with eq 4 to obtain a Kp,face comparison of the gas adsorption affinities toward PAHs of TMFs and the Portland ambient-backup (PAB) QFFs studied by Mader and Pankow (15). The results of the Kp,face calculations are summarized in Table 2 and Figure 1. The Kp,face values for fluorene and phenanthrene on the PAB QFFs at RH ) 37% are about the same as on TMFs at RH ) 2152%. The Kp,face values for pyrene and chrysene on the PAB QFFs at RH ) 37% are five to fifteen times larger than on TMFs at RH ) 21-52%. The Kp,face values on the PAB QFFs at RH ) 100% are as much as three times smaller than for TMFs at RH ) 21-52%. Thus, the differences in Sf (see Table 1) do not, by themselves, explain the differences in partitioning between the PAB QFFs and TMFs. The strength of sorption to PAB QFFs decreases as the RH increases. At least for the PAHs, though, the effect is smaller on the PAB QFFs (dlog Kp,face/dRH ≈ -0.011, this work) as compared to the clean QFFs (dlog Kp,face/dRH ≈ -0.025, (12)). On TMFs, there is little effect of RH on the partitioning of either PAHs or PCDD/Fs (14). Overall, the data suggest that for RH values below about 50%, the Kp,face values of PAHs are lower on TMFs than on the PAB QFFs. This is consistent with the generally held view that use of TMFs helps to minimize gas adsorption artifacts for SOCs.
Minimum Sample Volume Required To Attain Gas/Filter Equilibrium General. If a given gaseous compound has a significant affinity for a filter surface, a very high percentage of the gas can be removed from the air before the air leaves the filter, at least initially. This is due to the very high diffusion coefficients of gas molecules, a property which allows them to collide with a filter surface numerous times before exiting the filter. As a result, gaseous organic compounds can be collected with high efficiency using charcoal impregnated filters (16). When a gaseous compound passes through a QFF or TMF, the filter will continue to remove the compound from the sample stream until Kp,face equilibrium is reached between the incoming value of cg and the attained sorbed concentration on the filter. For a perfectly efficient sorption process, none of the compound will be able to pass through the filter until equilibrium through the entire filter is reached. The concentration of the compound in the gas leaving the filter remains at zero until the entire filter reaches equilibrium with the incoming gas-phase concentration. At that point, the exiting concentration “steps” up to the incoming concentration. While this type of ideal, step function chromatographic movement is never fully realized, it can be approached and is very useful as a conceptual model when developing a mathematical framework for predicting the magnitudes of filter adsorption artifacts. When backup filters have been used to correct for gas adsorption of SOCs to a front filter, the usual, implicit assumption has been that both filters have adsorbed the same amount of SOC of interest from the gas phase. An VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3423
TABLE 2. Values of log Kp,s (m3 m-2), log Kp,face (m3 cm-2), and Vmin,f+b (for 8 × 10 in. Filters) at 20 °C for Four PAHs on Clean QFFs, Portland Ambient Backup (PAB) QFFs, and TMFs log Kp,s (m3 m-2) @ 20 °C filter
fluorenea
phenanthrenea
pyrenea
chrysenea
ref
QFFs, clean, RH ) 25% QFFs, ambient backup, RH ) 37% QFFs, ambient backup, RH ) 100% TMFs, RH ) 21-52%
-0.59 -1.5 -1.8 -0.83
0.28 -0.61 -1.1 -0.020
1.6 0.86 0.015 0.78
3.1 2.4 1.2 1.8
(15) (15) (13) (14)
log Kp,face (m3 cm-2) @ 20 °C filter
fluorene
phenanthrene
pyrene
chrysene
QFFs, clean, RH ) 25% QFFs, ambient backup, RH ) 37% QFFs, ambient backup, RH ) 100% TMFs, RH ) 21-52%
-2.5 -3.4 -3.7 -3.3
-1.6 -2.5 -3.0 -2.7
-0.24 -1.0 -1.9 -1.7
1.23 0.56 -0.67 -0.61
Vmin,f+b (m3) @ 20 °C for two 8 in. × 10 in. filters
a
filter
fluorene
phenanthrene
pyrene
chrysene
QFFs, clean, RH ) 25% QFFs, ambient backup, RH ) 37% QFFs, ambient backup, RH ) 100% TMFs, RH ) 21-52%
3.4 0.37 0.20 0.54
26 3.2 1.1 2.3
593 97 14 22
18,000 3800 220 253
Log pLo (Torr) @ 20 °C ) -2.72, -3.51, -4.73, and -6.06 for fluorene, phenanthrene, pyrene, and chrysene, respectively.
TABLE 3. Values of log Kp,face (m3 cm-2) and Vmin,f+b (m3) (for two 8 × 10 in. Filters) at 20 °C for Selected PCDD/Fs on TMFs (14) compound
log Kp,face (m3 cm-2) @ 20 °C
Vmin,f+b (m3) @ 20 °C
2-MonoCDF 28-DiCDF 238-TriCDF 2378-TCDF 1-MonoCDD 28-DiCDD 124-TriCDD 1234-TCDD 12478-PeCDD
-2.6 -2.0 -1.1 -0.46 -2.4 -1.6 -1.4 -0.59 0.04
2.4 9.6 84 362 4.2 23 46 270 1100
attempt is then made to estimate the true P-phase mass of the SOC on the front filter as being equal to the total mass found on the front filter minus the mass found on the backup filter. There is a potential for a significant problem with this correction approach. This is because a nonzero volume of air is required to reach gas/filter equilibrium with the two filters, and the front filter will moreover reach equilibrium with the incoming G-phase compounds before the back filter. Thus, if sampling ends before equilibrium is reached on both filters, such a correction will still underestimate the extent of the gas adsorption to the front filter: the mass amount subtracted from the total front filter amount will be too small. When sampling using a single filter, for each given compound, the minimum air sample volume V required to reach gas/filter sorption equilibrium is given by
Vmin,f (m3) ) Kp,faceAf
(7)
Multiplying both sides of eq 7 by cg gives the mass delivered on the LHS and the mass sorbed at equilibrium on the RHS. Vmin,f is thus the volume that can deliver the G-phase mass amount required to achieve gas/filter adsorption equilibrium on one filter. The minimum V required to reach equilibrium for a front+backup filter combination is
Vmin,f+b (m3) ) 2Vmin,f 3424
9
(8)
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001
Again, we note that only when the adsorption process is perfectly efficient will equilibrium in fact be reached when V ) Vmin,f (one filter) or when V ) Vmin,f+b (two filters). The Vmin,f+b values at 20 °C of individual PAHs on the PAB QFFs and TMFs are presented in Table 2; the Vmin,f+b values at 20 °C of individual PCDD/Fs on TMFs are presented in Table 3. When sampling ambient atmospheric SOCs, values of V have typically been in the range 100-3000 m3. We first note that gas adsorption artifacts will in fact tend to be exacerbated when short sample times (and therefore sample volumes at the low end of that range) are used as a means of trying to reduce artifacts caused by temperature fluctuations and other temporal changes during air sampling. Second, based on the Vmin,f+b values in Tables 2 and 3, we conclude that some of the backup-filter-based corrections that are described in the literature have been carried out using V < Vmin,f+b for some of the SOCs of interest. Thus, even with efforts to correct for gas adsorption using backup filters, gas adsorption may have nevertheless influenced some of the values cg, cp, and Kp that have been reported. log Vmin,f+b vs log poL. Figure 2 contains plots of log Vmin,f+b vs log poL for PAHs on PAB QFFs and for PAHs on TMFs. The correlation equations given in Figure 2 for QFFs are specific for PAHs and the RH values considered. Vmin,f+b values for PAHs on PAB QFFs at other RH values can be interpolated over the range 25% e RH e 78% by assuming dlog Vmin,f+b/ dRH ) dlog Kp,face/dRH ) -0.011. Since partitioning to Teflon surfaces is believed to be rather nonspecific in nature (14), the correlation equation for TMFs can probably be used for other compounds besides PAHs by utilizing the value of log poL for the compound of interest at the temperature of interest. As noted above, partitioning of PAHs and PCDD/Fs to TMFs has not been found to depend in any significant manner on RH (14).
Prediction of the Gas Adsorption Artifact General. When using a filter/sorbent sampler, sampling can take place either with just a front filter or with a front+backup (f+b) filter combination. The following sampling periods can be defined: (1) front filter only samplingsperiod f1, prior to
TABLE 4. Equations for the Prediction of Gas Adsorption Artifacts on Measured Values of cg, cp, and Kp sampling period
V
f1 and (f+b)1 prior to front filter equilibration (no backup correction possible) f2 after front filter equilibration (no backup correction possible)
V < Vf,min
(f+b)2 after front filter equilibration, but prior to backup equilibration, with backup correction
Vf,min e V < Vmin,f+b
(f+b)3 after front+backup filter equilibration, with backup correction
V g Vmin,f+b
V g Vf,min
FIGURE 1. Log Kp,face vs log poL for the partitioning of PAHs to TMFs and for the partitioning of PAHs to PAB QFFs (Portland ambient backup QFFs). Line for PAHs on PAB QFFs at RH ) 37% based on Mader and Pankow (15). Line for PAHs on TMFs at RH ) 21-52% based on Mader and Pankow (14). Line for PAHs on PAB QFFs at RH ) 100% based on Luo and Pankow (13). front filter equilibration (V < Vmin,f) and period f2, after front filter equilibration (V gVmin,f) and (2) front+backup (f+b) filter samplingsperiod (f+b)1, prior to front filter equilibration (V < Vmin,f); period (f+b)2, after front filter equilibration, but prior to backup filter equilibration (Vmin,f e V < Vmin,f+b); and period (f+b)3: after front and backup filter equilibration (V g Vmin,f+b). Equations for predicting the effects of gas adsorption artifacts on the measured values of cg, cp, and Kp, referred to here as cg,meas, cp,meas, and Kp,meas, are developed below for each of the above five periods. When the measured values are free of any errors, then cg,meas/cg ) 1, cp,meas/cp ) 1, and Kp,meas/Kp ) 1. Error predictions can be made for measurements made both with and without the application of backup filter corrections. The equations are summarized in Table 4. They are not intended for use in actually correcting field data because of uncertainties in actual field values of the compound-dependent Kp,face and gas/particle Kp values, and because of the fact that the equations assume ideal stepfunction chromatographic movement of gas-phase compounds through the adsorbing filter. Rather, the main utility of the equations will be as guidance tools in designing field sampling efforts that utilize filter/sorbent samplers and in assessing the general quality of specific data obtained with such samplers.
cg,meas/cg
(
(
1-
) (
Vmin,f V
(
2 1-
1
cp,meas/cp
1 1+ KpTSP
0
)
Vmin,f V
1+
1+
)
Vmin,f KpTSPV
(
)
2Vmin,f 1 -1 KpTSP V
1
Kp,meas/Kp ∞
(
)
)
Vmin,f KpTSPV Vmin,f 1V 2Vmin,f 1 1+ -1 KpTSP V Vmin,f 2 1V 1 1+
(
(
(
)
)
)
FIGURE 2. Log Vmin,f+b vs log poL for the partitioning of PAHs to two TMFs and for the partitioning of PAHs to two PAB QFFs. (Vmin,f+b values are for two 8 × 10 in. filters.) Periods f1 and (f+b)1: Prior to Front Filter EquilibrationsNo Backup Filter Corrections Possible. Backup filter corrections are not possible during period f1 because no backup filter is being used. Backup filter corrections are not possible during period (f+b)1 because we are assuming here that no amount of the compound of interest is yet on the backup filter. When using a filter/adsorbent air sampler with a single filter
cg,meas ) mA/V
(9)
where mA ) mass (ng) of the compound collected on the adsorbent (e.g., polyurethane foam). If the sorption of a gaseous compound to a filter surface is a perfectly efficient process, none of the compound will pass through the front filter until equilibrium is reached. During this period, mA ) 0 and so cg,meas ) 0. The P-phase concentration of a given compound as measured using a filter/adsorbent air sampler is calculated from
cp,meas ) mF/Mparticles
(10)
where mF ) mass of a compound extracted from the particleladen filter, and Mparticles ) mass of particles collected on the filter. We further note that VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3425
cp,meas 1 )1+ cp KpTSP
(17)
For this case, in the atmosphere, we also note that the fraction of compound that is in the G-phase is given by
Rg )
cg 1 ) cg + cpTSP 1 + KpTSP
(18)
where the second half of the equality refers to the case of G/P equilibrium. The fraction of the compound that is in the P-phase is
Rp )
KpTSP cgTSP ) cg + cpTSP 1 + KpTSP
(19)
where again the second half of the equality refers to the case of G/P equilibrium. When none of the G-phase material passes through the filter, both the G- and P-phase material will be measured as P-phase so that the result in eq 17 can also be obtained according to
cp,meas Rp + Rg KpTSP + 1 1 1 ) ) ) )1+ (20) cp Rp Rp KpTSP KpTSP The measured G/P partition coefficient of a compound is calculated as
Kp,meas ) cp,meas/cg,meas
FIGURE 3. Predicted cg,meas/cg vs V for sampling of selected PAHs at 20 °C with a single 8 × 10 in. filter, assuming the properties of a. PAB QFF at RH ) 37%; b. PAB QFF at RH ) 100%; and c. TMF at RH ) 21-52%. (No backup filter corrections possible.)
mF ) mf,p + mf,a
(11)
where mf,p, ) mass of a compound present in/on the particulate material collected on the front filter, and mf,a ) mass of the compound adsorbed from the G-phase on the front filter surface. The value of mf,p can be calculated from
mf,p ) cpMparticles
(12)
Prior to gas/filter equilibrium, we are assuming that all of a given gaseous compound will adsorb onto the front filter so that
mf,a ) cgV
(13)
(14)
where cg and cp are the true G- and P-phase concentrations of the compound. Since
Mparticles ) TSPV
(15)
substituting eq 15 into eq 14 and the result into eq 10 leads to
cg cp,meas ) cp + TSP
9
Kp,meas ) cp,meas/0 ) ∞
(22)
and the error in Kp,meas can be very large. For period (f+b)1, we note again that the mass of the compound that is adsorbed from the gas phase onto the backup filter (mb,a) is assumed to be zero, and so no corrections for gas adsorption using backup filters can be made. Period f2: After Front Filter EquilibrationsNo Backup Filter Corrections Possible. After equilibrium is achieved with the front (only) filter, cg,meas is calculated using eq 9, but mA is given by the difference between the total mass of gaseous compound entering the sampler and the mass adsorbed on the front filter
mA ) cgV - mf,a
(23)
Since the front filter is assumed to have reached equilibrium with the gas phase
(16)
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001
(24)
Substitution into eqs 23 and 9 leads to
(
) (
)
V - Vmin,f cg,meas Vmin,f ) ) 1cg V V
(25)
The value of cp,meas is given by the sum of the mass of a compound in/on the collected particulate material plus the mass of the compound adsorbed from the G-phase onto the filter surface
cp,meas ) (mf,p + mf,a)/Mparticles
so that assumption of G/P equilibrium yields 3426
As discussed above, if it is assumed that no portion of the compound of interest passes through the front filter during periods f1 and (f+b)1, then cg,meas ) 0 so that
mf,a ) Kp,face Af cg ) Vmin,f cg
Substituting for mf,p and mf,a in eq 11
mF ) cpMparticles + cgV
(21)
(26)
As in the other cases, we will assume that mf,p is given by eq 12 and that during this period mf,a is given by eq 24. Substitution in eq 26 leads to
FIGURE 4. Predicted cg,meas/cg vs V for sampling of selected PCDD/ Fs at 20 °C with a single 8 × 10 in. TMF at RH 21-52%. (No backup filter corrections possible.)
(
Vmin,f cp,meas ) 1+ cp KpTSPV
)
(27)
The value of Kp,meas/Kp is obtained from the ratio of eqs 27 and 25
Kp,meas ) Kp
(
Vmin,f KpTSP V Vmin,f 1V
1+
(
)
)
(28)
Period (f+b)2: After Front Filter Equilibration, Prior to Backup Filter equilibrationswith Backup Filter Corrections. As in period (f+b)1, we assume that mA ) 0. If a backup filter is used to correct for gas adsorption artifacts, then
cg,meas ) (mA + 2mb,a)/V
(29)
wherein it has been incorrectly assumed that mf,a equals the mass of compound that is adsorbed to the back filter, i.e., it is assumed that
mf,a ) mb,a
(30)
This assumption is not correct because the backup filter is not yet at equilibrium with the incoming gas. We estimate mb,a here as the difference between the total mass of gaseous compound entering the sampler and the mass that is adsorbed on the front filter:
mb,a ) cgV - mf,a
(31)
Since the front filter is assumed to be at equilibrium with the gas phase, mf,a is given by eq 24. Substituting eq 24 into eq 31 and the result into eq 29 yields
(
FIGURE 5. Predicted cp,meas/cp vs V for sampling of selected PAHs at 20 °C with a single 8 × 10 in. QFF with the properties of a PAB QFF at RH ) 37% at selected TSP values. (No backup filter corrections possible.)
)
cg,meas Vmin,f )2 1cg V
(32)
For cp,meas, when corrections are made using a backup filter
Kp,meas ) Kp
1+
(
2Vmin,f 1 -1 KpTSP V Vmin,f 2 1V
(
)
)
(35)
Period (f+b)3. After Front and Backup Filter Equilibrationswith Backup Filter Corrections. The compound is now assumed to have reached gas/filter equilibrium on both the front and backup filters, and corrections for gas adsorption artifacts are being made under the now-correct assumption that eq 30 is valid. We have
cg,meas ) (mA + 2mb,a)/V
(36)
The mass of the compound that will be found on the adsorbent is given by the two-filter analogue of eq 23, namely
mA ) cgV - mf,a - mb,a
(37)
cp,meas ) (mF - mb,a)/Mparticles ) (mf,p + mf,a - mb,a)/Mparticles (33)
Substituting eqs 30 and 37 into eq 36 yields
Substituting for mf,p, Mparticles, mf,a,, and mb,a using eqs 12, 15, 24, and 31 yields
In other words, if both the front and backup filters have reached equilibrium with the incoming gaseous compound, then backup filter corrections can be made to remove exactly 100% of the negative bias from cg. As a result, if cg is being measured correctly, then so too is cp, so that
(
cp,meas 2Vmin,f 1 )1+ -1 cp KpTSP V From the ratio of eqs 34 and 32, we have
)
(34)
cg,meas/cg ) 1
(38)
cp,meas/cp ) 1 VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
(39) 9
3427
FIGURE 6. Predicted Kp,meas/Kp vs V for sampling of selected PAHs at 20 °C with a single 8 × 10 in. QFF with the properties of a PAB QFF at RH ) 37% at selected TSP values. (No backup filter corrections possible.)
FIGURE 7. Predicted Kp,meas/Kp vs V for sampling of selected PAHs at 20 °C with a single 8 × 10 in. QFF with the properties of a PAB QFF at RH ) 100% at selected TSP values. (No backup filter corrections possible.)
and
adsorption can cause significant negative bias in cg,meas for a PAH such as chrysene when V < 750 m3; for PAHs which are two or three times less volatile than chrysene, there would be significant negative bias for V values smaller than about 2000 m3. PCDD/Fs are typically sampled using V values in the range 1000-3000 m3. Figure 4 indicates that if a single TMF is used, gas adsorption may cause a significant negative bias in cg,meas for 1234-TCDD and 2378-TCDF for V < ∼3500 m3; for PCDD/Fs which are two or three times less volatile than 2378-TCDF and 1234-TCDD, there would be a significant negative bias for V values smaller than about 8300 m3. cp,meas/cp. Figure 5 gives plots of cp,meas/cp vs V for PAHs at 20 °C for a single 8 × 10 in. QFF at RH ) 37% and three different values of TSP. Analogous plots can be prepared for PCDD/Fs and TMFs using the data in Tables 2-4. In all cases considered in Figure 5 in this single filter configuration, prior to gas/filter equilibrium (period f1), cp,meas/cp > 1. After gas/ filter equilibrium is attained (period f2), cp,meas/cp begins its asymptotic approach to 1 according to eq 27. A relatively larger value of TSP reduces the magnitude of the artifact effect on cp,meas because it increases mf,p relative to mf,a (see eqs 17 and 27). Kp,meas/Kp. Figures 6-9 give plots of Kp,meas/Kp vs V for PAHs and PCDD/Fs at 20 °C for a single 8 × 10 in. filter as determined by filter type, TSP, and RH. For all of the cases considered, as discussed above, Kp,meas/Kp ) ∞ in the period prior to gas/filter equilibrium (period f1) because cg,meas ) 0 during that time frame. After gas/filter equilibrium is attained (period f2), Kp,meas/Kp begins its compound- and TSPdependent asymptotic approach to 1 according to eq 28. The magnitude of the artifact tends to increase as the volatility of the compound and/or TSP decrease. When using QFFs at
Kp,meas )1 Kp
(40)
Diagnostic Plots for Sampling without Backup Filters General. The above equations can be used to construct diagnostic plots to evaluate the influence of gas adsorption artifacts on cg,meas, cp,meas, and Kp,meas as functions of V, compound-dependent Kp values, compound- and filterdependent Kp,face values, and TSP. For the PAHs, plots are constructed for both QFFs and TMFs. For PCDD/Fs, plots are constructed only for cases when TMFs are used only because Kp,face values for PCDD/Fs on QFFs are not currently available, and not because PCDD/Fs are never sampled using QFFs. The gas/particle Kp values used to prepare the plots were obtained from Ligocki and Pankow (17) for the PAHs and from Eitzer and Hites (1) for the PCDD/Fs. For plots involving QFFs, it was assumed that the QFFs behave like the PAB QFFs studied by Mader and Pankow (15). cg,meas/cg. Figure 3 gives plots of cg,meas/cg vs V for PAHs at 20 °C for a single 8 × 10 in. QFF at RH values of 37% and 100%, and for a single TMF at RH values in the range 2152%. Prior to gas/filter equilibrium for the single filter (period f1), cg,meas/cg ) 0. After gas/filter equilibrium is attained (period f2), cg,meas/cg begins its asymptotic approach to 1 according to eq 25. Figure 4 provides analogous plots for PCDD/Fs at 20 °C sampled with a single 8 × 10 in. TMF at RH ) 21-52%. PAHs in the atmosphere are typically measured with high volume air samplers and V values of 100-1000 m3. Figure 3 indicates that if a single QFF or TMF filter is used, gas 3428
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001
FIGURE 8. Predicted Kp,meas/Kp vs V for sampling of selected PAHs at 20 °C with a single 8 × 10 in. TMF at RH ) 21-52% and selected TSP values. (No backup filter corrections possible.)
FIGURE 9. Predicted Kp,meas/Kp vs V for sampling of selected PCDD/ Fs at 20 °C with a single 8 × 10 in. TMF at RH ) 21-52% and selected TSP values. (No backup filter corrections possible.)
low RH values to sample for PAHs, we conclude that the effect of gas adsorption on Kp,meas can be very large for some PAHs at typical values of V. Similarly, when using TMFs to sample for PCDD/Fs, the plots suggest that at typical sample volumes, the effect on Kp,meas can be large, especially when TSP is low, e.g., 10 µg m-3.
Comparison of Predicted and Measured Gas Adsorption Artifact Effects Mass Percent on a Front Filter Due to Gas Adsorption. The percent Pf,a of a compound that is on the front filter due to adsorption to the filter is given by
Pf,a )
Diagnostic Plots for Sampling Using Backup Filters General. As in the preceding section, the gas/particle Kp values used to prepare the plots are from the literature (1, 17), and it is assumed that QFFs behave like the PAB QFFs studied by Mader and Pankow (15). As noted above, during the time when V < Vmin,f+b (i.e., periods (f+b)1 and (f+b)2), the ng cm-2 amounts of a given compound adsorbed on the front and backup filters are not equal. Gas adsorption artifacts on cg,meas/cg, cp,meas/cp, and on Kp,meas/Kp cannot therefore be eliminated during these periods by attempting corrections using backup filters. cg,meas/cg. Plots of cg,meas/cg vs V are given in Figure 10 for PAHs for three sampling cases at 20 °C. (Other cases can be considered using the information presented in Tables 2-4.) During all of period (f+b)1, cg,meas ) 0 because none of the compound has escaped the front filter. For period (f+b)2, cg,meas/cg begins to rise toward 1 according to eq 32; it reaches 1 once period (f+b)3 begins. cp,meas/cp. Plots of cp,meas/cp vs V are given in Figure 11 for PAHs for three sampling cases at 20 °C. Other cases can be considered using the information presented in Tables 2-4. Kp,meas/Kp. Plots of Kp,meas/Kp vs V are given in Figure 12 for chrysene with QFFs and RH ) 100% and for 2378-TCDF with TMFs at RH in the range 21-52%. Other cases can be considered using the information presented in Tables 2-4.
mf,a × 100% mf,p + mf,a
(41)
If we continue to make the assumption of ideal chromatographic movement of a gaseous compound through the filter for purposes of general guidance, then assuming G/P equilibrium as well allows the substitution of eqs 12 and 24 for mf,p and mf,a, respectively, so that
Pf,a )
(
)
100% KpTSPV +1 Vmin,f
(step function movement through filter, V g Vmin,f)
(42)
If we assume that gas/filter has been reached on both the front and backup filters, then eq 30 holds so that Pf,a can be estimated as
Pf,a )
mb,a × 100% mf,p + mb,a
(43)
Ligocki and Pankow (17) assumed the validity of eq 43 and used f+b sampling at a temperature of 8 °C, V ) 230 m3, and GFFs in Portland during rain events (RH ≈ 100%) to obtain Pf,a ) 38, 66, and 41% for phenanthrene, fluoranthene, and pyrene. These values can be compared with predictions based on eq 42 assuming (1) their Af, Mparticles, and backupVOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3429
FIGURE 10. Predicted cg,meas/cg vs V for sampling of selected PAHs at 20 °C with an 8 × 10 in., f+b filter combination using a. two QFFs with the properties of PAB QFFs at RH ) 37%; b. two QFFs with the properties of PAB QFFs at RH ) 100%; and c. two TMFs at RH ) 21-52%. (Backup filter corrections included in the calculations.) filter-corrected Kp values; (2) the Kp,s values for PAHs on PAB QFFs at 100% RH from Luo and Pankow (13); and (3) GFFs and QFFs exhibit similar adsorption properties at RH ≈ 100%, as adjusted for the difference in Mf/Af between the two filter types. Since the sampling by Ligocki and Pankow (17) occurred with V g Vmin,f+b for phenanthrene, fluoranthene, and pyrene, the predictions which can then be made using eq 42 yield Pf,a ) 56, 47, and 41%, respectively, for these three compounds, which are in very good agreement with the measured results. Kp,meas vs Kp. Figures 6-9 illustrate the influence of V, filter type, and TSP on the magnitude of the positive bias in Kp,meas due to uncorrected gas adsorption artifacts. These plots assume that a single 8 × 10 in. filter was used during high-vol air sampling (with no backup filter corrections therefore possible). The validity of the approach outlined in this paper can be evaluated by comparing such predicted Kp,meas/Kp values to values measured by Hart and Pankow (3). For V ≈ 500 m3, TSP ≈ 50 µg m-3, and RH ≈ 90-100%, Hart and Pankow (3) observed that the adsorption of gaseous PAHs to QFF filters appeared to yield Kp,meas/Kp ) 1.2-1.6. Figure 8b suggests very similar values, specifically 1.1-1.4 for such compounds under such conditions.
Implications Determination of SOC Concentrations and Partitioning Constants. The analysis conducted in this paper indicates that gas adsorption artifacts can significantly affect the values of cg, cp, and Kp which are measured for individual SOCs using a conventional filter/sorbent sampler. Such artifacts will be small only in those circumstances when the amounts 3430
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001
FIGURE 11. Predicted cp,meas/cp vs V for sampling of selected PAHs at 20 °C with an 8 × 10 in., f+b filter combination using two QFFs with the properties of PAB QFFs at RH ) 37% with selected TSP values. (Backup filter corrections included in the calculations.) of the SOCs adsorbed on the filters are small compared to both: (1) the amounts in/on the collected particles (so that the errors in the cp values are small) and (2) the amounts that entered the sampler in the gas phase (so that the errors in the cg values are small). This can be achieved when the gas/ filter Kp,face values for the compounds of interest are low and/ or the sample volume is large. If backup filter corrections are to properly eliminate gas adsorption artifacts, the ng cm-2 amounts of each SOC adsorbed on the front and backup filters must be equal. This can only occur when V is large enough such that lowest volatility compounds of interest have reached gas/filter equilibrium with the backup filter. Figures 2-12 can be used to guide the choice of sampling parameters. Under no conditions should it be advocated that a short sampling time, by itself, is a route to minimize sampling artifacts with filter/sorbent samplers. Determination of Organic Carbon (OC) Levels in the Atmosphere. The need to understand and regulate levels of PM in the urban atmospheric environment carries the need to determine both the levels and composition of such PM. Since organic compounds can contribute significant amounts of mass to atmospheric PM, determining particle-phase organic carbon (OC) in the atmosphere has been of interest for some time (18-22). Because some of the compounds making up gaseous OC in the atmosphere are SOCs, a portion of those SOCs will sorb to filters (19) and thereby cause the measured levels of both OC (µg m-3) and PM (µg m-3) to be too high (positive artifact). As in the measurements of individual SOCs discussed above, one approach that has been taken to deal with this problem has utilized correction with a backup filter. Interest in that correction method has continued to the present time (23, 24). A first observation of
values of TSP (and PM2.5) and decreases as TSP increases. The type of curvature present in Figure 5 of Kim et al. (24) would be reproduced by eq 46 when Kp ≈ 0.02. Compared to the Kp values of the PAHs and PCDD/Fs considered in this paper, Kp ≈ 0.02 is fairly low. It is not, however, a surprisingly low number given the fact that the gaseous compounds that are abundant enough in the atmosphere to cause several µg m-3 worth of OC artifact are sure to be fairly volatile in nature.
Nomenclature Roman af (m2 g-1)
specific surface area of the filter
Af (cm2)
face area of the filter
cg (ng
FIGURE 12. Predicted Kp,meas/Kp vs V. a, b, and c: Sampling of 2378TCDF with an 8 × 10 in. f+b filter combination using two TMFs at RH ) 21-52% and selected TSP values; d, e, and f: Sampling of chrysene at 20 °C with an 8 × 10 in. f+b filter combination using two QFFs with the properties of PAB QFFs at RH ) 100% and selected TSP values. (Backup filter corrections included in the calculations.) Kirschstetter et al. (23) is that different lots of QFFs can exhibit different tendencies to sorb gaseous OC. In terms of our current study, that result would be interpreted to be a consequence of lot-to-lot variability in the filter Kp,s and Kp,face values for the compounds responsible for the majority of the OC gas adsorption artifact in a given circumstance. Kirschstetter et al. (23) also noted that this artifact for particulate OC can be severe for short sampling times () low sample volumes): the amount of adsorbed OC on the backup QFF is lower than the amount adsorbed on the front QFF, and subtraction of the backup amount does not adequately correct for the artifact. Expressed in terms of our study, this is a result of the sample volume being too small relative to the type of Vmin,f+b values that typify the compounds responsible for the majority of the OC gas adsorption artifact. Last, a result from Kim et al. (24) regarding the OC gas adsorption artifact effects that bears discussion is that for their sampling conditions, the OC backup/front ratio (i.e., the ratio mb,a/(mf,a + mf,p) for OC) was largest for events when the PM2.5 levels were low. Assume for example that sampling takes place such that V ≈ Vmin,f+b for the compounds responsible for the majority of the OC gas adsorption artifact. Under the continued assumption of ideal chromatographic movement through the f+b filter system, we then have
mf,a ) mb,a ) 0.5Vcg
(44)
By eqs 12, 15, and 44
mb,a 0.5Vcg ) mf,a + mf,p 0.5Vcg + cpTSPV
(45)
which reduces to
mb,a 0.5 ) mf,a + mf,p 0.5 + KpTSP
(46)
The backup/front ratio for OC then approaches unity for low
m-3)
gas-phase concentration of a compound
cg,meas (ng m-3)
measured gas-phase concentration of a compound
cp (ng µg-1)
particle-phase concentration of a compound
cp,meas (ng µg-1)
measured particle-phase concentration of a compound
f1
period prior to front filter equilibration (V < Vmin,f) for sampling with just a single filter
f2
period after front filter equilibration (V g Vmin,f) for sampling with just a single filter
(f+b)1
period prior to front filter equilibration (V < Vmin,f) for sampling with a front and backup filter pair
(f+b)2
period after front filter equilibration but prior to backup filter equilibration for sampling with a front and backup filter pair (Vmin,f e V < Vmin,f+b)
(f+b)3
period after front and backup filter equilibration (V g Vmin,f+b) for sampling with a front and backup filter pair
Kp (m3 µg-1)
gas/particle partitioning coefficient of a compound
Kp,face (m3 cm-2)
partition coefficient of a compound expressed as [ng sorbed per cm2 of filter face area]/[ng in gas-phase per m3 of gas phase]
Kp,meas (m3 µg-1)
measured gas/particle partition coefficient a compound
Kp,s (m3 m-2)
surface-area-normalized gas/solid partition coefficient of a compound
mA (ng)
mass amount of compound collected on the adsorbent
mb,a (ng)
mass amount of gas adsorbed on the backup filter
mF (ng)
mass amount of a compound extracted from the particle-laden front filter
Mf (µg)
mass of the filter
mf,a (ng)
mass amount of gas adsorbed on the front filter
VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3431
mf,p (ng)
mass amount of a compound present in/on the particulate material collected on the front filter
Mparticles (µg)
mass amount of particulate material collected on the filter
OC (µg m-3)
organic carbon
PAB
Portland ambient backup filter
PM
particulate material
PM2.5 (µg
m-3)
particulate material 2.5 µm in diameter and smaller
Pf,a
percent of mass amount of a given compound on the front filter due to gas adsorption
poL(Torr)
subcooled liquid vapor pressure of a compound
RH
relative humidity
Sf
cm-2)
(m2
filter surface area factor
SOC
semivolatile organic compound
TMF
Teflon membrane filter
TSP (µg m-3)
total suspended particulate material concentration
V (m3)
air sample volume
Vmin,f (m3)
minimum air sample volume required to reach gas/filter sorption equilibrium for a given compound and a single filter
Vmin,f+b (m3)
minimum air sample volume required to reach gas/filter sorption equilibrium for a given compound with a front+backup filter combination
Greek Rg
fraction of a compound that is in the gas phase at G/P equilibrium
Rp
fraction of a compound that is in the particle phase at G/P equilibrium
Acknowledgments Funding for the project came in part from U.S. EPA Research Grant #R825376.
Literature Cited (1) Eitzer, B. D.; Hites, R. A. Environ. Sci. Technol. 1989, 23, 13891395.
3432
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001
(2) Foreman, W. T.; Bidleman, T. F. Atmos. Environ. 1990, 24A, 2405-2416. (3) Hart, K. M.; Pankow, J. F. Environ. Sci. Technol. 1994, 28, 655661. (4) Cotham, W. E.; Bidleman, T. F. Environ. Sci. Technol. 1995, 29, 2782-2789. (5) Gustafson, K. E.; Dickhut, R. M. Environ. Sci. Technol. 1997, 31, 140-147. (6) Simcik, M. F.; Franz, T. P.; Zhang, H.; Eisenreich, S. J. Environ. Sci. Technol. 1998, 32, 251-257. (7) Lohmann, R.; Lee, R. G. M.; Green, N. J. L.; Jones, K. C. Atmos. Environ. 2000, 34, 2529-2537. (8) Turpin, B. J.; Saxena, P.; Andrews, E. Atmos. Environ. 2000, 34, 2983-3013. (9) Andrews, E.; Saxena, P.; Musarra, S.; Hildeman, L. M.; Koutrakis, P.; McMurry, P. H.; Olmez, I.; White, W. H. J. Air Waste Manage. Assoc. 2000, 50, 648-664. (10) McDow, S. R.; Huntzicker, J. J. Atmos. Environ. 1990, 24, 25632571. (11) Turpin, B. J.; Huntzicker, J. J. Atmos. Environ. 1994, 28, 30613071. (12) Storey, J. M.; Luo, W.; Isabelle, L. M.; Pankow, J. F. Environ. Sci. Technol. 1995, 29, 2420-2428. (13) Luo, W.; Pankow, J. F. 2001, unpublished work. (14) Mader, B. T.; Pankow, J. F. Atmos. Environ. 2000, 34, 48794887. (15) Mader, B. T.; Pankow, J. F. Atmos. Environ. 2001, 35, 12171223. (16) Eatough, D. J. 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: 1999; Amsterdam, pp 233-285. (17) Ligocki, M. P. Ph.D. Thesis, Oregon Graduate Institute, Portland, OR, 1986. (18) Huntzicker, J. J.; Johnson, R. L.; Shah, J. J.; Cary, R. A. In Part. Carbon: Atmospheric Life Cycle; Proc. Int. Symp., 1982, Meeting Date 1980, Wolff, G. T., Klimisch, R. L., Eds.; Plenum: New York, NY, pp 79-88. (19) McDow, S. R.; Huntzicker, J. J. Atmos. Environ. 1990, 24A, 256371. (20) Turpin, B. J.; Huntzicker, J. J.; Adams, K. M. Atmos. Environ. 1990, 24A, 1831-1835. (21) Novakov, T. In Nature, Aim, and Methods of Microchemistry; Proc. 8th Int. Microchem. Symp. Austrian Soc. Microchem. Anal. Chem., 1981, Graz, Austria, August 25-30, 1980; Malissa, H., Grasserbauer, M., Belcher, K., Eds.; Springer-Verlag: pp 141165. (22) Novakov, T. In Part. Carbon: Atmospheric Life Cycle; Proc. Int. Symp., 1982, Meeting Date 1980; Wolff, G. T., Klimisch, R. L., Eds.; Plenum: New York, NY, pp 19-41. (23) Kirschstetter, T. W.; Corrigan, C. E.; Novakov, T. Atmos. Environ., 2001, 35, 1663-1671. (24) Kim, B. M.; Cassmassi, J.; Hogo, H.; Zeldin, M. D. Aerosol Sci. Technol. 2001, 34, 35-41.
Received for review August 16, 2000. Revised manuscript received April 26, 2001. Accepted May 7, 2001. ES0015951
