Human Exposure Assessment for Airborne Pollutants: Advances and Opportunities (1991)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Sampling and Physical Chem~cal Measurements INTRODUCTION This chapter considers sampling and physical-chemical measurement meth- ods available for assessing human exposures to airborne pollutants and em- phasizes recent advances. These advances could be used to improve e~osure- assessment methods. In assessing human exposures to airborne pollutants, numerous factors besides the contaminant must be measured especially if the assessment is based on fixed-site sampling or modeling. Accurate estimates in these instan- ces depend not just on concentration measurements from f~xed-site monitors in various locations, but also ore knowledge of numerous factors that influence the environments where the exposures occur (see Chapter 2~. Outdoors, these factors include temperature, humidity, precipitation, barometric pressure, wind speed and direction, turbulence, and mixing height. Insolation, as well as light scattering and absorbance, might also be important. Some of these factors also must be measured to mode! indoor environments. However, other fac- tors are unique to indoor environments such as: ventilation rates, pressure differentials across building shells and between building compartments, re- moval efficiencies of building filters, and contaminant deposition rates on indoor surfaces. Furthermore, modeling frequently requires measurements of source strengths. Outdoors, source-strength measurements include emission rates from a major point source (e.g., power plants). Indoors, source-emission rates could include volatile organic compound (VOC) emission rates derived from chamber studies of building materials, consumer products, and home furnishings (Tichenor, 1987~. These areas are too broad to be discussed com- prehensively in this chapter, whose focus is the measurement of airborne con- taminants. Nonetheless, measurement methods that produce information about environmental factors or emission rates should be accounted for in the 53

54 ASSESSING HUMAN EXPOSURE development of useful indirect methods to identify and control the factors most significant to human exposure. As outlined in Chapter 1, the choice of sampling and physical and chemical measurement methods to be used in an exposure assessment is driven by a studys specific aims. The analytical procedures should be chosen with atten- tion to the specific needs of the study. The "why," "what, ~when," and ~where" all influence the selection of the ~how,' discussed in this chapter. It is important for the analyst to ask, "What are we ultimately trying to accomplish?" Sampling frequency and duration are important elements of a sampling strategy. Certain analytical procedures provide real-time or instantaneous measurements of contaminant concentrations (e.g., long-path-length Fourier transform infrared spectrophotometers), while others provide an average value for the interval during which sampling occurs (e.g., collection of VOCs on the sorbent Tenax). Real-time measurements can be made consecutively to yield a continuous record of a contaminant concentration, or they can be made intermittently to yield a series of concentration "snapshots.- Integrated meas- urements can be made consecutively or intermittently, or they can be over- lapped, if more than one set of sampling apparatus is available. If monitoring is done for compliance purposes, the sampling frequency and duration likely are specified by regulation. However, rigid specification usually is not necessary for most types of exposure-assessment monitoring. If peak concentrations are important in assessing a potential health effect, then a sampling procedure should be integrated over a time scale no longer than that at which contaminant concentrations fluctuate. Furthermore, the sampling should be frequent enough to measure major fluctuations. Real-time continu- ous monitoring for a contaminant that causes a chronic health effect would be unnecessary, because the total contact is of concern. The time scale of the relevant biological effect for a contaminant must be considered in choosing the time scale of the sampling and measurement process (Lioy and Daisey, 1987~. Emissions of various airborne contaminants can be time dependent. For example, at a manufacturing site, time of day and day of week can influence emission rates and, consequently, the airborne concentrations of various spe- cies. Diurnal, weekly, and seasonal variations in emission rates can affect outdoor airborne concentrations. Such factors must also be considered when planning sampling frequency and duration. Spatial considerations are important in f~xed-site monitoring. As stated in the National Research Council (NRC) report, Complex Mixtures (NRC, 1988), The primary consideration should be the relevance of the sample site to potential human exposure." Selection of a sampling site can be purposive or probabilistic. Purposive sampling normally is conducted to answer questions

SAMPLING AND PlIYSICAL-CHEMICAL MEASUREMENTS 55 about a specific location (e.g., sampling near known emission sources, such as a power plant or a waste dump). Probability sampling seeks to provide an overall picture of an area. The choice of sampling sites should be influenced strongly by the nature of the potential human exposure. Table 3.1 is a sum- mary of designs that can be applied to the selection of sampling sites. It also includes a brief evaluation of when the different strategies are most useful. (For a further discussion of sampling sites, see NRC, 1988~. QUALITY ASSURANCE Using advanced techniques in exposure studies does not ensure the acquisi- tion of better qualitative data, but allows the potential of obtaining better data. Whether that potential is realized depends on the quality-assurance (QA) program that is designed into the study. The terms precision and accuracy often are used in quantitive studies. Precision is a measure of the agreement among individual measurements made of the same property of the sample. Accuracy refers to the degree of agreement of a measurement (or an average of measurements of the same property) with an accepted reference or true value. QA and its complementa- ry concept, quality control (QC), have many definitions. QA often is used to also include QC, and this report uses this convention. For environmental measurements, QC comprises operational activities that are carried out before and during the measurement process that are intended to ensure that data are of sufficient quality~ata whose precision and accuracy are known and are sufficient to meet the needs of ~ study. Examples of QC are calibration pro- cedures, maintenance of constant line voltage and temperature, use of blank and spiked samples, and use of traceable standard reference materials. QA also includes activities carried out to ensure that the collected data achieve the precision and accuracy required, such as interlaboratory comparisons, meas- urement system audits, and statistical procedures to highlight bad data or extreme values. These activities should be carried out by persons not involved in routine data-gathering operations. EPA has developed a comprehensive QA handbook that gives principles and recommended procedures for achiev- ing quality data in air-pollution measurement systems (EPA, 1976a,b). ERRORS In designing a QA program to meet the needs of a specific exposure study, it is useful to consider the four activities involved in any environmental meas- urement that can cause errors in the data obtained:

56 3 Cal - _' 4 - o C) At; a: C) S 3 I: es Cal ._ U) _ Ct Cat o Ct Us Cat . . o ._ CO ._ A: o - Ct ._ Ct m EM o _. CO .Q O I: O :~ A: O  ·= 0 c: ·= Cal Cat 1_' is, C) E it, e E ° E E ~.= ~A, ~U: S ~D _ ~, ~·= ~ ~o ~ 7 .= ~ ~ ~_ · _ ~ · O ~ ~ ~ , E ~ E ~E E ~ C E ~ ~ E ~: _ ._ _ _ - , ~ E E E E E E E ~ W ' ~E E c c o .e ~, ~-° P ~ V ~: z . . C) C~

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 57 · Selection of representative sampling sites. · Collection of the environmental sample. · Sample analyses. · Data handling. Site-Selection Errors The representativeness of the sampling site refers to selection for appropri- ate spatial and temporal definition. For example, a study to determine com- pliance with ~ ambient air~qu;ality steward for ~ given pollutant would re- quire air samplers to be placed at community sites that represent typical out- door air and to have the same measurement time as the standard. On the other hand, a study to determine total air exposure of a population to the same pollutant could require a sampling strategy that involves personal sam- plers or microenvironmental measurements combined with activity diaries. In many studies, site-to-site variability is the largest component of the total meas- urement error. EPA (1988b) guidelines for exposure studies provide general information on proper sitting of outdoor air-monitoring stations. Collection Errors The study of most air contaminants requires that the air sample be moved from the microenvironment into a collection device or analytical instrument. Errors can result from physical and chemical changes in the sample during and after sample collection. Air-collection procedures usually concentrate molecules that normally are diffuse and isolated, thus enhancing the possibility of concentrated molecules interacting with each other or with the collection medium or sampler components. These interactions can render some collect- ed molecules immeasurable by the chosen analytical procedure. Errors can also occur during handling, shipping, and storage of the samples. Pumps can be significant sources of artifacts in collected samples. For example, particle artifacts may arise because of mechanical wear or oil-droplet emissions. Gas-phase artifacts may arise as a result of emissions from hot pump oil or other pump lubricants. Also, the magnitude of the pump flow rate is an important consideration in microenvironmental sampling. The flow rate can be set so high that the sampling system decreases the contaminant concentration in the microenvironment being measured, and, near the end of the sampling period, the contaminant concentration in the microenvironment could be lowered artificially.

58 ASSESSING HUMAN EXPOSURE Another common source of error when sampling with a pump is a poorly defined flow rate. It is extremely important to calibrate and periodically check flow rates in any device that uses flow rate to quantify the volume of air sam- pled. Sample collection artifacts can arise from components other than the pump. These include tubing, improper sealants, and incompatible plastics. Collection artifacts can also be a problem: (a) during sample collection set up (potential sources include idling motor vehicles, smoking, and the use of insect repel- lents); (b) when using a sorbent, if the collection efficiency is poor (sorbent breakthrough); and (c) during vapor- and particle-phase sampling when the collection procedure itself may alter the distribution between the phases. Errors also can occur with samplers when study subjects do not wear or carry a personal sampler when they are expected to. Sometimes subjects are embarrassed by the noise or size of the samplers; subjects might change their activity patterns when wearing samplers to avoid embarrassing situations. A subject might wear the sampling device, but forget to turn it on. Other unin- tentional errors include accidentally sitting on the sampling line for a pump, effectively stopping the flow to the sampler. Although passive samplers might seem to alleviate many of these problems, an outer garment over the device seriously reduces sampling capability. Therefore, the design of personal sam- plers should include provisions to minimize their misuse and to ensure that they have been used properly. Analytical Errors Analytical errors are associated with the identification and quantitation of the chemical of interest in the sample collected. Qualitative errors can be minimized by increasing the selectivity of the analytical method that is used and by confirming compound identity by a second technique. This selectivity should minimize potential interferences, that is, the ability of chemicals other than the one of interest to interfere with the measurement process so as to give results either higher or lower than the true value. The metrics used to describe analytical quantitation errors are precision and accuracy. Methods that have good precision and accuracy can be used. However, methods that have poor accuracy but good precision often can be useful in studies that require understanding only of the relative differences among properties of environmental samples (Watson et al., 1983~. Critical operational procedures that can lead to analytical quantitative errors are poor- ly conducted calibration procedures and use of inadequate reference materials when carrying out calibrations. Even if these procedures are carried out care

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 59 fully, some errors are inherent in the analytical method-every method has a limit for precision and accuracy. Another factor that can lead to quantitative errors is the ~rugge~ess" of the method: its sensitivity to variations in the factors that affect the measurement (such as temperature, relative humidity, =d line voltage). Furthermore, many measurement methods cited in the literature glare reasonable precision and acewac,T when used by highly framed research staff, but lose their precision and accuracy when used by less well- trained personnel who might not maintain stable operating procedures. The routine use of field blanks and field spikes can reduce the occurrence and magnitude of analytic errors Data-Handling Errors Data-handling errors are among the errors that can occur during data ma- nipulation. These errors include mistakes in reading instrumentation, In trans- pos~ng data from one system to another (e.g., data-entry errors), and in cal- culating results in appropriate units. The EPA QA handbook (EPA, 1976a,b) gives guidelines for minimizing these errors. Many of these errors have been reduced through the extensive use of microprocessors that often collect and, ~ some cases, even analyze the data (Barrett, 1988; de Monchy et al., 1988), thus reducing human error in transcribing data and bias in data evaluation. Periodic human inspection of all steps of data collection, reduction, and re- porting is essential as one further QC measure (Taylor, 1987~. Inspection ensures that the automation of the data-analysis process does not obscure significant information not considered when the initial microprocessor programs were established. This is particularly true when programs are set to accept or reject data automatically. One common error in data reporting is the error of omission an omission of precision and accuracy information when reporting data. The literature is replete with misinterpretation and overinterpretation of data thought to be more certain than they really were. Expenmer~tal data always should be ac- companied by precision and accuracy information. Precision and accuracy are integral parts of the measurement. Designing a measurement strategy for a field study seems straightforward, but designing one that minimizes the errors almost always is difficult. Fur- ther, designing a good QA program for environmental field measurements might be more difficult than one designed for laboratory experiments. More people of varying skills usually are involved in environmental field studies than In laboratory studies: different members of a staff will design the study, col- lect the samples, analyze the samples, report the data, statistically analyze the

60 ASSESSING HUMAN EXPOSURE data, and interpret the data. Without an organized QA program, such sharing of responsibility for data quality can result in information of poor quality, or perhaps worse, data of unknown quality. Furthermore, environmental field measurements cannot be replicated, i.e., a measurement made today cannot be repeated tomorrow Obviously, QA is a critical part of exposure studies, and a QA program must be established as part of the initial study design. The plan should fit the specific aims of the study. The QA program also must be considered when establishing the budget, because an effective QA program costs about 15-25% of the measurement budget. From a practical point of view, this translates into significantly less data but data of a higher quality~han if QA were neg- lected. Those designing the study must take this reality into account when determining the statistical power of the study. Painful though it is to reduce the amount of exposure data, obtaining less data that are all good is much better than obtaining more data that are bad or unverifiable. AIRBORNE ANALYIES The nature of a given airborne pollutant its physical, chemical, and in some cases, biological characteristics~etermines the procedures appropriate to its sampling and measurement. A contaminant can exist in a vapor phase or particle-associated condensed phase, or it can be partitioned between these phases. The partitioning can result from the adsorption of a vapor-phase compound onto the surfaces of airborne particles, in which case the contaminant distribution is a function of the compound's liquid-phase (or subcooled liquid phase if the compound is a solid at ambient temperature) vapor pressure and also the surface area of airborne particles per unit volume of air (Pankow, 1987; Bidleman, 1988; Junge, 1977; Ligocki and Pankow, 1989~.p,p'-Dichlorodiphenyltrichloroethane is an example of an ambient contaminant commonly partitioned in this man- ner. Partitioning also can be due to dissolution of a vapor-phase compound in a liquid associated with airborne particles. An example is the dissolution of sulfur dioxide (SO2) in water associated with hydroscopic particles. Organic vapors also can dissolve in liquids associated with airborne particles. Still another partitioning process involves attached and unattached radon daugh- ters. Because a pollutant's dose to the lung can be very different in the vapor phase from that of the condensed phase, care must be taken that a sampling procedure does not alter the distribution between phases and present a false picture of the pollutant's physical state (Van Vaeck et al., 1984; Bidleman, 1988; Coutant et al., 1988; Ligocki and Pankow, 1989~.

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 61 If a pollutant is present in the condensed state, its distribution as a function of particle size is important information for assessing human exposure. Air- borne particles range in size from a few nanometers to hundreds of microm- eters (Finlayson-Pitts and Pitts, 1986~. Particles frequently are classified as After (~2~ am diem.) and "coarse" (>2.5 am diam.~. Fine particles are some- t~mes subclassified as nuclei mode (~-.~-1 '`m diem.) and accumulation mode (0.1-2.5 am diam.~. Fine and coarse particles tend to have different sources and, consequently, different chemical compositions. They also have different transport characteristics, such as settling velocities and diffusion coefficients, which lead to different atmospheric lifetimes. For these reasons, collection of s~ze-fractionated particles (at lead ~ fine and ~ coarse Faction) is usefo1 when sampling with the intention of future chemical analyses. The American Con- ference of Governmental Industrial Hygienists has established cut sizes ap- propriate for fractionation in relation to inhalation hazard (Phalen et al., 1986~. In addition to size, other physical properties of airborne particles, such as morphology and water content, can strongly influence their effects on living organisms and should be considered in any sampling methodology. Particle shape and surface texture are important morphological features and are char- acteristic of the nature of the particle; research is in progress to represent this information as a few characteristic numbers (FIopke et al., 1988~. Such re- search is concerned with the use of optical and electron microscopes as the major tools for determining these features with a primary focus on electron microscopes. Chemical species are not always distributed uniformly throughout a particle. For some species, the surface concentration is significantly larger than the bulk concentration. Examples include semivolatile organic compounds that have been adsorbed on particle surfaces and trace metals with low boiling points, such as lead, zinc, and cadmium, that are surface enriched by high- temperature processes. In such cases, bulk analyses would yield much lower concentrations than those actually in contact with environmental surfaces. Hence, surface analyses using techniques that provide elemental or chemical information such as SAM (scanning auger microprobe), SIMS (secondary ion mass spectrometry), XPS (x-ray photoelectron spectroscopy), total reflectance IR (infrared), LAM MA (laser microprobe mass analysis), and FTIR (Fourier transform IR) are integral to a thorough evaluation of the health effects of certain pollutants. The water content of airborne particles can affect partitioning of an inor- ganic gas between the vapor and condensed phases. The water content de- pends on the relative humidity (RH) of the air that contains the suspended particles, water-soluble salts associated with the particles, and different RHs

62 ASSESSING HU~4N EXPOSURE at which various salts deliquesce. When measuring the water content of parti- cles, it is important to remember that this value can depend strongly on RH. It is also important to ensure that the measurement procedure does not alter the water content. However, in some situations, protocols might require that the measurement be made at a specified RH. In considering the chemical nature of the analyte, some type of general classification scheme is useful in choosing sampling and analysis procedures. Such schemes can be as detailed as the outlines for inorganic and organic textbooks, or they can be fairly general. At the very least, the contaminant should be classified as organic or inorganic. If the contaminant is organic, subclassification into polar and nonpolar and as volatile, semivolatile, or non- volatile is useful. If it is inorganic, subclassification by periodic group, solu- bility, acidity, hardness, and radioactivity might be helpful. For biological analyses, an obvious distinction is that between viable and nonviable contaminants. The former include bacteria, viruses, spores, molds, and fungi. The latter include allergenic materials, such as arthropod frag- ments and insect excrement. Sampling for viable biological pollutants is com- plex, costly, and time consuming. A protocol for such sampling recently was developed by Morey and coworkers (1987~. CRITERIA FOR METHOD SELECTION This section evaluates the requirements under which a method must oper- ate, including sampling and analysis. The conditions for an ideal analysis are summarized in Table 3.2. However, optimal conditions for an analysis might require compromises. Sensitivity A method with adequate sensitivity is one in which an analyte can be de- tected at or below the level at which an adverse human-health problem is anticipated or observed. Ideally, a detection limit of at least an order of mag- nitude below the health-effect level is desirable. It also is desirable to have a broad linear range of O.lX-lOX the level of interest (i.e., a linear range of two orders of magnitude from the detection limit). Reproducibility of +2% for replicate analyses and stability of +5% during an 8-hour period also are desirable. Achieving high sensitivity during a continuous analysis is a very difficult task. Therefore, one of the first compromises made ir1 achieving high sensitivi

63 g ._ as o v g ·_ a: o : o Ct o  . V, 5 o o _ . _ ~ D X g O :> _, 'v, U. ~ Ct a: .; - _ . _ _ t~ ~ ~ Cal ~ C.) - ~ ~ ~.E _ Cal _, V Ct US I_ V ~ ~_ .5 ~ S: 's At O t~ 'U' ~ ~ ~lo: _ O _ . _ ~ 4_ _ 5 con ~ O ~O 4) t> Cal ·= O ~ O CO ~ .g _ ~ ~0 X Cal _4 ~ O .- v ~ 'v ~ · - v) c ~ - s: ~ . - , ~ :O .= · · - ) 's  o - e :> ~- '~ 5 o ._ ~ C~ C) ~ V, t~ U. Q v ._ ~D o C) .E ._ ~: C~ C~ C.) ·- `: ~ o t: ~: C~ :s :~ .- U: C~ o C) C) ~: O ~ c: =: Ct C ~ ·- `: O es  .g ~ _ _ O C~ ~C~ . _ _ ~C13 ._ ~ ~£ ~Ct c: 8 o c: ~ - V _ - ~o e ~_ ,`~: ~ _ V) ~: . ~ - 'e c~ ~ - - ~ ces .o O.4 :> ~ - · - ce ~: r - =5 O ._ ~ ~; ~ O ._ _ _ V C\S ~ .~ .~ _ :> ~D Ct _ C~ O . _ 3 ~ c: C~ o ~ o Ct ~ o - _4 t 1 ~D O - ~ g ~ ._  . - & ~ C) ~ ._ _ {,0 ~3 ~ g ~ .Um) ~ =: c ~ o C~ ~ ~ o o c: _ _ . oo- c: _ ~ & - =.~ C~ ._ C~ ~: C~ _d _ ._ ~._ o o . ~- ~: o C, - ~D - ._ C~ .4 Ct - C) .> .5 . - - o o ·- 4,,) V) .~;, C CO ~ CO es .5 ° ~ . _ ~ C~ OCO ~ C)~s ,~ U, 0 ~ C~ ~o o V

64 ASSESSING HUMAN EXPOSURE ty is to obtain a sample integrated over time, which reduces the temporal resolution in following the course of an exposure. This can have serious un- plications when short-term, high-level exposures have adverse health effects and could go unnoticed in an integrated exposure. Another area of compro- mise is to sacrifice selectivity to obtain sensitivity. An example of this occurs in the analysis of polychlorinated biphenyls (PCBs). Great selectivity can be achieved in analyzing the 209 different congeners using high-resolution gas chromatographic (GC) techniques; the individual concentrations of the con- geners can be summed to obtain a total PCB content. However, a technique with better sensitivity would be to chlorinate all PCBs to decachlorobiphenyl and analyze for the single compound. This improved sensitivity occurs at the expense of a detailed knowledge of which PCB congeners were present. Con- gener-specific information may be very important in an exposure-assessment perspective (McFarland and Clarke, 1989~. Sensitivity also can be improved by using a type of fixed-base monitoring sampler compared with a portable device that could be worn on the body. Fixed-base samplers often use more sophisticated detection methods (and are usually more expensive and bulky) than portable samplers and thereby im- prove the overall sensitivity of the analyses. Selectivity A method that is selective (or specific) is one in which the response ob- served for a desired analyte is due only to that analyte and is not from an interfering analyte or artifact produced during sampling or analysis. When making an analytical determination, the quantitative results obtained must be for the correct analyte and only that analyte. If other compounds interfere in the analysis, then it is important to understand their contribution to the results so that the data can be properly interpreted or so that other, more selective methods can be sought. High selectivity can be achieved using various high-resolution GC tech- niques coupled with appropriate selective detectors. However, chromato- graphic separations can be time consuming and require each analysis to be discrete. For example, in a continuous analysis for total aromatic compounds, the compounds can be fed directly into a photoionization detector (PID) with- out any chromatographic separation. Although the analysis is rapid, little information is gained about the chemical nature of the detected compounds. Separating the aromatic compounds on a high-resolution capillary column before PID detection reduces the rapid nature of the analysis, but greatly enhances the selectivity. Another means to increase selectivity of the analysis

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 65 is to use more sophisticated instrumentation, especially in the final detection of analyses. As an example, analyses can be carried out using GC in combina- tion with mass spectrometry (MS) as the detection device. However, MS is significantly more expensive than PID and, due to its size, reduces the portability of the analytical system. Rapidity A method is considered rapid if either the sampling or the analysis can be m~ie-d out on a time frame that is short compared with any adverse health response observed in an exposed individual. In this definition of rapid, real- time analyses are not necessarily carried out; the concentration of the analyte to which the individual is exposed could be determined in a short time frame. For example, several discrete samples could be taken on adsorbent media and subsequently analyzed. Understanding of the pharmacokinetics of the analyte in question is important in determining the sampling time. It is unnecessary to sample frequently if an analyte's half-life in the body is long. However, if an analyte is excreted or metabolized rapidly, it might be necessary to analyze within a very short time (Rappaport, 1988~. Sensitivity often is abandoned to achieve rapid analysis. If sufficient sensi- tivity is available (i.e., 10% of the level of interest), then shorter sampling times can be used to achieve the necessary temporal resolution. This ap- proach must be tempered with the knowledge that subsequent analyses are expensive, and the cost of the total analyses might be prohibitive. Selectivity also might be foregone to achieve rapid analyses. As with the earlier PCB example, if little discrimination is required in the analysis of aromatic com- pounds using PID, then the selectivity provided by the chromatography col- umn could be eliminated and the analyses carried out rapidly to give the total concentration of aromatic compounds. Comprehensiveness A comprehensive method often is desired for analyzing all analyses that might be responsible for an adverse health effect, particularly when a synergis- tic effect between analyses might exist. Comprehensive analysis can be par- ticularly important when trying to determine the ultimate source of a toxicant. It often is useful to track which compounds change concentrations in unison and use this information to identify their primary sources. One way of attaining comprehensiveness is to have multiple methods run

66 ASSESSING HUMAN EXPOSURE ning simultaneously. The difficulty with this approach is that it can become very expensive and time consuming. In many cases, a method can handle several analyses if they are similar in chemical nature; thus, the necessary comprehensiveness can be achieved with a single method. Portability Personal sampling or analysis devices must be sufficiently small, light, and quiet to be worn by individuals without causing them to modify their normal behavior. In many cases, samples are collected and then analyzed later. In situations that require real-time continuous (rapid) analysis, the analysis device might need to be worn by the individual. The power consumption of the device must be low to minimize the weight of the batteries and their need for recharging. The unit should also be rugged, because it might need to endure extreme conditions of heat, humidity, and shock. In addition to personal sampling, there is a continuing need for portable methods for air analysis that could be deployed in a variety of field settings (e.g., portable gas chromatography and mass spectrometers). Portability might be difficult when determining many analyses, especially when multiple methods are required. In these instances, a f~xed-based monitor would be needed to accomplish the desired analyses. Using a passive sampler rather than an active one sometimes precludes continuous monitoring; however, the simplicity and low cost of passive moni- tors make them ideal for portability. Table 3.3 presents the status of personal monitor development for selected contaminants. Cost The cost of sampling arid analyzing an analyte in a statistically sound man- ner should not be prohibitive. An excellent analytical technique might be available, but if the cost per analysis is too high, it could result in too few samples being taken to give a proper measure of exposure. Automation can reduce the costs; sharing capital equipment eliminates the need for costly equipment duplication. Many times the increased demand for a particular type of analysis encourages laboratories to devise simr~ler. cheaper ways of carrying out the analyses. One way to control sampling and analysis costs is to conduct a few prelimi- nary comprehensive analyses to survey actual conditions and then analyze a subset of compounds. Another method is to take fewer samples, which means --r---~ ~Or

67 4_ I: o o ._ a: o - Ct o U) C) o In Ct mom :~ ~ ~ I a ~ I ~ ~ I 1 ALL 1 . .. ~ _ _ _ _. _ . -A o1 1 1 1 ~ 1 1 1 1 ~  ~ R 'id R~L ~ ~ i:: ~: ld:~ ~0 0\ - C) - Ct o 11 Ct - ~: . Cot Q) - Ct U. o o ._ - o Cot ._ Ct ._ 11 - . Ct C) ._ 11

68 ASSESSING HUMAN EXPOSURE that fewer analyses must be performed. Conducting fewer analyses is a com- promise when continuous analyses are prohibitively expensive; such a compro- mise requires that the sampling resolution be established at the outset of the exposure study. If portability is a factor, it can be expensive, and a less costly alternative is to use a fixed-based system. METHODOLOGY The Measurement Process The measurement of an airborne contaminant can be visualized as a three- step process. First, the pollutant is sampled; then it is separated from other species also collected in the sample; finally, it is detected. In actual practice, these steps frequently overlap (Figure 3.1~. In this figure the individual rings for sampling, separation, and detection show areas of overlap, as well as areas where no overlap occurs. Different measurement processes have different combinations of overlap, ranging from none to complete. An example of a measurement procedure with no overlap is the following approach to the de- termination of the airborne concentration of benzoapyrene (BaP). First, res- pirable particles that might contain BaP are collected on a filter. Second, BaP is extracted from the particles and then separated from other compounds in the extract by thin-layer chromatography. Finally, BaP is detected using fluo- rometric techniques. Such measurement procedures (i.e., no overlap among the three steps in Figure 3.1) are usually labor-intensive. In many analytical procedures, two of the rings overlap; the overlap most frequently encountered is between separation and detection. For example, in GC with flame-ioniza- tion detection, one instrument combines the separation step (GC) with the detection step (flame-ionization detection). Less commonly, an analytical procedure involves overlap between sampling and separation. An example of such an overlap is a diffusion denuder, in which, as the vapor-phase pollutant is sampled it is also separated from the condensed-phase pollutant. Overlap between sampling and detection also occurs, as in a photometric ozone meter. Finally, in some analytical methods all three rings overlap (e.g., MS-MS). The following sections discuss sampling, separation, and detection as iso- lated steps. However, there are necessarily areas of overlap in each section, just as there are areas of overlap in Figure 3.1.

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 69 / / - - Detection - - - Sampling / - X / \ Separation 1 , -it' 1 - \ FIGURE 3.1 Steps in the measurement process. Different processes have overlap ranging from none to complete. Sampling Airborne contaminants are sampled actively or passively. Active sampling uses a pump to pull airborne contaminants through a collection device. Pas- sive sampling sometimes referred to as diffusive samplin~relies on diffusion to deliver airborne contaminants to the collection medium.

70 ASSESSING HUMAN EXPOSURE Passive Sampling The major advantage of passive sampling is that it does not require elabo- rate equipment. No pumps need to be maintained or calibrated; the sampling site does not need to be close to a power source; and pump noise, a frequent complaint of active samplers, is not a factor. Because pumps and accessories are not required, passive sampling is less costly and easier to implement than active samplers. Extensive passive sampling programs have been conducted by mail (Sexton et al., 1986)-sample collection devices can be sent and re- turned in this fashion. Passive samplers are favored for personal monitoring; they are lighter, smaller, and less likely to interfere in daily activities than are active samplers. Consequently, participants are more likely to cooperate in passive sampling programs. Furthermore, since the sampling device can be worn in the breathing zone, pollutants whose spatial and temporal concentra- tions vary extensively can be integrated appropriately. The major disadvantage associated with passive sampling is the long period required to collect sufficient material for analysis Indeed, these long integra- tion times mean that passive sampling is not suitable for pollutants whose health effects depend on peak exposures or pollutants that have immediate, acute health effects (e.g., hydrogen cyanide). Passive sampling also tends to be less accurate than active sampling. In addition, artifacts might arise from chemical transformations on the surface of the sorbent, a concern that in- creases as the length of the sampling period increases. The first widely used passive sampler was an NO2 monitor developed by Palmes (Palmes et al., 1976~. Passive samplers have since been used for for- maldehyde (Geisling et al., 1982), water vapor (Girman et al., 1986), nicotine (Hammond and Leaderer, 1987), and nonpolar volatile organic compounds (VOCs) (3M, 1982; Seifert and Abraham, 1983; Shields and Weschler, 1987~. A fixed-site passive sampler for ozone has recently been described (Monn and Hangartner, 1990) and an ozone passive sampler suitable for personal moni- toring has also been reported (Koutrakis et al., 1990~. A passive monitor for CO is under development at the Lawrence Berkeley Laboratory. However, few passive sampling devices are available for polar VOCs (e.g., acrylonitrile and selected amines), highly volatile compounds, and extremely reactive com- pounds. The nature of the sorbent varies with the nature of the analyte. Ideally, a linear concentration gradient exists from the open end of the sampler (ambi- ent concentration) to the sorbent surface (zero). An effective passive sampler requires an efficient sorbent that will keep the sampled pollutant concentra- tion near zero at the sorbent's surface. It must also be possible to desorb the pollutant or a derivative quantitatively for subsequent analyses.

SAMPLING AND PHYSICAL-CHEMICAL MEA SUREME[JTS 71 The potential exists to develop passive samplers specific to numerous gas- eous airborne pollutants. Many sorbents have not been tried or fully evalu- ated in passive devices; some materials recently introduced for GC and liquid chromatography might have application in passive sampling. Furthermore, promu~g new sorbents are being synthesized and developed. Some of these have been engineered at the molecular ferret and have binding sites or cavities specific to a given compound or class of compounds. To sample for extremely reactive compounds, the sorbent can be deliberately designed to react quickly and efficiently with an analyte to yield a stable, nonvolatile product that can be extracted and quantified. An interesting approach for selected analyses is to interface passive sam- plers with active devices. Passive sampling can be combined with sensors to make portable instruments and obtain a real-time readout of concentration. Such devices are available for CO, combustible gases, and other pollutants (Stetter and Rutt, 1980; Stetter et al., 1984; Penrose et al., in press). Active Sampling Although passive sampling is well suited to vapor-phase pollutants, active sampling can be used for this as well as condensed-phase contaminants. For vapor-phase pollutants, a known volume of air commonly is pumped through an efficient absorbent (Pagnotto, 1983~. Absorbents frequently used for or- ganic compounds include Tenax, XAD-2, activated charcoal, Ambersorb XE-340, polyurethane foam, or a combination of these absorbents in series. When using solid absorbents in active sampling, it is important to know the collection efficiencies of the sorbent for the analyses in question (Serum, 1981; Pagnotto, 1983; Bidleman, 1985~. A large body of literature exists on the retention and efficiency of numerous sorbents (e.g., Gallant et al., 1978; Figge et al., 1987; Maier and Fieber, 1988; Pankow, 1988~. When a sorbent has not been previously characterized, "backup" sections should be utilized as a measure of analyte breakthrough. An approach sometimes used with gas-phase acids or bases is to pump air through a filter impregnated with an appropriate base or acid that will scav- enge the airborne acid or base. Impingers with various absorbing solutions are commonly used for volatile pollutants. More recently, canister samplers have been designed in which an air sample is drawn into an initially evacuated stainless-steel canister. The internal walls have an inert chrome-nickeI oxide surface to decrease wall reactions. Commercial versions of these samplers can collect as much as 8 L of air for subsequent analyses. In each of these active approaches, some method is required to remove

72 ASSESSING HUMAN EXPOSURE airborne particles from the airstream before the volatile compounds are col- lected. The most common method is to place a filter upstream of the absor- bent, treated filter, impinger, or evacuated canister. However, potential arti- facts are associated with this approach. Semivolatile compounds adsorbed on coldected particles can evaporate from the surface of the filter and these com- pounds are sampled downstream by the collection device. An error in the opposite direction results when vapor-phase compounds adsorb or react with particles on the filter or the filter matrix itself (Van Vaeck et al., 1984; Cou- tant et al., 1988; Ligocki and Pankow, 1989~. In either case, the resulting measurements do not reflect accurately the partitioning of certain semivolatile compounds between vapor and condensed phases (see previous section on airborne analyses). This is potentially important information, because the phase of a pollutant can affect its chemistry, deposition site in the respiratory system, and toxicology. Use of a diffusion denuder eliminates this problem (All et al., 1989~. Air is pulled through a cylindrical tube whose walls are coated with an absorbent specific for the pollutant of interest. The interval dimensions of the tube and air velocity are such that only the vapor-phase pollutant diffuses to the walls; particles pass through to the other end (where they can be collected on a filter, if desired). Diffusion denuders can be configured using different absor- bents and geometries and are potentially applicable to a wide variety of com- pounds. They could be very useful to evaluate the partitioning of a compound between vapor and condensed phases (Coutant et al., 1988; Lane et al., 1988~. Many Improvements have been made in diffusion denuders; annular denuders collect reactive atmospheric gases 15-20 times more efficiently than tubular denuders (Possanzini et al, 1983; All et al., 1989~. However, these devices are too bulky for personal sampling. The annular denuder/f~lter-pack system to collect acidic gases and aerosols described by Koutrakis et al. (1988) is a promising monitoring device, but is more than 2 feet long. Even u ith diffusion denuders, the analyst must be wary of artifacts. Species absorbed on the treated walls of the denuder can be oxidized or reduced (less common) by other reactive gas-phase pollutants. For example, in laboratory studies, ozone has been demonstrated to oxidize nitrite ions on an alkaline denuder surface to nitrate ions (Febo et al., 1986~. In some active sampling methods, the airborne pollutant is not collected, but passes through a flow cell, where some physical property of the pollutant is monitored. From this measurement, a concentration is derived. Different examples of this approach include ozone meters and optical particle counters. The former compares the absorption (at 254 rim) of ambient air and ambient air from which ozone has been removed selectively. The latter uses light scattering to monitor the num- ber concentration versus diameter of airborne particles. In either case, ambi

SAAlPLING AND PHYSI=L-CHEMIC~ M~SU=ME=S 73 ent air is pumped through a flow cell where the measurement is made. Most real-time continuous analyzers sample in this way. Optical particle counters measure particle-size distributions based on light- scattering diameters. Measurements based instead on aerodynamic particle size can be made with a laser velocimetry technique. A commercial ~nstru- ment is available that uses a split laser beam to monitor the velocity of parti- cles leaving an accelerating nozzle. The device can monitor particles 0.5-15 Em In diameter for numerous size ranges. It provides not only data on num- ber concentration versus aerodynamic diameter, but also on mass concentra- tion and surface area concentration versus aerodynamic diameter. The same pling rate is 5.0 LJmin. For health studies, aerodynamic sizing is preferable to sizing based on light scattering, since respiratory deposition relates more directly to the former. Another instrument that provides data on the mass concentration of air- borne particles is the piezobalance. The particles are sampled by electrostatic precipitation onto the surface of a piezoelectric device. The change in res- onant frequency of the piezoelectric quartz crystal is used to monitor the mass gain. Electrostatic sampling is efficient for particles 0.01-10 Em in diameter. Commercial units sample at 1.0 L/min and can measure mass concentrations 10-10,000 ~g/m3 in real time. A similar instrument is the tapered element oscillating microbalance (TEOM). A variety of collection stages can be fitted to the narrow end of the oscillating tapered element. As mass is added to a particular collection stage, the frequency of oscillation decreases in relation to the amount of deposited mass. This instrument provides a particle sample that can be analyzed by other techniques. The TEOM can be used for personal monitoring and is more versatile than the piezobalance (Patashnick and Rupprecht, 1986), but it also is more expensive. Particles can be counted in flow cells, but a collection procedure is required if subsequent chemical analysis is intended. The simplest approach is to pump air through a filter. High-volume (hi-vol) samplers do this at very high flow rates~ypically 15 m3/min. However, the size of airborne particles is an important factor in estimating potential human exposures, which argues for including size-fractionation as part of any procedure that samples airborne particles. Furthermore, chemical composition and biological activities are related to particle size (Phalen et al, 1986~; these size ranges are based on the inhalation of particles by humans. The "inspirable mass fraction" comprises all particles that enter via the nose or mouth. The "thoracic mass fraction" is all particles that penetrate past the larynx. The "respirable mass fraction" is all particles that penetrate past the terminal bronchioles. Standard curves (the

74 ASSESSING HUMAN EXPOSURE percent penetrating to collector versus aerodynamic diameter) have been de- f~ned for each of these size fractions (Phalen et al., 1986~. To sample particles smaller than a certain diameter, size-selective inlets of various design can be placed upstream of the collecting filter. Size fractiona- tion can also be achieved with dichotomous samplers. Such samplers size- fractionate airborne particles into fine and coarse modes. The fractionation is based on the particles' aerodynamic sizes; particles are separated by virtual impaction and are collected on filters downstream from the separation device. Dichotomous samplers normally are fitted with a 10 or 15 Em diameter size- selective inlet, which places an upper limit on the size of the particles sampled in the coarse mode. Dichotomous units sample at a relatively low rate (typi- cally, 16e7 L/min) and, consequently, long sampling periods (1 day to 1 week) are required to collect sufficient material for chemical analyses. Sampling of airborne particles in more than two size ranges usually is con- ducted with a cascade impactor. This device is simply a set of impactors, operating in series, arranged in order of decreasing cutoff size. An impactor plate collects the particles in each size range and the final impactor is fol- lowed by a filter. The separation is again based on aerodynamic particle size. Commercial devices are available with as many as eight stages. If there are too many stages, the collection curves between successive stages overlap; such overlap already occurs with an eight-stage impactor. Furthermore, more stages mean more sampling time required to collect sufficient material in each size range for gravimetric and chemical analysis and more samples to analyze. An additional problem encountered with cascade impactors is that of "particle bounce"- particles fail to stick to their appropriate impactor plate and become re-entrained with smaller particles. Particle bounce can be reduced by greas- ing impactor plates, but this is only reasonable when examining inorganic constituents of a sample. A virtual impactor (such as the dichotomous unit) avoids this problem, because it has no impactor plates. There is a potential for sampling bias when collecting particles with a mass median aerodynamic diameter larger than 3 ~m. The combination of a small sampling inlet running at a high flow rate will result in undersampling large particles because of the inertia of these particles at the inlet (Breslin and Stein, 1975; Selden, 1975; Agarawal and Liu, 1980~. Sampling programs designed to assess human exposure to airborne pollu- tants frequently include indoor sampling. When active sampling is conducted indoors, care must be taken that the pumping rate is not so great (relative to the space being sampled) as to alter the indoor environment. In small rooms, this limitation precludes the use of hi-vol samplers or certain cascade impac- tors. To collect useful amounts of airborne particles with devices that sample at lower rates, lengthy sampling intervals are required. Because the volume

SAMPLING AND PHYSICAL-CNEMICAL MEASUREMENTS 75 of air cannot be increased without perturbing the sampled environment, the only way the sampling interval can be decreased in such studies is to improve the sensitivity of the subsequent particulate analyses. Recently, the indoor-air sampling impactor (IASI) was developed by Marple et al. (1987), which can operate at 10 or 4 L/min and has a sharp cut size at Act = 10 ,um and as<' = 2.5 ~m, respectively. Personal monitors that have a 2.5-pm dso cut size have been used in industrial hygiene (ACGIH, 1988a). Recently, personal impactors with a 10-pm d50 have been used in community-based field studies (Lioy, 1988~. Separation Chromatography Chromatography is a separation process often used to isolate airborne contaminants from other compounds that might interfere in detection of speci- fied contaminants. Chromatography involves the simple partitioning of ana- lytes between mobile and stationary phases. Using this approach, tens to hundreds of compounds are readily separated. Stationary phases can vary widely, including silicones, silica, alumina, florisil, various polymers, silica in which various materials have been bonded to the surface, and most recently, commercially available liquid crystals. An important feature of the stationary phase is that it must not be miscible with the mobile phase, and must not dissolve or volatilize in the mobile phase. Mobile phases also vary widely, ranging from gases to liquids, and include gases, which, when held above their critical pressure and temperature, behave as supercritical fluids. In most cases, chromatographic techniques derive their names from the nature of the mobile phase, e.g., gas chromatography (GC), liquid chromatography (LC), and supercritical fluid chromatography. How- ever, thin-layer chromatography (TLC) derives its name from the manner in which the stationary phase is configured. In TLC, the actual chromatographic process most often involves placing a sample on the head of a column packed with the stationary phase. The mo- bile phase then transports the components of the sample through the column, with the components separated based on the amount of time they spend in (adsorbed to or partitioned into) the stationary phase. 1d50 = diameter at which 50% of the particles penetrate to the collection medium.

76 ASSESSING HUMAN EXPOSURE Chromatographic techniques are often highly selective. However, this sep- aration process takes time, which gives chromatographic techniques the dis- advantage of being noncontinuous. The following discussion describes various chromatographic techniques and their application to analyzing airborne con- [aminants. It includes a discussion of each technique's advantages and disad- vantages, recent developments in the field, and new applications of the tech- nique. Gas ChromatograpI:y GC involves the separation of compounds based on their volatility and interaction with the stationary phase (Jennings, 1987~. The mobile phase is a gas, and the stationary phase can be quite variable, ranging from molecular sieves to synthetic organic polymers and liquid crystals. Because the separa- tion process involves compounds in the gas phase, GC is ideal for the analysis of many airborne contaminants. The National Institute for Occupational Safe- ty and Health (NIOSH) describes more than 500 methods to analyze gases and vapors. Of these methods, 51% are by GC (Saltzman, 19885~. Advantages. The most significant advantage of GC in the analysis of air- borne contaminants is that it combines the inherently high resolving power (high selectivity-the ability to separate individual components) of the chro- matographic column with easy interfacing to a variety of detection devices. The application of capillary columns, with internal diameters of 0.25-1 mm, has become increasingly popular during the past decade. Preparing these columns from fused silica has given them the inertness of glass columns with the flexibility of metal columns, making them particularly easy to use. Advan- ces In capillary-coating techniques have permitted thick film columns to be developed, which are ideal to separate highly volatile air contaminants (Jen- nings, 1987~. This provides a tremendous amount of selectivity in isolating an analyte of interest, even in complex matrices. However, the chromatographic column only separates compounds; it does not provide for their detection. Fortunately, GC is readily interfaced to a variety of detectors that provide either nonspecific (universal) detection of compounds eluting from the column or their highly specific and often highly sensitive detection. GC instruments can be made transportable and, in many respects, even portable. They are rugged enough to be transported in vehicles and taken to a site for field analysis. Many GC-detector system combinations are inexpen- sive and readily obtained. GC-detector combinations are also very flexible they can be configured to analyze for a very wide range of compounds. If sufficient sample is available and does not need to be analyzed immediately,

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 77 then a GC-detector system can be set up to analyze low-molecular-weight compounds and then reconfigured with a column suited for higher-molecular- weight compounds and, in this manner, achieve a more comprehensive analy- sis. If a high degree of specificity is needed, then the analysis of a specific analyte can be carried out using one type of chromatographic column, fol- lowed by reanalysis an a second column of a different character than the first, and the results from the two columns can be compared. This type of confir- mation is carried out to reduce the possibility that multiple compounds might coelute on a specific column, resulting in misleading results (Overtop et al., 1988~. GC is readily automated, which allows for unattended operation. This distinct advantage greatly facilitates the analysis of large numbers of samples and helps keep the cost per analysis low. Liquid autoinjectors for GC are available and are extremely useful in analyzing solutions such as those derived from the solvent Resorption of air-sampling devices (e.g., carbon disulfide desorption of charcoal tubes). Automated thermal desorption devices are also available for the analysis of air sampling tubes (e.g., Tenax sorbent trapping devices) (Kester and Zaffiro, 1987~. In addition, the recent popularity of "pas- sivated" air sampling canisters (e.g., SUMMA-treated canisters) has led to techniques to sample automatically from several gas containers (Blacha et al., 1988~. Furthermore, the combination of GC with mass spectrometry has made this a very popular tool for both the qualitative and quantitative analysis of a wide range of airborne contaminants. Disadvantages. The separation process that gives GC high specificity re- quires time. During this time, no other analyses can be conducted, which eliminates the continuous nature of any analysis. Many attempts have been made to work around this problem. These attempts include high-speed analy- ses in which thin-film capillary columns can separate 30-40 compounds in approximately 1 min (Hewlett-Packard, 1987). Another approach is to pro- ceed with the analysis until the analyte of interest elutes; the column is then backflushed to eliminate any late-eluting compounds from interfering in subse- quent analyses. If confirmational analyses involve a second column or differ- ent detector, then sufficient sample must be available, and it must be stable until the second analysis can be carried out, taking even more time. Some of these problems can be minimized by having two columns installed in the same chromatograph and splitting the sample between the two columns. In a simi- tar fashion, multiple detectors can be used simultaneously by splitting the eluant from the column between multiple detectors (Earp and Cox, 1984~. In all these instances, the selectivity can be increased, but the time needed to do the analyses prevents the technique from being continuous. New developments in GC also might help to resolve some of the limita ~.

78 ASSESSING HUMAN EXPOSURE lions. Multiplexing the analysis of samples has been used to decrease analysis time without greatly modifying the method (Phillips, 1980~. In this method, a second sample is injected onto the chromatograph before all the compounds have eluted from the first injection. Computer software tracks which peaks belong to which injection. Multiplexing can be an excellent technique for decreasing the analysis time without extensively modifying the method or pur- chasing additional hardware. Another disadvantage of GC is that it is not necessarily comprehensive. Certain highly reactive compounds are difficult or impossible to put through a chromatography column. These compounds either decompose at the tem- peratures required to be chromotographed or react with the stationary phase or the column material. In other instances, a compound decomposes on the column to form another, giving misleading results. An example of such a process is the on-column decomposition of N-nitrosodiphenyl amine to form diphenyl amine. In many instances, the chromatograph is operated in a tem- perature-programmed mode to increase the range of compounds that can be analyzed. This serves as an alternative to first analyzing for low-molecular- weight compounds and then reconfiguring for higher-molecular-weight com- pounds. This programming requires time for the analysis and for the chroma- tograph to recover to the initial starting point before beginning any subsequent analyses. Temperature programming also requires elaborate electronics to control the column temperature reproducibly. This increases instrument cost and, in some cases, increases its size and weight, making it less portable, al- though portable gas chromatography now are available that are capable of limited temperature programming. As chromatography become smaller alla smaller, a "gas chromatograph on a microchips could be fashioned (Angell et al., 1980; Overton et al., 1988~. Such a device might be so inexpensive that analysis time could be cut in half by doubling the number of chromatography: the sample would be sent to one GC and, while the compounds are elating from it, the next sample would be sent to the second chromatograph. In this manner, any reasonable time resolution desired could be obtained. Such small chromatography also would be more readily temperature programmed due to their small mass. They could thus be used to analyze a broad range of compounds while maintaining relatively short analysis times. Another technique that is gaining widespread use is multicolumn GC (Li- gon and May, 1984~. In this technique, the compounds first are separated on a primary column, and then fractions are cut from that column onto secondary columns. A nice application of this technique in air analysis involves the use of the primary column to separate water (the major component) from low- molecular-weight oxygenated hydrocarbons (tin et al., 1988~. These hydro

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 79 carbons then can be routed to a second analytical column without the deleteri- ous effects of water, and the detector can be run at a high sensitivity, because the major constituent has been eliminated. MulticoIumn GC has also been used recently to separate all of the PCB congeners in commercial mixtures (Schulz et al., 1989~. The multicolumn chromatograph can further be used to increase the: comprehensiveness of analysis without seriously increasing the analysis time. The primary column routes highly volatile compounds to a secondary column ideally suited to separate these types of compounds and then routes the low-volatility compounds to a second column designed for their analysis. The analysis for these two classes of compounds goes on simul- taneously, which shortens the analysis times The hardware for multicolumn chromatography is expensive and method development can be difficult, but the versatility of the technique should see its wider application in future air-analy- sis programs. Liquid Chromatography LC involves the separation of compounds when the mobile phase is a liquid (Snyder and Kirkland, 1979~. In many applications of LC the backpressure developed by the column requires the use of sophisticated pumping systems and has been named either high pressure or high performance liquid chroma- tography (HPLC). The mobile phases used with LC columns can vary quite widely depending on the nature of the analyses and the stationary phase. Typical mobile phases include hexane, methylene chloride, mixtures of water with acetonitrile or methanol, and various buffer solutions. The stationary phases are also quite diverse; they include silica, alumina, and florisil bonded phases in which an organic molecule has been attached to a silica surface, and a wide variety of ion-exchange resins. Use of ion-exchange columns has re- sulted in an LC class known as ion chromatography (IC), which uses a special type of ion-Is suppressor column to permit sensitive detection of the separat- ed ionic species using electrolytic conductance techniques (Small et al., 1975~. IC is used routinely to identify and quantify the major anions and cations associated with airborne particles, including sulfate, nitrate, chloride, ammoni- um, sodium, potassium, magnesium, and calcium. For all types of LC, the analyses of interest are dissolved in the mobile phase and then pass through a column packed with the stationary phase. The compounds elute from the column in the order determined by the extent of their interaction with the stationary and mobile phases. Again, as with GC, LC only separates compounds, it does not detect them. However, several

80 ASSESSING HUMAN EXPOSURE detectors can be interfaced to LC to provide for the sensitive and selective detection of eluted compounds. LC is not used as widely as GC to separate and analyze airborne con- taminants. LC is used to analyze such contaminants when the compounds are so thermally labile, highly reactive, highly polar, or nonvolatile that GC analy- sis is difficult or Impossible. Inorganic acid gases (e.g., HE, HCl, H3PO4, HN03, HBr, and H2SO4) are examples of these types of compounds. These gases can be collected from ambient air using water impingers or silica tubes, subsequently elated with water, and the ionic species (e.g., Fit and CI~) are then analyzed by IC (Eller, 1984a). A similar approach can be used to ana- lyze organic acids (Rosenberg et al., 1988~. LC also can be applied to the analysis of airborne contaminants that are difficult to trap or concentrate because of their reactive nature. An example of this class of compounds is the various aldehydes that can be collected in a solution of 2,4-dinitrophenyl- hydrazine (DNPH). DNPH reacts with the aldehydes, forming a hydrazone derivative that can be extracted and separated from interferences using LC. A particularly useful feature of the DNPH derivatization is that it forms a compound with a strong chromophore, which makes it amenable to sensitive ultraviolet (UV) adsorption detection (Riggin, 1984~. Thin layer chromatography (TLC) has direct ties to LC and often is used as a prescreening tool for in situ scanning of the TLC plate to determine the best mobile and stationary phases to use for a particular LC separation. TLC,s main application to the analysis of airborne contaminants is for the analysis of semivolatile compounds present on airborne particles that are small enough to be taken into the lung. A typical application would be for the separation of polycyclic aromatic hydrocarbons (PAHs) extracted from particulate matter, with the PAHs detected spectrofluorimetrically (Lioy et al., 1988~. Advantages. A major advantage of LC techniques is that the separation processes are done at lower temperatures than GC, allowing analysis of com- pounds that might be destroyed during GC analysis. For example, N-nitroso- diphenylamine can be separated from interfering compounds on an LC col- umn whereas, as stated before, this compound breaks down to diphenylamine during GC analysis. LC also is useful to separate organic compounds that are too polar or not sufficiently volatile for routine GC procedures. This applies to certain organic species associated with airborne particles. Thus, LC can be combined with GC techniques to give comprehensive analyses. Furthermore, since the columns and mobile phases used with LC can be tailored to par- ticular analyses, the comprehensiveness of LC techniques is enhanced. LC techniques are readily automated, which allows for unattended operation if many samples must be analyzed. The use of high-resolution columns also gives LC a high degree of selectivity. Recently, portable LC instruments have

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 81 been demonstrated (Berg and Buckley, 1987; Otagawa et al., 1986), which may lead to applications in field monitoring. A variety of LC detection techniques can be used to provide sensitive de- tection of eluants, including spectrophotometric, fluorometric, electrochemical, and refractivity detectors. In most instances, the original air contaminant is detected directly after separation. An example is the fluorescence detection of PAHs extracted from particulate matter (Wise et al., 1986~. In other cases, air contaminants such as aldehydes can be reacted with a reagent, such as dinitrophenyhydrazine (DOPE) before LC analysis to enhance their detection byUV. Another advantage :~f LC is that the column eluants are In a liquid media um; they can be manipulated chemically (oxidized, reduced, or derivatized) to enhance their detection. For example, nitrated PAHs from airborne, inhalable particles are reduced to amino-PAHs using a post-column, and the amino- PAHs are then detected using fluorescence techniques (Greenberg et al., 1986; Tejada et al., 1986~. This process is done continuously; no analysis time is lost, and the procedure can be automated. A slightly different twist to post- column derivatization and detection has been used to analyze aldehydes (Ta- keuchi et al., 1988~. In this technique, nicotinamide adenine dinucleotide (NAD) was included as part of the mobile phase. When an aldehyde eluted from the column, a post-column reactor of immobilized 3-alpha-hydroxysteriod dehydrogenase reacted with the NAD in the presence of the aldehyde to form NADH, the reduced form of NAD. The NADH was then detected by fluo- rometry. This technique further illustrates the versatility of detecting eluants from an LC column. Disadvantages. LC equipment tends to be more costly than GC and typi- cally less portable. LC shares the GC disadvantage of the time required for high selectivity, which prevents the analysis from being continuous. Mass transfer limitations prevent LC from attaining the high-speed analyses achiev- able with GC. Other techniques, such as multiplexing (multiple separations and detections occurring simultaneously), might be tried with LC to attempt to decrease the time between analyses. LC is used only if an analysis cannot be done by GC. Although several detectors are available for LC, the wide array of detectors available for GC make it the choice for analysis. Many GC detectors are difficult or impossible to use with LC. LC needs a universal detector comparable to the GC flame- ionization detector. Techniques are available to interface LC to mass spec- trometry, and their development and application will complement, but not re- place, GC as the primary chromatographic tool for the analysis of airborne contaminants.

82 ASSESSING HURON EXPOSURE Supercnfical Fluid C~romatograp~'y Future developments in the chromatographic analysis of airborne con- taminants win occur in the area of supercritical fluid chromatography (SEC). SFC uses supercritical fluids~est thought of as very dense gases with unique solvating properties as the chromatographic mobile phase. The chromato- graphic principles of SEC are very similar to those for GC and LC, and SEC can be considered as a bridge between GC and LC (Chester and Innis, 1985; Chester, 1986; White and Houck, 1986~. The primary factor that affects the solvating or elusion power is the density of the gas used as the mobile phase, which can be readily adjusted by varying the pressure and, to a lesser extent, the temperature. Density is analogous to temperature in GC or mobile phase composition in LC, affecting the chromat- ographic separation of a matrix. Various gases and liquids, such as CO2, Ar, freons, NH3, N2O, hydrocarbons, and even water, can be used as mobile phases to provide a wide range of polarities to affect separations. Modifying agents can also be added to the mobile phase to alter elusion characteristics. SEC uses the unique solvating and elusion power of supercritical solvents. Unlike GC, SFC allows for chromatography of relatively high-molecular- weight nonvolatile compounds. The technique is similar to LC in that many thermally labile components are chromatographable due to the low tempera- ture solvating power of the mobile phase. (Components with relatively high mobile phase densities that provide analyte elusion can be reached at low operating temperatures. The use of capillary columns, as opposed to packed columns, combined with an instrument capable of simultaneous density and temperature program- ming has greatly extended the usefulness of SFC. Packed columns suffered from limited resolution, and the high mobile phase flow rates made detector interfacing difficult. SFC has been successfully interfaced to numerous detectors, including GC- like detectors, such as flame ionization, flame photometric, thermionic, micro- wave plasma, and electron capture detectors, as well as mass spectrometers, and Fourier transform IR spectrometers. SEC also has been interfaced to LC- like detectors such as UV-absorbance and fluorometric detectors. The ability to interface SFC u ith so many detectors greatly enhances the versatility of the technique in comparison with LC methods. The one major problem confronting capillary SFC is the column's limited sample capacity, which in turn limits the ultimate sensitivity of the method. New injection techniques are being devised In which more analyte is intro- duced onto the column while limiting the amount of solvent (Andersen, 1988; Anderson et al., 1989~. These techniques should be useful in overcoming the

SAMPLING AND PHYSICAL CHEMICAL MEASUREMENTS 83 limitations of capillary SEC. When used with capillary columns, SFC obtains near capillary OC-like resolution. This resolution is invaluable since it permits additional compounds to be added to the sample matrix such as internal standards that are resolvable from the compounds already present in the sam- ple. Although GC is capable of higher resolution than SFC in a shorter time, capillary column SFC offers potentially greater resolution than LC in less time (Schoenmakers and Verhoeven, 1987~. If an analysis can be done by GC, that is the method of choice. However, if it cannot be done by GC or can be done only if extensive derivatization procedures are necessary, then use of SFC may be a useful alternative. Although SEC cannot replace GC or LC, it plays an intermediary, role between these techniques and Beep to be evaluate-d further. Permselective Merit bran es (Microfiltration ) Semipermeable membranes have been used to separate species collected in the environment in selected applications. Nafion is a fluoropolymer with a cation-exchange capability. It can be used to adsorb ions from solution or to scrub water from an air stream during sampling. The fluorocarbon, Teflon- like matrix gives Nafion the inert quality to remove humidity or add humidity to a sampling stream without altering the chemical nature of the sample. Porous Teflon membranes (e.g., Zitex and Gortex) can be purchased in a variety of pore sizes to separate particles from gases and vapors. These mate- rials often are used in particulate sampling devices, as protectors or limiters for passive sampling devices, and in electrochemical sensors to keep the elec- trolyte in but allow exposure to a gaseous analyte. Nonporous Teflon mem- branes are used in oxygen sensors to separate oxygen from a variety of poten- tial interferents. Also, nonporous silicone membranes are used to separate relatively nonpolar organic compounds from water vapor and other polar species. A new development, called solid-supported liquid membranes, has been used to remove and concentrate radioactive waste selectively from water streams. This same approach has been suggested as a method for cleaning breathing air in gas-mask applications. The basic principle consists of a bundle of porous and hollow fibers. The pores of the fibers are filled with a nonvolatile (or nonsoluble) liquid such as a high-molecular-weight alcohol. This liquid acts as a conduit to pass an analyte from an outside sample to a small volume inside a sampler, thus acting as a concentrating device.

&4 ASSESSING HUMAN EXPOSURE Sequential Solvent Extraction Sequential solvent extraction, also referred to as liquid-liquid partitioning, is used to separate complex mixtures of chemicals into smaller, more tractable groups. The approach is applied most commonly to mixtures of organic chemicals, such as the organic constituents of airborne particles or fly ash (Pelli~ri et al., 1978; Chrisp and Fisher, 1980; Daisey et al., 1980~. A set of solvents is chosen with different solvent properties (e.g., polarity or acidity), and the mixture is extracted, in turn, With each of the solvents. An additional analytical procedure (e.g., GC, high-pressure liquid chromatography (HPLC), or TLC) usually is brought to bear on the resulting extracts. This approach is especially useful in the hands of an imaginative and knowledgeable analyst. Many solvents and solvent properties can be used in innumerable combinations to separate compounds with different chemical and chvsical DroDerties. The Durified extracts reduce the sele~tivirv and .cn~ifi~i~v , ~ ~ ~ r ~ r ~ . ~ . . . _ restraints on further analytic techniques. Consequently, solvent separations can Neatly simplify subsequent analyses; in some cases, they are required. Furthermore, knowledge of the chemical classes present in each of the solvent extract fractions can be inferred from the chemical and physical properties of the solvent responsible for that fraction. This approach has several drawbacks It often requires a large amount of time and is difficult to automate. For each solvent used in the process, there is a separate extraction step. Each of these steps can be quite lengthy, espe- cially if Soxhlet extraction apparatus are used. Typically, large solvent vol- umes (>50 mL) are required for each extraction, requiring time-consuming solvent reduction steps prior to analysis. Solvent artifacts can also be a seri- ous problem, and both the acquisition of high purity solvent and their subse- quent disposal can be costly. During transfers and solvent reduction steps, material can be lost through volatilization or adsorption on the walls of glass- ware. Sequential solvent extraction does not Held clean cut separations of pollutants; spillover frequently occurs among the fractions. Finally, there is the potential for reaction between the extracting solvent and the pollutants of interest. Despite these disadvantages, sequential solvent extraction with new solvent combinations is a promising tool for separation and purification of complex mixtures. Supercntical Fluid Extraction Supercritical fluids can be used to desorb trapped airborne contaminants from various absorbents. VOCs and semivolatile organic compounds

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 85 (SVOCs) are often concentrated before analysis by trapping onto some type of adsorbent, typically charcoal or Tenax. These compounds then are stripped from the adsorbent using either solvent or heat. However, solvent desorption tends to dilute trapped analyses, and thermal desorption can decompose ther- mally labile compounds. Supercritical fluids can be used effectively to desorb analyses from these absorbents (Raymer ~ al., 1987; Wright et al., 1987; Haw- thorne et al., 1989a,b). When CO2 is used as the desorbing solvent, the com- pressed CO2 can be Repressurized on the head of a capillary GC column, where it deposits any analyses extracted from the adsorbent (Hawthorne and Miller' 1987~. During Repressurization, cooling also occurs, which further serves to retain even many Gentile compounds on the head of the capillary column. Thus, the extraction and concentration of analyses can occur ~ a single step. This technique could be very useful in the analysis of passive samplers. If these devices are to be used for long-term exposure monitoring (>24 hours), then highly retentive absorbents must be used (Mulik and Wil- liams, 1986) to minimize any reversible losses from the sampler (Hourani and Underhill, 1988~. Some common fluids have low critical temperatures (e.g., CO2, 31.3°C), and supercritical fluid extraction (SFE) at mild temperature could be used as an alternative to thermal Resorption for directly desorbing analyses from passive samplers into a GC column. This SFE-GC method avoids high thermal Resorption temperatures that can cause thermal degrada- tion of heat-sensitive compounds and generate sorbent-derived artifacts. SFE- GC can be configured to deposit all or a large portion of the analyses con- tamed on the sorbent into the GO column (Hawthorne et al., 1989a,b). This method can dramatically increase method sensitivity. SFE also allows for the Resorption of higher molecular weight (low volatility) compounds not amen- able to thermal Resorption. Since some fluids are also gases at ambient tem- peratures, off-line SFE allows for rapid concentration of the analyses by sim- ple venting of the gaseous extraction solvent. Another distinct feature of extractions using supercritical fluids is that the solvent strength is controlled by adjusting the pressure (density) of the fluid Therefore, if the adsorbent is first desorbed with low-pressure CO2 only the relatively nonpolar or lower molecular weight compounds are obtained. More polar and higher molecular weight compounds are then desorbed by increas- ing the pressure, adding a modifier to the fluid, or by changing to a different fluid (Hawthorne and Miller, 1987; King, 1989~. Thus, the Resorption process can also become a selective extraction process. The use of supercritical fluids for extraction and chromatography deserves greater research attention in the assessment of airborne contaminants.

86 ASSESSING HUMAN EXPOSURE Lipid Chromatography for Sample Preparation LC can be used strictly as a separation device for the fractionation of col- lected pollutants into some type of class groups before actual analysis. An example of this approach is the work of Lamparski and Nestrick (1989) during which stack gas effluents were analyzed for the presence of trace levels of polychlorinated dibenzo-p-dio~nns (PCDDs) and dibenzofurans (PCDEs). Samples were collected using a modification of EPA's Reference Method 5 sample collection train. The entire sample ire the train was extracted With benzene, and the crude extract was subjected to classical liquid column chrom- atography (LC on silica gel and silica gel treated with various reagents and alumina). These simple LC steps remove many interfering classes of com- pounds including pesticides and PCBs. The extracts were then subjected to two HPLC separations, which further segregate the PCDDs and PCDFs into chlorine classes. The extracts were then fractionated on a silica HPLC col- umn. The fractions were collected and further separated on a reverse-phase HPLC column and then analyzed for PCDDs and PCDFs by combined GC mass spectrometry (GC-MS), a highly selective and sensitive technique dis- cussed in a subsequent section. This method of sample preparation is ex- tremely thorough and shows the degree of sample preparation steps that may be necessary to analyze airborne contaminants. It also illustrates how simple LC and the more elaborate HPLC can be used for the preparation of samples for subsequent analysis. Detection In detection, an analyte typically is converted into some measurable signal that is indicative of the chemical identity or the amount of analyte that is present. Some detection devices might include some selectivity (not so much separation as discrimination), so overlap of detection and separation occurs. Most detectors can be used in a stand-alone mode, where an air sample is fed directly into a detector, and an analog signal is measured and related to a given concentration of analyte. This offers the advantage of being continuous but often at the expense of selectivity, particularly for nonspecific detectors, which respond to ~ broad range of compounds. Increases in selectivity can be achieved using more specific detectors, which discriminate against many com- pounds and only respond to a narrow range of compound classes or functional groups. These same detectors can be used in combination with some type of chro- matographic separation. The continuous nature of the analysis is lost, but the

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 87 specificity of the analysis increases. Selected detection devices are discussed in the following sections. Chromatography Defection Devices Nonspecific Detectors Foremost among the nonspecific (or universal) detectors is the flame-ioni- zation detector (FID) (S-CYCik, 1976~. This detector responds to any com- pound containing carbon-carbon or carbon-hydrogen bonds. It Is serf to inorganic species, such as N. O. CO2 or water, which makes it ideal to analyze trace-level organics in ambient air. It has excellent sensitivity and a dynamic range of 5-6 orders of magnitude. However, the near-universal detection capability of the FID can lead to interference problems, especially for complex matrices. The photoionization detector (PID) is another very popular detector (Dris- coll, 1977~. PID ionizes molecules via the absorption of a photon. The ener- gy of this photon must be greater than the ionization potential of the analyte. Various lamps can be used that generate photons with different energies, thus allowing some selectivity to be achieved, if desired. PID differs from FID In that FID combusts all of the eluant in the ionization process, while the PID leaves most of the molecules intact. PID is considered nondestructive and can be connected in series with other detectors. Although PID is not as universal as FID in its response to organic compounds (which depends on the particular lamp), its simple design makes it advantageous for field analyses. PID in combination with GC is particularly portable because it does not require spe- cial gases for its operation, as do many other detectors including the FID; PID con also use air as the mobile phase, greatly simplifying its operation. Selective Ionization Detectors Highly selective detectors include electron-capture detection (ECD) (Sev- cik, 1976), which frequently is used for selective analysis of halogenated com- pounds. However, a wide variety of nonhalogeIlated compounds also give an ECD response; in particular, oxygen interferes with ECD operation and thus limits its usefulness for direct analysis of air samples (Simmonds et al., 1976~. The thermionic ionization detector (TID) is an adaptation of FID. TID has high selectivity for nitrogen or phosphorous containing compounds, but has little response to other organics (Farwell et al., 1981~. An example of the

&; ASSESSING HUMAN EXPOSURE usefulness of this detector is for the selective detection of HCN or acryloni- trile in an air sample containing high quantities of other organics. Another TID application involves the derivatization of a nonnitrogen-containing analyte with a nitrogen-containing agent. The nitrogen hall respond to TID, but any Interfering compounds that are not derivatized uphill not respond to TID unless they originally contained nitrogen. This provides enhanced sensitivity and se- lectivity for the desired analyte. For example, acrolein can be collected using a tube containing a sorbent coated with 2-(hydro~ymethyl~piperidine. The acrolein reacts to form oxazolidine, which is desorbed from the tube and ana- lyzed by GC-TID (Eller, 1984b). Flame photometric detectors (FPDs) operate by combusting a sample and then monitoring for a specific wavelength of light that is emitted (Sevcik, 1976~. FED can be operated in a mode in which it is sensitive for phospho- rous-containing compounds, but is most widely used in air- monitoring studies to detect sulfur-containing compounds. Various other detectors based on chemiluminescence techniques have been used with GC for selective detection of nitrosamines and other nitrogen- (Britten, 1989) and sulfur-containing (Benner and Stedman, 1989) compounds. Thermal Conduction Devices The teal conductivity detector (TCD) is another universal detector (Sevcik, 1976~. This detector responds to all compounds with thermal conduc- tivity In the gas phase different from the gas (usually the mobile phase in GC) used to bring samples into the detector. It is extremely useful to analyze ambient gases that do not give a response to FID. However, because TCD responds to everything, background contamination is a problem that limits TCD's usefulness in trace analyses. Organ oleptic Methods An organoleptic or odor-appraisal method involves some type of separation of volatile compounds (typically by GC) followed by odor appraisal of the eluting compounds by humans (McGorrin et al., 1987~. This type of analysis can be very important: in that odor problems are a typical complaint in air- contaminant exposures. The previously mentioned detectors determine the nature and amount of a wide variety of compounds, but they cannot point out which compounds contribute to odors. This is particularly true if certain com- pounds have a low odor threshold and can be masked by the presence of other compounds.

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 89 Mass Spectrometry MS is one of the most widely used detection methods for organic com- pounds. The mass spectrometer is a tunable detector that can be set to moni- tor species characteristics of any analyte that can pass through a gas chro- matograph. (MS =n also ~ used to detect compounders that cannot pass through a gas chromatograph.) The species that are monitored are molecular ions or molecular ion fragments that can be produced by a variety of tech- niques, including electron impact ionization, positive and negative chemical ionization, and atmospheric pressure ionization (Watson, 1985~. Each tech- n~que has p-arti~r arivantages and disadvantages that depend on the analyte in question. The particular ion being monitored is selected by a mass-selec- tion process. This selection can be carried out in a nondiscriminating manner (low resolution), in which ions with unit mass difference are just resolved (e.g., ions with mass 28 are just resolved from ions with mass 29) or with high discrimination (high resolution) in which ions differing by just millimass units can be resolved (i.e., an ion with mass 28.00()0 is resolved from an ion with mass 28.0028~. The higher the resolving power, the higher the cost and so- phistication of the instrumentation. The high degree of sensitivity of this technique comes from two sources. First, because ions are formed and mass selected, ion-detection techniques can be used that are inherently very sensitive. Second, the high discriminating power of MS (i.e., rejecting everything but the mass of interest) removes a great deal of the interfering background signals and thus has good signal-to- noise ratio and sensitivity. Thus, the mass spectrometer can be used as a stand alone detection device for air monitoring. In this regard it has the dis- tinct advantage of giving instantaneous results. However, in many cases multi- ple compounds will give similar responses which limit the spectrometer's stand alone capability. Combining the power of MS with the separation capability of a chromatograph, especially GC, helps to eliminate many of these interfer- ences but with the loss of instantaneous results. GC-MS is particularly well suited to the analyses of airborne contaminants, but is limited by requiring that the analyses be thermally stable and volatile under GC conditions. This restricts the usefulness of GC-MS when analyzing metals, metal salts, high molecular-weight compounds and compounds suscep- tible to thermal decomposition. Since most airborne pollutants are thermally stable and volatile under GC conditions, GC-MS is a useful analytical tech- nique. Organic compounds sorbed onto particles can be analyzed, provided they can be desorbed (using either heat or solvent) and passed through a gas chromatograph intact (Weschler and Fong, 1986~. In some instances, a reac- tive molecule can be stabilized by derivatizing it and making it susceptible to .

90 ASSESSING HUMAN EXPOSURE GC-MS analysis. GC-MS can also be useful in analyzing biological marker compounds and metabolites to determine human exposure to airborne con- taminants (see Chapter 4~. Again, the criterion that the compounds be ther- mally stable in the gas phase must be met, and for many compounds (particu- larly me~bolites of the desired analyte), derivatization may be necessary. A particularly useful feature of GC-MS is the ability to use stable isotope- labeled isomers as surrogates in the analyses. Such surrogates are identical to the analyte except that they contain one or more isotonically labeled atoms. For example, hexadeuterobenzene (fully deuterated benzene) can be used as a surrogate for benzene. The mass spectrometer is capable of distinguishing these surrogates from the original analyte, because the surrogates differ in mass. The surrogates are particularly useful for accurate quantitation (Lewin et al., 1987~. If an analyte is being trapped on a sorbent such as Tenax, then the surrogate can be placed on the adsorbent before actively pulling air through the adsorbent (i.e., a field spike). Any losses that occur in the analyte breaking through the adsorbent or losses that occur in incomplete Resorption of the analyses from the adsorbent will be matched by the surrogate. In this fashion, an excellent degree of quality control can be achieved with virtually every sample that is taken. The usefulness of surrogates in the analysis of volatile compounds has been recognized by EPA in the use of surrogates in Method 1624, which is used in the analysis of water samples for purgeable volatiles. MS also offers another unique feature in environmental analyses through its ability to discriminate various isotopes of an element. In many cases, par- ticular isotopic ratios vary slightly, depending on the original source of the element or the environmental processes it has undergone. An example of this approach is the study of lead in California children, where the variability in the lead isotopic ratios from various products were used as a discriminator of the children's original lead source (Yaffe et al., 1983~. A similar discrimina- tion can be made in the various sources of sulfur oxide based on the oxygen isotope ratios (van Everdingen and Krouse, 19~. Thus, MS can be invalu- able not only in determining the existence of exposure to an airborne contami- nant but also in providing information as to its source. GC-MS has several disadvantages. First, most units are not very portable, and, in most cases, samples must be brought to the instrument rather than taking it to the sampling location. Newer, more compact units are being made and can readily be transported in specially built vans, but portability into indoor facilities often is limited. A GC-MS system was sent on the space craft to Mars. Designs are under way to have a mass spectrometer (and possibly a GC-MS system) compact enough that a soldier could carry it in a backpack to serve as a monitoring device for chemical warfare agents. Such a device

. SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 91 could be ideal for personal monitoring studies, providing the unit was light- weight and simple to use. Another disadvantage of MS in general and GC-MS in particular is that the systems are expensive. Samples from multiple sites usually are brought to an instrument rather than having units at multiple locations. Most G C-MS units are readily automated, which makes them capable of analyzing around the clock, which in turn allows many samples to be analyzed from numerous sites. Such automation includes the analysis of solvent desorbed traps, thermally desorbed traps, and most recently, the automated analysis of Summa syringes to determine highly volatile compounds that do not lend themselves to trap- ping. Summa syringes are samplers made of high passivated stainless ~eel, which renders them inert to many pollutants (Krasnec, 1986~. A sampling system with those syringes can be programmed through a microprocessor to collect samples at designated times and to dispel the collected samples into a GC-MS unit for analysis. Automated data reduction programs are also available, thus maximizing the output of the instrumentation and minimizing the cost per sample. The future for G C-MS lies in the simplification of the mass spectrometer so that it is easy to operate and is possibly controlled by a microprocessor- based "expert system." Future units should be compact, inexpensive, and easy to use. Many of these features are incorporated into a new type of mass spec- trometer, the ion trap mass spectrometer. This new and innovative device might resolve many of the disadvantages of present MS units. The ion trap detector is discussed in the next section.) It is also worth noting that simple benchtop" GC-MS systems are available for less than $55,000. Mass Spectronleby-Mass Spectrometry The technique of MS-MS shares many disadvantages of GC-MS. It is very expensive and not very portable. It also shares many of the advantages of GC-MS, such as good selectivity and sensitivity and the ability to use surro- gates. Good selectivity comes from the actual MS-MS process. A compound enters the mass spectrometer without passing through a gas chromatograph. The sample is then ionized, and particular masses selected that are character- istic of the analyses of interest. The selected ions undergo a decomposition process in which they are fragmented, and smaller ions are formed from the original ions. These decomposition ions are then analyzed by a second mass spectrometer, and fragment ions characteristic of the parent analyte are moni- tored. A two-step separation process is achieved, much like GC-MS, in which the first separation occurs on the gas chromatograph while the second occurs

92 ASSESSING HUMAN EXPOSURE with the mass spectrometer. In MS-MS, the process occurs in two successive mass spectrometers. The advantage of the MS-MS process is that the separa- tion processes occur at very high speeds (milliseconds or less). Thus, continu- ous monitoring is possible while retaining a high degree of specificity and sensitivity. This is not possible with a GC-MS system, because the continuous nature of the analyses are limited by the time required by GC for the separa- tion. However, it must be emphasized that MS-MS often is incapable of dis- criminating between simple compounds such as o-xylene and ethylbenzene, while such a separation by GC is straightforward (Sushan et al., 1987~. The GC process and the MS-MS analysis can also be coupled to give GC- &IS-MS. This approach combines the advantages Of both techniques, particu- larly where a high degree of resolving power is necessary to discriminate the desired analyte from other interfering compounds. Likewise, MS-MS analyses have also been combined with HPLC and SFC. A particularly useful feature of these combinations is their inherent flexibility. The separation processes can be juggled between the chromatograph and the MS-MS system to mini- mize both the analysis times and any interferents in the analyses. The greatest amount of experience with MS-MS for atmospheric analyses comes from the Sciex TAGA 600OE (Sushan et al., 1987~. This unit is made transportable by placing it in a van that is usually brought near a sampling site (e.g., a home). The sample (e.g., air from the interior of the home) is brought to the instru- ment via a long transfer line. This unit requires experienced operators to ensure the quality of the data. The ion trap detector (ITD) is a new type of MS system. The ITD is small and of simple construction and requires only a modest vacuum (Louris et al., 1987; Busch et al., 1988; SCIEX, 1989~. It contains a trapping cell consisting of two hyperbolic end electrodes and a single hyperbolic central electrode. The ionization and mass selection occur in this cell. The ITD can selectively trap and hold a specific ion, while rejecting other nondesired ions. In other words, the ITD serves as a trapping device collecting a specific ion (com- pound), while selectively rejecting others. This trapping process is limited only by space-charging occurring in the cell. The ITD can be operated in the MS-MS mode (Stafford et al., 1984~. In this mode, a single mass (m/z, mass/ charge) ion is selected and all others rejected from the ITD. A collision gas is then introduced, which causes the fragmentation of the selected ion. This fragmentation spectrum can then be examined to discern the structure of the original ion as in other MS-MS procedures. If desired, this process can be repeated on one of the fragment ions to allow MS-MS-MS procedures to be carried out as many times as necessary. The ITD has also been interfaced to an atmospheric-pressure ion source in similar fashion to that used on the TAGA MS-MS system (Asano et al., 1988~. The successful interfacing of this

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 93 particular source should make the ITD a valuable air sampling device deserv- ing greater research attention in the future. Ion Mobility Spectrometer The ion mobility spectrometer (IMS) first was known as a plasma chro- matograph (Karasek, 1974~. It operates on the principle of atmospheric-pres- sure ionization (the same principle as used on the TAGA MS-MS system) of any air contaminants, and the ions that are formed are analyzed by their drift time through a drift gas at atmospheric pressure. The technique-wa;s plagued by problems of ion clusters that varied with atmospheric conditions. A heated membrane device has been included that passes to the ion source only or- ganics that are permeable to the membrane (Carrico, 1986~. This has helped to minimize the clustering problem and has eliminated interferences from water, ammonia, and nitrous ondes. The IMS has the advantage of operating at atmospheric pressure, which eases the operational problems associated with mass spectrometers, which normally operate at reduced pressures. The IMS has shown sensitivities of less than 1 ppb and good response times (0.1 to 10 seconds), which would make it ideal for continuous analysis. However, it also has the disadvantage that the range of compounds that it can monitor is limit- ed by the necessity of the membrane inlet. Electrochemical Detectors Electrochemical detectors can be divided into three general types: potenti- ometric, amperometric, and conductimetric. Studies of human exposure to air pollutants have used electrochemical detectors as direct personal, microenvi- ronmental, and area monitors. Examples include the amperometric detectors for NO2 that were used in homes (Leslie et al., 1990) and the CO personal mowtor carried by people (Akland et al., 1985~. Potentiometric detectors such as pH and ion selective electrodes typically are used in the laboratory in supportive analytical procedures but also are available in portable kits for field use. Potentiomet~y Simple pH electrodes have been used widely in the laboratory since the introduction of the pH meter by Dr. Arnold Beckman in 1935. The typical

94 ASSESSING HUMAN EXPOSURE potentiometric detector for pH or selected ions operates by measuring a po- tential or voltage developed at the sample/solution interface. This interracial potential is ideally measured under conditions of zero current flow (infinite impedance) so that the potential reflects the Nernstian or thermodynamic potential of a specific chemical interaction at a membrane/solution interface. In this manner, the potential of the device can be made to reflect the concen- tration of the species of interest. In addition to pH, potentiometric probes have been developed for many different ions of interest (such as Na+, K+, Ca+ A. The attachment of enzymes and biological substrates to the sensing membrane has allowed potentiometry to be used to measure biological materi- aIs (Rechnitz, 1981; Arnold, 1983~. Miniature, inexpensive, microfabricated ion-sensitive devices have also been introduced (Lauks and Zemel, 1979), and combination of the field effect transistor (FET) and a micropotentiometric- sensor has facilitated the detection of gases and vapors in different media (Janata and Bezegh, 1988~. The most common, oldest, and best-understood ion-selective electrode is the glass pH electrode. The glass membrane can exchange protons with the solution with which it is in contact. On one side of the membrane is a solution with fixed proton activity (constant [H+] acid solution), and on the other side is the measurement solution. Each solution exchanges H+ with the glass to an extent determined by the chemical equilibrium. If the solutions are of differ- ent concentrations, a potential mill be developed, because the extent of proton exchange is different on each side of the membrane. This potential is detect- ed by having a reference electrode inside the membrane and a second elec- trode in the sample solution. The outside electrode must not be sensitive to changing tH+] so that only the membrane potential is observed. Selectivity is achieved because certain glasses will exchange only H+ and not other ions. Different glasses and materials that exchange different ions are substituted for the glass membrane to generate ion-selective electrodes and electrodes sensi- tive to a variety of analyses (Myerhoff et al., 1989~. Although potentiometric probes most frequently are used to determine various ionic/reactive analyses in solutions, the Sevringshaus electrode for gaseous CO2 is an example of the adaptation of the pH electrode for meas- urement of gases. The outside solution of the pH electrode is trapped by a gas-permeable membrane and held in place around the pH electrode. This external solution contains a buffer whose pH is altered by absorption of CO2. Thus, the more CO2, the growler the change in pH and the larger the signal from the pH electrode. The Sevringshaus electrode has been used for many years in the measurement of CO2 levels in blood. Similar potentiometric gas sensors exist that can be made to respond to ammonia and other gases in the air with reversibility. Selectivity generally is not great in such detectors, be

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 95 cause anything that affects the pH of the solution that surrounds the pH sen- sor will generate a response. Solid electrolytes have been used In high-temperature, potentiometric sen- sors for O2, SO2, and NOX, and some other species. Such oxygen sensors are used in stack gas monitoring and automobiles. The sensor response depends upon a specific reaction equilibrium and, therefore, can be extremely sensitive and selective. Recent work has contributed to the understanding of these de- vices and their microfabrication (Nicoloso et al., 1988~. Condu~imetry Conductimetry is used to detect gross ionic pollution in water (e.g., the YSI conductivity meter) and as a detector in chromatography (e.g., the Hall elec- trolytic conductivity detector-HECD) (Hall, 1974~. This detector operates on the principle of combusting the halogenated compound to form an HCl gas, which is then taken into a deionized, aqueous solution. The HC1 is a strong acid and dramatically increases the conductivity of the solution, which is de- tected by the solution passing through a conductance cell. The HECD also can be operated in a configuration in which it is selective for nitrogen-contain- ing compounds. It has the advantage of not being sensitive to oxygen, which makes it a useful detector for direct air analysis. Such devices could be used to detect and analyze for acidic and basic (or strongly ionic) materials in the field or laboratory Their use in exposure assessment is not widespread. Amperomet~y Amperometric defies have been used in exposure assessment studies as personal monitors and area monitors (Hartwell et al., 1984; Leslie et al., 1990~. The typical CO amperometric sensor operates by generating a current proportional to the amount of CO reacted (amperometrically) in the sensor. In such devices, the gas travels in a path that contacts a gas- porous mem- brane. On the inside of the porous membrane, the gas contacts an electrode and an electrolyte (the electrolyte cannot go through the membrane). The analyte dissolves in the electrolyte, migrates to the electrode surface, and reacts electrochemically, producing or using electrons. The electrons can be counted in an external electric circuit and the result displayed. The more ana- lyte present, the more current produced by the amperometric sensor; the current is proportional to the concentration. Such amperometric devices have been built in portable and fixed-site moni

96 ASSESSING HUMAN EXPOSURE tore for 02, CO, H2S, NO, NO2, hydrazines, HCN, NH3, formaldehyde, H2, EtOH, and many other compounds (Stetter et al., 1984; Janata and Bezegh, 1988~. The sensors have even been modified to be sensitive to hydrocarbons (Stetter et al., 1984), and the approach could be applied for many electro~nac- tive species. The galvanic (and electrolytic) sensors for oxygen measurement in the ambient air are well known in industrial hygiene and safety. The medi- cal application began with the Clark electrode (Clark et al., 1953) for 02, and modern microprobes have been built for measurement of the Or partial ores- sure in a single living human cell. -,~ r-I- r~~~ Many other amperometric techniques (e.g., anodic stripping voltammetry for metals and differential pulse polarography) are used in the laboratory for a variety of analyses but have not been applied in the field or used in exposure assessment to a significant degree. Electrochemical devices have the advantages of being extremely low-power and small size and hence, portable and reasonably inexpensive. These charac- teristics make electrochemical approaches ideal for field use and direct meas- urement of human exposure to airborne pollutants. The challenge for electro- chemical crevices is the attainment of adequate analytical performance in a practical field situation; combined sensitivity, selectivity, stability, lifetime, and ruggedness are needed. This combination has been demonstrated for several applications, including pH and CO, and could be used for many more applica- tions, such as NO2 in indoor air (Stetter et al., 1979; Chang and Stetter, 1990~. The selectivity of electrochemical detectors can be enhanced by using sensor arrays, selective filters, and GC techniques. The electrochemical gas sensor has been used successfully as a GC detector (Stetter et al., 1976; 1977; Blurton and Stetter, 1978) and can enhance portability, because no special carrier flame gases are required. Recent advances in microfabrication of structures and materials that can be used to construct amperometric (Buttner et al., 1990) and potentiometric detectors and the required circuitry (Turner et al., 1987) offer promise for lower cost, uniform, and exceptionally portable detec- tors. The combination of sensor arrays and chemometrics promises to in- crease the information content of existing systems (Zaromb and Stetter, 1984; Stetter et al., 1986~. Advanced electroanalytical methods that are effective in the laboratory, such as differential techniques (Borman, 1982), could be ap- plied to enhance the sensitivity, selectivity, and response time of electrochemi- cal field monitors. Spectroscopic Detectors Spectrophotometric methods of chemical analysis are among the oldest

SAMPLING AND PHYSICAL-CHEMIC~4L MEASUREMENTS 97 techniques that are used routinely to analyze airborne pollutants. A review of these procedures is presented in the third edition of the manual of methods adopted by the Intersociety Committee for a Manual on Methods of Air Sam- pling and Analysis (Lodge, 1989~. What follows are some recent advances that could extend the capabilities of selected spectrophotometric methods. Charge-coupled device (CCD) array detectors are an advance that promises to improve significantly the limits of detection in selected analytical areas (Chemical & Engineering News, 1989~. CCDs are arrays of p-doped silicon that are extremely sensitive to light falling on their surfaces. They respond over a wide spectral range (a quantum efficiency of at least 10% between 200 nm and 1000 nary), have ~ large dynamic range (1Os to 106), and are more rugged than conventional photomultiplier tubes. Furthermore, they are well suited to multichannel detection applications. In numerous analytical instru- ments (e.g., fluorometers, Raman spectrometers, emission spectrographs, and detectors for HPLC), CCDs are likely to replace photomultiplier tubes. The rectangular shape of currently available CCDs is a drawback in certain applications, because many spectrographs have astigmatic distortion. Howev- er, optics can and have been configured to use the CCD shape. In certain applications, the two-dimensional nature of the CCD has been an advantage. An example is TLC, where the plates are essentially two-dimensional. Entire TLC plates can be imaged onto a CCD detector, and the required response integration times can be less than a minute. CCDs are expensive (starting at approximately $30,000 for a functional system). However, as their applications expand and their production increas- es, the price is likely to drop significantly. Diode-array detectors use a one- or two-dimensional array of photodiodes to detect ultraviolet or visible radiation. Scanning time is very brief (approxi- mately 5 milliseconds) with such a device. Diode-array detectors have been used primarily to identify eluants of HPLC, although recently they have been applied to GC analyses (Kube et al., 1985~. A number of simple and inexpensive luminescence techniques have been developed recently and applied to the analyses of airborne pollutants. These include synchronous luminescence and room-temperature phosphorescence (RTP) to estimate the amount of polynuclear aromatic species in air particu- late extracts (Vo-Dinh et al., 1984a). RTP is part of a detection scheme that has been used in the development of a passive sampler to monitor personnel exposure to vapors of polyaromatic pollutants (Vo-Dinh, 1985~. The heavy- atom salts used as an adsorbent also serve as a direct sample medium (induc- er) for RTP detection. Surface-enhanced Raman spectrometry (SERS) is a technique with great sensitivity for compounds with a large Raman cross-section. The analyte also

98 ASSESSING HUMAN EXPOSURE must couple effectively with the substrate (typically silver, gold, or copper) used to achieve the effect. Within the past 5 years, SERS has been applied to the analyses of selected trace organic compounds (Vo-Dinh et al., 1984b) and chlorinated pesticides (Alak and Vo-Dinh, 1988~. Unfortunately, results obtained with SERS are difficult to quantify. This is due primarily to two factors: variability in electrostatic field effects and variability in chemical site effects (Moskovits, 1985). Infrared Detection The use of infrared (JR) detection with GC always has been limited severe- ly by low IR absorption coefficients, which results in poor sensitivity. Most IR techniques used to monitor for ambient levels of organics make up for this poor sensitivity by having long pathlength cells (typically 1 to 20 m). These types of cells are incompatible with GC detection, because the large volume of these cells produces such a large dead volume that all analyte resolution achieved during the chromatography is lost during detection. One IR tech- nique that has gained popularity in volatile organic analysis involves matrix isolation. The sample eluting from the GC column is frozen in an argon ma- trix onto a cryogenically cooled disc which revolves while the various analyses are elating. The eluted compounds are frozen in place so that they can be subsequently analyzed by a variety of techniques, including Fourier transform IR (Bourne et al., 1984~. These frozen chromatograms can be stored for several weeks without loss in resolution. Since the sample molecules are froz- en in place, extended signal averaging of the IR spectra is permitted, resulting in excellent sensitivity. Another advantage of the matrix isolation is that the spectra are very sharp, resulting from the elimination of band broadening due to molecular rotations and intermolecular interactions. GC-matrix-isolation/FTIR is being used for the qualitative analysis of air- borne contaminants (Childers et al., 1987~. The technique complements quali- tative GC-MS analyses by providing additional structural information about matrix-isolated compounds. The matrix-isolation interface is expensive, which limits its usefulness in ambient air analysis to specialized qualitative analyses. The actual collection of spectra can be very time consuming, and therefore, the technique should not be considered when massive numbers of samples are to be analyzed or when the real goal of the analysis is to quantify the various analyses.

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 99 M~crosensors For the purpose of this discussion, "microsensors. are defined as tiny detectors typically manufactured by microfabrication processes; included in this definition are small detectors created using wire electrodes and the like. The major types of mi=osensors are summarized In Table 3.4. Most of these kinds of sensors have been fabricated using microelectronic process technology. The commercial potential of these devices depends upon their characteristics and their state of development. A passive microsensor often will operate in a diffusion-limited mode' be- cause the analyte of interest will be at a very low partial pressure (few ppb). Even at 10~ torr (about 1 ppbv), a clean surface will be contaminated in about a second. Thus, the response time of a microsensor might be limited at very low concentrations. However, for human exposure monitoring, a response time of a few seconds almost always will be acceptable. The response magni- tude will be limited by the size of the effect produced in the microsensor. The effect can be a chemical or physical change, and this change can be moni- tored by spectroscopic, thermal, electrochemical, dielectric, electronic, or other variable properties of the microsensor. Chemiresistors are simple devices and have been studied and used for many years. The chemiresistor interacts with a pollutant and changes conduc- tivity (or more correctly, impedance). The amount of change of resistance is a measure of the gas/vapor analyte concentration. These devices typically are not very selective but can be very inexpensive and are used around the world (e.g., Figaro detectors) for detection of hydrocarbons, combustible gases, CO, and alcohol. When used in an array with the application of chemometrics, selectivity can be increased. The hot wire/catalytic bead detector uses a plati- num resistance thermometer to detect explosive levels of combustible gases. It can be a very portable device but, in its present form, is not sufficiently sensitive for ambient measurements of hydrocarbons. The FET (field effect transistor) and other devices, such as capacitors and Shottky diodes, have been used as the basis for chemical detection (NRC, 1984~. A very sensitive and selective H2 sensor can be made from an FET with a Ed gate, and preparation with an ultra-thin film gate can lead to reac- tivity for other gaseous species. Capacitive sensors for humidity are also pos- sible using a device called the charge flow transistor. The SAW (surface acoustic wave elements) (Wohltjen, 1984), lambda wave, and the piezoelectric balance are all similar devices. The SAW is smaller and more sensitive than a simple piezoelectric balance, and the lambda wave de- vice can be used at low frequencies. In a typical gas-sensitive device, the sur- face of the device is coated with a sorbent layer that reacts with the analyte

100 ASSESSING HUMAN EXPOSURE TABLE 3.4 Mi~osensors Potentially Applicable to Airborne Contaminants Biosensors Electrochemical sensors Potentiometer Amperometnc devices Contact potential sensitive elements I hermal sensory Thermistor and resistance thermometer elements Thennoelectric/bolometric sensors Semiconductor-based elements Piezoelectric oscillator thermal sensitivity Pyroelectric sensors Black body measurements Stress and pressure sensors Photoacoustic effect Mass sensitive elements Bulk piezoelectric elements (thickness monitors) Surface acoustic wave elements (SAW) Plate mode oscillators Interface impedance Elastic constant sensitive fiber optic elements Electromagnetic properties sensors: passive Solid state conductivity measurements (chemiresistance) Dielectromet~y Dielectric proper",' measurements Absorptivity Index of refraction Phase shift and interface impedance (e.g., ellipsomet~y) Spectral "fingerpnnt" Surface-enhanced Raman spectromet~y Electromagnetic properties sensors: active Nonlinear behavior, including frequenter doubling Fluorescence Source: Developed from a presentation made by J. Zemel at the "Workshop on Advances in Assessing Human Exposure to Airborne Pollutants," Yale Universi- ty, October 19-20, 1988.

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 101 gas of interest. Weight changes in the adsorbent layer are detected by the piezoelectric substrate. Sensitivity depends upon the weight of the film. Thicker films and strongly interacting compounds are most readily detected; however, thicker layers have slower response times. SAW devices and piezo- electric balance sensors for gas detection me commercially available for some analyses, and detection is typically In the ppm range, but ppb detection can be achieved for selected analyses. Fiber optic sensors have some application in light-pipe devices for PNAs and for specific analyses, such as O2 and ammonia in air. ~ addition to chemical sensing, advances ~ microengineering Include me- cllanical devices such as pumps, balances, gears, wheels, springs, valves, mo- tors, and similar mechanical structures that are micron-dimension. The ap- plication to small field samplers and exposure assessment in the future seems appropriate, but these new technologies have not been applied toward use in exposure measurements. The possibility for microsampling and microanalyt- ical devices exists, and their potential application to exposure assessment could yield some exciting results. Microdetectors, including MS, PID, FID, and TCD, could be commercially available. Several microsensors have been reported (NRC, 1984~. Some devices are available; others are in various stages of development and commercialization. The physical microsensors, e.g., for pressure, temperature, and flow, are better developed than the chemical sensors. Microsensors for detection of certain physical properties are readily available. However, chemical microsensors need further development to make a substantial impact on applied science. For example, the microthermal conductivity detector for GC applications has been developed, but no comparable micro-FID or micro-FPD exists. Also, microelectrochemical sensors have been reported but are just now being intro- duced commercially. Electron Microscopy Electron microscopy has been an important tool in the examination of submicron features since it became commercially available in the 1960s. How- ever, in the analysis of ambient particle samples, it could only serve as a quali- tative or, at best, semiquantitative tool because of the time necessary to find particles manually, focus the microscope, and obtain image and x-ray spectral data for a given particle. During the past decade, an alternative approach that has shown the ability to provide quantitative microscopic examination of par- ticle samples for source identification and apportionment is computer-con- trolled scanning electron microscopy (CCSEM). The CCSEM system has

102 ASSESSING HUMAN EXPOSURE been described in detail by Casuccio et al. (1983) and Hopke (1985~. The system automatically steps across a field of particles using the backscattered electron intensity to determine when it has struck a particle. When a particle is detected, the control system decreases the step size and determines the geometric center of the particle by constructing diagonals across the particle. The diagonals are then reconstructed to converge at the particle centroid. The electron beam is focused at the centroid, and the fluoresced x-ray spec- trum is obtained with a windowless Si(Li) detector that permits detection of light element x-rays. With the high vacuum in an electron microscope and new ultrathin vacuum windowed detectors, it is possible to observe the x-rays from carbon to uranium. In this manner, CCSEM can perform elemental analysis from carbon through uranium on extremely small quantities of matter. However, the cross-sections for vacancy production are low for electrons, and the total excited volume is small. It is generally only possible to observe ele- ments present at a few tenths of atom percent levels or higher. Trace ele- ments generally are not observable. It must be noted also that the presence of a high vacuum in the microscope causes particles formed of moderate to high vapor pressure materials, including NH4NO3, to evaporate, and thus be unobservable (Leon" et al., 1983~. From the pattern of observed x-rays, a density can be assigned. From the physical size and density, an aerodynamic diameter can be calculated. Thus, the CCSEM provides size, shape, and elemental composition data for each particle. Because of the low-fluorescence yields for the light elements, the x-ray spectra typically are accumulated for 15 seconds each. The use of CCSEM can greatly increase the information available on the physical and chemical characteristics of ambient or collected source-emitted particles. A recent improvement in the computer control of the microscope focusing now permits individual particle images to be captured at 256 by 256 pixels per image In 256 gray levels to accompany the elemental composition data. This secondary electron image provides a three-dimensional image showing the particle morphology and surface texture. Thus, information on the size, shape, and composition are directly available from the microscopic examination. These properties then can be used to group the particles into different classes (Kim and Hopke, 1988a). The mass fractions associated with each of these classes become the variables on which the source apportionment will be made. A particle-class balance assumes a linearly additive sum of class-mass fractions times a fractional contribution of aerosol mass by the particulate sources analogous to the chemical mass balance used for bulk sample com- position data. The ability to assign particles unequivocally to well-defined particle classes will provide a better set of variables for the apportionment

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 103 study, because these better-defined classes will have more specificity, and thus less noise in the mass-balance-fitting process (Kim and Hopke, 1988b). The image data have not been fully exploited in receptor modeling because of the lack of quantitative methods to interpret them in an automated process. Hopke et al. (1988) suggested the use of fractal dimensions calculated from these single particle images as an approach to characterize the particle texture. Their preliminary results were encouraging, but clearly showed the need for further study. Instrumental Neutron Activation Analysis (INAA3 The basic concept of instrumental neutron activation analysis (INAA) is that a small fraction of the stable nuclei present in a sample becomes radioac- tive when bombarded with neutrons in a reactor. Neutrons can penetrate into the nucleus because of the absence of any repulsive forces; capture probabil- ities are relatively large. Thus, for most elements, stable nuclei can be cor- rected to ones that are radioactive in a reproducible manner. Following many of the subsequent beta-decay processes that occur in the induced radioactivity, one or more high-energy photons or gamma rays are emitted. A high-resolution semiconductor detector interacts with a gamma ray and yields an electronic pulse, the maximum voltage of the pulse being proportional to the gamma-ray energy. This pulse is amplified, shaped, and then sorted by a pulse-height analyzer so that gamma rays of different ener- gies result in counts in different locations in a computer memory. The energy of a gamma ray is unique to a particular isotope of a specific element so that a qualitative analysis can be made by observing which peaks are found in the spectrum of gamma-ray energies emitted by the activated samples. A quanti- tative analysis can be made by relating the number of gamma rays emitted by the sample relative to a standard containing a known amount of that element. The standard is irradiated either simultaneously with the sample or with flux monitors (typically thin wires containing elements that are easily activated to gamma emitting radionuclides) that are used to measure the neutron flux to which the sample and the standard have been exposed. The gamma-ray spec- tra from a series of samples and standards are recorded. Appropriate com- puter codes can then reduce the gamma-ray counting data to elemental con- centrations. Samples can include biological materials (plants, proteins, enzymes, insect blood, human blood, and hair), geological materials (soils and ores), and en- vironmental samples (water, airborne particles, street dust, and fly ash). The analysis does not require extensive sample preparation and chemical marlipu

104 ASSESSING HUMAN EXPOSURE rations. If the sample can be packaged in a container that permits irradiation in the reactor and subsequent gamma-ray counting, it can be analyzed by INAA. Although there is some low-level residual radioactivity associated with the samples, the technique is relatively nondestructive of the sample. Neutron activation analysis can provide information on as many as 40 elements with sensitivities ranging from picograms to micrograms of analyte element. There are elements, such as lead, that do not produce gamma-ray emitting products. These elements cannot be analyzed by INAA. Neutron activation is a well- established method that is now widely used to characterize environmental trace elements (De Soete et al., 1972~. X-ray fluorescence (XRF) is another common multielement method for aerosol samples. A good discussion of XRF applied to the analyses of air- borne particles is presented in Malissa and Robinson (1978~. The resolution of XRF is such that particles <100 am diameter are effectively analyzed as if they were homogeneous. Radon and Radon Progeny Measurements Radon is a naturally occurring gaseous radioactive element that is found ubiquitously throughout the environment and is typically found in higher con- centrations in indoor atmospheres than in outdoor air. It decays to a series of four short-lived decay products that can be deposited in the human respira- tory tract either directly or after attaching to pre-existing ambient particles. EPA reported in a September 1988 press conference that the decay of these radioactive progeny can induce lung cancers and might be responsible for 20,000 lung cancer deaths per year in the United States. Measurements can be made of radon and decay products, and methods for each of these meas- urements are presented. Radon Radon can be measured directly at environmental levels using the ability of the emitted alpha particle to excite a ZnS(Ag) scintillator to produce meas- urable emitted light. Hemispherical (Lucas, 1957) and simple right-circular cylindrical (George et al., 1976) detector cells with optically clear, flat win- dows, and interior walls coated with ZnS(Ag) have been used. For simple grab measurements, the cell is evacuated. A valve into the chamber is opened, and the chamber is filled with ambient air. The cell is then placed on a photomultiplier tube, and the count rate of light pulses is measured.

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 105 With proper calibration, this count rate can be related to the ambient radon concentration. However, the concentration of radon In a building is highly variable. Thus, grab samples of radon rarely reflect the long-term, average indoor concentration. An alternative approach is an active system to pull air through the scintilla ration chamber using inlet and outlet connectors. The counts in a given tune (e.g., 15 minutes) can be converted to an approximate radon concentration. Because of the accumulation of decay products in the chamber, the actual radon concentration is related to the activity counted through complex cal- culational methods. As in all active systems, pump failure is possible and careful flow control is needed. Therefore, this system is not convenient for long-term monitoring. Two methods are available to obtain an integrated measure of radon con- centration. For short intervals (2-7 days), canisters containing activated car- bon can be set in a room. The radon is adsorbed on the carbon and ac- c~nulates over time. After the sampling period is over, the canister is sealed, and the decay products build up in the container. Because equal activities of the shorter-lived decay products will develop after about 4 hours, a measure- ment of the emitted gamma radiations from the progeny can be used to deter- mine the radon concentrations. However, the carbon canisters, dependent on diffusion for eventual contact between the radon and the sorbent, are useful only for a limited period and can have difficulties because of water-vapor adsorption reducing the amount of adsorbed radon. A better, long-term, integrated measurement can be made using a track- etch detector. In these detectors, a small piece of a special type of plastic is placed In a small plastic cup and sealed with a moisture-proof plastic film. The radon can diffuse through the plastic film. Its decay and those of its decay products will cause the plastic detector to be bombarded with alpha particles. The alpha particles produce a track of radiation damage in the plastic that can be more easily dissolved than the bulk material. Thus, when the plastic detector is etched in a mild alkaline solution, microscopic holes can be seen where the number of holes per unit area can be related to the radon concentration. These detectors can be used for several weeks to as long as one year and thereby provide more accurate annual average values for the radon exposure. A recent development is the use of an electret for passive radon measure- ments (Kotrappa et al., 1988~. The concept is similar to the track-etch detec- tor in that the detector is placed in a cup so that the radon can enter it. However, the measure of radon exposure is the decrease in surface electric charge on an electret made of Teflon FOP. Such a system has a simple read- out device that only has to measure the residual surface charge on the electret

106 ASSESSING HUMAN EXPOSURE and provides a useful alternative measurement method for periods of several weeks to several months. Radon-Decay Products The measurement of the radon-decay products is difficult, because the concentration and size distribution must be determined for the decay products. The decay-product behavior is quite complex because of the ability to attach to airborne particles, as well as to indoor surfaces such as walls, ceilings, and furnishings. The particle size determines the ability of the radioactivity to be deposited in the room and In the respiratory system. The respiratory deposi- tion provides the dose to the critical tissues; the sizes of most concern are those less than 10 nm. These highly diffusive particles can most easily deposit in the body, whereas only about 20% of the particles with diameters around 100 nm are retained in the respiratory tract (James, 1988~. Measurement of the total concentration of alpha-emitting particulate matter is relatively easily determined. Air is drawn though a membrane filter, and the collected alpha activity can be measured using either a ZnS(Ag) scintil- lator for total alpha counting or a solid-state detector that provides alpha- spectroscopic determination of the specific decay products. From multiple sequential counts and the known decay kinetics of the radon progeny, the concentrations of each of the decay products (hippo, 2~4Pb, rabbi) can be cal- culated. The proper choice of filter can provide quantitative collection of activity and an adequate radioactive source for accurate measurement of the collected activity. Size measurement of the decay products is a difficult problem. George (1972) proposed a method for measuring the "unattached" fraction ~ which air is drawn through a filter that is covered with a 60-mesh screen. The highly diffusive activity attaches to the screen. The activity counted on the screen is the unattached activity, and that which passes through the screen to the filter is the attached activity. Activity-size distributions have been measured with diffusion batteries down to sizes of 10 to 15 rim (Knutson, 1988), because conventional diffusion batteries have relatively little resolving power for ultra- fine particle sizes. Ramamurthi and Hopke (1988) reviewed several measure- ment systems for unattached fractions in the context of the improved under- standing of penetration of particles through single screen. It is now understood that the unattached activity is not a single species with one fixed diffusion coefficient, but rather an ultrafine mode in the particle size spectrum between 0.5 and 5 nm. Improvements in measurement methods (Reineking and Porstendorfer, 1986; Holub and Knutson, 1987) have made

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 107 it possible to determine the full range of sizes for grab samples. There re- mains a need to develop a system that will provide a series of size and con- centration measurements. Radon concentrations are determined as a measure of exposure rather than decay-product concentrations' because integrated radon measurements are relatively simple, inexpensive, and accurate, and dosimetric models suggest that radon is a reasonable surrogate for the decay products. However, the development of new methods to measure decay product size and con- centrations would permit direct long-term measurement of the species that are Me proumate cause of the health effects. Chemometrics Chemometrics is a new field of chemistry that deals with the extraction of maximal information from chemical data that have been produced using the techniques described in the previous sections. Chemometric methods have been extensively described by Sharaf et al. (1986) and by Massart et al. (1988~. It has applications to analytical chemistry in that often information content of a detector system is used only partially by simple data-reduction techniques. In addition, many of the newer analytical methods are truly multivariate tech- niques. To optimize the sensitivity and selectivity of such methods, it is neces- sary to use a proper multivariate design for calibration (Deming and Morgan, 1987~. It is impossible to fully optimize many modern analytical instruments with a one-variable-at-a-time approach. Similarly, interferences also must be treated in a multivariate manner Lath a proper statistical design if the value of multiple-species-measurement instruments are to be used fully. One immediate application is in the analysis of data from several sensors, where qualitative and quantitative analysis can be performed on the air sample (Stetter, 1986~. These methods can be used to resolve and quantitate the components of mixtures that the separation methods do not separate fully. Often the patterns obtained from the chemometric analysis of the analytical data can reveal the origin of the contaminants or information on their reac- tivity in the air. In addition, time-dependent response of a detector such as a drifting cali- bration can contain information that can be extracted easily using techniques such as Kalman filtering (Brown, 1986~. Thijssen et al. (1984a) developed a model for random drift, and a criterion was developed for deciding when to recalibrate or to measure the next unknown. Such methods have been ex- tended to include optimization of the calibration scheme (Thijssen et al., 1984b) and to incorporate a generalized standard additions approach to cali

108 ASSESSING HUMAN EXPOSURE bration (Vandeginste et al., 1983~. In many instances in the past, drifting calibration data would not have been recognized as such or would have been discarded because of the observed drift. Considerable work has been completed on developing structure-activity relationships so that the behavior of some of the easily measured compounds might be useful in predicting the atmospheric or physiological behavior of other structurally related compounds. Although much of the prior work has been on compounds of possible pharmacological activity, structure-activity studies also could help predict other biological responses, such as toxicity, bioconcentration potential, and other aspects of ecosystem behavior. Chemometric methods have been used to identify the sources of mutagen- icity observed in collected airborne particulate matter (Daisey et al., 1986; Wallace, 1987; Lewis et al., 1988) and are likely to be more widely used in the future to help relate observed atmospheric composition to various toxicological responses. When applied to resolving sources of airborne particulate matter or applications to interpreting airborne mutagenicity, chemometric methods commonly are called receptor models and are discussed more fully in Chap- ter 6. SUMMARY In assessing human exposures to airborne pollutants, numerous factors besides the contaminant must be measured, especially if the assessment is based on fixed-site sampling or modeling. Accurate estimates in these in- stances depend not only on concentration measurements from fixed-site moni- tors In various locations, but also on knowledge of numerous factors that ~nflu- ence the environments where exposures occur. The measurement of an air- borne contaminant can be visualized as a three-step process. First, the pollu- tant is sampled; then it is separated from other species also collected in the sample; finally, it is detected The choice of the measurement methods to be used in an exposure assessment is driven by a study specific aims and by the nature of a given airborne pollutant. Quality Control/Quality Assurance Quality assurance is a critical part of exposure studies and must be es- tablished as part of the initial study design, at which point it should be decided what precision and accuracy are needed to test the study hypothesis. An ef

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 109 fective quality assurance program is costly (approximately 15-25% of total expenses) and should be considered when establishing a project's budget. The use of field and lab spikes and blanks should be a routine practice. Simple techniques for generating or supplying analytical standards in the field need to be developed. Particular attention should be given to standards for highly reactive compounds, which often are of the greatest concern from human health perspective and for which stable standards often are difficult to prepare. Techniques could include the use of permeation devices; novel on- site generation of standards (e.g., reactors that quantitatively produce an ana- lyte); and compressed-gas standards, where containers are made of highly deactivated materials. High passivated, sta~nless-stee] canisters might be useful for this purpose. Validation studies of samplers and analytical instruments used in exposure assessments are required. Validations should be performed in settings similar to those used in exposure studies. Sampling Techniques and Strategy The choice of a sampling strategy and a measurement method hinges on the study specific aims and hypotheses. Physical, chemical, and biological characteristics of the pollutant dictate the method chosen to sample and meas- ure airborne concentrations. A contaminant can have very different health effects in the vapor phase and in the condensed phase. Care must be exercised that a sampling procedure collects all of the appropriate phases and does not present a false picture of a contaminant's physical state. Personal monitoring (active or passive) is the most direct approach for assessing human exposure to airborne pollutants. However, the portability requirement of this technique typically decreases method sensitivity compared with stationary microenvironmental monitoring. Personal monitors need to be developed for many contaminants, including certain metals, PAHs, other semivolatile organic compounds, polar VOCs, and radon decay products. In some cases, personal monitors already exist, but need to be refined, reduced in weight and size, and validated (e.g., airborne particles, certain pesticides). More general instrument needs include personal sampling pumps with im- proved reliability, stability, and wide ranges of accurate flow rates and low noise levels. Lighter batteries with longer lives will contribute to improved personal monitoring. For pollutants whose effects might be related to peak exposures, personal samplers are needed that will monitor continuously for only short-term peak

110 ASSESSING HUMAN EXPOSURE exposures. Samplers based on electrochemical principles potentially could meet this need. In such an application, Bath the use of an electronic discrimi- nator, the signal characteristic of the analyte could be recorded only when it exceeds a preset threshold value. For monitoring devices using other chemical or physical principles, the measurement of short-term peak exposures is a more difficult problem and requires additional research. Quiet and unobtrusive microenvironmental samplers (active or passive) are needed if they are to be used more widely; such samplers should be available for the majority of atmospheric contaminants. Passive samplers are well suited to the collection of long-term integrated samples collected over days or weeks and can be extremely useful in a per- sonal or microenvironmental study. However, long-term sampling with a passive monitor places great constraints on the sorbent; it must retain the analyses without promoting unintended reactions among adsorbed analyses during the sampling period. Improved sorbents are needed that would permit long-term sampling for a wide variety of analyses. New sorbents are also required for polar organics, highly volatile compounds, and very reactive spe- cies. Ideally, passive samplers should be easily desorbed with the proper sol- vent or thermal techniques. New Resorption processes need to be developed. Procedures that permit desorption with a minimum of dilution, such as super- critical fluid extraction, would be especially useful. Such improvements would avoid the compromise in analytic sensitivity that often results from the large volumes used in liquid Resorption techniques. Instrumental Techniques Numerous advances have been made in instrument design, operation, and experimental deployment during the past 10-15 yearse LC techniques are being used to analyze for compounds not amenable to GC. In particular, IC (ion chromatography is being used increasingly to analyze for highly polar air contaminants. In addition, LC-MS is developing into a productive technique to complement GC-MS. The development of new MS instrumentation (e.g., ion trap mass spectrometers) has made MS a valuable air-analysis device. Microsampling and microanaIytical devices have been developed but are not yet widely applied. A summary of attributes of different measurement tech- niques is presented in Table 3.5. (Table 3.5 does not summarize all the techniques discussed in this chapter; it does include some established tech- niques that were not discussed in this chapter.) Some contaminants are not distributed uniformly over the surfaces of air- borne particles. To evaluate health effects properly, particulate analyses

111 :5 a) a: c 'e · - o v, D .c o m TE ~ ~ ~ ~ ~ ~ ~ on at >.1 i cl =1~;~1;^ S r X x xx x | xx x; x i x x . ~ of =1 of of ^1 =1 ^1 ~ ~ t r W 1 in: 4 1~ *0 ',> ~ :~ :~

112 As I: ._ a C.) - o - ~D Ct o O~ in U)  o ~ ·C4) :E - ._ 2 Cal C, 5 ._ S C) Xx X X X O .s I: O iF~ O C-> Cal 3 ~ , 6 - 6~ XX X _ U:  Cal : - . - =: s o ._ U) C5 V) o x o D 5: c0 ~ Ct At, D Ct S e ,` ~ ~ 0 CO ~xxx 2 ~ ~ o, ~ :~` ~ _ ~ ~ Q) USA ~ Cal s o ~ * .° 3 cut ._ o ct - ~ ~ e :^ o ) e~ ~ .> ~ ·u, o c) · - u) 'e c~ ~ u, c) .c c.> o .u - ·> t' 8 o ~ Ct ~ ~ ~V =, & C)  s: CO o U) t> ._ C) oe :^ CtS ~ ~ o C) Ce 1o ~s h7 o .~ cO~ ~ .O o~ o ~ Ct £ C~ Ct - Co, ~ ~ CO ._ o cs s:~ ~ CO :^ X X ~, ~ _' ·= o 4_ C~ ~ ·§

SAMPLING AND PHYSICAL-CHEMICAL MEASUREMENTS 113 should include surface techniques that provide elemental or chemical infor- mation. Field-Study Instruments Research and development should focus on better instrumentation for field studies. These include more portable, reliable, and rugged gas chromato- graphs, gas chromatograph/mass spectrometers, and ion trap mass spectrum eters. Sampling methods, instruments, and software that interfaces with such instruments are also needed to permit unattended sample collection and analyses in field settings. A sensitive, highly specific detector applicable to numerous compounds is needed for the liquid chromato~aph; continued improvements in LC-MS are beginning to fill this gap.