Atmospheric Chemistry and Physics on the Volatility and Production Mechanisms of Newly Formed Nitrate and Water Soluble Organic Aerosol in Mexico City

Measurements of atmospheric gases and fine particle chemistry were made in the Mexico City Metropolitan Area (MCMA) at a site ∼30 km down wind of the city center. Ammonium nitrate (NH 4 NO 3) dominated the inorganic aerosol fraction and showed a distinct diurnal signature characterized by rapid morning production and a rapid midday concentration decrease. Between the hours of 08:00–12:45, particulate water-soluble organic carbon (WSOC) concentrations increased and decreased in a manner consistent with that of NO − 3 , and the two were highly correlated (R 2 =0.88) during this time. A box model was used to analyze these behaviors and showed that, for both NO − 3 and WSOC, the concentration increase was caused primarily (∼75–85%) by secondary formation, with a smaller contribution (∼15–25%) from the entrainment of air from the free troposphere. For NO − 3 , a majority (∼60%) of the midday concentration decrease was caused by dilution from boundary layer expansion , though a significant fraction (∼40%) of the NO − 3 loss was due to particle evaporation. The WSOC concentration decrease was due largely to dilution (∼75%), but volatilization did have a meaningful impact (∼25%) on the decrease, as well. The results provide an estimate of ambient SOA evaporation losses and suggest that a significant fraction (∼35%) of the fresh MCMA secondary organic aerosol (SOA) measured at the surface volatilized.


Introduction
Secondary organic aerosol (SOA) is an important component of ambient particulate matter. It frequently comprises a large fraction of the total organic carbon aerosol (OC), often greater than 50% on a carbon mass basis (Kanakidou et al., 2005). Despite its 20 high relative abundance and ubiquitous presence in the atmosphere, many characteristics of SOA are not well understood and require better characterization. These include chemical composition, relative contributions from anthropogenic and biogenic sources, and major formation mechanisms. These uncertainties preclude the accurate modeling of SOA concentrations (e.g. Henze and Seinfeld, 2006 Interactive Discussion pounds (VOCs), which leads to less volatile products that may condense into the aerosol phase. Many of the condensed species are subsequently semi-volatile (i.e. they exist in both the gas and particle phase at typical atmospheric conditions) and contribute significantly to organic aerosol mass (Seinfeld and Pankow, 2003). Though this broad conceptual framework is generally accepted, the semi-volatile nature of SOA 5 is not well constrained. SOA chemical composition determines its volatility. In general, low molecular weight compounds are more volatile than high molecular weight compounds (e.g. oligomers). Multiple smog chamber experiments have analyzed SOA volatility, including Offenberg et al. (2006), Kalberer et al. (2004), and An et al. (2007). These studies have pro-10 vided information about the volatility of specific secondary products (i.e. the volatility of SOA from α-pinene/ozone and α-pinene/NO x experiments . Thermal properties of ambient OA and SOA have been examined by analyzing the data (thermograms) from certain organic carbon analyzers (Yu et al., 2004), but the analyses were conducted at temperature heating profiles such that elucidating information on 15 volatility in conditions relevant to the atmosphere was not possible.
Numerous studies focused on ambient aerosol have examined a semi-volatile fraction of the bulk organic aerosol using a sampling-based operational definition (e.g. Modey and Eatough, 2004). Semi-volatile organic matter (SVOM) is defined as those compounds which penetrate a quartz filter but are collected on a backup charcoal im-Introduction are reported at ambient temperature and pressure. Inlets for aerosol measurements were located approximately 10 m above the ground and were fitted with cyclones (URG, Chapel Hill, NC) to select particles with aerodynamic diameters less than 2.5 µm (PM 2.5 ). Water-soluble organic carbon (WSOC) aerosol was measured with a particle-into-liquid sampler (PILS) coupled to a total or-15 ganic carbon analyzer (Sievers, Boulder, CO) (Sullivan et al., 2004). Aerosol inorganic ionic components were measured by a PILS coupled to dual ion chromatographs (IC) (Metrohm, Herisau, Switzerland) (Orsini et al., 2003) to quantify a suite of both cations and anions. A Sunset Labs OCEC field analyzer (Sunset Laboratory Inc., Tigard, OR) following Bae et al. (2004) recorded particulate organic carbon (OC). Hydroxyl (OH) 20 and peroxy (RO 2 ) radical concentrations were measured with a chemical ionization mass spectrometer (CIMS) (Sjostedt et al., 2007). Measurements of nitric oxide (NO) and photolysis rates (J values) were made with a chemiluminescence detector (Ryerson et al., 2000) and an actinic flux spectroradiometer (Shetter and Muller, 1999) et al. (1994).

Results and discussion
Measurements at T1 were performed from 7-31 March 2006, however the present analysis focuses only on the period of 27-29 March. The selected period is fairly 5 representative of the month, but with the greatest consistency in the diurnal behavior of PM 2.5 mass and the highest daily maximum NO − 3 concentrations (Fig. 1). It was also the period least impacted by biomass burning (Stone et al., 2007), which along with SOA is a source of fine particle WSOC (Sullivan et al., 2006). an increase of 13.8 µg m −3 (more than 300%) over pre-sunrise concentrations. A box model was constructed to investigate the contribution of secondary nitrate formation to this AM increase on 3/27 and 3/28, the days for which boundary layer height (BL) data were available. From Seinfeld and Pandis (1998), the governing equation for an Eulerian box model with a well-mixed and expanding boundary layer is

NH 4 NO 3 Formation
where c i is the species i concentration, c 0 i is the species i background concentration, c a i is the species i concentration aloft, Q i is the emission rate, R i is the chemical 4815 Introduction

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion production rate, S i is the loss rate, u is the wind speed (with constant ∆x direction), and H(t) is the BL height. For the purpose of modeling NO − 3 at the T1 site during the hours of 07:30-11:00, a number of assumptions were made to simplify the model. The removal rate, S i , was neglected from the model because the dry deposition velocity for 0.1-2.5 µm particles (0.01-0.2 cm s −1 (Seinfeld and Pandis, 1998)) is not important 5 on the timescale of several hours. Wind speeds, which generally ranged from 1-2 m/s during the period of interest, were too low to make advection an important process on the timescale of several hours, either. Each day before 11:00, the ISORROPIA-II thermodynamic equilibrium model (Nenes et al., 1998;Fountoukis and Nenes, 2007) predicted that greater than 90% of total nitrate (NO − 3 + HNO 3 (g)) partitioned in the 10 particle phase due to thermodynamic conditions that favored the condensed phase, including low T , high RH, and excess ammonia gas (NH 3 ) (Fountoukis et al., 2007). Thus all HNO 3 (g) produced was assumed to contribute to the observed aerosol nitrate loading. The HNO 3 (g) formation rate (µg m , where k is the temperature-dependent kinetic rate constant (Sander et al., 2006), NO 2 15 was calculated from observations of NO, O 3 , J values, and RO 2 assuming photochemical steady state, and OH was measured. Applying the stated assumptions gives The model used BL depth data from Shaw et al. (2007) and NO results indicate that approximately 75% of the observed nitrate concentration increase was due to secondary photochemical production while approximately 25% was due to entrainment from the free troposphere. Salcedo et al. (2006) observed similar temporal behavior in the MCMA, with morning production of HNO 3 (g) immediately followed by NH 4 NO 3 formation, and the trend appears to be a common phenomenon in the MCMA. Introduction

Nitrate loss
The NO − 3 peak was short lived, as concentrations began to drop precipitously at approximately 11:00 (Fig. 1). Over the period between ∼11:00 and ∼12:45, the NO − 3 concentration decreased by an average of 14.9 µg m −3 , an 82% drop from the peak concentration. This decrease occurred while the HNO 3 (g) production rate remained 5 relatively high. Causes of the dramatic NO − 3 concentration decrease following the late morning maximum were likely some combination of boundary layer dilution and chemical processes (i.e. particle evaporation). Precipitation data showed that wet deposition was not important on any of the days in question. We first investigate the role of chemical processes, followed by the effects of meteorological processes on ambient 10 concentrations.
The effect of a thermodynamic phase shift on NO − 3 was examined with ISORROPIA-II. The model uses inputs of ambient temperature, relative humidity, and total concentrations (gas + particle) of inorganic species to predict the equilibrium state of each species along with the aerosol water concentration. Model results are shown in Fig. 2. 15 Based on rising ambient temperature and decreasing relative humidity, the model predicted a rapid phase shift of NO − 3 to HNO 3 (g). NH 4 NO 3 is semi-volatile and is sensitive to changes in RH and temperature (both of which directly impact aerosol water content), with higher temperature (lower RH) favoring the gas phase and lower temperature (higher RH) favoring the particle phase. It is notable that the predicted equilibrium shift 20 coincided with the large observed decrease in NO − 3 concentration (Fig. 2). The model predicted an average morning aerosol water concentration maximum of 15.4 µg m −3 , followed by a rapid concentration decrease to 1.2 µg m −3 that preceded the nitrate concentration drop by approximately 1 h. Given additional time for aerosol evaporation following the water loss, the timescale of this observed delay is consistent with the 25 analysis of Fountoukis et al. (2007). The observational data and model results agree well and suggest that, despite continued production of HNO 3 (g) and available NH 3 (g), changing ambient conditions produced a rapid shift in the equilibrium of the system Interactive Discussion towards the gas phase and were responsible for some fraction of the observed NO − 3 concentration decrease. Dilution must also be considered because the boundary layer underwent rapid expansion during the nitrate decrease. The box model with BL depth data from Shaw et al. (2007) was used to assess the impact of dilution on 3/27 and 3/28, the two 5 days for which BL depth data were available. On these two days, dilution accounted for approximately 66% of the observed nitrate concentration decrease. There were two major differences between the assumptions used to generate Eq. (2) and those used for the concentration decrease analysis. Though the photochemical production of HNO 3 (g) continued while the nitrate concentration fell, this term (R i ) was excluded 10 from the model because, unlike the morning hours, the fraction of total nitrate in the aerosol phase was not constant and was rapidly shifting to favor the gas phase. Additionally, between the hours of 11:00 and 12:45, the NO − 3 concentration aloft, c a i , was assumed to be zero. Both of these simplifications result in an overestimation of the dilution impact. Thus, the estimated 66% dilution loss is an upper bound and indicates 15 that at least 34%, but likely more, of the observed nitrate loss was due to particle evaporation. Because of the observed consistency in the diurnal trends of meteorological parameters and pollutants over the three day period, it can be assumed that the losses from dilution on 3/29 were similar to those on 3/27 and 3/28. In further support of this analysis, an approximation of BL dilution on 3/29 was made using two independent 20 chemical tracers, water vapor and carbon monoxide (CO). Assuming aloft CO and water vapor concentrations of 100 ppb and 6500 ppm, respectively, the nitrate losses from dilution and evaporation were estimated at 70% and 30%, respectively. Though there is a high amount of uncertainty in this estimate, the results on 3/29 were consistent with those obtained using the measured BL depth and suggest that the behavior on 25 3/29 was similar to that on 3/27 and 3/28.

Photochemical WSOC production
The PILS-WSOC method provides an approximate measurement of SOA mass (Sullivan et al., 2006;Kondo et al., 2007). Biomass burning is also a source for particulate WSOC and may account for some of the measured background WSOC observed in Mexico City, however, the biomass burning tracer acetonitrile (de Gouw et al., 2006) 5 did not follow the diurnal WSOC trend on these selected three days and so we conclude that the consistent diurnal profile is mainly due to SOA formation. For the three days, the WSOC concentration increased an average of 1.6 µg C m −3 (a 50% increase) over the three-hour period from 08:00 to 11:00 each day and had an average peak of 4.8 µg C m −3 before noon (Fig. 1). Similar to NO − 3 , the WSOC con-10 centration experienced a daily decrease shortly after the morning maximum. Between ∼11:00 and ∼12:45, the WSOC concentration decreased an average of 2.7 µg C m −3 , a 56% decrease from the peak concentration.
Overall, WSOC and NO − 3 were highly correlated (R 2 =0.80), indicating similar sources and atmospheric processing. However, between the hours of 08:00-12:45, 15 a period that included the concentration increase and decrease, the correlation was even stronger (Fig. 3). During this time the coefficient of determination (R 2 ) was 0.88 (all three days), 0.92 (3/27), 0.85 (3/28), and 0.98 (3/29) (Fig. 3). The high correlations show that during these periods the sources and atmospheric processes affecting both species were likely the same. Thus, similar to NO − 3 , it is probable that a large fraction 20 of the WSOC concentration increase was due to secondary formation. Multiple studies have observed increased WSOC:OC ratios during periods of SOA enhancement (Sullivan et al., 2004;Kondo et al., 2007). WSOC comprised, on average, 44% of OC for the three-day period, however, the ratio increased throughout the morning and was 52%, on average, by the late-morning peak. The correlation with solar radiation and 25 the increasing WSOC:OC ratio also provides evidence for photochemical SOA formation. Because the contributions from biomass burning were presumed to be small, it is likely that anthropogenic sources were responsible for the majority of the observed − 3 , and WSOC. Furthermore, the strong WSOC-NO − 3 correlation is also suggestive that OH was the primary oxidation pathway in the observed SOA formation. Similar photochemical production of SOA has been observed previously in the MCMA (Salcedo et al., 2006).

NO
The high correlation between WSOC and NO − 3 indicates that the processes inducing the WSOC and NO − 3 concentration decreases were also similar and that the volatility of WSOC was similar to that of NO − 3 under the thermodynamic conditions encountered during these observations. This implies that roughly two-thirds (1.8 µg C m −3 ) of the observed WSOC concentration decrease was due to dilution, while roughly a 10 third (0.9 µg C m −3 ) was due to evaporation losses. If the same reasoning is applied to both NO − 3 and WSOC losses, then the estimate also represents a lower bound on the evaporation losses and indicates that at least 20% of the WSOC (at peak concentration) partitioned back to the gas phase. It is possible that most of the WSOC produced in the early morning was semi-volatile, but the rapid dilution due to boundary layer 15 expansion limits the extent to which this process can be investigated by this approach.
These findings provide insight into the semi-volatile nature of ambient SOA, an area in which knowledge is currently limited. Specifically, fresh SOA with a large anthropogenic component in fairly non-acidic aerosol possesses similar semi-volatile properties to NH 4 NO 3 at the T and RH recorded here (Fig. 1). Grieshop et al. (2007) found 20 the partitioning of fresh biogenic SOA generated in a smog chamber to be reversible when subjected to isothermal dilution which, when combined with the present results, suggests that fresh SOA (age <4-5 h) of both biogenic and anthropogenic origin may partition back to the gas phase to maintain equilibrium. Overall, these results support the findings of Wilson et al. (2006)  The results also suggest that, similar to nitrate, the loss of aerosol water did impact WSOC concentrations. A study by Loeffler et al. (2006) found that when aqueous particles containing glyoxal and methylglyoxal were evaporated, there was minimal loss of particulate organic mass due to oligomer formation. If the observed evaporation of liquid water did induce oligomer formation, then it occurred here on a small scale that 5 did not have a substantial effect on the stabilization of semi-volatile SOA species. The present results do not provide evidence for or against significant oligomer formation, but they do suggest a minimum time delay of several hours before potential oligomer formation and overall SOA stability occurs on a large scale. When considering the question of oligomer formation, the observed aerosol acidity is also noteworthy. Studies 10 suggest that oligomer formation is enhanced in the presence of acidic particles (Gao et al., 2004). In this case if oligomer formation did occur, it was in the presence of a fairly non-acidic aerosol, as evidenced by an abundance of available NH 3 (g) and the simultaneous presence of significant SO 2− 4 and NO − 3 concentrations in the aerosol phase. 15 Finally, airborne measurements made over the sampling sites during MILAGRO have also been used to investigate the production of SOA and NO − 3 by changes in concentrations with respect to CO as plumes were advected away from the urban region (Kleinman et al., 2007). They found that NO − 3 and SOA production were not similar, as observed on the surface; SOA was still produced relative to CO when net NO − 3 20 production had apparently stopped. They concluded that this was due to the more semi-volatile nature of NO − 3 than SOA. These differences between the surface measurements and aloft may be due to lower temperatures and higher RH aloft, and high but different volatility dependences of SOA and NO − 3 . These types of differences show that extrapolating observations of SOA production rates recorded in one study directly 25 to other environments may be problematic.

Conclusions
In 2006, gas phase and PM 2.5 aerosol measurements were made in the Mexico City Metropolitan Area as part of the MILAGRO field study. For three days (27)(28)(29), NO − 3 and WSOC were highly correlated (R 2 =0.80). This correlation was even stronger (R 2 =0.88) during the morning and early afternoon hours when concentra-5 tions of these species grew quickly and then fell even more rapidly. The morning NO − 3 concentration increase was predominantly from secondary photochemical production. The NO − 3 concentration peaked before noon and subsequently experienced a rapid concentration decrease. Box model and aerosol equilibrium model results indicate that at least one-third of the nitrate concentration decrease was due to particle evaporation 10 while the rest was due to dilution from boundary layer expansion. Assumptions used to generate the box model make the evaporation estimate a lower bound, and indicate that the impact of evaporation was likely higher. Due to the high correlation between WSOC and NO − 3 , enhanced during the morning and early-afternoon hours, it is likely that the sources and chemical processes affecting both species were similar. This 15 means that the WSOC concentration increase, while less dramatic than that of NO − 3 , was caused predominantly from SOA formation. The source of this SOA was likely anthropogenic, with OH reaction being the initiator. Similarly, the WSOC concentration decrease was likely due partly to boundary layer dilution and partly to aerosol evaporation, with evaporation accounting for at least one-third of the decrease. The results 20 provide novel insight into the semi-volatile nature of fresh, anthropogenic, ground-level SOA, however their translation to other environments requires further examination. Weingartner, E., Frankevich, V., Zenobi, R., and Baltensperger, U.: Identification of polymers as major components of atmospheric organic aerosols, Science, 303 (5664)