The Nano Aerosol Mass Spectrometer (NAMS) was deployed to the California Nexus Los Angeles ground site in Pasadena, California during May–June 2010 to study nanoparticles in the 20–25 nm size range. NAMS gives a quantitative measure of the elemental composition of individual particles, and molecular apportionment of the elemental data allows the O/C mole ratio of carbonaceous matter in each particle to be determined. Abrupt increases in nanoparticle number concentration were observed in the afternoon on sunny days, and coincided with a shift in the wind direction from the southeast to the southwest. Nanoparticles analyzed during these time periods were found to contain enhanced levels of sulfur and silicon relative to particles analyzed earlier in the day, and the O/C ratios of carbonaceous matter changed from a distribution dominated by primary motor vehicle emissions (O/C ratio < 0.25) to one dominated by “fresh” secondary organic aerosol (O/C ratio between 0.25 and 0.65). The wind direction and chemical composition dependencies suggest that the afternoon increase in number concentration originated from motor vehicle emissions in the downtown Los Angeles area that were photochemically processed during transport to the measurement site. It is likely that photochemical processing led to both a change in the composition of preexisting particles and to the formation of new particles.
 Ambient nanoparticles, defined here as particles having physical diameters below 100 nm, significantly influence human health [Li et al., 2010; Oberdörster et al., 2005; Terzano et al., 2010] and climate [Bhaskar and Mehta, 2010; Charlson et al., 1992; Lohmann and Feichter, 2005]. Nanoparticle number concentrations are highly variable especially in the urban environment where they can exhibit strong spatial and temporal dependencies. For example, within the first 100 m of a freeway, the particle number concentration has been found to decrease almost exponentially with increasing distance downwind of the freeway [Sioutas et al., 2005]. Adjacent to an intersection, intense spikes in number concentration occur at regular time intervals that are related to the stoplight cycle governing traffic flow through the intersection [Klems et al., 2010]. As one moves away from the roadway microenvironment, these variations decrease and eventually blend into the ambient background, which is influenced by longer time frame processes such as diurnal dependencies of human activity (e.g., rush hour patterns of motor vehicle emissions) [Zhang et al., 2004] and meteorology (e.g., photochemical particle formation) [Kulmala et al., 2004].
 Understanding the factors that influence ambient particle number concentrations is facilitated by measurement of nanoparticle chemical composition. Toward this end, we have developed a nano aerosol mass spectrometer (NAMS) that gives the elemental composition of individual, size-selected nanoparticles in the 10–30 nm size range [Wang and Johnston, 2006; Wang et al., 2006]. This method helps bridge the gap between what is known about particle composition above 100 nm in diameter, which can be probed with a variety of single particle techniques [Nash et al., 2006; Prather et al., 2008], and what is known about gas phase species and ambient nuclei. NAMS has been used in previous studies to determine the composition of ambient nanoparticles in urban [Klems et al., 2011; Zordan et al., 2008] and rural/coastal [Bzdek et al., 2011] environments. In these studies, nanoparticles emitted directly from motor vehicles were found to contain a significant amount of unoxidized carbonaceous matter, whereas particles formed photochemically during new particle formation (NPF) events tended to have elevated levels of nitrogen, sulfur and oxygen relative to particles analyzed off-event.
 In this study, NAMS was deployed to the California Nexus (CalNex) Los Angeles Ground Site in Pasadena, California during May and June 2010. While fine particles in the Los Angeles basin have been studied extensively in the past, relatively little is known about the composition of ambient nanoparticles. Hughes et al.  collected 24-h PM0.1 samples from the rooftop of a building in Pasadena during winter 1996. For all days studied, organic matter was the most prevalent chemical component, representing almost half of the ultrafine mass. Sulfate, nitrate and elemental carbon were also found in significant amounts, and several trace metals were detected. Ning et al.  collected morning and afternoon 3-h PM0.18 samples in downtown Los Angeles during summer 2006. Organic matter was found to be the major constituent, again representing approximately half of the collected mass. Sulfate, nitrate and ammonium represented another quarter to third of the collected mass. The sulfate and nitrate mass fractions were approximately equal in the morning, but sulfate increased relative to nitrate in the afternoon, presumably due to photochemical processing. Several individual organic compounds and trace metals were also characterized. Relative to these earlier studies, NAMS allows smaller particles to be analyzed with better time resolution and the oxidation state of organic matter can be followed on a particle-by-particle basis. The results provide insight into nanoparticle formation and transformation in the Los Angeles basin.
2. Experimental Methods
 NAMS has been described in detail elsewhere [Wang and Johnston, 2006; Wang et al., 2006]. It consists of four parts: an inlet containing an aerodynamic lens and digital ion guide (DIG) to focus particles, a digital ion trap (DIT) to size-selectively capture particles for analysis, a high energy pulsed laser to generate atomic ions from the captured particles, and a reflectron time-of-flight (TOF) mass analyzer for atomic ion analysis. Prior to entering the NAMS inlet, particles are charged with a unipolar charger [Chen and Pui, 1999; Smith et al., 2004]. Inside NAMS, particles are size selected by the frequency of the square wave potential applied to the DIT and the voltage applied to a field adjusting lens (FAL) upstream of the DIT that assists particle injection and trapping in the DIT [Pennington and Johnston, 2012]. The DIT traps particles based on the mass normalized diameter (dmn), which is related to the mobility diameter (dm) through the particle density, ρ, as shown in equation (1) [Johnston et al., 2006; Wang and Johnston, 2006]:
where the reference density, ρ0, is 1 g/cm3. During the CalNex campaign, the DIT frequency was set to 10 kHz with a square wave potential of +505 V/−507 V, the FAL was set to −500 V, and the DIG was set to 50 kHz with a potential of +500/−500 V. The parameters chosen allowed for the trapping and subsequent analysis of 20–25 nm (dmn) particles. Each particle mass spectrum was processed by zeroing the baseline, mass calibrating the time axis, and integrating the ion signal across m/z peaks of interest (multiply charged ions of C, N, O, S, and Si). An in-house method [Zordan et al., 2010] was used to deconvolute the overlapping signals at m/z 4 and 8, which correspond to the isobaric ions C+3/O+4 and O+2/S+4, respectively. Although H+ was observed in the mass spectra, previous work has shown that hydrogen cannot be reliably quantified by NAMS and is therefore excluded from the analysis.
 Ambient nanoparticle measurements were performed at the CalNex ground site in Pasadena, California between 21 May and 10 June 2010 during which time over 130,000 particle mass spectra were acquired. Particles were sampled from a height of ∼4 m above the ground through a ∼5 m length of 0.5 in (OD) tubing at a flow rate of 5.0 L/min. A second sampling line was connected to a Scanning Mobility Particle Sizer (SMPS; Electrostatic Classifier, model 3080, Condensation Particle Counter, model 3025a, TSI, Inc., St. Paul, MN) to provide concurrent measurement of the particle number concentration and size distribution.
3. Results and Discussion
 The NAMS data set for the CalNex campaign consists of over 130,000 mass spectra from which the elemental composition of individual particles, expressed as the mole fraction of each element detected in a particle, are determined. The entire data set represents a continuum of particle compositions, which can be visualized by construction of a “carbon mole fraction” plot as shown in Figure 1. In Figure 1, the elemental composition of each particle in the data set is plotted, beginning with the particle having the highest carbon mole fraction and ending with the particle having the lowest carbon mole fraction.
 Several important features of ambient nanoparticle chemical composition in Pasadena are illustrated by Figure 1. First, carbon and oxygen are the most prevalent elements, together accounting for over 80% of the atoms (not including hydrogen) in nearly all of the particles analyzed. The large carbon mole fraction in most particles highlights the dominant contribution of carbonaceous matter to ambient nanoparticle composition and is consistent with a previous measurement of the bulk nanoparticle chemical composition in this location [Hughes et al., 1998].
 Second, about 7% of the particles contain a carbon mole fraction greater than 0.9 with oxygen making up most of the balance. These particles are composed of either elemental carbon or hydrocarbon-like molecular species (O/C elemental ratio less than 0.1). These possibilities cannot be distinguished because NAMS does not give a quantitative measure of the hydrogen content. In a previous study of urban nanoparticles with NAMS, particle compositions of this type were shown to originate directly from motor vehicles and were especially evident during “spikes” in particle number concentration that arose from vehicles accelerating at an intersection adjacent to the measurement site [Klems et al., 2010]. The relatively small number of these particles and the general lack of spikes in number concentration in the CalNex data set underscore the relatively small contribution of freshly emitted particles from motor vehicles to the ambient aerosol at the ground site in Pasadena, Ca. There are no major roadways in close proximity to the measurement site. Likely sources for these particles are local traffic, delivery trucks to the site and vehicles at a firehouse adjacent to the site.
 Third, particles in the middle of the plot have oxygen, nitrogen and sulfur mole fractions that are highly correlated with each other but anti-correlated with carbon. The oxygen, nitrogen and sulfur mole fractions increase almost monotonically as the carbon mole fraction decreases. These particles represent a continuum of compositions, differing in the relative amounts of inorganic and organic species as well as in the degree of oxidation of the organic matter (to be discussed later). Fourth, on the far right of the plot are particles having zero or trace amounts of carbon. These particles tend to have a disproportionally large amount of silicon. Some also have a very large amount of the oxygen, higher than can be explained by the presence of sulfate, nitrate, silica and oxidized carbonaceous matter (if carbon is present) based on the corresponding mole fractions of S, N and Si. This “excess” oxygen in these particles suggests the presence of particulate water that did not fully evaporate prior to analysis, or possibly some oxidized transition metals and/or heavy metals that are not easily detected by NAMS.
 Characteristic diurnal variations of the aerosol size distribution and particle number concentration were observed on many days during the measurement campaign. These variations are illustrated in Figure 2a, which shows a sequence of size distributions for a 5-day period in late May and early June. The number concentration was generally low overnight and during the morning hours. In the afternoon, an abrupt increase in number concentration was observed across a wide range of particle sizes. This feature typically lasted a few hours and dissipated in the evening. Also shown in Figure 2a is the approximate particle size range analyzed by NAMS, which overlapped the lower edge of the afternoon features. Figure 2b shows the average particle concentration (dN/dlogdm) measured by SMPS in the approximate size range analyzed by NAMS. Figure 2c shows the NAMS hit rate (particles analyzed over a 10 min period), which is similar to that expected from Figure 2b. Minor differences between Figures 2b and 2c are related to the difference between mobility (SMPS) and mass normalized (NAMS) particle selection and minor changes in the alignment of NAMS over time.
 The afternoon feature illustrated in Figure 2 was consistently observed on sunny days, but conspicuously absent on cloudy days. During the measurement period, NAMS and SMPS were operational from midnight to midnight for a total of 10 days that exhibited an afternoon feature similar to Figure 2. The average diurnal variations in particle concentration, wind direction and chemical composition are shown in Figure 3 for these days (May 21, 22, 26, 28, 29, 31 and June 1, 2, 4, 9). The increase in number concentration during the afternoon (Figure 3a) coincides with a shift in wind direction from the southeast to the southwest (Figure 3b) and a change in particle composition (Figure 3c). Median values and ranges for the mole fractions of individual elements that give rise to the plot in Figure 3c are shown in Figure 4. In the morning, particles are dominated by carbon and oxygen, with a small but significant amount of nitrogen and hardly any sulfur or silicon. In the afternoon when the wind direction shifts and the particle concentration increases, the sulfur and silicon contents both increase. While the elemental fractions of these species are small, the mass fractions of the corresponding molecular species are much larger. For example, if one assumes that all of the sulfur exists as sulfate, then the increase in elemental mole fraction from 2% in the morning to 5% in the afternoon, Figure 4c, represents an increase in sulfate mass fraction from 13% to 32% of the total nanoparticulate mass.
Figures 3c and 4 show that carbon is the most prevalent element in ambient nanoparticles at this site. In order to gain further insight into the character of this carbonaceous matter, the elemental mole fractions are apportioned to likely molecular species so that the oxygen to carbon (O/C) mole ratios can be determined. The method assumes that sulfate, nitrate, ammonium and organic matter make up the majority of the particle mass. This assumption is consistent with previous nanoparticle composition measurements in Pasadena and Los Angeles [Hughes et al., 1998; Ning et al., 2007] as discussed in the introduction section. Molecular apportionment is based on the following additional assumptions: (1) all sulfur is assumed to exist as the sulfate anion; (2) all silicon is assumed to exist as silicon dioxide; (3) nitrogen is assumed to exist as either nitrate or ammonium ions; (4) in a single particle, the sulfate anions must be fully neutralized (i.e., two ammonium cations for each sulfate anion) before any nitrate can exist in the particle; (5) nitrate is assumed to exist only in neutralized form, i.e., ammonium nitrate. Note that while the apportionment scheme will “pair” sulfate with an appropriate amount of nitrogen to neutralize it (as ammonium sulfate) the two species are apportioned separately. In Figure 3c, the N/S ratio is greater than or equal to 2 at all times, suggesting that sulfate is fully neutralized and a significant amount of ammonium nitrate is present. We note that both nitrate and ammonium have been shown to exist in nanoparticles, previously [Smith et al., 2004]. After the inorganic species are apportioned, the carbon and remaining oxygen are grouped together to give an indication of the oxidation state of the carbonaceous matter based on the apportioned O/C ratio. Note that the apportionment of silicon as silicon dioxide as opposed to a species with a different O/Si mole ratio does not significantly affect the O/C ratio calculation for most particles because substantial amounts of silicon are only observed in particles with little or no carbon (see Figure 1). The method, as well as potential artifacts and inaccuracies, are discussed in detail elsewhere [Bzdek et al., 2011; Klems et al., 2011]. In particular, the method does not explicitly account for species such as organosulfates, organonitrates and amines. If these species are present in significant amounts, apportionment will split them into an organic fraction that is grouped with other organic matter in the particles and an inorganic fraction that is grouped with the appropriate species (sulfate, nitrate, ammonium).
 We emphasize that the apportionment method is not meant to serve as a quantitative determination of molecular species, but rather as a qualitative interpretation of the elemental composition on the basis of major molecular species known to exist in ambient nanoparticles in this location. Laboratory standards have shown that while most elemental ratios are accurate to less than 10%, some can be off as much as 15%. If we assume an error of 10%, then error propagation for an average, multicomponent particle observed in Pasadena gives an uncertainty in the O/C ratio of 30%. The uncertainty can be much higher for the ∼15% of particles on the right end of Figure 1 that have little or no carbon.
 Since molecular apportionment is applied to individual particle compositions, the distribution of O/C ratios within a group of particles can be determined in addition to the average O/C ratio of the entire group. This distribution is unique to single particle analysis and provides additional insight beyond what is obtained from bulk analysis methods such as the aerosol mass spectrometer (AMS). Figure 5 shows the mass fraction of carbonaceous matter in each particle plotted in order of decreasing carbon mole fraction, similar to Figure 1. The mass fractions are color coded depending on oxidation state: low O/C ratio (particles having carbonaceous matter with an O/C ratio smaller than 0.25), mid O/C ratio (particles having an O/C ratio between 0.25 and 0.65), and high O/C ratio (particles having an O/C ratio greater than 0.65). Particles having an O/C ratio greater than 2.0 (ratio can be infinite if there is no carbon in the particle) are not included in these plots since they necessarily must contain some particulate water or undetected heavy metals, which are not accounted for in the molecular apportionment method. Less than 5% of the particles were removed because of an O/C ratio greater than 2. It should be noted that laboratory particle standards (sucrose, ammonium sulfate, NaCl), generated with an electrospray particle generator do not show any evidence of particulate water within the uncertainty of the elemental composition measurement. Particle standards generated with an atomizer under humid conditions may contain up to 10% excess oxygen, presumably due to particulate water. While particulate water can influence the estimation of O/C ratios of ambient particles, its influence is expected to be small.
 The low, mid and high O/C ratio ranges roughly correspond to HOA, SV-OOA and LV-OOA fractions of organic matter identified by AMS measurements at various locations around the world [DeCarlo et al., 2010; Ng et al., 2010; Raatikainen et al., 2010]. There are several important features in Figure 5. First, a relatively small number of particles (∼10% of the data set) contain a mass fraction of carbonaceous matter near 100%, and in all cases they are composed of low O/C material. The relatively small amount of low O/C material in the data set highlights the small contribution that primary emissions make to Pasadena nanoaerosol. Second, the great majority of particles contain partially oxidized (mid O/C) carbonaceous matter, while a smaller number of particles contain highly oxidized (high O/C) material. If it is assumed that mid and high O/C carbonaceous matter roughly correspond to fresh and aged secondary organic aerosol (SOA), the relative number of particles of each and their mass fractions indicate that Pasadena nanoaerosol is composed mainly of fresh SOA. Third, particles containing mid and high O/C material have mass fractions of carbonaceous matter that are significantly smaller than 100%. There are no “pure” SOA particles in the data set – all SOA exists within internally mixed particles containing both organic and inorganic matter.
Figure 6 shows histograms of the number of particles versus O/C ratio for three time periods that encompass the diurnal variations illustrated in Figure 4. Figure 6a shows the histogram of particles analyzed in the morning between 6 A.M. and 12 P.M. Figure 6b shows the histogram of particles analyzed in the afternoon between 12 P.M. and 8 P.M. when particle concentrations tend to be highest. Figure 6c shows the histogram of particles analyzed in the evening and nighttime from 8 P.M. to 6 A.M. All three histograms are partitioned into low, mid and high O/C carbonaceous matter. These plots provide insight into the origin of the afternoon increase in nanoparticle concentration. The morning time period (Figure 6a) contains the greatest fraction of particles with low O/C ratio carbonaceous matter, almost half of all nanoparticles in ambient air. A previous study of urban nanoparticles [Klems et al., 2011] has linked low O/C ratio carbonaceous matter to motor vehicle emissions. The distribution in Figure 6 suggests that nanoparticle formation in the morning is dominated by primary motor vehicle emissions near the measurement site. Another distinguishing feature of this time period is that the nitrogen mole fraction is more than twice the sulfur mole fraction (see Figures 3 and 4), which suggests nitrate is a significant molecular species in the particles.
 In the afternoon, the wind shifts from the southeast to the southwest and when the weather is sunny the particle number concentration increases dramatically (Figure 2). The concentration increase is accompanied by two important changes in chemical composition. First, the sulfur mole fraction more than doubles (Figure 3). Second, the O/C ratio shifts to much higher values (Figure 6b). While the fraction of high O/C particles remains relatively constant from morning to afternoon, the low O/C fraction decreases and the mid O/C fraction increase. Both of these changes are consistent with photochemical processing of the aerosol.
 In the evening, the wind direction shifts again and the number concentration drops, presumably because a new air parcel is transported to the site that has been influenced differently by photochemical processing. While the elemental mole fractions remain largely unchanged from the afternoon, the O/C ratio distribution shifts substantially (Figure 6c). The mid O/C fraction decreases while both the low and high O/C fractions increase. The increase in high O/C particles most likely reflects continued oxidation of mid O/C carbonaceous matter. The increase in low O/C particles suggests that primary motor vehicle emissions become a more important nanoparticle source as photochemical processing subsides.
Figure 7a shows the diurnal variation of carbonaceous matter O/C ratio (averaged over 30 min intervals) during the 10 days where particle features were observed. Figure 7a shows that the O/C ratio drops slightly in the morning hours when the particles are more likely formed via primary processes. Figure 7b shows how the fraction of particles classified into each of the three O/C bins (high, mid, low) changes across each 30 min period. Comparison of Figures 7a and 7b show that the relatively small change in average O/C ratio masks a much larger change in the particle-by-particle distribution.
 The southwest wind direction suggests that the nanoparticle laden air in the afternoon has been transported from the downtown Los Angeles area. HYSPLIT analysis of the back trajectories confirms this conclusion. Unoxidized gas and particle phase carbonaceous matter from motor vehicle emissions in downtown Los Angeles during the morning rush hour are oxidized leading to both the growth of pre-existing particles and the formation of new particles as the air parcel is transported to Pasadena in the afternoon. Photochemical particle formation is indicated by a large increase in nanoparticle number concentration across a wide range of particle sizes and the change in chemical composition. Sulfuric acid is thought to drive nucleation [Sipila et al., 2010; Yu et al., 2005] and oxidized organics are thought to contribute to nanoparticle growth [O'Dowd et al., 2002; Smith et al., 2010]. NAMS composition measurements show that both of these species become enriched as a percentage of particle mass during the afternoon increase in number concentration.
 Nanoparticle size distributions and chemical compositions in downtown Los Angeles have been reported previously by Sioutas and coworkers [Moore et al., 2007; Ning et al., 2007]. In summer 2006, they found that the diurnal variation in number concentration showed two features – a peak in the morning rush hour from 6 to 8 A.M. local time and another in the late morning and early afternoon from about 10 A.M. to 3 P.M. The morning peak in number concentration is not observed in Pasadena (Figures 2a and 3a). This difference highlights the greater impact of motor vehicles on particle number concentration in Los Angeles. In Pasadena, NAMS composition measurements show that ambient nanoparticles become enriched with unoxidized carbonaceous matter and nitrogen (presumably nitrate because there is so little sulfate) during the morning rush hour. Similar enrichments in composition were observed in the Los Angeles study. Based on chemical composition, primary motor vehicle emission is the main source of ambient nanoparticles at both locations during this time period, but the source strength is not sufficient to raise the number concentration at the Pasadena measurement site. An increase in nanoparticle concentration is observed in both locations during the afternoon hours, with a 2–3 h time shift between the two locations that is consistent with transit of the air parcel from one location to the other (10 mile distance; average wind velocity 3–4 mph). An increase in sulfate mass fraction accompanies the afternoon number concentration increase at both locations. In addition, the NAMS measurements show clear evidence for atmospheric processing of carbonaceous matter.
 The diurnal dependence of the N/S ratio is also of interest. As discussed previously, the nitrogen and sulfur traces in Figure 3c show that the N/S ratio is always greater than 2, indicating full neutralization of sulfur to ammonium sulfate. Additionally, the ratio reaches a peak value greater than 3 during the morning hours, indicating the presence of other species such as nitrate. A morning increase in nitrate has been reported previously [Yu et al., 2005] and may be related to higher NOx emissions during this time period.
 Last, silicon is observed primarily in the afternoon, evening and nighttime hours, as shown in Figure 4e. While silicon has been reported previously in urban nanoparticle studies [Phares et al., 2003; Zordan et al., 2008], its origin is uncertain. Figure 1 shows that in Pasadena particles containing large amounts of silicon also tend to contain little or no carbon. This observation and the diurnal dependence observed in Figures 2 and 3 suggest that motor vehicles are not the main source of these particles.
 An abrupt increase in nanoparticle number concentration was observed on sunny days during the CalNex measurement campaign. During this time period, the wind direction shifted from the southeast to the southwest. Chemical composition measurements show that the change in number concentration was accompanied by an increase in sulfur mole fraction (indicating an increase in the sulfate mass fraction) and an increase in the O/C ratio of carbonaceous matter. The wind direction and chemical composition dependencies suggest that the increase in number concentration originated from motor vehicle emissions in the downtown Los Angeles area that were photochemically processed during transport to the measurement site. It is likely that photochemical processing led to both a change in the composition of preexisting particles and to the formation of new particles.
 NAMS measurements were supported by National Science Foundation grant CHE0808972. Infrastructure at the CalNex ground site was made possible by the efforts of Jochen Stuz, John Seinfeld, and support staff at California Institute of Technology, and funding was provided by the California Air Resources Board (CARB). The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and/or READY Web site (http://www.arl.noaa.gov/ready.php) used in this publication. The authors thank Richard Flagan for technical assistance during the measurement campaign.