Mass spectrometry represents the heart of many of the analytical instruments used for real-time aerosol measurements. Many of the recently developed instruments evolved from instrumentation developed several decades ago made possible by relatively recent technological advances in electronics and computers. Mass spectrometry allows the measurement of the full range of inorganic and organic species within aerosols without requiring sample pretreatment and with very fast time resolution and high sensitivity. Simultaneous measurement of particle size allows the chemical composition as a function of size to be determined as well. Furthermore, single-particle instruments permit the distribution of aerosol chemical components within an aerosol population (mixing state) to be directly measured (e.g., particles that are composed of identical internal mixtures of EC and sulfate can be distinguished from an external mixture of pure EC and pure sulfate particles). The major difference between the mass spectrometers used for characterizing aerosols lies in the method used to produce ions from the aerosol sample, which includes laser desorption/ionization (LDI), thermal desorption followed by electron ionization (EI) or chemical ionization, and two-step LDI.

Laser desorption/ionization offers the advantage that it can be used to desorb, ionize, and detect essentially all aerosol chemical components from an individual particle. Thermal desorption followed by electron ionization can be more quantitative for a limited subset of components (i.e., organic species as a whole, ammonium, nitrate, sulfate) as the ionization occurs after the chemicals have been desorbed from the aerosol matrix. However, EC and refractory components such as mineral dust and sea salt do not evaporate with this method, and thus, this ionization scheme can only measure the more volatile fraction of aerosols. Additionally, thermal desorption instruments typically measure an ensemble of aerosol particles, and thus, single-particle information is not obtained. Thermal desorption followed by chemical ionization has also been developed and benefits from the selectivity and softer ionization of CI compared to EI, which is particularly important for the detection of organic compounds. The ionization method is one of the single most important considerations in interpreting the data collected using real-time mass spectrometric techniques. Particle mass spectrometry is most commonly performed using time-of-flight (TOFMS), but quadrupoles (QMS) and ion traps (ITMS) are also frequently used (23).

Instruments that measure the chemical composition of single particles in real time have been developed by numerous groups over the past two decades. For recent reviews of the development of single-particle mass spectrometers (SPMS), the reader should consult these recent papers (17, 20, 25). These SPMS permit the chemical characterization of aerosols in real time with excellent temporal and size resolution. Most are transportable and PALMS has sampled from an aircraft platform. All instruments described below share some of the same basic design and operational principles. Aerosols are drawn into an inlet and through a series of plates or skimmers to pump away most of the background gas while pulling particles into the instrument. One or two lasers are used to detect the particle's arrival by measuring the scattered light from the particle. Some single scattering laser instruments determine the particle's size from the intensity of scattered light. More commonly, instruments measure the time of flight between two scattering lasers to obtain the aerodynamic diameter of each particle. Subsequently, an ultraviolet laser is fired at the appropriate delay time to intercept each particle as it enters the ionization region, desorbing and ionizing compounds from each particle. The ions produced are then extracted to a flight tube for detection via time-of-flight mass spectrometry. Some instruments utilize dual flight tubes to detect both positive and negative ions from the same particle. These instruments include the aerosol time-of-flight mass spectrometer (ATOFMS), laser mass analyzer for particles in the airborne state (LAMPAS), particle analysis by laser mass spectrometry (PALMS), and rapid single-particle mass spectrometer (RSMS). The ATOFMS is commercially available from TSI, Inc. (Shoreview, MN). The major differences between the various single-particle LDI instruments include the laser wavelength used to perform the LDI and also the method used to determine the size of each particle. Most instruments described in the next section detect particles between roughly 0.2 and 5.0 m (fine and coarse size modes) while the RSMS detects particles from 0.045 to 1.250 m and the UF-ATOFMS detects particles from 50 to 300 nm.

(1) Recent Findings Using One-Step Laser Desorption/Ionization. Instruments employing one-step LDI schemes are used to address a wide range of questions regarding the role of aerosols in atmospheric chemistry involving ambient studies of aerosols, source apportionment and characterization, laboratory studies and heterogeneous kinetics, transformations of aerosol composition during atmospheric transport, and detection of bioaerosols. The use of these instruments for the analysis of ultrafine and organic particles has been recently reviewed (22).

Characterization of Ambient Particle Composition. The primary application of SPMS has been to determine the spatial and temporal variations of the chemical composition of ambient particles as a function of particle size. This information is then used to develop a picture of the regional variability of particles and to conduct source apportionment studies to link the measured particle composition to particle emission sources. Comparisons of the results from three SPMS for ambient measurements conducted during the Atlanta Supersite Experiment are presented in the Instrumental Comparison: Atlanta Supersite Experiment Section.

Fine and Coarse Aerosols. Taking advantage of the high sensitivity and time resolution of SPMS, ambient measurements of aerosols are being performed to increase our understanding of the extent and impact of long-range transport in our atmosphere. Measurements conducted in Mace Head, Ireland, showed strong correlations between particle composition and the history of the air masses as indicated by back trajectories, indicating long-range transport of aerosols from Europe, America, and Africa (56). Dust particles believed to have originated from the Saharan desert showed an aluminum/silicon signature while dust from a local source had a calcium-rich signature. Sulfate was more associated with carbon in the submicrometer mode than the supermicrometer mode. Ammonium was detected on the same particles as sulfate in both the sub- and supermicrometer size modes. Nitrate and ammonium were also associated with one another in the supermicrometer mode, particularly on dust particles.

Owega et al. (57, 58) report the long-range transport of forest fire particles and mineral dust to downtown Toronto. Back-trajectory analysis revealed these to be from Northern Canada and the Saharan desert, respectively. These short-term events occurring on an hourly time scale would likely not be detectable by semicontinuous PM_2.5 or gas-phase measurement techniques as they represent a very small fraction of the ambient air whose signal would be further diluted by long sampling times. This highlights the value of real-time single-particle analysis for the detection of small numbers of unique particle types from specific sources.

Correlations between the chemical composition of the gas and particle phases at a given location are achieved by the use of colocated SPMS and gas-phase measurement instruments. Changes in the composition of ambient particles in Riverside, CA, according to the time of day (diurnal cycle) were detected, and the different particle types were scaled to atmospheric concentrations (46). In the mornings, relatively high levels of supermicrometer mineral dust particles were present, while afternoons were dominated by submicrometer particles composed of organic carbon and ammonium nitrate, which were correlated with elevated gas-phase ozone concentrations. These changes in particle size and composition are consistent with the relative amount of time the particles spent over land as indicated by the air mass trajectories reaching Riverside. This study also provided further evidence that a cut point at 1.0 m would be more appropriate for distinguishing between organic- and inorganic-dominant particle types in urban air.

Attribution of particular particle types in ambient air from specific pollution sources to determine the potential for long-range transport of continental pollution has also been accomplished using SPMS. Outflow of pollution from the Indian subcontinent during the winter monsoon was measured during INDOEX by ATOFMS (59). Gas-phase acetonitrile (a long-lived tracer of biomass/biofuel burning), number concentrations of submicrometer carbon- and potassium-containing particles, and particulate mass concentrations of submicrometer nss-potassium were all highly correlated, suggesting they were emitted from the same or related sources. These measurements are evidence of the long-range transport of emissions from biomass/biofuel burning, demonstrating that pollution from the Indian subcontinent can impact areas downwind of south Asia. Carbonaceous aerosols produced by biomass/biofuel burning accounted for ~75% of carbon-containing particles in areas far removed from sources, up to 7 days according to back-trajectory analysis. Air parcels from the Arabian peninsula were enriched in carbon particles containing potassium and lithium, indicative of coal-burning emissions. Carbon-containing particles that did not contain potassium were also measured in these air masses, indicative of fossil fuel combustion as opposed to biomass/biofuel combustion. These results demonstrate the importance of combining simultaneous gas- and particle-phase measurements to achieve accurate source apportionment of specific particle types.

Measurements in the free troposphere by PALMS over the eastern Pacific Ocean and western coast of North America revealed discrete layers of aerosols (60). Evidence of long-range transport of layers of biomass/biofuel and urban/industrial/crustal aerosols from anthropogenic and biomass burning sources in eastern Asia were found over vertical extents of 0.3-1.5 km and horizontal dimensions of at least 950 km. Substantially enhanced particulate sulfate mass was formed during transport through a mid-latitude cyclonic system by gas-to-particle conversion. Such cloud systems appear to scavenge most existing particle mass while allowing gas-phase precursors of particles to be transported, thus altering the downstream chemical and physical properties of the particles. The presence of high amounts of pure or nearly pure sulfuric acid particles in the lower troposphere was detected at distances corresponding to a two-week transport time from possible sources.

Rapid changes in the size distribution, number density, and chemical composition of aerosols reaching a remote site in Scotland were caused by the arrival of two distinct air masses (61). One was a polluted marine air mass that had passed over land, the other a cleaner marine air mass that had been transported directly from the sea. Particles in the polluted marine air mass had accumulated amines and ammonium as they passed over agricultural areas. There was also evidence of chloride depletion from these marine particles and a similar but less distinct depletion observed for bromine and iodine. The lighter elements lithium, boron, and berillyium, though infrequently detected, were typically found in the same particles, suggesting a new unknown source of these particles.

To determine the density of ambient particles, the scattered light intensity was compared to the aerodynamic diameter of organic and sulfate particles sampled by PALMS above 11 km over Florida (62). There was a distinct relationship between light scattering and geometric diameter when the diameters were recalculated using an assumed density for organic and sulfate aerosols. By sorting particles using optical, aerodynamic, and chemical composition data, the density and thus geometric diameter of individual particles could be calculated. In another example of quantitative measurements made using SPMS, Held et al. measured the vertical flux of nitrate from a forest on a summer day using LAMPAS (63).

Ultrafine Aerosols. The RSMS efficiently detects particles below 300 nm; the most recent generation is described by Lake et al. (64). An aerodynamic lens is used to focus the particles, reducing the divergence of the particle beam and increasing the detection efficiency for ultrafine particles. A 193-nm excimer laser is free fired at the particle beam to generate positive and negative ions, which are detected in dual flight tubes. To determine composition as a function of size, the pressure upstream of the critical orifice is varied to alter the aerodynamic diameter of particles that are most effectively focused and thus ablated by the laser pulse. Particles between 45 and 1250 nm are typically measured in nine size bins by stepping through different discrete sizes of particles.

Ultrafine particles from 14 nm to 1.3 m were measured by RSMS during the Atlanta Supersite Experiment in 1999 (discussed in more detail later) (65). Particles with diameters smaller than 100 nm were dominated by carbon-containing particles. Larger particles had varied compositions but were generally mixtures of organic and inorganic components such as crustal materials and metals. Particles classified as organic and containing significant nitrate accounted for 74% of all the particles detected. These results are very different from those obtained in Houston (66). The sources near the measurement site in Houston were dominated by refineries, chemical plants, and incinerators. The major particle type in Houston contained potassium, silicon, and, to a lesser degree, carbon. Transient plumes of particles were also detected. While the Atlanta measurements were dominated by carbon-containing particles and had a larger diversity of particle types, measurements in Houston had fewer particle classes and were dominated by inorganic species.

Particles from 30 to 1100 nm in Pittsburgh, PA, were dominated by particles composed of carbon and ammonium nitrate (~54% of all particle fit this class) (67). Carbon was detected in ~79% of all particles detected throughout the study, in all size bins and wind directions, indicating numerous sources throughout the area. Metal-containing particles dominated the remainder of particle types observed. By examining the mass spectra, size range, wind direction, and indicators of atmospheric aging, many of these metal particle types could be associated with specific sources in Pittsburgh. The silicon-potassium-iron-gallium particle class was attributed to two coal-fired power plants. Sodium-potassium-zinc-lead particles came from the direction of a company specializing in zinc and other nonferrous metals while the lithium-sodium-potassium-iron classes were attributed to a steel and iron manufacturer housing blast furnaces and steel mills. The chromium-molybdenum-tungsten particle class came from the direction of a plant that manufactures metal alloys using electric arc furnaces. Emissions from a high-temperature furnace explain the very small size of this particle class whose number size distribution peaked at ~75 nm. All the metal particle classes discussed above were found in smaller size bins, from 75 to 300 nm. These results indicate that high-temperature furnaces are the single source of metal-containing ultrafine particles in Pittsburgh.

Measurements of metals in fine and ultrafine particles in Baltimore found that iron and lead were commonly found in the same particles, suggesting a common source (68). Arsenic and lead, however, were found in separate particles, suggesting different sources for these heavy metals, while vanadium is found in all wind directions, likely reflecting fossil fuel combustion as a source of this metal. A instrument similar to the RSMS, the single-ultrafine-particle mass spectrometer, detected particles from 30 to 300 nm in College Station, TX (69). Nitrate was found in most ultrafine particles, likely due to sampling near agricultural areas. Detection of nitrate varied with temperature and relative humidity (RH). An amine particle class was present only during periods of low relative humidity.

Modification of the ATOFMS to increase the detection efficiency and extend the lower size limit to include ultrafine particles (UF-ATOFMS) has been performed (70). Using an aerodynamic lens, the particles are focused into a tightly collimated beam and improvements to the light scattering detection region allow single particles between 50 and 300 nm to be sized and detected. The ability to size and chemically analyze each particle allows the chemistry of the complete size range to be sampled simultaneously without the need to step through different size ranges. The initial results from this instrument for the detection of ultrafine particles in La Jolla, CA, have been reported (70). Characterization of studies of light-duty vehicle particle emissions and measurement of the effect of fine particle concentrators on fine and ultrafine particle composition have been performed with the UF-ATOFMS and are discussed below under Source Characterization and Laboratory Studies. The instrument has also been deployed in field studies in Atlanta, GA, and Rochester, NY, and these results will soon be published.

Probing Atmospheric Aerosol Transformation Processes. Beyond performing ambient characterizations and source apportionment studies of ambient aerosols, SPMS instruments are providing valuable information about physical and chemical changes that particles undergo as they are transported through the atmosphere. These can be the result of physical processes such as coagulation with other particles, scavenging and subsequent chemical processing by cloud/fog droplets, and heterogeneous reactions with trace gases. Along with the excellent size and temporal resolution provided by SPMS, the ability of these instruments to measure the mixing state of various components in individual particles is a key factor that allows important transformation processes to be elucidated from ambient measurements. The observations that are described below help to both confirm predictions based on laboratory and modeling studies of atmospheric processes and give rise to new questions that warrant further investigation.

Atmospheric Aging of Particles. During a flight over Utah, the PALMS instrument intercepted a forest fire plume during the ITCT mission (71). Large increases within acetonitrile and CO concentrations and particle number concentrations were observed in the plume. The particles contained carbon, potassium, ammonium, and organics, but no pure soot particles were observed in the plume. The mass spectra of these particles were used to identify forest fire-produced particles detected at other times and locations during this and other campaigns when gas-phase measurements were not always available for comparison. During the ITCT, 33% of particles sampled in the North American troposphere and 37% of particles transported from Asia were identified as biomass burning particles. In other missions, ~7% of stratospheric particles were identified as originating from biomass burning. During CRYSTAL-FACE over the Florida area, the number of stratospheric particles from biomass burning increased to 52% due to strong influences from active fires in Colorado and Canada that were transported to the stratosphere in this region. The presence or absence of nitrate or sulfate on these particles can be used as an indication of atmospheric aging. For example, typically particles in fresh plumes have lower peak areas for sulfate than aged plumes do. The biomass-burning particles detected in the stratosphere in the forest fire plume during CRYSTAL-FACE had essentially no sulfate. This suggests that SO_2(g), the precursor to particulate sulfate, was not available to be taken up on the particles in the stratosphere. As the biomass plume was very dense and little mixing occurred in the stratosphere, any SO₂ that was present originally in the plume would have been widely distributed and thus diluted over a large number of biomass particles. Other biomass-burning particles measured in the stratosphere but not in the forest fire plume did contain large sulfate signals, suggesting that these particles were transported to the stratosphere more slowly and thus had the opportunity to accumulate sulfate during transport.

Two types of ultrafine nitrate particle events were detected in Baltimore in 2002 by the RSMS: large nighttime bursts of "pure" nitrate particles from 50 to 90 nm and a smaller and less frequent daytime burst of "pure" particles from 50 to 90 nm that grew to 110 to 220 nm with time (72). During these events, the number of other particle types that were mixed with nitrate also increased, likely due to the condensation of ammonium nitrate onto preexisting particles. Virtually all particles larger than 300 nm contained some nitrate at all times, and during these nitrate events, all particle types and sizes contained nitrate. These events were observed to occur when ambient temperatures were low and relative humidity was high, conditions that favor the partitioning of ammonium nitrate to the particulate phase. Ammonium was rarely detected in ultrafine particles due to a low relative sensitivity factor, which results in a low ion signal for ammonium compared to other secondary species such as nitrate, making detection more difficult.

The air mass sampled by ATOFMS during ACE-Asia reaching the R/V Ronald Brown in the Sea of Japan showed dramatic differences before and during a major dust event (73). Before the dust front, 50-60% of supermicrometer particles containing nitrate were aged sea salt particles that had been depleted of chloride. During the dust event, 60-80% of supermicrometer nitrate was associated with mineral dust that had variable NO₃^-/Al mass ratios, suggesting that the nitrate is a secondary product formed by the uptake of acids by dust. This agrees with the rapid uptake of nitric acid on dust observed in laboratory studies (74). Further evidence for the uptake of nitric acid by dust was found in measurements made with LAMPAS-2 at Jungfraujoch, Switzerland, a site in the free troposphere influenced by remote continental and marine air masses (75). Detection of nitrate on mineral dust particles without the presence of ammonia suggests that nitric acid could have directly reacted with the dust.

Hydroxymethanesulfonate (HMS) was detected by ATOFMS in particles measured in Bakersfield, CA, during periods of fog dissipation (76). This species serves as an indicator that these particles had undergone fog processing that produced HMS via oxidative reactions occurring in the fog droplets. The HMS remained in the particles as they evaporated during fog dissipation. Over 90% of particles with HMS also contained (organic and elemental) carbon, ammonium, nitrate, and sulfate. The presence of HMS and EC in the same particle was unexpected based on the typical hydrophobic nature of EC. These particles were found to also contain OC, providing direct evidence for fog processing of aged EC particles, which then accumulate OC, HMS, and other secondary species, likely as a coating around the EC core. Particles containing HMS were not detected after more than 12 h after the fog dissipated, indicating that HMS is a short-lived marker for fog processing and may not be detected by filter samples, which are often collected for over 24 h.

Characterization of Particle Ice Nucleation. A method that separates particles which freeze to form ice particles from those that do not has been developed, permitting the analysis of the dependence of ice formation mechanisms on particle composition by coupling it with PALMS (77). A continuous flow diffusion chamber (CFDG) allows the control of the temperature and relative humidity that the ice nucleation occurs under. A counterflow virtual impactor then permits only those particles that have grown by freezing to enter the PALMS's inlet for detection. Using this method, the chemical compositions of particles that froze were measured at a stationary site in Mt. Werner, CO, which is frequently exposed to air masses from the free troposphere. Ice nuclei that underwent heterogeneous freezing were dominated by mineral dust/fly ash and metallic particles (78). This suggests an important anthropogenic influence on ice nuclei formation. Background aerosols that froze homogeneously were dominated by sulfate and did so at temperatures less than -38 C and at relative humidities expected for sulfate aerosols. Organic compounds internally mixed with sulfate aerosols appeared to inhibit ice nucleation, requiring higher relative humidities for freezing to occur than for nominally pure sulfate aerosols. Further evidence that particles with large organic signals are inefficient at ice nucleation is provided by Cziczo et al (79). In measurements using the CFDG method described above to study the freezing properties of ambient aerosols and in direct measurements of actual cirrus ice crystals in the upper troposphere during CRYSTAL-FACE, particles with large organic signals preferably remained unfrozen. There was an unequal partitioning of the organic aerosol component between the frozen and aqueous phases. Using the CFDG, particles that froze readily had almost no organic content, while those with a strong organic signal only froze at higher water saturations. The mechanism that causes this decreased freezing efficiency for organic particles is unknown and warrants further investigation. Also, the molecular identification of the organic compounds that caused this effect could not be determined due to excessive ion fragmentation. It is quite possible that only particular types of organic compounds produce this effect, and thus the use of another SPMS instrument, which allows one to further probe the organic fraction, is merited.

A counterflow virtual impactor was used to separate particles that had frozen in the upper troposphere/lower stratosphere to form cirrus from the "interstitial" background aerosol, which did not freeze during CRYSTAL-FACE (80). Sea salt was often a component of cirrus indicating a homogeneous freezing mechanism had occurred. Large particles preferentially froze over small particles. During a dust transport event from the Sahara, mineral dust was the major component of ice residue indicating a heterogeneous freezing mechanism occurred on the dust particles. These dust particle residues did not contain nitrate or sulfate signals; however, dust coated with sulfate is the typical system used in laboratory studies of heterogeneous ice nucleation (81). It appears that both mechanisms, heterogeneous and homogeneous freezing, can occur in the same geographical locations and at the same time of year. Meteoritic material was frequently found in ice particle residues, highlighting the impact that stratospheric particles can have on tropospheric cirrus cloud formation.

Source Characterization and Laboratory Studies. In addition to the wealth of field studies being conducted with SPMS instruments as discussed above, these instruments are also providing valuable information about atmospheric processes through laboratory experiments of aerosol chemistry relevant to the atmosphere. Many of these laboratory studies are motivated by findings from field measurements made using some of the state-of-the-art instruments discussed in this review. Rudich's review of laboratory investigations on organic aerosol transformations includes some results from particle mass spectrometers and presents the current research issues in this field (29). SPMS instruments are also used to characterize the composition of particles emitted from particular sources so that these fingerprints can be used to identify the sources of similar particles detected in ambient measurements.

Kinetics, Products, and Mechanisms of Heterogeneous Reactions. The kinetics of wet sea salt particles between 100 and 220 nm reacting with nitric acid in a flow tube was determined by RSMS (82). Quantification of the nitrate and chloride peaks was performed by analyzing standard solutions of NaCl/HNO₃ to generate a relative ion response for these species. The reactive uptake coefficient of nitric acid was found to increase linearly with increasing particle diameter and agreed very well with previous experiments that used EI-MS to measure changes in the gas-phase nitric acid concentrations (83). It was concluded that the reactive uptake of nitric acid was limited by the rate of formation of HCl_(aq) in the droplet that then partitions to the gas phase.

There is a great deal of interest in the kinetics and mechanisms that produce SOA and also in the identification of SOA in ambient aerosols (26). A few experiments using SPMS to study this chemistry have been published, and many more will likely be published soon as new techniques and intense research in this area provide further insights. SOA formation is also investigated using several other instruments discussed elsewhere in this review. The formation of oligomers from the photooxidation of 1,3,5-trimethylbenzene with NO_x in the absence of seed aerosols has recently been reported (84). Their particle characterization method involved collecting particles on a steel plate via an impactor and subsequent analysis without pretreatment by LDI-TOFMS. While this method is not strictly on-line or single-particle analysis, the information obtained is similar to that from SPMS instruments discussed throughout this section. A series of peaks from 400 to 900 m/z separated by mass differences of 14, 16, and 18 Da indicated the formation of oligomers in the particles formed by homogeneous nucleation. The intensity of the higher mass peaks increased as the reaction was allowed to continue for up to 20 h. Oligermerization in organic aerosols had been studied previously by Jang et al. (85), but they used inorganic acid seed aerosols to induce heterogeneous nucleation of organic aerosols and catalyze the hydration and polymerization reactions by the acidity in the seeds. The formation of oligomers in organic aerosols is significant as it will increase the partitioning of organic species to the gas phase beyond that predicted by thermodynamic models and also change the chemical composition and thus chemical and physical properties of organic aerosols (2, 85).

The first observations of SOA formation in a smog chamber using SPMS were obtained by ATOFMS for the photooxidation of secondary and tertiary alkylamines (86). The particles that formed contained peaks corresponding to alkylammonium salts and other N-containing organic compounds. A peak at m/z +86 was found to be a good general marker for amines, as suggested by previous laboratory studies (87). These peaks were then used to identify amines in ambient particles measured in Atlanta and during a vehicular exhaust source study (88) that commonly contained similar peaks in their mass spectra. This permitted the first detection of the oxidation products of alkylamines in the particulate phase of ambient aerosols. The formation of potentially toxic imines from the reaction of particles containing amines with gas-phase aldehydes has been reported by Haddrell and Agnes (89). Glycerol, sea salt, or salt particles containing 1,8-diaminonaphthalene and suspended in an electrodynamic balance (EDB) were exposed to various aldehydes and then impacted on a steel plate for analysis by LDI-MS one particle at a time. Imines were detected as the reaction products in both wet and dry particles. By plotting the normalized peak area of the imine product as a function of reaction time, the pseudo-first-order rate constant for the reaction of benzaldehyde with glycerol particles spiked with 1,8-diaminonaphthalene was determined. Using this value in an atmospheric chemistry model indicated that the rate of imine formation from amines is comparable to the loss of amines from reaction with O₃ or OH under tropospheric conditions. These results suggest that imine formation in particles is likely under certain realistic tropospheric conditions. The authors suggest that environments with high concentrations of both amines and aldehydes, such a cigarette smoke, are good candidates for imine formation. As the potential toxicity of particle-bound imines remains under investigation, this reaction could play an important role in determining the health effects of aerosols. Furthermore, the formation of imines will tend to decrease the hydrophilicity of organic compounds in aerosols, in contrast to most atmospheric aging processes, which oxidize and thus increase the hydrophilicity of organic aerosol components.

Characterization of Aerosols during Health Effects Studies. The health effects occurring from exposure to ultrafine and fine aerosols remains an important and poorly understood topic (13-16). Studies of the health effects of aerosols typically require that the particle number be concentrated so that acute effects of the aerosols on exposed humans can be observed. To achieve this, particle concentrators employing virtual impactors are typically used. However, the small sizes of fine and ultrafine particles preclude the normal use of a virtual impactor for concentration of submicrometer particles. The particles must first be grown by the condensation of water under highly supersaturated conditions to supermicrometer sizes before being concentrated by the virtual impactor and then dried back to their original size. This process involves the creation of an aqueous phase around/in the particles that could alter the chemical composition of the ambient particles used in exposure studies. As the chemical composition of particles is believed to be a principal factor determining their toxicity, any changes in the chemical composition of the particles caused by the concentrators must be determined.

A PALAS generator is commonly used to create "sootlike" EC particles for use in particle exposure health studies. The UF-ATOFMS characterized the chemical composition of PALAS-generated fine and ultrafine (50-300 nm) soot to assess the variability and reproducibility of the chemical composition of these model aerosols (90). Highly reproducible single-particle mass spectra were obtained; over 96% of particles produced from pure graphite (¹²C) rods contained a distinct carbon fragment (C_1-3) ion envelope. Particles generated from ¹³C graphite rods contained more OC than the ¹²C-generated particles. This is likely due to different surface properties of the two particle types and the greater adsorption of organic vapors from the particle-free dilution air by the ¹³C particles. Homogeneous particles were also generated with Fe-¹²C and welding particles with almost all (92 and 97%, respectively) of the spectra containing reproducible Fe/¹²C and Fe/Mn/Cr ion ratios in their respective individual particle mass spectra.

The first report of changes in the chemical composition of fine (100-300 nm) and ultrafine (50-100 nm) particles caused by passing through a fine particle concentrator has been made by Su et al. (50) using the UF-ATOFMS to characterize the single particles in real time. Fine ambient particles were concentrated using various versions of the Harvard Ultrafine Concentrated Ambient Particle System (HUCAPS) (91) and the Versatile Aerosol Concentration Enrichment System (VACES) (92) in urban and rural sampling sites in the eastern United States. There was a significant reduction in the number of EC particles relative to mixed EC-OC and OC particles after particle concentration. Concentrated ultrafine particles showed a 30% increase in the amount of OC on EC particles for the same particle size. The number of aromatic and polycyclic aromatic hydrocarbon (PAH)-containing fine and ultrafine particles also increased after concentration. These changes can be explained by partitioning of (water-soluble) organic compounds present in the ambient gas phase to the liquid phase when the particles are grown by the condensation of water. Repartitioning of OC from larger to smaller EC particles during this growth process could also be occurring in the concentrator. The single-particle spectra showed ion markers indicative of sulfur chemistry occurring in the morning hours, which can be explained by partitioning of SO_2(g) to the dilute water particles, analogous to cloud and fog processing of aerosols observed in the troposphere. The changes in the chemical composition caused by fine-particle concentrators must be considered when such systems are employed for aerosol exposure health effects studies. In contrast, no composition changes were observed using ATOFMS to sample ambient coarse particles (>2.5 m) before and after a coarse particle concentrator (that does not require particle pregrowth with water) (93). There was no change in the 10 major particle classes upstream and downstream of the concentrator, and no new particle classes were detected after passing through the concentrator.

Source Apportionment and Characterization. As discussed in the Fine and Coarse Aerosol section, unique marker combinations or ion patterns in the mass spectra of specific particle types can be used to apportion ambient particles to particular sources. To achieve this, the unique fingerprints of particles from specific sources must be obtained under controlled conditions. The emissions from 28 light-duty gasoline vehicles were characterized using the UF-ATOFMS for particles between 50 and 300 nm (94). These vehicles represented a wide range of catalytic converter and engine technologies (1953-2003 model years). In the ultrafine mode (50-100 nm), EC particles combined with calcium, phosphate, and sulfate were dominant (close to 85% of particle number), with a lower abundance of OC. The most likely source of calcium and phosphate in the particles is from lubricating oil. The relative amounts of EC particles decreased with increasing size, and OC particles became more prevalent above 100 nm. The fast time resolution of the ATOFMS permitted the measurement of the composition of the emissions under several driving conditions on the dynamometer and investigation of the effect of cold and hot engine starts. The first 2 min of the cold engine start produced more than five times the number of particles than any other driving cycle. These particles were dominated by ultrafine EC with calcium, phosphate, and OC. The dominance of EC particles by number, particularly from the newer, lower emitting gasoline vehicles, suggests that EC by itself is not a suitable unique tracer for diesel emissions. These results are being used as size-resolved mass spectral signatures for source apportionment of ambient measurements made by UF-ATOFMS.

Kirchner et al. compared diesel- and graphite spark-generated soot particles to evaluate the use of graphite-generated particles as a proxy for diesel soot in laboratory studies (95). Key differences between the mass spectra of these two soot particle types were observed, particularly for the positive ions. The positive mass spectra of diesel engine particles were quite simple with strong signals from Na⁺, K⁺, and Fe⁺. Spark-generated particles contained many organic fragments in their positive mass spectra such as ⁴³[C₂H₂O]⁺, ⁵⁵[C₃H₃O]⁺, and ⁷³[C₃H₅O₂]⁺ as well as inorganics such as Na⁺ and K⁺. The negative mass spectra of diesel particles had major peaks from CN^-, ⁴³[C₂H₃O]^-, NO₂^-, C₅^-, C₆^-, and C₇^- carbon clusters and HSO₄^-. The spark-generated particles' negative mass spectra were different, however, with principal peaks from Cl^-, CN^-, NO₂^-, and NO₃^-, as well as carbon clusters C₅^-, C₆^-, and C₇^- and a small HSO₄^- peak. The key differences appear to be that diesel particles contain oxygenated carbon fragments in the negative spectra while spark-generated particles have these fragments in their positive spectra while their negative spectra are dominated by Cl^-. These two different types of soot particles were also coated with the reaction products from the oxidation of -pinene with O₃. The mass spectra of the diesel- and spark-generated particles showed major differences in the peaks that grew in after the particles were coated with the low-volatility reaction products, particularly for the negative ions. These differences could be caused by a matrix effect that produces different ionization and fragmentation patterns of the same species depending on the surface they are adsorbed to. The differences between these two types of soot were also confirmed by QMS-SIMS, TOF-SIMS, and XPS. For example, XPS found a more pronounced layer of acid derivatives of -pinene on the diesel soot than the spark-generated soot. These results indicate that spark-generated soot particles have different surface compositions than diesel soot particles, and the authors suggest caution should be exercised in using spark-generated particles as proxies for diesel engine soot in laboratory experiments.

The relative contribution of diesel particles to total particles was estimated for various sites in Germany by Vogt et al. by attributing particles classified as EC and EC + secondary species to diesel particles (55). The relative contribution of diesel particles to 0.5-m particles was 23% at a rural site, 35% beside an Autobahn, and 10-35% at an urban site impacted by heavy-duty construction vehicles.

Soot particles from combustion were collected at two different heights from an incineration plant burning wood and then resuspended for on-line analysis (96). A PAH signature was detected in the m/z 250-500 range, as observed by laser microprobe mass spectrometry. Using reduced laser fluence, ATOFMS could detect nonfragmented organic species in particles. These results confirmed that particles sampled from the bottom of the first air supply zone are PAH rich while those sampled at the third air supply zone contain no detectable PAHs as they have been oxidized.

Mainstream cigarette smoke aerosols were analyzed by ATOFMS and produced positive mass spectra with peaks at lower masses formed from the fragmentation of PAHs (97). Negative ions contained carbon clusters, CN^-, and nitrate fragments. This supports a reaction mechanism wherein NO is adsorbed by soot and reacts with methanol, producing nitric acid and methyl nitrite, whose ion signals increased as the smoke aerosol ages. The molecular ion for nicotine, however, decreased with time, reflecting its volatile nature.

On-Line Detection of Bioaerosols. The rapid real-time detection and speciation of aerosols containing or composed of biological organisms (bioaerosols) is of major importance for national security and military concerns over attacks involving the potential release of bioweapons. Naturally emitted bioaerosols also represent a potentially important but as yet poorly characterized component of the atmosphere, possibly playing roles in cloud formation and chemical transformations occurring in the atmosphere. They have been detected throughout the troposphere and also in the stratosphere (98).

The real-time and reagentless detection of individual airborne cells with ATOFMS and analysis using a modified clustering program has been reported (99). This method was able to distinguish two species of Bacillus spores from one another and also from a variety of background mixtures of powders, soil, and fungal spores by matching the mass spectra with fingerprints taken from pure samples. Using the same method, Russell et al. (100) report an increased detection efficiency for biomolecules when a light-absorbing chromophore is present such as an aromatic system. A series of one-component solutions of amino acids, peptides, dipicolinic acid, and mixtures of these were analyzed and produced protonated and deprotonated molecular ions in the positive and negative mass spectra, respectively, and numerous fragmentation and neutral loss products of these ions. The detection efficiency of nonabsorbing compounds mixed with absorbing aromatic compounds such as dipicolinic acid was higher than pure solutions of these nonabsorbing species. This enhancement produced by absorbing components in the matrix may explain why it is possible to detect nonabsorbing biomolecules in the core of Bacillus spores as the core contains large amounts of dipicolinic acid. The ions observed in both polarities were a product of the acidity and basicity of the biomolecules in the mixture, evidence for competitive protonation/deprotonation occurring in the ion plume similar to that observed in matrix-assisted laser desportion/ionization experiments.

The addition of 3-nitrobenzyl alcohol to particles containing biomolecules produced mass spectra with an intact molecular ion and little fragmentation (101). The particles, containing erythromycin and gramicidin S, were passed through a saturator and condenser, which added the liquid alcohol matrix to the particles, before undergoing LDI by a 351-nm laser and detection in a two-stage reflectron TOF-MS. Mass spectra of the same particles without the added matrix, however, did not produce mass spectra containing any high-mass peaks.

(2) Two-Step LDI and Other Instrumental Developments in SPMS. A major area of recent development in SPMS has been the use of two laser pulses to decouple the desorption and ionization steps of the LDI process. The common principles of two-step LDI involve the desorption of chemicals from the particles using an infrared laser pulse and then ionization of the resulting vapor plume with an ultraviolet laser pulse a short time delay after the IR laser is fired. This technique was first used for on-line single-particle analysis by Morrical et al. (102) and has been further developed by the Baer and Prather groups. By separating the desorption and ionization steps, reduced matrix effects on the ionization and fragmentation patterns are observed and depth profiling (103) of the chemical composition of the particle can be performed. Most importantly, the laser power required for the second ionization laser is much lower than that required for one-step LDI. This results in reduced fragmentation of fragile organic compounds and an increased signal from their molecular ions, permitting the identification of organic aerosol components on a molecular level and improved detection limits (22, 104, 105).

The reaction kinetics of 680-nm to 2.45-m pure oleic acid particles with ozone was measured using two-step LDI of single particles and TOFMS (104). The reaction was followed by monitoring changes in the signal from oleic acid's molecular ion, which represents ~15% of the entire ion signal due to reduced fragmentation under two-step LDI conditions. The increase in the ion signal for the molecular ion of a major reaction product, 9-oxononanoic acid, could also be monitored. Reactive uptake of ozone was found to decrease with increasing particle size. It is suggested that this is due to the rate of diffusion of ozone within the oleic acid particle limiting the reaction kinetics. This diffusion limitation could explain why the lifetime of oleic acid observed in ambient particles is much longer than that predicted by previously known kinetics.

Mass spectra of particles containing pesticides produced by two-step LDI have a more intense molecular ion and less fragmentation, thus providing lower detection limits for these toxic pollutants (105). The laboratory-generated particles were analyzed using the usual one-step LDI process at 266 nm as well as the two-step LDI using a 10.6-m CO₂ desorption laser followed by a 266-nm Nd:YAG laser pulse 0.6 s later to ionize the desorbed vapor. Detection limits in single particles using two-step LDI range from less than 1-15 amol. This is significantly lower than detection limits of GC/MS or HPLC methods typically applied for the quantification of pesticides in environmental samples.

The internal energy of the molecules in the vapor plume decreases with time, and thus, the degree of fragmentation can be reduced by increasing the time delay between the firing of the desorption and ionization lasers (106). The degree of fragmentation of oleic acid aerosols was decreased by using a lower energy ionization laser. Vaporization of the organic particles with a heater followed by laser ionization showed more uniform heating of the particle compared to two-step LDI and little fragmentation. However, the total ion intensity was lower with the heater as it cannot be placed very close to the extraction plates because it would distort the electric fields. Increased fragmentation of oleic acid is observed by delaying the ion extraction and thus providing the ions more time to undergo fragmentation within the ionization region.

The large size of the CO₂ laser typically precludes the use of two-step LDI in transportable SPMS instruments, though Hauler et al. (107) describe a mobile instrument that uses two-step LDI and REMPI-TOF detection to measure gas-phase and particulate aromatic compounds. A unique source allows rapid switching between gas- and solid-phase detection. Solid samples are introduced on the surface of a sample probe for desorption by the infrared CO₂ laser followed by ionization of the vapor plume by a 266-nm laser pulse. Single photon ionization of aliphatic compounds using a 118-nm laser was also achieved. Another transportable instrument that detects PAHs by two-step LDI-TOFMS first collects particles in three size bins on a strip of aluminum foil and then automatically advances the strip into the instrument for laser analysis (108). Though this method does not detect single particles or operate in real time, both the sampling and analysis are completely automated, can achieve 20-min time resolution between samples, and can operate for up to 3.5 days. An important finding was the lack of adsorption of gas-phase compounds to the aluminum foil, which would have caused a positive sampling artifact.

Another newly developed transportable instrument collects particles of less than 300 nm, which are focused by an aerodynamic lens onto a probe in the source region and then ionized using two-step LDI with detection by a reflectron TOF-MS (109). Photoionization aerosol mass spectrometry (PIAMS) is focused on measuring organic aerosol components, which are typically enriched in submicrometer particles. The small size and thus smaller mass of the particles can make detection of single particles using two-step LDI difficult as fewer ions are produced. By first collecting the fine and ultrafine particles on a probe, the detection limit can be increased but single-particle analysis is sacrificed. Samples were typically collected for 1 min and then immediately analyzed. The 1064-nm fundamental of a Nd:YAG laser was used for the desorption laser pulse. As most compounds are transparent at this wavelength, vaporization is thought to occur by the rapid heating of the metal probe's surface. The signal intensity was found to be independent of particles sizes between 80 and 250 nm; however, the molecular ion signal of oleic acid particles is not as high as that produced by two-step LDI of single particles (110). Relative quantification of oleic acid and pyrene in two-component particles was also possible. The reduced fragmentation caused by two-step LDI allowed PIAMS to distinguish between organic aerosols in meat cooking, wood burning, diesel, and gasoline exhaust. Current detection limits are in the 5-50-pg range with 1-min sampling. To detect compounds found at typical atmospheric levels, the sensitivity must be increased by a factor of 100 and will be addressed in future designs.

Large differences in the mass spectra obtained by SPMS for identical particles are primarily caused by inhomogeneities in the laser pulse used in one-step LDI due to "hot spots", which are localized regions of very high laser power densities (49, 111). As the micrometer or smaller sized particles are much smaller than the laser spot, each particle can receive dramatically different laser energies depending on where it is located in the laser beam. The use of a homogeneous flat-top laser beam should eliminate much of the variation observed between detected particles and has been employed by Wenzel and Prather for the ATOFMS (112). Using a fiber optic to transmit the laser power at 266 nm from a Nd:YAG solid-state laser and produce a homogeneous flat-top laser spot for LDI reduced the relative standard deviation of the total positive ion signal from 110 to 31% for 2,4-dihydroxybenzoic acid (DHB) particles. Furthermore, only one particle class was obtained by the ART-2a clustering algorithm for vigilance factors up to 0.86 for identical DHB particles detected using the fiber optic. The same particles produced ~25 different particle classes when the fiber optic was not employed. The use of fiber optics to homogenize the laser pulse energy is an important step toward increasing the quantitative abilities of SPMS.

To investigate the chemical composition of particles as a function of their hygroscopicity, a humidified tandem differential mobility analyzer (25) selected particles according to their size and growth in the humidified region before being sent to the SPLAT-MS for analysis (113). This utility of this method was demonstrated by separating an external mixture of hygroscopic ammonium nitrate (AN) particles and polystyrene latex (PSL) spheres with variable coatings of AN. The particles were size-selected with the first DMA, passed through a humidified region where hygroscopic particles grow by water uptake, and then the resulting size distribution was scanned with a second DMA. AN particles had significant growth factors due to the high water solubility of NH₄NO₃. PSLs with AN coatings grew to a lesser extent, and the amount of growth increased with increasing thickness of the water-soluble AN coating. The second DMA then selected the center of each particle size mode for analysis by SPLAT-MS. The mass spectra confirmed that the more hygroscopic particles were AN with no carbon features as seen for PSL particles. The peak area from NO⁺ produced by NH₄NO₃ was correlated with the thickness of the AN coating on the PSLs, and this signal was thus largest for the most hygroscopic coated PSLs. The results from this method for the analysis of ambient aerosols on Cheju Island, South Korea, during ACE-Asia were also briefly reported. Ambient 250-nm particles were selected and humidified, and then the more and less hygroscopic modes, with growth factors of 1.4 and 1.1, respectively, were analyzed as above. The more hygroscopic particles were internal mixtures of sulfate and organics and of sulfate-enriched sea salt particles. Roughly 82% of these 61 particles contained water-soluble compounds. The less hygroscopic particles were primarily crustal in nature, having signals from K, Al, Fe, Si, and Na, indicating they were primarily composed of insoluble mineral components. The authors conclude that it is the relative amounts of soluble and insoluble components that determines a particle's hygroscopicity. These preliminary results demonstrate that this system's sensitivity and time resolution are suitable for on-line field measurements of ambient particles as a function of their hygroscopic growth.

The first use of photoelectron resonance capture ionization (PERCI) for organic aerosols has been reported by LaFranchi et al., and the procedure produces soft and sensitive ionization by the attachment of low-energy photoelectrons to organic analytes (114). PERCI-MS was used to study the products of the ozonolysis of oleic acid aerosols in real time by using a tunable UV laser to generate photoelectrons (<1 eV), which ionize the vapors produced by the particles as they impact a heated filament. The molecular ion for oleic acid in unreacted particles was nearly 100% of the total ion signal, demonstrating the high sensitivity and minimal fragmentation produced by this ionization method. Relative decreases in the oleic acid ion signal accompanied by increases in ion signals from the oxidation products were observed as the reaction time increased. The detection of products with masses greater than that of oleic acid suggests reaction pathways in addition to those previously proposed, such as the formation of secondary ozonides. Quantification is not yet possible as characterization and reproducibility studies of the vaporization and PERCI sources are still required.

(1) Ambient Measurements. The thermal desorption chemical ionization mass spectrometer (TDCIMS) is capable of measuring the chemical composition of ambient aerosols from 6 to 20 nm using chemical ionization with on-line sampling and 20-min resolution (115, 116). Ambient ultrafine particles are collected for 6 min by first charging and then electrostatically precipitating them onto a collection wire. The wire is inserted into the source region of the CIMS, and chemicals are desorbed from the wire by heating it to 300 C. Cryogenic N₂ is used to form H₃O⁺, O^2-, and CO^3-, and their higher clusters for ionization of the vapors, and either the positive or negative ions are then detected by a QMS. Time limitations created by selective ion monitoring only permit the detection of six different ions while achieving adequate characterization of the pulse of ions produced when the wire is heated. Preliminary ambient measurements of 6-20-nm particles in Boulder, CO, found NH₄⁺ to be the most dominant positive ion while NO₃^- and HSO₄^- were the dominant negative ions. However, volume-normalized concentrations of ammonium, nitrate, and sulfate were anticorrelated with ultrafine aerosol number concentrations. During peaks in aerosol loading, sulfate accounted for only 11% of the total observed aerosol volume. This and other results indicate that sulfate, nitrate, and ammonium are not a significant fraction of particles detected during time periods with high levels of total aerosols. Numerous other peaks with smaller ion signals were also detected but have not yet been identified.

Measurements of particle nucleation events in Atlanta using the TDCIMS produced much different results with the new particles being completely composed of ammonium sulfate (117). A new inlet that combines the size-selected aerosols from three Nano DMAs connected in parallel allowed the TDCIMS to collect enough size-resolved particle mass at 6 or 13 nm while maintaining 10-min time resolution. Although much effort was spent trying to detect other chemical components of the new particles, no species besides ammonium and sulfate were observed in measurable quantities, not even nitrate. Prior to the nucleation events, the particle volume fraction of ammonium sulfate was low, ~0.15-0.6, but this rapidly increased up to values of 1 during these events, indicating that the new particles were essentially pure ammonium sulfate. Simultaneous measurements of the aerosol hygroscopcity and volatility made by a Nano TDMA system also suggested the particles were composed of ammonium sulfate.

Gas-phase and particle-phase nitric acid concentrations in cirrus clouds were directly measured using the NOAA CIMS (118) during CRYSTAL-FACE (119). A two-channel inlet system on the aircraft was used to measure gas- and particle-phase HNO₃ from the forward-facing inlet and only gas-phase HNO₃ from the downward-facing inlet. Heated tubing evaporated any condensed-phase HNO₃ and then mixed it with SiF₅^- reagent ions for detection with a QMS. The fraction of condensed-phase HNO₃ increased strongly with increases in particle surface area density. Maximum levels of condensed-phase HNO₃ exceeded the gas-phase equivalent of 0.8 ppbv during some cirrus events, and the percentage of total HNO₃ in the condensed phased ranged from near zero to 100%. A mean surface coverage of HNO₃ was determined to be 1.9 × 10¹³ molecules/cm³ but coverages as high as 1.4 × 10¹⁴ molecules/cm³ were occasionally encountered. Based on these findings and the large diameters of cirrus cloud particles encountered during CRYSTAL-FACE, the redistribution of HNO₃ from the gas to particle phase could be very effective in some clouds in the upper troposphere.

(2) Laboratory Studies. The aerosol CIMS permits the on-line detection of organic aerosols in seconds by thermally vaporizing the particles in a heated tube as they are drawn into the mass spectrometer (120). Varying the temperature of the inlet allows the composition of the particles as a function of volatility to be explored. Chemical ionization of the organic vapors has been performed using a variety of positive and negative reagent ions including low-specificity ions such as NO⁺ and protonated water clusters, and more specific ions including H⁺(CH₃OH)₂, O₂^-, SF₆^-, and F^-. A linear ion response to the mass of organic aerosol was found, and detection limits for pure oleic acid particles were (1.0-10) × 10⁹ and (4.5-10) × 10⁹ molecules/cm³, obtained with the H⁺(H₂O)₂ and O₂^- reagent ions, respectively. The lower detection limit corresponds to ~10 particles/cm³ for 1-m particles, typical of particle concentrations in urban areas. This instrument is well suited for laboratory experiments and will provide unique insight into the organic species in aerosols. However, ambient organic aerosols are typically complex mixtures of numerous organic compounds all individually present in small amounts, and this must be taken into consideration if such an instrument is to be used for ambient measurements.

The reaction kinetics and products of several different pure and mixed organic aerosols with ozone have been reported using aerosol CIMS (121, 122). The first direct evidence for the reaction of Criegee intermediates with oleic acid was found and explains previous disagreements between the kinetics of this reaction measured by monitoring the gas phase (123, 124) versus aerosol phase (104, 125); the latter two studies are discussed under Two-Step LDI and Other Instrumental Developments in SPMS and Thermal Vaporization with Electron Ionization, respectively. The particle-phase reaction identified by Hearn et al. (122) causes a faster loss of oleic acid, which produces faster observed kinetics when the rate is determined by monitoring changes in [oleic acid]_particle by the particle mass spectrometers. Measuring the kinetics by following changes in [ozone]_gas however will obtain the true rate of ozonolysis of oleic acid. Hearn et al. found the reaction of oleic acid with Criegee intermediates to account for 36% of the loss of oleic acid in the particles. By correcting for this additional loss, the calculated true reactive uptake for ozone is in excellent agreement with those obtained by the gas-phase flow tube techniques of Thornberry and Abbatt (124), and Moise and Rudich (123). Hearn et al. (122) conclude that the reaction of ozone with pure oleic acid occurs solely on the particle's surface and attribute this to the inability of ozone to penetrate through the proposed highly ordered structure of oleic acid. The reaction of ozone and methyl oleate, however, occurs in the bulk and can be explained by the reduced molecular order expected in these particles as methyl oleate cannot form hydrogen-bonded dimers while oleic acid can. This series of results is a powerful demonstration of the important and unique information that can now be obtained using state-of-the-art particle mass spectrometers for heterogeneous reactions.

The reaction products in SOA generated from the oxidation of biogenic organic vapors by ozone in the laboratory were determined by directly injecting the aerosols into the ion source of a commercial ion trap MS (126). Negative-ion atmospheric chemical ionization was employed to detect the particulate products of the ozonolysis of gas-phase -pinene and limonene with a 1-s time resolution and low-pptv detection limits. Highly oxidized carboxylic acids were detected including high-mass multifunctionalized compounds whose detection had not previously been reported. Detailed reaction schemes that rationalize the observed products were developed using information generated by collision-induced dissociation (MSⁿ) and the subsequent loss of neutral fragments such as CO₂, H₂O, and C₂H₂O₂ from the major product ions.

(1) Source Characterization and Ambient Aerosol Measurements. The TDPBMS measured the composition of nanoparticles (D_m < 50 nm) and accumulation mode particles emitted from the tailpipe of a heavy-duty diesel truck (129). By comparing the mass spectra obtained from the truck with those of used oil and diesel fuel samples, it was determined that at least 95% of the particulate mass was unburned engine oil. Small contributions from unburned fuel, oxidized organic combustion products, and sulfuric acid were also detected. Volatility measurements made using a TDMA indicated a core of low-volatility components, which were not measured by the TDPBMS, and is likely composed of soot, metal oxides, or low-volatility organics.

Measurements made using an AMS onboard a Twin Otter aircraft during ACE-Asia detected distinct layers of submicrometer aerosols whose nonrefractory composition consisted of sulfate, ammonium, and organics separated by layers with much lower particle concentrations (130). The organic mass signals indicated that primarily oxygenated and saturated hydrocarbons were present while unsaturated organics were not a significant amount (<3%) of the organic aerosol mass. Pollution layers influenced by emissions from China (using back-trajectory analysis) were found up to ~3700 m altitude while pollution layers with influence from Korea were found below 2000 m. Large amounts of sulfate and organics were typically present in all the layers though a few had a high organic concentration but low sulfate loading. The mass distributions of sulfate, ammonium, and organics as a function of particle size were often well correlated, suggesting an internal mixture of ammonium sulfate and organics. In other atmospheric layers, an external mixture of ammonium nitrate and ammonium sulfate aerosols was hypothesized. Concentrations of sulfate and organic species as high as 10 and 13 g m^-3, respectively, were measured in these polluted layers, while nitrate and ammonium reached up to 4 and 3 g m^-3, respectively.

Emissions from motor vehicles in New York City were measured under realistic conditions with high time resolution in a "chase" study (131) in which an AMS and several other particle and gas-phase instruments followed vehicles in a mobile sampling platform (132). The ability to directly sample emissions from vehicles in traffic under real-world operating and emission dilution conditions provides a more realistic picture of vehicular emissions and also permits the emissions to be characterized on an individual vehicle basis, providing insight into the range of emissions variability in the fleet. The focus of this study involved characterizing new and existing bus engine technologies used by the NYC Metropolitan Transit Authority (MTA) and contrasting them with the rest of the city's heavy-duty vehicle fleet. The organic carbon fraction of particles in diesel exhaust was dominated by unburned lubricating oil under most operating conditions. This was determined by comparing the spectra of sampled diesel particles to spectra obtained from pure lubricating oil and diesel fuel samples similar to the method used by Sakurai et al. (129) discussed above. The mass distributions of organic and sulfate peaked at D_va ~90 nm, and their colocation suggests they were internally mixed. This ultrafine mode of organics could be used as a signature for motor vehicle-related particle emissions, and efforts are underway to use it to quantitatively estimate the contributions from vehicle emission to total ambient aerosol loadings. MTA diesel buses with newer engines emit less nonrefractory PM₁. Buses equipped with Continuous Regenerating Technology (CRT) particle filters and buses using compressed natural gas (CNG) emit about 60% less nonrefractory PM₁. However, CRT-equipped bus emissions had higher gas-phase NO₂/NO_x ratios and CNG-fueled buses had higher CH₄ and CH₂O emissions. It should also be noted that refractory material such as EC can account for ~50% of emitted diesel PM mass and this fraction is not measured by the AMS.

Significant (nontraffic related) nucleation events were observed on 2 of 16 days of measurements in Pittsburgh, PA, performed with an AMS and two SPMS systems (128). All the events began in the morning with a significant increase in the number of nuclei mode particles (D_m = 3-10 nm measured by SMPS). The mass concentrations of sulfate, ammonium, nitrate, and organics for D_va = 33-60 nm increased 20-90 min after this initial spike, the lag being caused by the time required for the nuclei particles to grow into the accumulation mode and be detected by the AMS (D_va > 33 nm). This rapid rate of particle growth indicates that condensation was the primary growth mechanism, not coagulation. Particulate sulfate increased first and fastest, followed by ammonium, indicating that the nuclei mode particles were initially very acidic and then became more neutralized by reacting with NH_3(g); it took about 2 h for the particles to become completely neutralized. The concurrent increase in the ion signal from m/z 44 (an oxidized organic fragment) in the early afternoon indicates that photochemically produced secondary organics also contributed to the growth of the ultrafine particles (D_m = 100-1000 nm) later in the growth event. Nitrate was always only a minor component of the aerosols and contributed least to particle growth.

Ambient measurements conducted with an AMS during the Atlanta Supersite Experiment are presented under Instrumental Comparison: Atlanta Supersite Experiment and compared to the results from SPMS and semicontinuous soluble ion instruments.

(2) Laboratory Studies. The density of polystyrene latex spheres coated with a layer of oleic acid was found to increase upon oxidation by O₃ by monitoring changes in the particle's vacuum aerodynamic diameter (D_va) and organic mass with an AMS and changes in its mobility diameter (D_m) using an SMPS (133). A decrease in D_m with a subsequent increase in its D_va indicated that the particle was becoming smaller and denser as the oleic acid reacted with ozone. The change in particle/layer density was determined using two independent methods that agreed to within 2%. Concomitant with the increase in particle density was an increase in the oxygen content of the reacted organic layer, while a decreased particle size is explained by the evaporation of volatile products such as 1-nononal. The density and oxygen content of the layer continued to increase slowly even after all the original oleic acid has reacted, suggesting that the oxidation products continued to react with ozone or themselves to some extent. The kinetics of ozone reacting with oleic acid has previously been shown to be size dependent by using the AMS to monitor the changes in the oleic acid ion signal as a function of ozone exposure for different sizes of pure oleic acid particles (125). A short reacto-diffusive length of ~10 nm determined in that work confines the reaction to a thin shell near the surface of the particle, as suggested by studies discussed above.

The self-reaction of organic compounds in particles to produce organic oligomers and increase the partitioning of organics from the gas phase to the particle phase has been proposed as an important yet poorly understood formation mechanism for SOA (2, 85). Formation of dimers and trimers of glyoxal that partitioned to various seed aerosols were observed using an AMS to characterize the evolution of particle composition in a smog chamber (134). The ion signal from smaller organic fragments increased first and then declined with reaction time while the signals from larger mass fragments produced by at least glyoxal trimers increased. Increases in the chamber's relative humidity delayed the rise in the organic ion signals, which is explained by a dilution of the reactants and slowing of the rate of self-reaction between the organic reactants. A mild acid catalysis was observed when acidic seed aerosols were used, and evidence for the formation of organic sulfates was reported when seed aerosols containing sulfate were present. Their proposed detailed reaction mechanism involving the hydration of glyoxal followed by the self-reaction of the gem-diol products, further hydration and cross-reactions, successfully explained the appearance and growth of major ion fragments as the reaction proceeded.

The reaction of a series of gas-phase cycloalkenes (C₆-C₈, C₁₀) with ozone in the presence of alcohol vapor revealed the formation of alkoxyhydroperoxy aldehydes produced by the reaction of stabilized Criegee intermediates with alcohols (135). Mass spectra of the SOA products measured by TDPBMS show that these alkoxyhydroperoxy aldehydes can isomerize in the gas phase to form a cyclic peroxyhemiacetal which then partitions to the particle phase. The products of Criegee intermediate reactions with alcohols may contribute up to 25% of the SOA mass in these chamber studies and also provided valuable information for the interpretation of SOA experiments conducted using alcohols as radical scavengers (136).

Condensable iodine-containing vapors produced by the photolysis of diodomethane (137) have been suggested as an explanation of new particle formation bursts observed in coastal areas and were investigated in a smog chamber using the AMS (138). The production of new particles, which then grow via condensation, was only observed when ozone was present and the ultraviolet lights were turned on. The AMS detected mass fragments indicating the presence of a variety of iodine oxides, iodine oxyacids (e.g., HIO₃), and water in the new particles. The condensation of these iodine-containing vapors onto sulfate clusters in coastal areas would increase the size of the sulfate clusters, thus reducing their rate of scavenging by coagulation with larger preexisting particles and increasing the lifetime of the new nuclei particles. This new source of nuclei produced by the photolysis of iodine compounds that are emitted by the ocean could have important impacts on marine CCN concentrations, cloud formation, and thus global climate change.

(3) Recent Developments. A new data analysis technique has been developed to apportion the mass concentration and temporal evolution of the organic aerosol measured by the AMS into POA and OOA components (139). The use of just two ion signals, m/z 57 for POA (mostly C₄H₉⁺, from hydrocarbons typically associated with combustion exhaust) and m/z 44 for OOA (mostly CO₂⁺, from oxygenated organic compounds), could account for 98% of the variance in the organic spectra signal and time variance for 16 days of measurements in Pittsburgh, PA (the AMS results for this campaign have been previously discussed above (128)). Using m/z 57 and 44 as the initial principle components, the complete mass spectra of the POA and OOA components could be distilled and their temporal signals refined using an iterative multiple multivariate linear regression that considers all the contributions from organic species to the entire mass spectra up to m/z 300. This extended algorithm produced dramatic improvements to the fit of the sum of POA + OOA mass concentrations to total measured organic mass concentrations. These improvements are primarily caused by changes to the POA component spectral signature, indicating that the contribution of m/z 57 to POA is somewhat variable across sources and time, though it is still valid as a POA tracer. The few periods when the POA + SOA components did not capture a small (<10%) fraction of the organic mass are likely caused by sources that are not well-represented by these derived components, such as biomass burning or meat cooking organic aerosols.

The POA component mass spectra strongly resemble those from diesel bus exhaust aerosols measured during vehicle chase studies conducted in New York City (132) (R² = 0.98). There are also strong resemblances to the more volatile fraction of lubricating oil and fuel aerosols. Thus, the POA component is primarily derived from combustion aerosols. The derived OOA component is very similar to the mass spectra of aged/oxidized organic aerosols in urban and rural areas and qualitatively similar to humic and fulvic acids. The presence of OOA components can result from primary aerosol emissions, the oxidation of POA through atmospheric aging processes, and the formation of SOA. The mass concentration and size distribution of the OOA component track those of sulfate very well, peaking in the afternoon when photochemistry is maximized, and thus, it is concluded that OOA is primarily due to SOA in Pittsburgh (139). The use of two tracers, m/z 57 and 44, for POA and OOA could account for most of the variance in the mass concentrations and mass spectra of the measured organic aerosols, though m/z 44 is a better tracer for OOA than m/z 57 is for POA. The mass spectra of the two components were remarkably invariant in time.

The Aerodyne AMS has recently been modified, replacing the QMS (Q-AMS) with an orthogonal acceleration reflectron TOFMS (TOF-AMS), to increase the sensitivity of the instrument and provide the option of measuring mass spectra from single particles (140). The instrument can operate in three modes: summing all measured spectra into time averages, using a chopper to measure averaged mass spectra as a function of particle size, and a single-particle mode in which size-resolved spectra are not averaged together. In the single-particle mode, a complete mass spectrum is obtained every 24 s and a single particle is typically measured over 2-5 spectra (48-120 s). The first field trial of the TOF-AMS, conducted in Queens, NY, demonstrated the instrument's ability to acquire spectra from single particles when this mode was used for short periods of time during the PMTACS-NY Winter Intensive in January 2004 (140). The measurements of nitrate, sulfate, ammonium, and organics in single particles revealed the most common fine particle type to be a mixture of ammonium, nitrate, sulfate, and organics, which is consistent with previous SPMS measurements. Particles containing little nitrate but a large amount of sulfate, and vice versa, were also detected, and all particles containing nitrate, sulfate, or both were well correlated with ammonium, suggesting the particles are composed of ammonium nitrate, ammonium sulfate, or both. The vaporization/ionization method used is the same as that of the original Q-AMS (discussed above), and thus, the refractory portion of aerosol mass including dust, sea salt, and EC remains undetectable. The instrument was only run in single-particle mode intermittently for short periods of time due to the large amount of data generated and the computer processing time required, resulting in a low duty cycle of 5%. Improvements in the data acquisition software are planned to increase the duty cycle.

Mass concentrations of ammonium, nitrate, and sulfate measured by the new TOF-AMS compared well (R² ~ 0.9) with simultaneous measurements made with a standard Q-AMS (140). The correlation for mass concentrations of particulate organics were not as high, with R² varying from 0.7 to 0.81 on different days. The TOF-AMS detected 70-80% of the mass of nitrate and sulfate that the Q-AMS did, while ~60% of the organic mass was detected by the TOF-AMS. These differences demonstrate that improvements to the data analysis routines used to extract quantitative measurements require improvement and that the ion extraction and transmission efficiencies in the TOF-AMS still need to be characterized. The size distributions obtained by the single-particle mode are biased toward larger particles as these generate larger ion signals that exceed the set threshold for a particle detection event. Size distributions of particle mass concentrations generated by the TOF-AMS (5-min averages) in the chopped mode are much smoother and contain less noise than those obtained by the Q-AMS (10-min averages), due to the greater statistics obtained by the more rapid collection of mass spectra in the TOF-AMS. Several smaller features in these size distributions of particle mass were clearly evident in the TOF-AMS data, while they were almost missed by the less sensitive Q-AMS.

Instruments that are capable of measuring the soluble ion fraction of aerosols are continuing to be developed. They are capable of quantitatively measuring a wide variety of ions, including NH₄⁺, K⁺, Na⁺, Mg²⁺, Cl^-, NO₃^-, SO₄^2-, and some short-chain organic acids, down to 4-min sampling times, allowing semicontinuous on-line measurements. This suite of measured ions permits the estimation of the acidity of aerosols. Currently these instruments typically measure the entire aerosol population below a certain size cutoff, and thus, size-resolved composition information is not obtained. As these instruments measure an aggregate of aerosols and not single particles, conclusions regarding the mixing state of aerosols rely heavily on assumptions. While these instruments are often used to characterize aerosols on their own, they are frequently coupled with particle mass spectrometers at the same sampling sites to allow the different results to be compared and validated, and intercomparisons of these instruments are frequently performed (42, 60, 141-145). Some of these instruments, such as the particle-into-liquid sampler (PILS), are aircraft deployable and the PILS is commercially available.

The Atlanta Supersite Experiment in 1999 (below) provided an excellent opportunity for the comparison of five state-of-the-art instruments that measure the soluble nitrate and sulfate fractions of particulate matter in a semicontinuous manner with time resolutions of 1 h or less. The design and operating principles of the five instruments have been summarized by Weber et al. (143) and are only briefly described here. All the instruments used particle size cuts such that only fine aerosols (D_m < 2.5 m) were sampled and employed various denuders to eliminate contributions from gas-phase nitrogen and sulfur species. Two of the instruments, from Georgia Tech./Brookhaven Laboratory and The Netherlands Energy Research Foundation, use steam to grow the particles, which are then impacted, washed off, and the soluble ions analyzed by ion chromatography. The Georgia Tech. instrument was the prototype for the PILS discussed below. Another instrument that also uses ion chromatography, from Texas Tech., impacts the particles on a glass fiber filter and then elutes the particles from the filter with water (146). This method avoids the use of steam due to concerns that any nitric oxide gases that are not removed by the denuder could be converted to nitrate and nitrite by the steam and cause positive sampling artifacts. Aerosol Dynamic Inc.'s instrument impacts and vaporizes the particles to measure the evolved NO_x and SO₂ gases using two separate commercial gas analyzers. The Atmosphere Research and Analysis instrument measures nitrate by first collecting it on a Mo catalysis and converting it to NO. Particulate nitrate is determined by the difference between NO measured from particles converted by the catalyst and NO from particles simply collected on a Teflon filter, using a commercial NO_x analyzer. The Aerosol Research and Analysis, Georgia Tech./Brookhaven Lab, and Texas Tech instruments were only recently developed before this campaign.

By the end of the one-month study, the instruments were in good agreement; the R² values between instrument pairs ranged from 0.7 to 0.94. Nitrate concentrations (overall average ~0.5 g m^-3) were much lower than sulfate (overall average ~14 g m^-3) as expected due to the larger measurement challenges for semivolatile nitrate. The semicontinuous instruments agreed to within 20-35% for nitrate and 10-15% for sulfate, and their agreement improved over the course of the campaign. These results were compared with those obtained for nitrate and sulfate from eight different 24-h filter measurement methods. The agreement between sulfate measurements made by the filter methods (1 = 8%) was better than for the semicontinuous measurements, while the nitrate measurements made by the filters had more variability (1 = 22%). This may suggest that the semicontinuous instruments, with their much shorter sample collection time, are less susceptible to sampling artifacts due to aerosol volatility (i.e., ammonium nitrate) than the 24-h filter collection methods are. Further results obtained by these soluble ion instruments and real-time particle mass spectrometers in Atlanta are discussed below.

The PILS can also measure a few soluble organic acid components commonly measured in aerosols including formate, acetate, and oxalate, by extending the elution time of the IC to 15 min (147). By coupling the PILS to a commercial total dissolved carbon analyzer, the unspeciated WSOC concentration of PM_2.5 could be measured in 6-min integrals (148). The ionic composition of fine aerosol sampled from aircraft flights during TRACE-P over the western Pacific Ocean was measured by PILS with 4-min time resolution (142). Relatively pure biomass burning plumes were frequently encountered below latitudes of 25N. In these plumes, fine particulate mass potassium, nitrate, and ammonium were all well correlated along with gas-phase CO and light-absorbing aerosol measurements. Data from satellite fire maps and back-trajectory analysis indicated biomass burning in southeast Asia as the likely source. These associations confirm the use of fine water-soluble potassium as a unique tracer for biosmoke, and K⁺ was used to estimate the influence of biomass or biofuel burning as ~20% of the measured Asian pollution outflow plumes. Surface-level pollution layers were also encountered in the marine boundary layer, characterized by high sulfate, low amounts of nitrate, and almost no potassium. At latitudes above 25N, highly mixed plumes of urban/industrial and biosmoke plumes, indicated by high SO₄^2- and K⁺, respectively, were encountered. These mixed plumes were also the most heavily polluted; the fine particle ionic mass and particle volume were higher for plumes with any contribution from biosmoke than plumes with no evidence of biosmoke.

The mass concentrations of soluble Na⁺, K⁺, NH₄⁺, Ca²⁺, Mg²⁺, Cl^-, NO₃^-, and SO₄^2- in aerosols of <1.3 m measured by PILS during ACE-Asia and TRACE-P were used to identify likely sources of particulate matter (149). In addition to a sea salt origin, Cl^- showed an urban emission source possibly related to biofuel combustion. Both sea salt and dust contributed to Mg²⁺. During both campaigns NH₄⁺, NO₃^-, SO₄²⁺, and gas-phase CO were strongly correlated, indicating that combustion was the dominant source for these species and their precursors. As H₂SO₄ is a stronger acid and much less volatile than HNO₃, the presence of both SO₄^2- and NO₃^- in particles along with NH₄⁺ is only expected for nonacidic particles. As >90% of the samples collected below 3 km during both campaigns contained some nitrate, Lee et al. conclude that NH₃, which appeared to be emitted from combustion and urban sources, was comparable in supply to H₂SO₄ and its precursor SO₂ such that ~70% of the total aerosol acidity was neutralized by NH₃. Ammonia could not account for all of the acid neutralization, suggesting there are other alkaline species to consider that are not measured, such as amines. The magnitude of the electrical charge imbalance produced by summing all the measured positive and negative ions was >20% and likely caused by unmeasured ions such as organic acids and bases, CO₃^2-, and the hydronium ion. Ca²⁺ played an important role in the neutralization of nitrate when [NO₃^-] was low (~1.5 pptv), likely through the reaction of HNO₃ with CaCO₃ in Asian dust. However, at very high aerosol nitrate concentrations (>200 pptv), neutralization by calcium bases could not explain this high nitrate loading. The strong correlation of high nitrate concentrations with elevated gas-phase CO levels suggests that HNO₃ formed by photochemistry in these polluted (CO-rich) air masses could be adsorbing to mineral dust surfaces, as proposed by laboratory studies (74).

The ability to quantify the soluble ionic fraction of aerosols with short time resolution is a powerful tool for providing further insight into the chemical composition and associations of important components in aerosols. However, caution must be used when using one or several ionic species for source apportionment from aggregate measurements of aerosol composition. As Lee et al. (149) discuss, the use of Ca²⁺ as a tracer for mineral dust may be impacted by urban sources of Ca²⁺ such as construction sites. Another important source of particulate calcium in fine and ultrafine aerosols, discussed previously in this review, is lubricating oil in combustion engine emissions, which contain significant amounts of Ca²⁺ (94, 129, 132). Furthermore, the use of Cl^- as a sea salt marker can be complicated by other Cl^- sources such as the combustion of biofuel suggested by Lee et al. and also HCl_(g) that has adsorbed to mineral dust particles (150). PILS measurements have also recently been used to infer the mixing state of aerosols (151). Such extrapolations from data that do not measure the composition of single particles are limited by uncertainties and crucial assumptions. This analysis is also complicated by the fact that all particles below a certain size threshold are analyzed (there is no size segregation), and key species such as CO₃^2- in this case, are not measured.

A specially designed inlet permitted the simultaneous continuous sampling of gas-phase acids (HNO₃, HONO, HCl, SO₂) and NH_3(g) and their corresponding aerosol ions at a rural site in the Amazon Basin (152). Biomass burning was a strong source of HNO₃, HCl, HONO, and especially NH₃, whose concentration was 10 times higher than the acidic gases during these times. The mixing ratio of NH₃ was ~3 times higher during the burning season than the wet season while the acids were at least 2 times higher. The enhancement of trace gases during the burning season significantly increased the amount of corresponding water-soluble inorganic species in the aerosols. Aerosol NH₄⁺ was at least 4-10 times higher than anionic species, corresponding to the higher NH₃ levels, and declined substantially from the dry burning season to the wet season. During the burning season, NO₃^- was the dominant anion and showed the largest decrease in the wet season (~70%). Nitrogen-containing gases and aerosol ions displayed a diurnal pattern. The partitioning between the gas and aerosol phases of these species was dependent on relative humidity and ambient temperature, resulting in enrichment of aerosol NH₄⁺ and NO₃^- at nighttime. Aerosol SO₄^2-, which is nonvolatile, did not have a diurnal variation and remained stable during day and night. HONO_(g) had a typical diurnal cycle during biomass burning with a nightly maximum but was not completely depleted during sunlight hours, suggesting a continuous source of HONO in addition to heterogeneous formation from NO₂ and surface water.

The high time resolution of these semicontinuous instruments allows the equilibrium of ammonia/ammonium and nitric acid/nitrate between the gas/particle phases to be evaluated. PM_2.5 measurements with a PILS and gas-phase measurements of NH_3(g) and HNO_3(g) were measured with 15 min or shorter time resolution during the 1999 Atlanta Supersite Experiment (153). The measurements of aerosol NO₃^-, SO₄^2-, Cl^-, NH₄⁺, and Na⁺ along with ambient temperature, RH, and pressure were used to predict the corresponding gas-phase concentrations of NH₃ and HNO₃ using the ISORROPIA model under thermodynamic equilibrium. The partitioning of ammonia/ammonium and nitric acid/nitrate between the gas and particle phases was highly dependent on the particle acidity, which was derived from the PILS data. By comparing the predicted concentrations of NH_3(g) and HNO_3(g) with those measured, Zhang et al. concluded that the particle-phase species were in equilibrium with the gas phase if one or more of the following is true: (1) the PILS measurements of PM_2.5 SO₄^- concentrations were systematically overestimated; (2) PM_2.5 concentrations of alkaline components were systematically underestimated; (3) the ISORROPIA model systematically overestimated the particle acidity. The use of semicontinuous aerosol measurements with high time resolutions is a major improvement over previous attempts to evaluate the equilibrium of aerosol components using 6-h to 1-day filter measurements.

Atlanta, GA, is the EPA's first designated PM_2.5 Supersite. In 1999, a large campaign was undertaken to investigate the nature and causes of Atlanta's chronic air pollution problems, leading to frequent violations of the air quality index for ozone and particulate matter. This experiment also provided an ideal opportunity to compare the results from simultaneous measurements made by four different particle mass spectrometers, the PALMS, ATOFMS, RSMS, and Aerodyne AMS (Real-Time Mass Spectrometry of Aerosols), as well as semicontinuous analyzers of soluble ions in aerosols (Semicontinuous Aerosol Soluble Ion Measurements). This comparison and the design and capabilities of the four real-time instruments have already been well described by Middlebrook et al. (154), and thus, only a summary of the key points is presented here. The design and operation of the semicontinuous analyzers have been described by Weber et al. (143) and discussed above. Detailed results from individual instruments are also presented below. The three laser-based SPMS instruments found similar individual particle classifications; primary ones were organics mixed with sulfate, sodium/potassium sulfate, soot/hydrocarbon, and mineral dust. Comparing measurements of particles between roughly 0.25 and 0.5 m on overlapping days revealed good broad agreement of the particle detection frequency for the four main particle classes between the three laser-based mass spectrometers. Measurements of particulate nitrate made by the PALMS and AMS were compared to semicontinuous measurements. Correlation was better for the AMS data than the PALMS though the PALMS measurements tracked relative changes in the nitrate concentrations well. Also, the PALMS data were better correlated with the semicontinuous measurements when the particles were dried before entering the instrument's inlet.

In measurements conducted in Atlanta by ATOFMS, supermicrometer particles were primarily dust and sodium-containing particles with secondary species (47). The size distribution of carbonaceous species was bimodal with peaks in the sub- and supermicrometer modes. The unexpected supermicrometer mode of carbon-containing particles was due to partitioning of organic vapors onto sodium and dust particles. Bursts of carbonaceous particles containing lithium were frequently detected in air masses from the west and were identified as coal-burning particles produced from coal-fired power plants lying in this direction. Secondary species were found on all particle types, indicating that all the particles were highly transformed.

During certain periods in Atlanta, large amounts of detected particles scattered light but did not produce a mass spectrum in the ATOFMS, decreasing the instrument's hit rate (54). This "missing particle class" primarily occurred in the smallest size range, from 0.35 to 0.54 m. By comparison with semicontinuous measurements, these time periods were found to coincide with increased levels of ammonium sulfate. Particles that are primarily ammonium sulfate are not ionized easily, which explains why mass spectra were not produced. A method was developed to account for the instrument's sampling bias by scaling the ATOFMS data to simultaneous measurements made with laser particle counters to obtain atmospherically relevant particle number concentrations.

Measurements made by PALMS during the Atlanta Supersite provided interesting information on the behavior of nitric acid (155). Nitrate in particles peaked in the morning when relative humidity was highest. Nitrate also had a small local minimum in the afternoon when gas-phase nitric acid was highest. The afternoon maximum in nitrate was more pronounced on hydrocarbon/soot or aluminosilicate mineral particles than the organic/sulfate particles. The RH dependence of nitrate in particles was strongest for the organic/sulfate particle class. The soot and aluminosilicate particles, however, still had a strong nitrate signal even during periods of low RH. These results suggests that the afternoon maximum of nitrate in the mineral and soot particles was caused by the uptake of nitric acid (or NO₂), which also peaks in the afternoon and reacts with basic species in these particles. Oxidized organics had a diurnal cycle similar to that of nitrate except that the afternoon maximum was associated with the organic/sulfate particle class. Particles with oxidized organics were also associated with hydroxymethanesulfonate, which suggests that aqueous-phase oxidation produced some of these oxidized organics.

Measurements of ultrafine particles from 14 nm to 1.3 m by RSMS found that particles smaller than 100 nm were dominated by carbon-containing particles (65). Larger particles had varied compositions but were generally mixtures of organic and inorganic components such as crustal materials and metals. Particles classified as organic and containing significant nitrate accounted for 74% of all the particles detected. The major measured components of the Atlanta urban aerosol measured by the AMS were sulfate and organics with a minor contribution from nitrate (156). A diurnal cycle of aerosol nitrate was observed with maximums in the early morning and minimums in the afternoon. This suggests that the partitioning of nitrate to the particle phase is favored by lower temperatures and higher RH, as observed and suggested by other simultaneous SPMS measurements above. Sulfate does not have a diurnal cycle, reflecting its low volatility and inability to partition to the gas phase. AMS measurements of sulfate agreed well with those made by two on-line PM_2.5 ion chromatography instruments (Semicontinuous Aerosol Soluble Ion Measurements); the R² was 0.79 for AMS and PILS sulfate concentrations. The correlation for nitrate was lower (R² = 0.49) due to the lower overall aerosol nitrate concentrations and thus a reduced signal-to-noise ratio measured by the AMS.

The 1-h or shorter time resolution of the semicontinuous instruments allowed the detection of diurnal cycles in aerosol nitrate concentrations during the Atlanta study (143). Minimum nitrate concentrations occurred at mid to late afternoon while maximum values were observed in the early morning when temperatures were lowest and relative humidity highest. Sulfate concentrations did not exhibit this diurnal behavior but showed short- and long-term trends that may reflect local and regional conditions, respectively. The lowest sulfate concentrations (<1 g m^-3) were observed during periods of precipita