Appendix 2. Additional Information About Proposed Approach

2.1 Estimated Annual Average PM concentrations

Pittsburgh is an area influenced by reasonably well-defined local, remote, and biogenic sources (Figure A.1). This will make source-receptor relationships much easier to quantify compared to areas closer to the eastern seaboard where the urban centers interact with each other strongly. The estimated annual average PM_2.5 concentration for the Supersite location is approximately 20 µg m^-3 (Falke, 1999). The maximum daily average measured PM₁₀ concentration in the area in 1998 was approximately 150 µg m^-3.

 

Figure A.1 Annual average PM_2.5 concentrations (1994-1996) derived from PM₁₀ aided spatial estimation, Falke (1999).

2.1 Acronyms A list of definitions of acronyms for instrumentation (see Table 2) follows:

¹ Ultrafine SMPS: Scanning Mobility Particle Spectrometer (TSI model 3936N25) for 0.003-0.150 mm

² SMPS: Scanning Mobility Particle Spectrometer (TSI model 3934L) for 0.01-1.0 mm

³ APS: Aerodynamic Particle Sizer Spectrometer (TSI model 3320) for 0.5-1.0 mm

⁴ ELPI: Electrical Low Pressure Impactor (Dekati)

⁵ Ultrafine CPC: Ultrafine Condensation Particle Counter (TSI model 3025A) for 0.003-10 mm

⁶ FRM: Federal Reference Method PM_2.5 Sampler

⁷ TEOM: Tapered Element Oscillating Microbalance PM_2.5 Sampler

⁸ CMU Sampler: Carnegie Mellon University system shown in Figure 2

⁹ LPI: Low Pressure Impactor developed by Susanne Hering

¹⁰ MOUDI: Micro-Orifice Uniform Deposit Impactor

¹¹ IC: Ion Chromatography

¹² ICPMS: Inductively Coupled Plasma Mass Spectrometer

¹³ OC and EC: Organic Carbon and Elemental Carbon

¹⁴ PC-BOSS: Particle Concentrator-Brigham Young University Organic Sampling System. Analysis is conducted for several species: sulfate and nitrate (IC), NH4+ (spectrophotometry), OC/EC (Temperature Programmed Volatilization analysis), pH and acidity.

¹⁵ R&P sampler: Ruprecht and Patashnik sampling system

¹⁶ GC-MS: Gas Chromatography-Mass Spectrometry

¹⁷ FTIR: Fourier Transform Infrared Spectroscopy

¹⁸ MOUDI-PUF system: MOUDI-Polyurethane Foam plug preceded by a cyclone and denuder

¹⁹ HPLC: High Performance Liquid Chromatography

²⁰ RSMS-II: University of Delaware Single Particle Mass Spectrometer

²¹ SEM: Scanning Electron Microscopy

²² HFAS/GFAA: High Frequency Aerosol Sampler/Graphite Furnace Atomic Absorption Spectrophotometry

²³ TDMA: Tandem Differential Mobility Analyzer

²⁴ CCN Counter: Cloud Condensation Nuclei Counter manufactured by DH Associates

²⁵ GC-FID: Gas Chromatography-Flame Ionization Detector

²⁶ CSU collector: Colorado State University cloudwater collector

2.2 Speciation of Organic Aerosol

The material collected in the denuder, the filter and the PUFs will be analyzed by an improved GC/MS technique developed by the Rogge group for the quantification of the concentrations of individual compounds. The proposed improved technique (for details see section 4.3) will allow the increase of the pool of organic compounds that can be chromatographically separated and made elutable from the gas chromatographic column used. Daily filter samples will be collected throughout the sampling campaign. Collected filter samples will be grouped into biweekly composites and analyzed together, yielding biweekly average ambient fine particle organic compounds concentrations. Each 29th day, XAD-Denuder/Filter/PUF sampling will be conducted and the entire set will be analyzed separately to obtain additional detailed information about the partitioning of semi-volatile organic compounds between the gas- and particle-phases. During the intensive sampling periods the sampling frequency will be increased to two samples per day (12 hour samples) and the XADs, filters, and PUFs each will be analyzed separately. For 2 days of two intensive campaigns, 4 hours sampling will be conducted and all XADs, filter, and PUFs analyzed separately. The proposed schedule will provide a unique dataset with high temporal and chemical resolution that will broaden the knowledge about the chemical composition of organic particulate matter. This dataset will allow the Rogge group to derive source/receptor relationships using their published organic source profiles together with the collected ambient data.

Each of the samples (XAD, filter, PUF), and composited filter samples will be spiked with 7 deuterated recovery standards before extraction. XAD will be extracted successively four times with a mixture of dichloromethane/acetone/hexane (2:3:5) (Gundel et al., 1995). Quartz fiber filters will be extracted with dichloromethane three times using mild sonication. Similarly, PUF samples will be extracted three times with dichloromethane in a specially designed apparatus that allows to repeatedly compress the PUF foam plug. After extraction, the sample volume will be reduced to about 0.5 to 2.0 ml and divided into two aliquots. One aliquot will be used for derivatization with diazomethane to convert organic acids to their methyl ester analogues and acidic hydroxy compounds to their respective methoxy analogues. After the derivatization, both sample aliquots will be stored in a freezer until analysis by GC/MS. In order to make other polar organic compounds accessible to chromatography, a method will be developed and refined during this project that uses BSTFA [bis(trimethylsilyl)trifluoroacetamide] for compound derivatization by silylation (Simoneit et al., 1999). Furthermore, for selected samples, silica gel column chromatography will be conducted with 8 different eluting solvents, fractionating the sample extract according to compound polarity. Solvent elution sequence includes: (F1) hexane, (F2) hexane/toluene (3:1), (F3) hexane/toluene (1:1), (F4) hexane/ethyl acetate (19:1), (F5) hexane/ethyl acetate (9:1), (F6) hexane/ethyl acetate (6:1), (F7) hexane/ethyl acetate (4:1), and (F8) methanol. In order to obtained quantifiable fractions, a method will be developed that allows to quantitatively track the fractionation and conversion of the polar fractions into GC chromatographicable derivatives.

Extract analysis will be accomplished using a Hewlett-Packard Model 5973 quadrupole mass spectrometer coupled with a HP 6890 series gas chromatograph. At least the following compound classes will be searched for and when identified also quantified: n-alkanes, n-alkanoic acids, n-alkanols, n-alkanones, furans, furanones, resin acids, levoglucosan, sterols, PAHs, compounds identified in smog chamber experiments as possible secondary organic atmospheric reaction products, and others. As far as available, high purity standards will be used to aid quantification. Beyond the listed compounds, every effort will be made to identify in a few selected samples everything possible, applying the extract derivatization and fractionation methods to be developed.

2.3 Size-resolved FTIR spectra with solvent rinses

Size-segregated samples will be collected on five impaction stages with cut points between 0.05 and 2.0 mm for 24 hours on ZnSe disks at 1 lpm using a Hering Low Pressure Impactor (LPI) during the three intensive periods. Spectra will be taken of a clean area of the ZnSe disk and of the deposit using a Mattson Research Series 100 FTIR spectrometer, and spectral results will be presented as the percent of incident light absorbed versus wavenumber. Each spectrum will be an average of 200 spectral scans at 2 cm-1 wavenumber resolution. Functional groups will be identified based on spectral libraries and standards. Instrumental sensitivity will be evaluated by daily analysis of a standard thickness polystyrene film and weekly analysis of KBr pellets comprised of mixtures of organic and ammonium sulfate standards. Following initial analysis samples will be gently rinsed with spectral-grade hexane, spectral-grade acetone, and deionized water and reanalyzed after each rinse. The rinse is performed by holding the substrate at a 45° angle and applying 1 ml of solvent to the deposit from a pipet. The solvent is allowed to flow slowly over the deposit for approximately 5-10 seconds. This process will be repeated twice. The substrates are then air dried and reanalyzed by FTIR. Acetone is applied to the sample after the final water rinse to facilitate drying. Spectra are integrated using GRAMS 32 peak integration software and functional group absorbances are quantified relative to sulfate.

2.4 LIBS

Laser-Induced Breakdown Spectroscopy (LIBS) is an optical technique that can be used to detect trace inorganic species and particles at femtogram-picogram levels in combustion products and in the ambient atmosphere. LIBS has been used as an analytical technique for gases, liquids, and solids (Song et al., 1997; Schechter, 1997). Applications of LIBS employ a pulsed laser with a high peak power to form a spark (breakdown) in the medium to be examined.

Figure A2: Conceptual illustration of the LIBS system and a sample spectrum.

 

In gases, the temperature of the resulting plasma is in the range of 10,000 - 15,000 K, hot enough to fragment all molecules into their constituent atoms, and to excite the electrons in the atoms out of the ground state and into excited electronic states. As the plasma cools, excited electrons relax back into their ground states, emitting light at characteristic atomic frequencies. Identification of the atoms present in the sample volume occurs using well-known atomic emission lines, and quantification of the atomic species concentration occurs via quantification of the intensity of the atomic emission lines. A conceptual diagram and sample spectra are illustrated in Figure A2. Sensitivities range from femtograms for some species, such as Be, to well above the picogram level for other species, such as Sb.

LIBS is extremely well suited as a diagnostic for trace inorganic species. The Buckley group has recently been directing a project at Sandia National Laboratories to apply LIBS as a real-time, in situ monitor for toxic metals in combustion systems (Buckley et al., 1999; Hahn et al., 1999). A field detection system has been assembled that can run in an unmanned mode for continuous emissions monitoring (Figure A2). Laboratory experiments suggest that LIBS can be used to measure particles as small as 100-150 nm depending on the species present. With the capability in the software to detect single particle hits, it is possible to do joint composition and sizing analysis of single particles. This ability makes LIBS an excellent diagnostic for trace species in the ambient atmosphere. Laboratory measurements to validate this technique have been accomplished. Ambient measurements have been made at industrial sites and in Livermore, CA.

2.4 Single Particle Mass Spectrometry

There are now a number of groups world-wide who have built single particle instruments and employed them in field campaigns (e.g., Noble and Prather, 1996; Murphy et al., 1998). All of these measurements were performed using the light scattering type of instrument so only examined particles greater than a few hundred nanometers in diameter. From experience with single particle instruments and a large body of literature on laser ablation mass spectrometry, we can summarize the capabilities of the current RSMS-II design that are especially suited for exploring atmospheric aerosols.

1. External mixing properties -- atmospheric particles are emitted by a wide range of sources and undergo disparate processing in the atmosphere. This results in particles of a given size having a range of compositions. These compositions can often be grouped into categories or classes of similar composition (e.g., Song et al., 1999).

2. Composition of fine and ultrafine particles -- Although there are large numbers of fine and ultrafine particles in the atmosphere, the mass of particles in these size ranges is quite small. As a result of this small mass loading, using conventional techniques to analyze the particles in these size classes necessitates sampling a large volume of air either at a high volume flow rate or over a long sampling period. The single particle technique described here analyzes individual particles so the overall particulate mass in a given size is not a constraint. Wexler and Johnston have demonstrated that single particle analysis is effective down to about 10 nm (Carson et al., 1997) and has good sensitivity to impurities at 50 nm (Ge et al., 1998). This particle size range is extremely important to atmospheric processes such as cloud dynamics and radiative balance.

3. Size distribution of composition -- In all the experiments proposed here, RSMS-II will be run in parallel with independent particle sizing instruments (a TSI SMPS and APS) so that the particle hit rate can be normalized to the actual particle concentration that is sampled. Thus particle-sizing instruments will record the size distribution while RSMS-II will be used to establish the composition of particles analyzed at a given size. The RSMS-II hit rate will be calibrated beforehand using particles that are easy to analyze. Thus the discrepancy in the particle number distribution obtained from RSMS-II and the pure sizing instruments will indicate the percentage of particles either not analyzed or not analyzable.

4. Speciating inorganic compounds -- The speciation abilities of the instrument are currently not as varied or quantitative as bulk instruments, but Wexler and Johnston have shown that for instance different sulfur-containing compounds, such as sulfate, methane sulfonates, and hydroxymethane sulfonates, can be distinguished from each other (Neubauer et al., 1996). Elemental and metals analysis has also been demonstrated (Murphy et al., 1998; Ge et al., 1998).

5. Distinguishing elemental from organic carbon -- From the laser ablation literature, it is known that elemental carbon particles can be distinguished from organic carbon particles by the pattern of carbon-hydrogen peaks in the mass spectra (DeWaele et al., 1983). The usual laser desorption/ionization single particle analysis is only effective for speciating aromatics and distinguishing organic from elemental carbon. More in-depth single particle organic analysis techniques are described later in this proposal.

6. Temporal resolution -- Atmospheric particles often contain labile compounds whose concentration is affected by reactions or condensation/evaporation. Due to the presence of point sources, fronts, and turbulent eddies, the concentrations and compositions of atmospheric particles may exhibit sharp spatial gradients. Single particle analysis is fast compared to bulk analytical techniques. The transit time from sampling to analysis is measured in seconds so that a) rapid gradients in number or composition can be identified and b) labile compounds are not significantly transformed between sampling and analysis. The anticipated hit rate of hundreds of particles per minute will enable rapid transients in composition to be resolved.

Focusing of Aerosol Particles. It has been known for a number of years that passing an aerosol through a sharp or conical orifice where the flow is choked (that is, sonic) focuses certain particles (Dahneke, 1982; Fernandez de la Mora and Riesco-Chueca, 1988 and references therein). The focussed particles have a Stokes number value around one, the exact value depending on the nozzle geometry and the distance from the nozzle to the focal point. The U. of Delaware team employs this principle in their current single particle instrument, a second-generation version called RSMS-II (Rapid Single-particle Mass Spectrometry-II). The instrument focuses only a narrow range of particles sizes to the source region of the mass spectrometer, but this size range can be selected by adjusting the pressure upstream of the nozzle. If we consider a given nozzle geometry, let us term the Stokes number that is focused by Stkf. Thus we can write Stkf = (Dp2rpuo/18mDn)Cc where Dp is the particle diameter that is focussed, rp is its density, uo is the velocity through the orifice (which is sonic since the flow is choked), m is the viscosity of air, Dn is the orifice diameter and Cc is the Cunningham non-slip correction factor. All the terms except Dp and Cc can be combined into an effective diameter, Dp,max = (18mDn Stkf/ rpuo)1/2 giving (Dp/Dp,max)2 = 1/Cc. Since the maximum value of Cc is 1, Dp,max is the maximum particle aerodynamic diameter than the nozzle can focus. Cc is a function of the particle diameter and pressure. The usual function for the Cunningham correction factor is difficult to manipulate analytically (Seinfeld and Pandis, 1998), but a simple linear approximation, Cc = 1 + 1.66 (2l/Dp), is close to the more precise equation and has a maximum error of only 10% at Kn=1. Using this approximation for Cc gives a quadratic in Dp whose solution is

Dp = ((3.32l)2 + Dp,max 2)1/2 - 3.32l

where l, the mean free path, is a function of pressure via l = lopo/p, where po and lo are the pressure and mean free path at standard conditions. Since the mean free path is a function of pressure, the particle size focussed in the mass spectrometer can selected by adjusting the pressure as long as this size is smaller than Dp,max.

The maximum size that can be focussed, Dp,max, is a function of properties of air (m, uo) and the particle (rp) which are not adjustable. The only adjustable parameter is the nozzle diameter, Dn, and geometric considerations such as the cone angle and distance from the orifice to the focal point, which are incorporated in Stkf.

It is best to run the orifice choked because this focuses the smallest particles, but this requirement places limits on the orifice diameter in that the vacuum pump driving the orifice must be large enough to choke the flow. A 600 lpm pump is large but reasonably transportable and leads to a 3 mm orifice. The result is that for unit density particles, Dp,max is about 2 micron. At low pressures, the focussed diameter is a linear function of pressure but this sensitivity reduces as the focussed diameter approaches Dp,max. By scanning the pressure from about 300 torr to about 1 torr, the nozzle is able to focus particles ranging from 1 micron to 10 nm in aerodynamic diameter - selecting the pressure is the next step.

One challenge to the design of the instrument is controlling the pressure upstream of the nozzle over a wide dynamic range while efficiently transmitting particles. We use a bank of 10 critical orifices whose flow is controlled by a rotary valve (Valco Instruments http://www.vici.com). The area of each orifice is selected to evenly distribute the 10 particle diameters logarithmically over the 10 nm to 2 micron size range. Each flow-control orifice is a sharp hole minimizing particle deposition and associated clogging. After the orifice bank is a section of straight tube where the flow straightens and becomes laminar before it enters the focussing orifice, giving more reliable focussing characteristics. The transmission efficiency of the inlet and the instrument's ability to size and analyze atmospheric particles has been submitted for publication (Mallina et al., 1999).

Standard software in the University of Delaware laboratory will be used to convert the time-based raw spectra to mass-based spectra that can then be interpreted. The pressure will be converted to particle aerodynamic diameter via equation 2. The parameter Dp,max is a function of the geometry of the nozzle, which is fixed during each experiment. The pressure governs the mean free path, l, which then determines the particles size that is focussed. The pressure diameter relationship will be calibrated with standard aerosols selected by a DMA both before and after each sampling period. When aerodynamic techniques are employed to size particles, as RSMS-II does, the particle density and aerodynamic properties contribute to the uncertainty in sizing. The density can be estimated from the particle composition reducing this uncertainty. Likewise, from the relative humidity and composition, the particle shape (e.g., spherical liquid, facetted crystal) can be estimated and used to reduce the uncertainty in converting aerodynamic to physical diameter.

Particle Detection. Particles that are focussed to the source region of the mass spectrometer are detected in one of two ways. Larger particles are detected by light scattering. A CW doubled Nd-YAG laser passes through the particle beam just upstream of the center of the source region. Two PMTs in the source region detect forward scattered radiation at angles of +/- 30 degrees and signal coincidence from these two PMTs is used to trigger the ablation laser (Ohigashi et al., 1994). Tests show that light scattering is effective down to about 200 nm.

For smaller particles, light scattering is not effective so the ablation laser is free fired (Reents et al., 1994; Carson et al., 1997; Ge et al., 1998). Currently we use an MPB PSX-100 laser, which fires at up to 100 Hz. To maintain single particle analysis, it is important to operate the instrument such that most laser shots miss a particle. For instance, if one in 10 shots result is a hit, then 1 out of 100 hits is a double hit, that is, two particles were in the source region. Here the nozzle flow rate characteristics become useful. Generally, the atmosphere contains a great many ultrafine particles and relatively fewer fines. This is somewhat counterbalanced by the flow rate through the nozzle. Larger particles are focussed at higher pressures where the sampling flow rate is higher whereas smaller particles are focussed at lower pressures where the sampling flow rate is lower. Nevertheless, if the hit rate is too high, dilution of particle stream is necessary to maintain single particle analysis. Particles are "detected" by the presence of a spectrum. The technique has been used in the laboratory to analyze particles down to 10 nm (Carson et al., 1997) and we have used it to demonstrate sensitivity to 0.1 mass percent impurities in 50 nm particles (Ge et al., 1998).

Since some particles may not be readily analyzed by RSMS-II and for the smallest particles, the spectrum is used to identify the presence of the particle, the measurements contain some uncertainty as to the number of particles present. As a result, the particle number will be measured simultaneously with an independent instrument (a TSI SMPS). The RSMS-II and standard number concentration measurements will be compared to determine the fraction of the particles analyzed.

Particle Analysis. The source region of the mass spectrometer is specially designed to maximize the probability of a particle hit when the laser is free-fired. First, the excimer laser beam propagates collinear to and in the opposite direction from the particle beam. In most designs these two beams are normal to each other so the overlap region is only a few hundred microns. With collinear beams, the overlap is very large and the source region size is governed by the ion optics in the mass spectrometer. In RSMS-II the ion optics have been designed to allow a 4 cm source region (Carson et al., 1997). The excimer laser beam is focussed to the center of the source region where it has a waist about 0.6 mm in diameter. This widens to about 2mm at the top and bottom of the source region. The particle beam passes through a 1 mm skimmer before it enters the source region of the mass spectrometer. In the center of the source region, 18 cm from the primary orifice, the beam is very well defined and about 4 mm across (measured by impacting oleic acid particles on a glass slide). It is the overlap between the hour-glass-shaped laser beam and the conical particle beam which determines the particle hit-rate probability. The overall particle hit rate is the product of this probability and the volume flow rate sampled.

Data System. The data system consists of a 500 MHz 8-bit A/D converter mounted in a PC (Precision Instruments 9847). The board was customized by the manufacturer so that valid spectra could be detected in firmware on the board -- PCs are not yet fast enough to check in software each spectrum for validity at 100 Hz. Valid spectra are stored on the PC disk and archived on writeable CD-ROM.

Individual spectra can be viewed in real time in the field and can be used provide feedback to other measurements. Ultimately, a more statistically significant analysis of the data will be performed. Specifically, spectra from each particle each size will be grouped into categories of similar particle type. Two methods will be employed. We have used the CART (Classification And Regression Tree) algorithm successfully to bin laboratory generated particles by composition and quantify the amount of different sulfur compounds in a particle (Neubauer et al., 1996). To quantify, the algorithm must be trained with standards of known composition, which may be impractical for atmospheric applications. Alternatively, a neural network based algorithm has been tested on atmospheric particles and is able to classify them without a priori training (Song et al., 1999). This is clearly advantageous for atmospheric applications. Alternatives to these algorithms will be explored if they fail to classify the spectra.

RSMS-II has a number of advantages over the more conventional designs currently in use.

1. Particle size selection. Particles can be sized and selected over a much wider size range and the particle transmission rate for the selected size is much higher than for nozzles that transmit a broader range.

2. Particle sampling. Instruments that transmit a range of sizes must cope with the dramatic increase in particle number as the particle diameter decreases. Since the vast majority of atmospheric particles are small, significant time will be spent waiting for the larger particles to appear. With the new design, a given size is sampled until a statistically significant number are obtained and then another size is selected.

3. Particle detection. For smaller particles where the atmospheric number concentrations are typically higher, free firing is practical for analyzing particles. The particle number concentration is so low for larger particles that the hit rate becomes prohibitively low so the particles are detected by light scattering.

4. Particle size limits. In free-firing mode, the instrument is not limited by light scattering for particle detection. Thus ultrafine particles can be sized and sampled.

5. Particle analysis. Impactors can be used to obtain aerosol composition but the mass of smaller particles is very low in the atmosphere. Thus a large sampling time is needed to obtain the composition of fine and ultrafine particles for typical atmospheric loadings. Single particle analysis is well suited to analyzing small particles because the analysis is proportional to number not mass.

6. Volume flow rate. The volume flow rate into the instrument is proportional to the pressure upstream of the focussing orifice. As a result this flow rate is higher for larger particles (which are fewer) and lower for smaller particles (which are more plentiful) helping to equilibrate the hit rate over particle size.

2.5 Cloud and fog water collection

The Colorado State University's CASCC2 collects fog/cloud drops by drawing droplet-laden air at 5.8 m³ min-1 over six banks of 508 µm diameter Teflon strands. Drops are collected by inertial impaction on the strands. The 50% size cut for the CASCC2 corresponds to a drop diameter of 3.5 µm. Collected drops coalesce and are carried by gravity and aerodynamic drag down into a Teflon sample trough and through a Teflon sample tube to a polyethylene sample bottle. The collector is cleaned prior to each fog/cloud event and blanks are taken. Operation of the CASCC2 will be automated for this study. Fog/cloud presence will be detected using a Gerber Scientific Model PVM-100 Particulate Volume Monitor. This instrument monitors forward light scattering to determine cloud/fog liquid water content (LWC). When the LWC is sufficiently high to indicate fog presence, the CASCC2 will automatically be turned on. At the same time the PVM-100 will activate a modem to place a call to a pager carried by the individual conducting the fog measurements, notifying her/him that sampling has commenced. The operator will come to the site immediately to ensure equipment is operating properly and to change samples as needed.

2.6 Determination of semi-volatile material

The combination of the technology used in the BIG BOSS sampling system (Tang, 1994) and the Harvard particle concentrator (Sioutas, 1994a,b) has resulted in the Particle Concentrator-Brigham Young University Organic Sample System (PC-BOSS) (Ding 1997, 1999a; Eatough 1999) (Figure A3). The PC-BOSS has five advantages over diffusion denuder systems previously used to determine fine particulate semi-volatile material.

1. Both semi-volatile ammonium nitrate and semi-volatile organic compounds are determined with the same sampler (Ding 1999b, Eatough 1999).

2. The use of a particle concentrator eliminates the need for the independent measurement of the efficiency of collection of gas phase organic compounds by the diffusion denuder (Ding 1999a; Eatough 1999).

3. The elimination of the independent measurement of denuder breakthrough (Cui 1998; Eatough 1995; Tang 1994) results in a sampler which can be used in routine field operation.

4. The collection of the concentrated particles in a low-volume flow stream simplifies the power requirements of the sampler.

5. The effective high sample flow obtained with the particle concentrator allows for the collection of samples over time periods as short as one hour (Eatough 1999; Pang 1999a,b).

The PC-BOSS sampler has been validated in sampling programs completed in Tennessee (July 1997), in Riverside CA (August-September 1997), in Bakersfield CA (February-March 1998) and in Provo UT (November-December, 1998) (Pang 1999a; Ding 1999b; Pang 1999b). In the experiments completed to date the efficiency of the denuder for the removal of gas phase organic material, nitric acid and SO₂ was shown to be near 100%. Most gas phase ozone, NO2 and ammonia is also removed by the denuder (Ding 1999b; Eatough 1999; Obeidi 1999). The agreement for fine particulate mass, carbonaceous material, sulfate and nitrate between collocated PC-BOSS samplers was ±8% and agreement for these species between the PC-BOSS and comparable diffusion denuder samplers was ±5-10% (Ding 1999b, Pang 1999a,b). For all studies, the results obtained with the PC-BOSS sampler indicated that from 10 to 50% of the fine particulate mass was not measured with the PM_2.5 FRM sampler due to the loss of semi-volatile organic material and ammonium nitrate during sampling. The majority of the loss at Riverside and Provo was due to semi-volatile organic material, but at Bakersfield the major species lost was ammonium nitrate.

During the intensive periods samples five samples will be collected and analyzed daily with a PC-BOSS sampler by the BYU team to establish the diurnal pattern of fine particulate chemical species, including semi-volatile nitrate and organic material. Approximately half of the samples collected during the three two-week intensive field sampling programs will be selected for analysis. PM_2.5 mass will be determined from the material collected on the Teflon filter in the minor channel of the PC-BOSS (Figure A3). This mass will be corrected for the loss of ammonium nitrate and semi-volatile organic material (determined from analysis of the Nylon and charcoal impregnated filter, CIF, in the minor flow channel of the PC-BOSS). Constructed mass will also be determined from the various analyzed species. The concentration of fine particulate sulfate and of nitrate retained by a particle collection filter will be determined by ion chromatographic analysis of ultrasonic aqueous extracts of each quartz and Teflon filter (a total of three filters/sample) of every PC-BOSS collected sample. The concentration of nitrate lost from particles during sample collection will be obtained by ion chromatographic analysis of IC eluent extracts of both the Nylon and CIF filters of the PC-BOSS. The concentrations of ammonium ion in extracts of the minor flow Teflon filter for each PC-BOSS sample will be determined spectrophotometrically. The pH of these extracts will be determined by pH measurement and the acidity obtained as previously described (Koutrakis et al., 1988). The concentrations of organic carbonaceous material and soot retained on each PC-BOSS quartz filter (Figure A3), will be determined by temperature programmed volatilization analysis (TPV) from ambient to 800 EC in a N2/O2 atmosphere (Eatough 1993, Ellis 1982). The concentrations of semi-volatile organic material lost from the particles during sampling will be obtained from TPV analysis of the material collected by each CIF of the PC-BOSS in a N2 atmosphere from ambient to 400 EC (Eatough 1993, Tang 1994).

2.7 Continuous Nitrate, Sulfate, and Carbon Measurements

For particle nitrate the ADI approach has been compared to denuder-filter methods at three different California locations, namely Riverside, Mira Loma and Bakersfield, and has yielded regression slopes of 0.96, 1.00 and 1.07 with correlation coefficients of 0.96. Example data from Bakersfield, California is shown in Figure A4 (Dutcher et al., 1999). Laboratory tests show collection efficiencies of greater than 95% for particles between 0.1-1 µm. Interference testing shows less than 1% response to nitric acid or ammonium ion.

Data from the combined sulfate, carbon nitrate system collected in December 1998 in Marshall, Colorado are shown in Figure A5 (Hering et al, 1999). The data show that the time

Figure A4. Comparison of nitrate concentrations measured by the automated nitrate monitor with measured by a filter sampler.

 

Figure A5. Time series of the automated sulfate, nitrate, carbon measured at Marshall, Colorado, with comparison of the mass sum of species (left hand axis) to optical particle counter volume (mm³/cc, right hand axis). Sum of species is calculated as 1.29*nitrate + 1.375*sulfate +1.4*carbon.

 

variation in the sum of the nitrate, sulfate, and carbon species measured by these systems tracked the total particle concentration, as indicated by physical size distribution measurements. Additionally, the mass sum is comparable to the measured aerosol volume, with a mass to volume ratio just over 1 g cm^-3. Note that the relative proportions of carbon, sulfate, and nitrate vary. Such data are extremely valuable to the proposed atmospheric aerosol modeling for the Pittsburgh Supersite. The combined modeling and measurement effort will provide valuable insights into the sources and formation processes for these particle constituents which comprise the majority of the fine particle mass.

2.8 Bioaerosols

The Hernandez group using direct microscopy and molecular biological analyses will develop and apply rapid, quantitative, aerosol assays to characterize the identity, distribution and activity of microbiological components present in outdoor aerosols. This aerosol microbiology study will focus on bacteria, fungi and their spores. The definition of "microbiological" here does not dictate the necessity to be isolated or quantified by culture. In response to the limitations of conventional microbiological analysis, the Hernandez group will refine accepted (and novel) epi-fluorescent microscopy practices with genetic-based, molecular biology assays into rapid, non-culture based techniques that can directly measure and classify microbiological bioaerosol material in outdoor air. Our sampling approach for characterizing outdoor bioaerosols is to use two different types of common samplers in conjunction with each other: cyclone wet samplers and high volume PM₁₀&2.5 filter samplers. The wet samplers will be used for culturing, microscopic and genetic analyses, while the total PM samplers will be used for biomass determinations (vs. total mass). Sensitive biological stains and image analyses will be used to directly characterize microbiological aerosols collected by both types of samplers. Image analysis will be used to accurately measure the size and morphology of stained biological particle recovered from outdoor air. Using widely accepted molecular methods, genetic probes, polymerase chain reaction (PCR), and denaturing gradient gel electrophoresis (DGGE) will be used to classify airborne microorganisms on a phylogenic, rather than a culturable basis.

2.9 Meteorology

Atmospheric trajectory analysis has long been used to identify the location of upwind regions that contribute material to samples collected hundreds and even thousands of kilometers downwind. Information contained in modeled trajectories includes the upwind transport pathway corresponding to air arriving at the sampling site and the temperature at each point in the trajectory. This information identifies the transit time from upwind sources, thus aiding in the determination of en route chemical transformation and possible interaction with clouds, among other things. The trajectory model developed by Harris and Kahl (1994) will be applied by the Kahl group to calculate trajectories twice daily for a 1.5 year segment of the project. This model calculates 10-day transport along isentropic surfaces, thus accounting for the adiabatic vertical motions that characterize the basic features of atmospheric transport. The model switches from isentropic to mixed-layer mode when trajectories come close to the earth's surface. In this mode, air parcels are advected using winds averaged throughout the lowest 500 m of the atmosphere. In the isentropic mode the parcels are advected using winds found along a surface of constant potential temperature. The model also features a dynamic "theta-chooser" algorithm, which calculates the potential temperature surface that corresponds to the desired altitude of trajectory arrival at the receptor.

2.10 Gas-phase measurements

Individual VOC concentrations will be measured chromatographically using the GC-FID and GC/MS techniques. Samples will be collected into 32 L Summa canisters that will be pressurized during sampling to approximately the equivalent of 96 L of air at one atmosphere. The GC-FID method will use an HP gas chromatograph and integrator system with a J&W Scientific DB-1 column. The sample will be loaded through a gas-sampling valve onto a Supelco Carbotrap 300 sorbent trap, at 200 to 300 mL min-1. Residual water will be removed by passing He through the trap at 100 mL min-1 for 7 min. The trapped sample is then desorbed at 340oC for 10 min through a second gas-sampling valve into a loop filled with glass beads. The organics are vaporized by heating the loop at 100oC and cryo-focused onto the head of the column held at -60oC. The column-oven temperature is held for 5 min at -60oC and then increases at 4oC min-1 to 200oC. Using a similar technique Lewis et al. (1999) reported a species quantification limit of about 0.1 µg m^-3 allowing them to quantify the concentrations of roughly 130 compounds in Atlanta air.

The GC-MS analysis will be similar to the GC-FID. Ambient samples will be loaded onto a multibed Tenax TA-Ambersorb-Charcoal sorbent trap in a Dynatherm Analytical Instruments model 890. The sample will be desorbed at 300oC for 6 minutes to a freeze-out loop immersed in liquid nitrogen. The rest of the analysis will follow the GC-FID protocol.

The Colorado State University (CSU) team will make gas phase measurements of hydrogen peroxide and soluble organic peroxides using a continuous monitor based on the method of Lazrus et al. (1986). This dual-channel instrument, constructed at CSU with the assistance of Drs. Greg Kok and Teresa Campos of NCAR, features a detection limit of ~ 0.1 ppbv. Air is drawn into the instrument through a Teflon sample tube. The first instrument channel scrubs soluble hydroperoxides (including hydrogen peroxide and soluble organic peroxides) from the air into solution where they are reacted to produce a fluorescent dimer, the concentration of which is measured by an on-line fluorimeter. The second instrument channel is identical to the first except that catalase is added to the sample stream to destroy hydrogen peroxide, permitting measurement of residual, organic peroxide concentrations. The hydrogen peroxide concentration is given by the difference between the two channel concentrations. The peroxide analyzer is calibrated by automated injection of aqueous standards. Routine checks are also made to verify efficient transmission of gaseous peroxides through the Teflon sample tube and efficient destruction of hydrogen peroxide by catalase in the second channel. Data will be recorded at 10 minute intervals and averaged to longer periods as needed for use in the cloud/fog study and for tests of the influence of hydrogen peroxide and organic peroxides on health. Expertise and instrumentation for the peroxide measurements will be provided by CSU with on-site assistance from CMU personnel.

2.11 Polar Organics

Polar organic compounds will be detected by an aerosol MALDI experiment. Matrix-assisted laser desorption ionization (MALDI) is widely used to obtain mass spectra of highly polar compounds. The analysis is performed by mixing the analyte with a matrix compound that strongly absorbs the laser radiation. The matrix-analyte mixture is then irradiated with a pulsed laser beam to eject intact analyte ions without fragmentation. The U. of Delaware group has performed MALDI of bulk and single-particle samples for several years (e.g., Mansoori et al., 1996).

On-line single particle MALDI will be performed as follows. Ambient aerosol will be mixed with a small flow of the hot vapor matrix quickly cooling it to ambient temperature and reducing the vapor pressure. The matrix vapor will condense on existing aerosol particles producing a matrix/analyte mixture like that used in a conventional MALDI experiment. This principle is similar to that of a "mixing-type" condensation nucleation counter (Okuyama et al, 1984) except that a much smaller amount of vapor will be condensed on the particles. The increase in particle size due to matrix condensation is controlled by the relative flow of the matrix vapor and the aerosol. A large body of literature has established the operating parameters (e.g., Willeke and Baron, 1993).

The U. of Delaware team has built a prototype apparatus to condense a very thin film of benzoic acid matrix on ultrafine particles. The goal of this experiment was not to perform MALDI of aerosol organics but rather to study the effect of a thin, absorbing coating on the yield of ions from non-absorbing particles. With this apparatus, it was found that a benzoic acid film corresponding less than a 5% increase in the particle size (the smallest size increase that could be measured with our experimental setup) was sufficient to increase the number of particles analyzed by as much as a factor of two. This increase in the "hit" rate arises from particles that pass through the edge of the laser beam. Because particles near the edge of the laser beam experience a lower laser irradiance than particles in the center of the beam, they may yield too few ions to be detected. Adding a small amount of benzoic acid increases the amount of laser radiation that is absorbed by the particle and increases the yield of ions from the core material. In this way, particles near the edge of the laser beam are detected and the hit rate increases. This experiment underscores the dramatic effect that even a small amount of an absorbing material can have on the laser ablation process.

Based upon conventional MALDI experiments, we expect that no more than a 1:1 mass ratio of matrix to analyte is needed to enhance formation of organic molecular ions. Since particle mass is proportional to the cube of the diameter, a doubling of the particle mass by condensation of matrix vapor will result in a 26% increase in particle diameter. This size change is small relative to the density uncertainty and corresponding size uncertainty for aerodynamically sized particles and therefore should not significantly alter conclusions drawn from correlated size-composition measurements of the transformed particles.

2.12 Semi-continuous metals

The semi-continuous metals sampler of the Ondov group consists of a high-frequency aerosol sampler (HFAS) and state-of-the-art true simultaneous multi-element Graphite Furnace Atomic absorption spectrometer. The HFAS consist of a state-of-the-art dynamic aerosol concentrator. PM 2.5 is sampled at 200 L min-1 and delivered to GFAA system or to a sample fraction collector for on or off-line analyses, respectively. In <10 (for urban air) to 20 (rural air) minutes the U. of Maryland (UMCP) group typically collects enough slurry to permit 4 suites of 4 or 5 elements to be determined, each in triplicate. At the 200 L min-1 sampling rate, analyte masses delivered to the GFAA at the smallest concentrations observed by Ondov in rural Maryland air exceed instrumental sensitivities (rivaling far more expensive ICP-MS but with far less sample volume requirement) by factors of 2 to >400 for Al, As, Cd, Cu, Fe, Mn, Ni, Pb, Se, V, and Zn. In addition to high temporal resolution, it is of paramount importance that analytical concentration measurements be accurate. Tests with NIST Standard Reference Material 1648 ("Urban Particulate Material") confirm these results for Cd, Pb, Zn, Se, As, Cr, Mn, Cu, and Ni. Tests for the remaining elements are in progress. Additional field tests of the instrument are being made in College Park, MD, and at the Atlanta Super Site in August, 1999.

The Ondov group will construct and deploy two HFASs, i.e., one at the main site and one at the satellite site. These will be deployed for 2 weeks during each of 3 two-week intensives and will be used to collect samples continuously at a sub-hourly interval to be selected during a pilot field test prior to the first scheduled intensive. To save costs and permit a maximum number of analyses to be performed, samples will be collected in fraction collectors (each holding up to 288 samples) and returned to UMCP for analysis on a (truly) simultaneous Graphite-Furnace Atomic Absorption Spectrometer for 18 elements in the following four groups:

Group 1: As, Cu, Mn, Ni, Cr
Group 2: Cd, Se, Ag, Pb
Group 3: Al, Fe, Zn, Ca, and Bi (Bi is an internal standard),
Group 4: V, Ti, Be, Ba

The elements selected include criteria pollutants (Pb and Be), hazardous air pollutants (known as air toxins; i.e., Cd, Cr, Cu, Ni, As, and Se), first series transition metals either known (V and Zn) or suspected to elicit respiratory inflammation (Ti, V, Cr, Mn, Fe, Ni, Cu, Zn; i.e., all but Sc and Co); essential nutrients (Fe, Zn, Se, Cr), and the aquatic toxin, Al. In addition, most are primary source marker species (i.e., As and Fe, steel; Zn, Cd, Cr, Pb, Ag, incinerators; Ni and V, residual fuel oil combustion; Se, coal combustion; Ti, paint manufacture and applications (fine particles) or crustal dust (coarse particles); Ca limestone/construction material; and Cu, Cd, Zn, respective smelter emissions and/or metals processing.

Until more experience is gained, 3 replicate analyses will be made for each suite of elements. Analysis of each suite about 5 minutes per replicate, so that the time to complete 3 replicates is 15 minutes per suite. The initial target collection rate will be 6 samples per hour, pending ambient concentrations. As the collection rate will far exceed the analytical rate the UMCP group will analyze selected series of samples based on observations made with continuous aerosol mass and particle concentrations, SO₂ concentration, and wind direction. At 200 samples from each intensive will be analyzed. Many more samples can be analyzed if fewer suites are selected, however, the best strategy may be to remove replications as 10 minute sampling times will likely provide for replication, in which case >600 samples would be analyzed per intensive. Note that each fraction collector can accommodate 288 samples, i.e., about 2 days worth.

2.13 Three-Dimensional Modeling Study

In collaboration with Ted Russell in Georgia Tech., the CMU groups have created a comprehensive three-dimensional model for the study of PM_x in the region shown in Figure A6. The model includes a state-of-the-art description of PM processes and describes the complete aerosol size/composition distribution using user-selected chemical and size resolution. The model is also coupled to a sensitivity analysis module so it can calculate directly the sensitivities of PM concentrations to small changes in source strength. The response to larger changes can be calculated with additional simulations.

Figure A6. Modeling region used in current simulations of the size/composition distribution of PM in the central and eastern US by the CMU group. The model grid can be adjusted for different applications. High resolution (5x5 km) not shown here will be used in the area surrounding the Pittsburgh Supersite for additional simulations.