Topic Leader: Lara A. Gundel
Contributors: Douglas A. Lane, John Volcken

Abstract

1a. How are SVOC and PM-associated OC defined?

i. Theoretical Definitions

SVOC

Borrowing from Van Vaeck et al (1984) and others (for example, Finlayson-Pitts and Pitts, 2000, p 412) we define semi-volatile organic compounds (SVOCs) as organic compounds that show significant gas and particulate concentrations in the atmosphere. Pankow (1993) generalized this by including other surfaces. He observed that SVOCs can have non-negligible fractions of their environmental masses in both the atmosphere and partitioned to surfaces such as soil, plants, building materials and sampling media. SVOCs have vapor pressures between about 10-4 or 10-5 and 10-11 atm (100.1 and 10-6 Pa; 10-1 and 10-8 Torr) over the ambient temperature range. Compounds with higher vapor pressures are present primarily in the gas phase, whereas low volatility compounds are found on or within particles.

SVOCs are ‘sticky’ or multi-phasic; they partition their mass between the gas phase and any surfaces that afford a degree of sorption, such as fine particles in air or sampling media. The degree of gas-particle partitioning affects the transport, deposition, and atmospheric fate of these compounds, since SVOCs, once airborne, can deposit onto vegetative surfaces, windows, carpets, soils, or even the human body.

Nearly all classes of organic compounds contain semi-volatiles: alkanes, PAHs, PCBs, PCDDs, PBDEs, nitro-aromatics, terpenes, acids, carbonyls, and lipids, to name a few. SVOCs enter the atmosphere by direct emission, frequently as byproducts of incomplete combustion. Polar SVOCs are also produced by oxidation of precursor unsaturated compounds, and they can be incorporated into PM as secondary organic aerosols (SOA). Precursor organics are emitted from transportation, industrial and biogenic sources. Also, many of the potentially carcinogenic organic compounds found in the atmosphere are semi-volatile.

PM-associated OC

Roughly half the mass of urban fine particles in the US can be attributed to carbonaceous components. From 10 to 30% of this is elemental or black carbon, and carbonate-containing compounds have negligible contributions. Here we define PM-associated organic carbon as the complex mixture of organic compounds that are incorporated into airborne particles by direct emission, abrasion, condensation and surface reactions. The term PM-associated OC may be useful when considering the influence of the complex mixture of particle-associated organic compounds as a whole, for example, when investigating aerosol properties like organic film thickness, hydrophilicity, and optical absorption.

Relationships between SVOCs and PM-associated OC: gas/particle partitioning

SVOCs and PM-associated OC are related through the partitioning of SVOCs onto PM. However, less than half of PM-associated OC is semi-volatile under temperate conditions. The partition coefficient Kp is the most commonly used parameter for describing gas/particle partitioning, primarily because of its log-linear relationship to compound vapor pressure, poL. Although a compound’s vapor pressure at the temperature of interest has the greatest influence on partitioning, the interaction between compound structure and the sorptive medium plays an important role (e.g., compound size and polarity vs. adsorptive affinity or absorptive capacity). Plots of log Kp vs. log poL can provide information on the nature of the partitioning and may indicate whether sampling artifacts have impacted the measurement. Gas/filter partitioning coefficients can also be used to improve sampler and field study design (Mader et al., 2001).

Using Pankow’s nomenclature (as presented in Ch. 3 of Lane, ed. 1999), the equilibrium partitioning of a semi-volatile compound to an environmental surface S can be represented most simply by

G + S = P (1)

where G and P represent the gas and particulate phases of the SVOC. Since the gas and particulate phase concentrations are usually collected on an adsorbent and filter, respectively, their concentrations have been conveniently represented by A and F in the literature. If the sorbing surface is total suspended particulate matter, its concentration can be represented by TSP. (G/P theory takes the same form for size segregated particles, but TSP is used here to be consistent with Pankow’s development.) At equilibrium the gas/particle partitioning constant Kp for adsorption of the SVOC compound i onto the solid surface of a particle can be expressed as in (2):

Adsorption to a solid surface (2)

Pankow (1987) showed that Langmuir adsorption theory predicts that Kp (at constant temperature) is inversely proportional to the vapor pressure of i. If i is a solid, the sub-cooled liquid vapor pressure poL is used.

adsorption (3)

Ns and atsp are terms for the number of adsorption sites per unit area and the surface area of the particles, respectively. For compounds of the same class, with similar enthalpies of desorption and vaporization Q among the members, plots of log Kp versus log poL for will be linear slope of –1, as shown in (4).

adsorption (4)

SVOCs can also absorb into liquid particles such as environmental tobacco smoke or liquid (organic and/or water) films on particles with solid cores, as Pankow and colleagues have shown. Gas/particle partitioning of SVOCs in urban areas is better explained as absorption than adsorption. The absorptive partitioning of SVOC i into a liquid organic layer on a particle is like a gas dissolving in a liquid (Finlayson-Pitts and Pitts, 2000, p 417), and the measured partitioning coefficient for absorption takes the same form as (2).

absorption into a liquid film or droplet (5)

Fi,om represents the particle- associated concentration of i in air as measured from a filter, with explicit recognition that i has dissolved in liquid organic material, om, on the particle. For absorption into liquid films on particles, Pankow (1994) showed that Kp is proportional to the weight fraction of om to TSP.

adsorption (6)

Kp is inversely proportional to the product of poL and the activity coefficient g of i in the liquid phase. Vapor pressure is the most important factor influencing Kp, followed by activity coefficient and molecular weight. If the activity coefficient does not vary much across members of a class of SVOCs, plots of Kp vs log poL will have a slope of -1 for both adsorption and absorption, as in equations (4) and (7).

(7)

Recent contributions to gas/surface partitioning theory address observed deviations from the predictions of equations (4) and (7). Jang, Kamens et al. (1997) applied a comprehensive thermodynamic approach to calculate group contributions to activity coefficients for adsorption of SVOC into non-ideal organic films. This allows calculation of activity-normalized partitioning coefficients, Kp,g. Goss and Schwarzenbach (1998) argued that slope deviations from –1 in log Kp vs log poL plots do not necessarily indicate non-equilibrium conditions, and they indicated how these deviations can be used to identify types of sorbate/sorbent interactions and thus characterize sorption processes. For example, they showed how acid/base interactions can influence gas/surface partitioning polar SVOC. Harner and Bidleman (1998) demonstrated that using laboratory-derived octanol/air partitioning coefficients circumvents the need to estimate activity coefficients for compounds absorbed in the organic films that coat urban particles. Mader et al. (2000, 2001) expanded partitioning theory to quartz and Teflon filter materials that are used to collect particles, thus tackling the sticky problem of SVOC adsorption artifacts in PM sampling on filters.

ii. Operational Definitions

Partitioning constants are calculated from measured concentrations of individual compounds in each phase. The terms Ai and Fi in equation (2) must be determined for individual compounds. For some classes of compounds, good estimates of vapor pressures can be calculated from observed molecular weight-vapor pressure relationships for members of the class. Whereas concentrations of PM-associated OC can be determined experimentally as the collective concentration of ‘organic carbon’ by programmed evolved gas analysis, at present there is no way to determine total airborne concentrations of ‘semi-volatile organic carbon’ in an analogous way. PM-associated OC includes temperature dependent PM-associated SVOC of largely uncertain composition. Thus, at present it is not possible to distinguish SVOC and PM-associated OC quantitatively and accurately without introducing technique-dependent bias.

As Turpin et al. (2000) point out, the term SVOC is usually operationally defined, or undefined, as below:

= concentration of i measured from filter F, and (8a)

= concentration of I measured from adsorbent A (8b)

Because the typical meaning of SVOC is rooted in sampling strategy, Turpin et al. (2000) prefer to use the term ‘condensable’ for airborne compounds that are found in both the gas and particulate phases.

The total carbon content of PM-associated OC can be determined by thermal analysis, as discussed at the OC/EC workshop in 2003. At present, measured concentrations of PM-associated OC depend on both the sampling and analytical methods. Assuming negligible particulate carbonate,

Particulate C = PM-assoc. organic carbon (OC) + elemental carbon (EC). (9)

However, even the widely used thermal-optical differentiation of PM-assoc. organic carbon from elemental carbon depends on operationally defined algorithms. If i and j represent individual semi-volatile and non-volatile organic species,

(10)

At present, the SVOC term in equation (10) can be estimated from thermal methods for particulate C (Fan et al., 2003) and possibly from source characterization (Schauer et al.), but accurate determination of the organic carbon content of real-world SVOC (as a class) remains elusive.

Fig. 1 has schematic representations of air sampler configurations that lead to different operational definitions of particulate-associated organic carbon. Component descriptors are given on the left, while the labels on the right indicate the function of each section. F denotes a particle-collecting section (usually, but not always a filter; sometimes a sorbent that collects SVOC evaporating from the particles during sampling). A stands for the section that collects gas phase SVOC (sorbent-coated denuder, sorbent such as XAD resin, or backup filter for estimating the gas phase SVOC adsorption artifact that occurred on F).

Figure 1. Sampler designs with different operational definitions of particulate-associated organic carbon. The arrows show the direction of flow. The usual component descriptors are shown on the left, but the descriptors may not accurately describe the original phase of airborne SVOC that collect there. Origins are indicated using the notation of equation (8).

Below are the operational definitions of the gas and particulate concentrations of SVOCi for the sampler designs in Fig 1.

(FA, DFA and EA) [SVOCi]g = Ai (11a)
(FFA) [SVOCi]g = A1i + A2i (11b)

(FA and EA) [SVOCi]p = Fi (12a)

(FFA) [SVOCi]p = Fi - A1i (12b)
(DFA) [SVOCi]p = F1i + F2i (12c)

Partitioning coefficients derived from these designs differ, as discussed in Topic 1c.

1b. What are the potential bias or problems associated with different PM and SVOC sampling techniques?

Filter-Adsorbent (FA) with configuration FA; equations (11a) and (12a).

This is often referred to as the ‘conventional’ sampler design, since it has been used for at least two decades for speciation of semi-volatile and particulate organics in airborne PM. The sorbent A is typically polyurethane foam, XAD resin or a foam/XAD/foam sandwich.

Figure 2. Filter-Adsorbent or conventional sampling configuration for sampling airborne semi-volatile and particulate organic compounds.

At equilibrium SVOC adsorb and desorb from particles. The upper part of Figure 2 shows adsorbed SVOC as small dots associated with the irregularly shaped particles. Particulate SVOC are trapped on the filter medium F, and gases pass through the filter F for adsorption on A. Positive sampling artifacts occur when gaseous SVOC adsorb to the filter medium, as shown by small dots on the filter. Negative artifacts result from particulate SVOC evaporating from the filter deposit during sampling. The volatilized SVOC become trapped on the adsorbent A.

McDow and Huntzicker (1990) found that sampling artifacts for OC depend on face velocity and sampling duration. Turpin et al. (1994) showed that the quartz filter positive artifact decreased as a fraction of the total OC, with increasing sampling time, at constant face velocity and particle mass concentration. After reviewing studies available at the time, Turpin et al. (2000) concluded that positive artifacts for OC are usually larger than negative artifacts when sampling urban air. Gas/particle partitioning measurements of individual semivolatile compounds typically indicate that negative artifacts dominate (Van Vaeck et al., 1984; polycyclic aromatic hydrocarbons (Fan et al., 19xx); a variety combustion-generated organics, mostly non-polar or moderately polar species, Schauer et al., 19xx, Schauer et al., 19xx).

Filter-Filter Adsorbent (FFA) with configuration FAA;equations (11b) and (12b)

Two filters in series are often used as a simple sampling approach to correct for positive artifacts in sampling particulate-associated OC. The FFA design with configuration FAA has not been used widely for gas/particle partitioning of individual species. Assuming that the upstream filter has collected a negligible fraction of the total airborne SVOC, the downstream filter, used as the adsorbent A1 for SVOCg, has the close to same amount of adsorbed SVOC as the upstream, particle-laden filter. The remainder of the airborne SVOC will be trapped if an adsorbent component A2 is used.

An improvement is to use co-located filter pairs. The upstream filters are Teflon and quartz, but both downstream filters are quartz. This approach is sometimes called tandem sampling with the pairs referred to as TQ and QQ. The OC determined from the quartz behind Teflon is subtracted from the OC on the upstream quartz filter. This should represent the positive artifact more accurately, since Teflon filters do not have the adsorptive capacity of quartz. Turpin et al. (2000) concluded that artifact OC measured from A1 could be up to 40% higher if determined from the quartz behind Teflon rather than quartz behind quartz.

Denuder-Filter-Adsorbent (DFA) with configuration AFF; equations (11a) and (12c)

When gas phase SVOC are collected upstream of filters, positive artifacts can be minimized. This process denudes the gas stream of its SVOC, and the adsorption occurs in a denuder. Because the gas/particle partitioning equilibrium is disturbed during sampling, particulate SVOC become more susceptible to volatilization. A post-filter adsorbent, F2 in Fig. 1, traps semivolatile species that have evaporated from particles on F1.

How denuders work: DL slide 5

Lane et al. (198x, reviewed in chapter x of Lane, 1999) and Eatough et al. (199x) pioneered application of denuder difference methods for sampling airborne semi-volatile and particulate organics. To remove SVOC upstream of filters, Lane et al. used Tenax-impregnated GC-stationary phase gum on glass annular denuders, while Eatough et al. use parallel strips of activated carbon-impregnated paper. Sorbents (F2) were used downstream of the filters. These designs require co-located conventional samplers (FA) for gas/particle partitioning measurements because SVOC can not be determined from the denuders. Gundel et al. (1995) introduced extractable XAD-coated denuders for direct determination of gas and particle concentrations without the need for co-located conventional samplers. XAD-coated denuders and filters have been incorporated into several sampler designs (Gundel et al., 1999; Lewtas et al., 2001; Mader et al., 2001c; Hammond et al., 2003).

So far, the DFA (AFF) sampling approach has worked better for gas/particle partitioning of individual species than for determination of aggregate particulate SVOC as OC.

Electrostatic precipitator (EA) with configuration FA

Dr. Douglas Lane will discuss how characterization of particle-associated organic species can be affected by the sampler geometry. Collected PM should be representative of the PM composition at the time of sampling, and not influenced by volatilization of SVOCs from the collected particles or sorption of SVOCs to the particles or sampling media. From the perspective of particle composition, these processes lead to negative and positive sampling artifacts, respectively, and they will be illustrated for samplers that collect particles onto filters upstream of sorbents for SVOC. Multi-channel annular diffusion denuders are being used to minimize positive PM artifacts. They have also been incorporated upstream of filters in the filter/sorbent geometry for determination of the gas/particle partitioning coefficients of source and atmospheric SVOCs. Data from recent field measurements using the Integrated Organic Gas and Particle Sampler (IOGAPS) will be presented to demonstrate the uses and limitations of denuder-based samplers. The application of denuders in smog chamber studies will be illustrated with gas/particle partitioning of the products of reactions of PAH with OH. The importance of understanding the limitations of air samplers and the proper selection of a sampler for a particular sampling objective will be emphasized.

1c. Advances in sampling and analysis of SVOC

Dr. John Volckens will 1) highlight recent advances in sampling and analysis of SVOCs, 2) discuss techniques to interpret artifact-biased data, and 3) recommend strategies for future research based on needs of the community.

As mentioned in section 1a, the ratio Kp is the most commonly used parameter for describing gas-particle partitioning, primarily because of its log-linear relationship to compound vapor pressure, poL. Gas-particle partitioning ratios (Kp) are defined at equilibrium, when the rates of mass transfer between phases are equal and at steady state. However, the conditions governing SVOC equilibrium are easily disrupted, especially when trying to measure gas-particle phase distributions in air. However, care must be taken with the use of Kp because this ratio is easily corrupted by even minute artifacts. Such sampling artifacts are widely known but difficult to prevent, predict, or account for after the fact (Volckens and Leith, 2003).

Plots of log Kp vs. log poL can provide information on the nature of the partitioning and may indicate whether sampling artifacts have impacted the measurement. Furthermore, most time-integrated sampling techniques (i.e. filter) cannot provide a true representation of ‘average Kp’. A need exists to develop improved sampling techniques with less perturbation of semi-volatile equilibrium, shorter sampling periods, and lower limits of detection.

References: