Particulate organic nitrates observed in an oil and natural gas production region during wintertime

Organic nitrates in both gas and condensed (aerosol) phases were measured during the Uintah Basin Winter Ozone Study from January to February in 2012. A high degree of correlation between total aerosol volume at diameters less than 500 nm and the particulate organic nitrate concentration indicates that organic nitrates are a consistent, if not dominant, fraction of fine aerosol mass. In contrast, a similar correlation with sub-2.5 μm aerosol volume is weaker. The C :N atomic ratio inferred from field measurements of PM2.5 and particulate organic nitrate is 34 : 1. Calculations constrained by the observations indicate that both condensation of gas-phase nitrates and heterogeneous reactions of NO3 /N2O5 are responsible for introducing organic nitrate functionality into the aerosol and that the source molecules are alkanes. Extrapolating the results to urban aerosol suggests organic nitrate production from alkanes may be a major secondary organic aerosol source.


Introduction
Submicron-sized aerosols affect the global radiative balance directly as a result of variation in their optical properties and indirectly via their effect on clouds.They modulate atmospheric composition by scavenging gas-phase material, including oxidants.The aerosols impact visibility and public health (Hallquist et al., 2009;Went, 1960;Kleinman et al., 1995).Observations have shown that submicron-sized aerosols typically contain ∼ 50 % organic material by mass (Zhang et al., 2007), and the sources of these organic aerosols have been the subject of considerable debate (Bahreini et al., 2012;Gentner et al., 2012;Shilling et al., 2013;Worton et al., 2013).Field observations have found organic aerosol in urban areas to correlate with anthropogenic emissions, and that they are formed in the atmosphere through chemical reactions of gaseous precursors (Hallquist et al., 2009).However, it has also been shown that urban organic aerosol contains significant amount of carbon that is "modern" (Lewis et al., 2004;Szidat et al., 2004;Zhang et al., 2013), i.e., from a biological source, such as biogenic emissions or biomass burning.It remains challenging to explain secondary organic aerosol (SOA) produced from modern carbon but controlled by anthropogenic emissions (Weber et al., 2007).Nitrogen oxides (NO x ) are possible candidates for modulating aerosol formation as they are primarily anthropogenic, and so an aerosol formation pathway mediated by NO x may explain this seemingly counterintuitive phenomenon in terms of radiocarbon age.One potential tracer for this process is organic nitrates in aerosol.
Recently, methods to identify organic nitrate in ambient aerosol have become available.Observations in chamber studies of the NO + / NO + 2 peaks in aerosol mass spectrometer measurements (Rollins et al., 2010a(Rollins et al., , 2009;;Farmer et al., 2010), by Fourier transform infrared spectroscopy (FTIR) of ambient aerosol (Day et al., 2010;Mylonas et al., 1991;Garnes and Allen, 2002) and by thermal dissociation-laser-induced fluorescence (TD-LIF) of ambient aerosol (Rollins et al., 2012(Rollins et al., , 2013(Rollins et al., , 2010b)), indicate that organic nitrates in ambient aerosol are observable, and that there are mechanisms to produce them in significant yields from common organic precursors.For example, in Bakersfield, California, during summer, Rollins et al. (2012) found that aerosolbound organic nitrate production contributed to as much as 30 % of the aerosol growth rate at night, with the nitrooxy group representing 8.4 % of the growth mass.
Recent studies of organic matter in ambient aerosol have focused on anthropogenic emissions of gasoline, diesel (Gordon et al., 2013;Jathar et al., 2013;Gentner et al., 2012;Bahreini et al., 2012), motor oil (Worton et al., 2014) and biogenic VOCs (Brown et al., 2009;Paulot et al., 2009;Froyd et al., 2010) as precursors.For long-chain aliphatics, laboratory experiments and simulations have demonstrated substantial contribution of organic nitrates in the resulting particulate matter (Jordan et al., 2008;Lim and Ziemann, 2009;Matsunaga and Ziemann, 2010), but there exist few field observations capable of assessing whether these mechanisms are representative of the ambient processes.Here we describe observations of organic nitrate aerosol observed in the Uintah Basin, Utah, during winter 2012.The measurement site is heavily influenced by oil and gas drilling operations and has negligible input of biogenic emissions, providing an excellent opportunity to explore the role of organic nitrates formed from aliphatic compounds in the production of ambient aerosol.

Methods
Observations of NO 2 , total organic nitrates and particulate organic nitrates (p ANs) were made as part of the Uintah Basin Winter Ozone Study (UBWOS) in January and February of 2012.The instruments were installed on a 19 m tower located on an operational oil and gas well pad containing a wellhead for water injection with a nearby unpaved access road.The measurement site (40.14370• N, 109.46718• W) is approximately 30 miles (∼ 50 km) south of Vernal, the nearest town in Utah.The aerosol measurements that are the focus of this manuscript were made from inlets 9 m above the ground.
NO 2 , total organic nitrates and p ANs were measured by TD-LIF with coupling to an inlet denuder as described in Rollins et al. (2010).TD-LIF is described in detail elsewhere (Day et al., 2002;Thornton et al., 2000).Briefly, in these experiments a CW 408 nm solid-state diode laser (Power Technology Inc., LDCU12/7639) was used to excite NO 2 molecules.The laser light was directed sequentially into three multi-pass white cells and total fluorescence due to NO 2 at wavelengths longer than 700 nm was detected using a red-sensitive photomultiplier tube (Hamamatsu H7421-50) behind dielectric filters that set the transmission window.The cell pressure was maintained at 3 Torr.
Simultaneous detection of organic nitrate species was accomplished by quantitative conversion of alkyl nitrate (-ONO 2 ) and peroxy nitrate (-OONO 2 ) moiety to NO 2 molecules through thermal decomposition at 17 ms residence time at 380 • C in a 0.25 in.OD quartz tube.The sampled air contained both gas-and aerosol-phase organic nitrates which, upon passing through the thermal dissociation (TD) oven, were converted into an excess NO 2 signal compared to the ambient NO 2 concentration monitored simultaneously in an unheated channel.To distinguish the particulate-phase component, an activated charcoal multi-channel denuder of 10 cm in length (MAST Carbon Inc.) was placed ahead of another 380 • C thermal dissociation oven to remove the gasphase organics (Rollins et al., 2010b) as well as NO 2 so that only aerosol-phase components remained.The particle transmission efficiency of the denuder was calculated to be 60 % for 20 nm diameter particles and over 90 % for particles larger than 70 nm diameter, ensuring detection of the vast majority of aerosol mass.To reduce intake of dust, a 2.5 µm cyclone was placed on the main inlet with a bypass pump maintaining the necessary total flow rate of 5 L min −1 .In addition to these TD-LIF measurements, N 2 O 5 (Wagner et al., 2011), peroxy acetyl nitrate (PAN) (Williams et al., 2000) and ClNO 2 (Roberts et al., 2009) were independently measured.Total alkyl nitrate ( AN) is defined as the measured TD-LIF signal at 380 • C, corrected for ozone effects (Lee et al., 2014) and with NO 2 , N 2 O 5 , PAN and ClNO 2 subtracted.The particle organic nitrate observations require no correction as gas-phase molecules are scrubbed by the denuder.We expect alkyl nitrates to be the dominant component of the particulate organic nitrate signal observed, because the peroxy nitrate concentration and the concentration of its precursors are much lower than the corresponding alkyl nitrates.
The TD-LIF instrument was calibrated hourly using locally generated zero air mixed with an NO 2 standard to give five different concentration levels, spanning a range from 0 to 20 ppb.The instrument zero was monitored twice per hour.Concentration data were reported to the NOAA archive (http://esrl.noaa.gov/csd/groups/csd7/measurements/2012ubwos/) at a time resolution of 1 min, averaged from 1 Hz raw data.The detection limit for the instrument at 1 min averaging time was 24 ppt for NO 2 and particulate nitrate and 34 ppt for total organic nitrate, defined as the 1σ value of the noise.The charcoal denuder was occasionally checked for saturation by introducing the calibration NO 2 gas mixture before, rather than after, the denuder section in a calibration routine, and no NO 2 breakthrough was observed.
Colocated aerosol instrumentations include measurements of the particle size distribution from 10 to 500 nm diameter range using a scanning mobility particle sizer (SMPS, TSI Inc.), from 0.7 to 10 µm diameter range using an aerodynamic particle sizer (APS, TSI Inc.), and an aerosol mass spectrometer (AMS, Aerodyne Inc.).Sub-2.5 µm aerosol filter samples were collected twice daily, one covering daytime and the other covering nighttime.Properties derived from these filter samples include total aerosol mass, total organic carbon (OC), total elemental carbon (EC) and cation concentrations (ion chromatography).A particle-into-liquid sampler (PILS) was colocated with the filter sampler.Meteorological conditions were recorded at the top of the 19 m tower, including wind direction, wind speed, temperature, pressure and relative humidity.Gas-phase measurements used in this analysis include gas chromatography with mass selective detector (GC-MS) and proton transfer mass spectrometry (PTR-MS) for VOC speciation, cavity ring-down spectroscopy (CRDS) for N 2 O 5 and NO 3 , chemical ionization mass spectrometry (CIMS) for ClNO 2 , and gas chromatography with electron capture detection (GC-ECD) for PAN.(For a comprehensive list, see: http://esrl.noaa.gov/csd/groups/csd7/measurements/2012ubwos/instruments.html.)

Observations
The concentration of total organic nitrates (including the contribution from ClNO approximately 300 ppt.PAN was about 250 ppt at night, increasing to 400 ppt in the late afternoon.A median value of 45 ppt p ANs was observed with a peak value around 150 ppt (Fig. 1).The level of p AN varied more slowly than ANs, except at times of pristine air intrusion, during which its concentration decreased promptly.From a multi-day perspective, p ANs were observed to accumulate during stagnant periods, as did long-lived trace gases including VOCs, NO 2 and ANs.Local VOC composition consisted predominately of alkanes, which accounted for ∼ 77 % of total OH reactivity (6.5 s −1 ) due to VOC.Alkenes, alkynes and aromatics accounted for only 2.3, 0.2 and 8.9 % of OH reactivity, respectively.
p AN was correlated with other aerosol measurements, the strongest of which (R 2 = 0.72) was with aerosol volume at diameters below 500 nm (Fig. 3).In contrast, the correlation with total aerosol volume up to 2.5 µm particle diameter is weaker (R 2 = 0.23).We believe this is due to the presence of mineral dust in the larger size fraction.The source of this dust is likely exposed dry soil on the numerous unpaved roads in the surrounding area.Correlations between p AN and gas-phase species such as propane (R 2 = 0.15) and NO 2 (R 2 = 0.1) are weak.

Aerosol composition inferred from observations
The inorganic components observed in aerosol during UB-WOS consist of mineral dust, salts and elemental carbon.
Since p AN data represent condensed-phase organic nitrooxy groups, the strong correlation of p ANs concentration with PM 0.5 volume is suggestive of the existence of both a significant organic component and a persistent nitrooxy functionality in this size range.Complementary evidence for this observation is obtained by correlating the concentration of aerosol Ca 2+ , typically found in minerals, with the total aerosol volume at diameters between 500 nm and 2.5 µm.The resulting high correlation (R 2 = 0.78) is in contrast to the one obtained for Ca 2+ with the PM 0.5 volume (R 2 = 0.03), again lending support to the assumption that the organic particles dominate the particle size range under 500 nm while inorganic components dominate the size range from 500 nm to 2.5 µm.
To obtain a quantitative estimate of organic/inorganic mass in PM 2.5 , linear decomposition of aerosol specific volume was used (Appendix A).The specific volume was calculated as the ratio of PM 2.5 aerosol volume measured by SMPS and APS divided by PM 2.5 aerosol mass from impactor filter samples.The use of filter data limits the number of independent estimations to two values daily, one during the day and one during the night.This method assumes that inorganic and organic components have distinctive values in density, and they form external mixture in the aerosol phase.The resulting equation relating the observed PM 2.5 specific volume ( v) to aerosol organic nitrate group mass fraction (f ONO 2 ) is shown in Eq. ( 1).
where v org and v dust represent the specific volume of aerosol organic and inorganic components, respectively, while γ is the mass ratio of the non-nitrate-containing organics to the organic nitrate group.Note here that the p AN measurement is insensitive to inorganic nitrate ions and allowed us to use it as an unambiguous tracer for organic components in the aerosol phase.Equation ( 1) predicts a linear relationship between the aerosol specific volume and organic nitrate mass fraction given that organic nitrates represent a constant fraction in the organic mass, a condition satisfied as demonstrated in Fig. 4. The y intercept of a plot of aerosol specific volume versus organic nitrate mass fraction gives the specific volume of the inorganic component directly.We obtained a value of 0.168 cm 3 g −1 , corresponding to a nominal density of the inorganic component of 5.95 g cm −3 , a value similar to iron(III) oxide (d = 5.24 g cm −3 ).Organic molecules with moderate oxygenation have a density ( v −1 org ) of approximately 0.85 g cm −3 (for example, 1-butyl nitrate (d = 0.882), tertbutyl nitrate (d = 0.867), nonanol (d = 0.827) and butanol (d = 0.81)).Using this estimate, we obtain a γ value of 11.This constrains the organic mass associated with organic nitrate group in aerosol to approximately 680 amu.This associated mass estimated from the light component of the ambient aerosol typically accounts for 53 % of the observed aerosol volume and 14 % of the aerosol mass.
It is possible to account for the contribution of soluble inorganic salts that are measured by PILS by subtracting out an inorganic salt component defined by the sum of NH + 4 , NO − 3 and SO 2− 4 groups and their average density of 1.76 g cm −3 .The sum of PILS ions can at times account for up to 50 % of PM 2.5 mass.The adjusted aerosol volume and mass after the subtraction of salt component is then analyzed using the same technique as above to extract a heavy and a light component.The R 2 value of 0.6 for the correlation of spe- cific volume to organic nitrate mass fraction is identical to the correlation presented above.A γ value of 9.6 is derived from this analysis We can now constrain the molecular formula of particulate organic components using this corrected γ value derived above.Together with aerosol mass spectrometer observations during high aerosol loading periods of an O : C value of 0.2 (S.Murphy, personal communication, 2013) and a generic chemical formula containing carbon, oxygen, hydrogen and organic nitrate (-ONO 2 ) of the form (CH 2 ) n O m (HONO 2 ) for a fully saturated molecule, we derive an elemental ratio of C : H : O : N = 34 : 69 : 10 : 1.Note that any linear combination of organic molecule mixtures giving the same average C : H : O : N ratio can satisfy this constraint, and may consist of both nitrates and non-nitrates.
The range of C : H : O : N ratios consistent with the observations can be estimated from the confidence interval associated with the slope and intercept of the linear regression in Fig. 4. At 95 % confidence interval, we estimate the uncertainty of the γ value to be ±17 %, given an organic matter density of 0.85 g cm −3 .We point out that the large uncertainty in the y intercept does not contribute significantly to the uncertainty of the result because the difference in specific volume is dominated by the organic component.Propagating this range gives C : H : O : N ratio between 28 : 57 : 9 : 1 and 40 : 81 : 11 : 1.Note that, although the estimated carbon number appears to be high relative to the implied carbon number for organic aerosol generated from the Deepwater Horizon plume (de Gouw et al., 2011), our estimate is relative to the organic nitrate functional group in the aerosol phase.The implication is that not all organics responsible for organic aerosol formation during UBWOS contained organic nitrate groups.

Daytime production
In the following section we attempt to close the daytime p AN budget using gas-phase oxidation of aliphatic molecules followed by partitioning of oxidation products into the aerosol phase.
p AN is thought to be exclusively secondary.Consider daytime processes in the alkane-rich environment observed during UBWOS: the oxidation of an organic molecule R starts with a reaction with OH radical.For a simplified schematic (Reactions R1 and R2) of a single oxidation step in the presence of NO, two generic products are formed with relative yields governed by the organic nitrate yield α.
The simple alkyl nitrate R(ONO 2 ) and products formed from the subsequent reactions of alkoxy radical RO are assumed to partition into the aerosol phase as a function of their respective vapor pressure.If the partitioning follows ideal solution behavior within the existing aerosol organics, the fraction of the organic products expected to end up in the condensed phase is represented as K p in Eq. ( 2), where P * represents the saturation vapor pressure of the organic molecule, N org the amount of organic molecules in the condensed phase in mol m −3 and k B as Boltzmann's constant.
The largest alkane reported during UBWOS 2012 was undecane (C 11 H 24 ).Using a power law we extrapolate the OH reactivity of larger alkanes by fitting a linear relationship (R 2 = 0.99) to the observed C 9 ∼ C 11 alkane reactivity in the log space.This approximation combines the decay in gasphase concentration due to reduction in vapor pressure and the increase in alkane OH reactivity with alkane size to generate a complete set of alkane consumption rates due to OH reactions.We then estimate the properties of the OH oxidation products from alkanes using a simplified scheme of three species for each carbon number group: an alkyl nitrate, a hydroxy nitrate and a hydroxy carbonyl, with branching ratios of α, (1 − α)α and (1 − α) 2 , respectively, as detailed in Appendix B. The absolute contribution of each type of oxidation product to aerosol formation is therefore the individual formation rate weighted by K p .The total aerosol yield attributable to this mechanism is obtained by summing the yield over all carbon number groups above C 6 .We used α values estimated with the method by Carter and Atkinson (1989) updated by Arey et al. (2001) at the appropriate temperature and pressure.The saturation vapor pressure P * of the reaction products follows the parameterization of SIM-POL.1 (Pankow and Asher, 2008).
Using this procedure, we calculate the particulate nitrate formation rate for the daytime period of 30 January, dur- ing which time the aerosol loading peaked at 2 µg m −3 .Using an approximate daytime OH concentration of 10 6 cm −3 (Edwards et al., 2013), the total production rate of aerosolphase organic nitrate groups is 3.4 ppt h −1 .The condensable molecules responsible for carrying the nitrooxy groups into the aerosol phase have carbon chain lengths in the range of C 10 to C 17 , as shown in Fig. 5.The representative C : H : O : N ratio of aerosol organics predicted by this alkane-only estimation is 18 : 37 : 5 : 1, or a non-nitrate to nitrate mass ratio of 4.5 : 1.This is about half of the value inferred directly from observations (γ = 9.6).One possible explanation for the difference is that approximately half of the non-nitrate carbonaceous component originates from oxidation of aromatic compounds (∼ 10 % of alkanes OH reactivity) or primary sources.Uncertainties in the saturation vapor pressure and deviation from ideal solution behavior may also contribute.The result obtained above only utilizes the alkane composition data.Independently, the total production rate of p AN can be calculated using time series of the measured p AN concentration by solving the following mass balance equation (Eq.3): Briefly, the total production rate of p AN (p(p AN)) is the rate of change of p AN concentration plus the loss rate.The loss is represented as a first-order process with a loss rate constant k mix (s −1 ) which can be estimated using the known production rate and concentration of n-propyl nitrate as a tracer (Lee et al., 2014).Turbulent mixing-out of the basin was the dominant driver of k mix .The inferred production rate during the same daytime period as the above analysis is 3.6 ppt h −1 , nearly identical to the estimate using alkane composition.It is noted that, by assuming the same loss characteristics as n-propyl nitrate, the effect of dry deposition is likely underestimated.Alternatively, the p AN formation rate calculated using loss characteristics of HNO 3 yields a daytime production rate that is 21 % higher (4.4 ppt h −1 ).This is likely an upper limit.

NO 3 / N 2 O 5 chemistry
In addition to daytime source of p ANs, nighttime chemical production may also be important.The dominant reactions are typically those initiated by NO 3 and N 2 O 5 radicals, either through gas-phase oxidation followed by condensation or through heterogeneous reactions on the surface of existing organic aerosol.Due to cold temperatures that make gasphase reactions of NO 3 less important by shifting the equilibrium towards N 2 O 5 , as well as the scarcity of unsaturated hydrocarbons observed, the condensation pathway is likely unimportant.Multiple lab studies on both environmental and synthesized aerosol particles (Gross et al., 2009;Zhao et al., 2011a, b;Xiao and Bertram, 2011;Bertram et al., 2009) have demonstrated that the reactive uptake of NO 3 or N 2 O 5 can be significant, and for certain class of organics molecules (e.g., alkenes and alcohols) can give high yield of organic nitrates as condensed-phase products.As opposed to daytime analysis, it is difficult to directly estimate the net nocturnal p AN production with Eq. ( 3) due to the lack of concentration variation and difficulty in estimating the loss term.However, we found that inclusion of heterogeneous production is necessary to explain the nighttime concentration of p AN and we characterize the heterogeneous reactions in the following modeling section.

Modeling the aerosol time series
A box model incorporating the above daytime and nighttime mechanisms was used to simulate the organic nitrate content of the aerosol.
The observed AN is assumed to represent the total concentration of organic nitrates produced due to photochemistry as described in Lee et al. (2014), with an effective saturation vapor pressure as a tuning parameter to determine the effective partitioning based on Eq. ( 2).The bidirectional exchange is calculated according to gas kinetic theory and detailed balance derived from the saturation vapor pressure accounting for the Kelvin effect.
For the nighttime chemistry, we introduce a new parameter, the retaining coefficient (ζ ), defined as the probability of reactive uptake yielding condensed-phase organic nitrates given a gas molecule-surface collision.This differs from the reactive uptake coefficient by excluding non-nitrate forming channels.The p AN production is calculated from gas molecule-surface collision rate corrected for diffusion transport and then multiplied by ζ .The organic nitrates formed through this pathway are assumed to remain in the condensed phase.
Loss of p ANs is assumed to follow the loss of aerosols.The most important process is turbulent mixing, with a lifetime of ∼ 8 h during daytime periods.The full time series of first-order loss rate is calculated using tracer methods (Lee et al., 2014) and applied to the simulation.We used HNO 3 to estimate the loss rate which gives a more realistic loss profile during the evening, when the boundary layer is stable.
The resulting time series of predicted concentrations are shown in Fig. 6, where the top panel describes the two consecutive accumulation periods and the lower panel extends the simulation to all observations.We find an effective saturation vapor pressure of 26 ppb for the ensemble of total organic nitrate species and NO 3 reactive uptake coefficient ζ NO 3 of 0.1 for the full length of simulation.The model is highly correlated with the observations (slope of 0.98 and R 2 value of 0.72) over the 2-week period shown in the top panel of Fig. 6.A reduced correlation with a slope of 0.72 and an R 2 of 0.66 is observed for the full data (bottom panel of Fig. 6).This fitted effective saturation vapor pressure is consistent with one calculated from the VOC speciation (28 ppb ANs).The major discrepancies in the extended simulation arise when organic aerosol mass becomes more abundant in the super-0.5 µm size range (up to 36 % of total p ANs), particularly during the period from 12 to 21 February.It is possible that organics were present as external coatings on the inorganic minerals of large size particles, leading to an enhanced surface to volume ratio relative to a pure particle of the same organic mass.The result is underestimation in predicted organic nitrate content due to discrepancies in the aerosol mixing state.Also, the possibilities of secondary chemistry in the presence of inorganic salts cannot be ruled out.
The ζ NO 3 value of 0.1 should be interpreted as a projection of the overall reactivity onto NO 3 reactions, since NO 3 and N 2 O 5 interconvert rapidly and both species may contribute to heterogeneous reactions.Given an observed median NO 3 / N 2 O 5 ratio of 0.007, the heterogeneous chemistry may be equally satisfied with an N 2 O 5 -based ζ N 2 O 5 of 8 × 10 −4 , or any linear combination of the two chan-nels.However, we point out that due to the enhancement of N 2 O 5 lifetime as a result of the cold temperature and high NO x concentrations encountered, the heterogeneous reaction cannot be dominated by NO 3 as this requires a value of ζ NO 3 larger than even aerosols made of pure unsaturated organic molecules such as solidified oleic acid (γ NO 3 = 0.076) and conjugated linoleic acid (γ NO 3 = 0.08) as observed in laboratory studies (Gross et al., 2009).Therefore, N 2 O 5dominated heterogeneous reactions with hydroxy groups in the aerosol phase are a more likely source of nighttime p AN production, and are within range of reported reactive uptake coefficient measured on surface of glycerol particles (γ N 2 O 5 = 8.14 × 10 −4 ) and wintertime aerosol in Colorado (γ N 2 O 5 ∼ 0.01) (Wagner et al., 2013).More laboratory experiments with emphasis on condensed-phase products under low-temperature conditions are necessary to verify the values obtained here for saturated organic aerosol systems.Finally, it is noted that the nocturnal production of p ANs (6 ppt h −1 ) does not constitute a significant local sink of NO 3 (production rate of 150 ppt h −1 ).

Implications
Our particulate organic nitrate measurements during wintertime in the Uintah Basin, Utah, represent a unique opportunity to characterize the chemistry of alkane-derived SOA under ambient (albeit cold) conditions.This is relevant to environments when anthropogenic activities heavily influence the VOC composition.According to the study of Gentner et al. (2012), both gasoline and diesel fuel sampled at various locations in California were primarily alkanes, with an overall longer chain length in the diesel fuel.Tailpipe emissions of unburnt diesel fuel as well as motor oil (Worton et al., 2014) may represent the predominant source for large alkanes observed in cities (Boynard et al., 2014).For example, VOC enhancement ratios observed in Los Angeles showed a non-decreasing trend from n-nonane to undecane, the largest alkane reported (Borbon et al., 2013).Similar mixing ratios have also been observed in Sacramento, California (Steiner et al., 2008).This is different from the decreasing trend observed in Utah, where the VOC sources are evaporative, but rather consistent with the composition from direct emissions of liquid diesel for which the distribution of C 10 to C 20 alkanes are relatively flat.
We calculated the potential organic nitrate aerosol formation from alkanes using the liquid fuel composition tabulated in the supporting information of Gentner et al. (2012) and the partitioning method detailed in Sect.4.2 for OH-initiated oxidation.OH reaction rates for alkanes larger than dodecane are parameterized (Kwok and Atkinson, 1995).For an organic aerosol loading of 2 µg m −3 , the organic nitrate aerosol yield is 14 wt % for diesel fuel and 0.004 wt % for gasoline, both calculated at 298 K. Since it was estimated by Gentner et al. (2012) that both gasoline and diesel vehicles emit a similar amount of VOC by weight in the Bakersfield region (Kern county), a ratio of 44 % (gasoline) to 56 % (diesel), the diesel emission dominates the source strength of particulate organic nitrates.We further estimate the potential p AN concentrations according to the mass yield calculated above and tailpipe VOC emission of 7 µg m −3 from diesel vehicles, a number we derived based on the SOA production estimation in Gentner et al. (2012).The result is 46 ppt p AN due to diesel traffic emissions in the Bakersfield region, a value accounting for 77 % of the observed daytime ambient concentration by Rollins et al. (2013), who reported local p AN concentration of ∼ 60 ppt during daytime periods of the Cal-Nex 2012 campaign.Contributions from local biogenic precursors as well as fugitive losses from oil and gas productions may account for the remaining particulate nitrates.However, we point out that photochemical aging is required to achieve the yield from our estimation and the above value should be interpreted as an upper limit.
The above example of particulate organic nitrate production around Bakersfield region illustrates the possibility of distinct pathways responsible between daytime and nighttime periods.The saturated but much heavier alkanes will contribute during the day from OH oxidation and partition more efficiently into the aerosol phase, while the more reactive but generally lighter biogenic emissions may dominate nighttime production due to NO 3 and N 2 O 5 chemistry.This is of particular interest in regions with representative VOCs consisting of a mixture of anthropogenic and biogenic contributions.

Conclusions
We present PM 2.5 particulate organic nitrate concentration measurements obtained in wintertime Utah using the TD-LIF technique.Of the median 1 µg m −3 organic aerosol estimated, we found organic nitrate to be a consistent portion of the organic mass within the sub-0.5 µm particle size ranges with an average C : H : O : N elemental composition of 34, 69 : 10 : 1, likely as a mixture of C 10 to C 17 organic nitrates and oxygenates.With the help of a box model, we demonstrate that the particulate organic nitrate concentration observed can be reproduced by gas-phase condensation and heterogeneous chemistry of N 2 O 5 .Both channels contribute almost equally, consistent with the lack of day-night change observed in condensed-phase organic nitrate content.By applying our analysis to the California central valley region, we demonstrate that diesel tailpipe emissions can potentially contribute to a significant portion of ambient particulate organic nitrates observed.
In order to obtain a converging estimation with respect to the long-carbon chain end of the VOC spectrum, it is necessary to extrapolate the contribution of heavy VOCs beyond the measurement which terminates at undecane.Using a linear fit in the log space of the grouped VOC reactivity with specific carbon number, we obtained an estimation, shown in Eq. (B1), for the 30 January accumulation period: ln (kx) = −0.5893× n + 3.9223, (B1) where kx is the total reactivity in units of s −1 of alkanes with carbon number n.The total aerosol source of each molecule type within each carbon number class is then calculated in the same way shown for the hydroxy nitrates of size n (Eq.B2).Note that K p is the fraction of the species in the aerosol phase, calculated using Eq. ( 2) in the main text.The total nitrate groups incorporated into the aerosol phase are then calculated by summing over all carbon groups of alkyl nitrates and hydroxy nitrates.Other functional groups are calculated similarity with application of appropriate weightings.For example, the total CH 2 group contribution is calculated according to Eq. (B3).
2 and N 2 O 5 ) and particulate nitrates (p ANs) are shown in Fig. 1.After correcting for PAN, ClNO 2 and N 2 O 5 , ANs account for an afternoon peak of 40 % NO y and exhibit a strong diurnal pattern, reaching a median value of 2.2 ppb at local noon as shown in Fig. 2. At night high concentrations of N 2 O 5 and ClNO 2 (∼ 0.6 ppb combined) were present and ANs decreased to

Figure 5 .
Figure 5. Source strength of functional groups for molecules contributing to aerosol formation on 30 January categorized by carbon number.

L.
Lee et al.: Particulate organic nitrates observed during UBWOS 2012 bon chain length in the R group at 273 K. α R + OH → R(ONO 2 ) (BR1)

nn
× (S n {R(ONO 2 )} + S n {R(OH)(ONO 2 )} +S n {R(=O)(OH)}) (B3) Hourly time series from measurements of total organic nitrates plus N 2 O 5 and ClNO 2 , particulate organic nitrate (p AN), and total aerosol volume measured by SMPS and APS with filter-sampled PM 2.5 mass (red dots, secondary axis) and local wind speed.
Time series of predicted particulate organic nitrate concentration from box model simulation for two accumulation events (top panel) and the campaign period (bottom panel) as stacked areas showing contribution from each of the two mechanisms responsible.The observations are plotted in yellow.