Secondary organic aerosol production from modern diesel engine emissions

Secondary organic aerosol production from modern diesel engine emissions S. Samy and B. Zielinska Division of Atmospheric Sciences, Desert Research Institute, 2215 Raggio Pkwy., Reno, NV 89512, USA Received: 24 July 2009 – Accepted: 13 August 2009 – Published: 27 August 2009 Correspondence to: B. Zielinska (barbara.zielinska@dri.edu) Published by Copernicus Publications on behalf of the European Geosciences Union.


Introduction
Oxidation of hydrocarbons from biogenic sources contributes to the major fraction of secondary organic aerosol (SOA) in the global atmosphere (Griffin et al., 1999;Szidat et al., 2004).Yet anthropogenic SOA is thought to contribute significantly to the total aerosol mass in urban/downwind urban areas, and may contribute up to one third of global Correspondence to: B. Zielinska (Barbara.Zielinska@dri.edu)SOA (Volkamer et al., 2006).It has recently been suggested that underestimation of SOA formation in models is in part due to the oxidation of uncharacterized semivolatile organic compounds (SVOC) from vehicle emissions (Robinson et al., 2007;Weitkamp et al., 2007).Primary diesel exhaust (DE) is a complex mixture, composed of hundreds of organic and inorganic compounds in both gas and particle phases (Sagebiel et al., 1996;Zielinska et al., 2004a, b).The physical and chemical transformations of primary DE in the atmosphere are a concern for both human health and climate regulation purposes (Seinfeld and Pandis, 1998;Zielinska, 2005).
Although a number of laboratory studies have been carried out to elucidate the photochemical and oxidative processes that could lead to SOA formation (Kamens et al., 1981(Kamens et al., , 1982;;Stern et al., 1987;Pandis et al., 1991;Forstner et al., 1997;Kleindienst et al., 2002Kleindienst et al., , 2004;;Kroll, 2006), most of these studies have focused on single precursor compounds or simplified chemical mixtures investigated in simulation chamber experiments with or without inorganic seed particles (such as ammonium sulfate).For example, the reaction of aromatic and polycyclic aromatic hydrocarbons (PAH) that are important constituents of motor vehicle emissions, have been studied in a series of photochemical chamber experiments (Zielinska et al., 1989a, b;Atkinson and Arey, 1994;Arey, 1998;Volkamer et al., 2001;Johnson, 2004Johnson, , 2005;;Martin and Wirtz, 2005;Hamilton et al., 2005;Gomez Alvarez et al., 2007).A comparison of these laboratory studies with results obtained from analysis of ambient aerosol particles is not straightforward, and discrepancies in results (e.g.aerosol yields and partitioning) may be attributed to the more simple chemical mixture and aerosol matrices in which these reactions occur, compared to those in the real atmosphere (Graber and Rudich, 2006).Thus, there is a need to investigate transformation processes of real mixtures arising from various combustion processes such as DE (Kroll and Seinfeld, 2008).
Published by Copernicus Publications on behalf of the European Geosciences Union.
S. Samy and B. Zielinska: Secondary organic aerosol production from modern diesel engine emissions Atmospheric transformations of DE and associated SOA production was examined in a set of experiments carried out in Valencia, Spain, with the aid of the European Photoreactor (EUPHORE) outdoor simulation chamber.The specific aims of this study were to characterize the gas and particle phase products of atmospheric transformations of modern diesel engine emissions under the influence of sunlight, ozone, hydroxyl radicals (OH), and nitrate radicals (in the dark); and explore the changes in biological activity of DE before and after the atmospheric transformations take place (Zielinska et al., 2009).
Modern light-duty diesel engines emit relatively low amounts of volatile organic compounds (VOC) and particulate matter (DPM), while NO x emissions range from 300-500 ppm.At these high levels of NO x (>1 ppm in-chamber mixing ratio for 30 µg m −3 DPM in-chamber mass concentrations), the formation of OH radicals and ozone will be extremely low and any OH or ozone produced will be rapidly removed from the system through reactions with NO to form NO 2 and nitric acid (Finlayson-Pitts and Pitts, 2000).The photochemical transformation of organic compounds will be effectively shut down, making the photochemical experiments carried out in the sunlight not efficient or relevant to atmospheric conditions.Therefore, it was necessary to remove excess NO x by developing a NO x denuder (Zielinska et al., 2009).In addition, the modern diesel engine emits very low amounts of VOC, which resulted in typical concentrations of less than 10 ppb of VOC in the chamber.In most urban ambient atmospheres, VOC excess due to various source emissions, such as gasoline vehicles, is chemically available (Finlayson-Pitts and Pitts, 2000;Zielinska et al., 2004a).Therefore, in some experiments additions of VOC and/or OH radical to the chamber atmosphere were made to enhance the reactivity of the mixture and simulate urban atmospheres.
This article focuses on the SOA production and characterization of polar organic compounds (POC) from aged DE in a series of experiments carried out at EUPHORE in 2006.A more comprehensive discussion of the project is available in Zielinska et al. (2009).

EUPHORE facility and experimental design
The EUPHORE chamber is currently one of the best equipped and largest (∼204 m 3 ) outdoor simulation chambers in the world (Becker, 1996), which allows investigation of atmospheric transformation processes under psuedorealistic ambient conditions (dilution in the range of 1:300-400).The chamber is made of a fluorine-ethene-propene (FEP) foil, with a thickness of 127 µm (Du Pont).This foil transmits more than 80% of the radiation in the wavelength range of 280-640 nm, and the surface-to-volume ratio of the chamber is 0.96 m −1 (Becker, 1996).All experiments were conduct at a relative humidity <1%.
A modern diesel engine (i.e. a common rail direct injection, turbocharged, inter-cooled engine) was obtained from Ford Motor Company in England.
The 1.8 L, Lynx V277 90PS Stage 3, Delphi Fuel System, Fixed Geometry Turbo engine is used in Ford Focus and Transit Connect automobiles.The dynamometer system is equipped with a continuous gas analyzer (Horiba), which monitors the engine-out gaseous emissions (CO, CO 2 , NO x , THC, O 2 ).The system also records fluid temperatures, RPM, and load.The engine was operated under a load of 60 Nm, which represents approximately 50% of total engine power.A low sulfur diesel fuel (47 ppm, with 15% total aromatic content) was utilized, and steady-state conditions were achieved (30 min warm-up) prior to each DE injection.
Before starting each experiment the chamber was flushed for a few hours with purified air (continuous monitoring of gas and particle constituents was performed during flushing), and the DE was introduced to the chamber through a split valve system when the water and oil temperatures reached approximately 88 and 90 • C, respectively.A mixing fan (4000 m 3 h −1 ) within the chamber was used to assure the rapid mixing of constituents.The fan was switched off after approximately 5 minutes to avoid excessive loss of particles to the chamber wall (i.e.impaction due to continued turbulence).If additional reactants were introduced to the chamber (e.g.VOCs), the fan was switched on for a short duration (1-2 min) to achieve fast mixing of the reactants.
The initial two series of experiments were carried out in January 2005 (winter campaign) and May 2005 (summer campaign), followed by a second summer campaign in May-June 2006.The test matrix provided experiments that examine the effects of aging and NO 3 radical (obtained from N 2 O 5 decomposition) reactions in the dark, photo-oxidation, and OH radical reactions in the light on the composition of DE.
Table 1 shows the combined dark (D) and light (L) experimental matrix.All D runs were done with the chamber protective cover on, and all L runs with the protective cover removed.The purpose of the light exposures was to examine the effects of photolysis reactions on the exhaust composition, while the dark experiments provide a baseline dataset for modern DE aging without the influence of sunlight.
VOC concentrations in the chamber atmosphere following DE injections were typically less than 10 ppb (confirmed by in-situ GC/MS and canister sampling followed by GC/MS analysis for 71 compounds ranging from C 2 -C 11 ; see Zielinska et al., 2009).Therefore, light experiments with addition of formaldehyde (as a source of OH radical from the photolysis reaction), toluene as a representative VOC (2190±259 µg m −3 ; see Table 4 for specific values), or a mixture of several aromatic compounds including naphthalene, were conducted.Two different VOC mixtures were added to the chamber for three experiments in Atmos.Chem. Phys., 10, 609-625, 2010 www.atmos-chem-phys.net/10/609/2010/2006 to supplement the gaseous atmospheres.The compositions of both mixtures and the resulting initial concentrations in the chamber are listed in Table S1 (Supplemental): http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf.
In the initial phase of these experiments (Winter 2005) it was recognized that the specific aims of this work could not be achieved without effective denudation of NO x (NO+NO 2 ) prior to infusion of exhaust into the reaction chamber.Therefore, a newly developed NO x denuder was used in the 2006 experiments discussed in this article (unless otherwise stated).For more information on the NO x denuder development and application see Zielinska et al. (2009) and Samy (2009).

Sampling instrumentation and analysis
Determination of particle size number and volume concentrations were accomplished using a Scanning Mobility Particle Sizer (SMPS) (TSI-3936).NO x and NO y species were monitored using chemiluminescence (ECO Physics CLD 770AL, Teledyne API 200AU, Advanced Pollution Instrumentation, Inc., San Diego, CA) and Fourier Transfer Infrared (FTIR; NICOLET Magna 550 w/liquid N 2 cooled MCT detector) instruments.The White mirror FTIR system (553.5 m optical path) monitored toluene and formaldehyde in experiments with specified additions, along with other gas phase species (glyoxal, methylglyoxal) with a spectral resolution of 1 cm −1 and a time resolution of 5 min.See Gomez Alvarez et al. (2007) for description of FTIR-DOAS determination of glyoxal and methylglyoxal at EUPHORE.An ozone monitor (Monitor Labs ML9810) and J NO2 radiometers (for direct and reflected sunlight) were interfaced with the EUPHORE data acquisition system.The volume of the chamber was supplemented with the dry, clean air to keep the chamber at a slightly positive pressure and the dilution rate was monitored by SF 6 addition to the chamber at the beginning of each experiment.Several past studies have addressed chamber specifications and systems (e.g.Koltz et al., 1998;Volkamer et al., 2001;Martin and Wirtz, 2005).
In-situ VOC measurements were periodically made with the EUPHORE GC/MS system (Varian 3400 interfaced with a MS Saturn 2000), which incorporates a cryogenic sample pre-concentration trap (SPT).The GC/MS system is located on the platform below the chamber, and is interfaced with a sampling port (2.5 m, 1/8 PTFE sampling line).The GC was equipped with a 30 m DB-624 (Cyanopropenyl Polysiloxane) fused silica capillary column (J&W Scientific, 0.32 mm id, 1.8 µm film).A 15 min sample collection time with an EMF controller (40 ml/min; total sample volume =600 ml) was used, and desorption was performed at 270 • C.
The particle-associated SVOC were collected from the chamber at the end of the exposures, using an XAD-4 (Aldrich Chemical Company, Inc.) coated annular denuder followed by a 90 mm Teflon-impregnated glass fiber (TIGF) filter and XAD-4 cartridge with a 100 SLPM flow (Gundel et al., 1995).The XAD-denuder strips the gas-phase species by molecular diffusion out of a laminar flow stream, before collection of the particles on the second stage (filter/cartridge) (Peters et al., 2000).DPM was collected on quartz fiber filters following each experiment and analyzed for organic and elemental carbon (OC/EC) by thermal/optical reflectance (Chow et al., 2004).Sulfate and nitrate analysis was performed by ion chromatography (Dionex, DX500).All filter/XAD samples were stored in a freezer at the EUPHORE facility and shipped to the Desert Research Institute (DRI) at www.atmos-chem-phys.net/10/609/2010/Atmos.Chem.Phys., 10, 609-625, 2010 the conclusion of the campaigns.Samples were packed in coolers with frozen blue ice and shipped using an overnight delivery service (which resulted in 2-4 days transfer time).
Prior to the campaign, TIGF filters (90mm) were cleaned by sonication for 10 min in dichloromethane (CH 2 Cl 2 ) twice, with the solvent replaced and drained, and sonicated for 10 min in methanol twice with the solvent replaced.Amberlite XAD-4 (Aldrich Chemical Company, Inc.) was washed with liquinox soap and rinsed with hot water, followed with DI water and technical grade methanol (3-4 times).The XAD-4 was then extracted using a Dionex Accelerated Solvent Extractor (Dionex ASE 300 Accelerated Solvent Extractor) for 15 min/cell with CH 2 Cl 2 at 1500 psi and 80 • C, followed by acetone.It was then dried in a vacuum oven at 50 • C.
The SVOCs collected on each denuder-filter-XAD sampling train were extracted separately with high-purity, HPLC grade solvents (Fisher Sci.).Filters and XAD were extracted with ∼170 mL 1:1:1 of CH 2 Cl 2 , hexane and methanol by accelerated solvent extraction (ASE) at DRI.The ASE method pressurizes and heats the media for 15 min/cell at 1500 psi and 80 • C. All media was extracted twice under the same conditions and then combined for each sample to assure completed extraction of analytes.Samples from both light and dark exposures were analyzed for a suite of SVOC, which includes POC listed in Table S2: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf, and are classified by compound groups including: alkanoic acids, alkanedioic acids, and aromatics acids.The following deuterated POC internal standards were added to the samples prior to extraction: hexanoic acid-d 11 , benzoic acid-d 5 , succinic acid-d 4 , adipic acid-d 10 , suberic acid-d 12 , homovanillic-2,2 acid-d 2 , myristic acid-d 27 , oleic-9,10 acid-d 2 , tetradecanedioic acid-d 24 , eicosanoic acid-d 39 , cholesterol-d 6 .The extract underwent a trimethylsilyl (TMS) derivatization (Nolte et al., 2002;Rinehart et al., 2006), utilizing a mixture of bis-(trimethylsilyl)-trifluoracetamide (BSTFA) +1% (wt) trimethylchorosilane (TMCS) with pyridine (silylation grade) to enhance detection sensitivity.Derivatized samples were analyzed by electron impact (EI) GC/MS at DRI.A Varian CP-3800 GC equipped with a CP-8400 Autosampler and interfaced to a Varian 4000 Ion Trap Mass Spectrometer was used to perform splitless injections onto a 30 m 5% phenylmethylsilicone fused silica capillary column (DB-5MS, Agilent Technologies), with an integrated 10 m of deactivated guard column.Calibration checks were analyzed following sample analyses.All individual compound calibration checks were within 20% (usually <10%).Field blanks underwent extraction and analysis in the same manner for blank subtraction.
The XAD coated glass annular denuder was extracted at the EUPHORE facility immediately following sampling, with the same solvent ratio previously described.Approximately 200 mL of solvent was poured into the denuder section, which had been capped at one end.The other end was capped, and the denuder was manually inverted about 30 times.The solvent was drained and the procedure repeated two more times.Extractions of field blanks were performed on-site and all extracts were concentrated to ∼3 mL by rotary evaporation.All annular denuder extracts were refrigerated during the campaign, followed by shipment (with filter/XAD samples) to DRI for analysis by GC/MS.

Chamber wall loss correction and percent yield
SOA mass production was determined following a sizesegregated, first-order wall-loss correction.First-order wallloss beta (β) coefficients as defined by Fuchs (1964), were calculated using diesel in dark (no addition) experimental SMPS data for selected bins (14 total) throughout the size range (17.5-1015 nm), followed by extrapolation to determine all bin values.In this approach, the diesel in dark (no addition) experiment can be viewed as a "baseline" chamber atmosphere when evaluating SOA mass production.The size-segregated β values ranged from 1-15×10 −5 s −1 and displayed good agreement with diameter specific values and theoretical trends discussed in previous work (Crump and Seinfeld, 1981;Park et al., 2001;Johnson et al., 2005).Particle loss for each size bin was accounted for and added to SMPS number concentrations for each 5 min time-step before conversion to total particle volume concentration assuming spherical particles.This procedure was performed for all reported experiments, and the conversion to mass assumes an aerosol density of 1 g cm −3 .Toluene (as a representative VOC) was added to the chamber atmosphere, following DE injection in four separate experiments (Sect.3.3).SOA percent (%) yield [SOA mass formed (µg m −3 )/toluene mass reacted (µg m −3 ) multiplied by 100] for toluene addition experiments (L-3) was determined by monitoring toluene (FTIR, 5-min average) data for changes in mass concentrations (corrected for dilution) during aging.Glyoxal and methylglyoxal yields were determined following correction for secondary loss (photolysis [J(glyoxal)=J(NO 2 )*0.0109] from Klotz et al., 2001;[J(methylglyoxal)=J(NO 2 )*0.019] from Plum et al., 1983; and OH-reaction K OH from Meller et al., 1991as described in Volkamer et al., 2001).A plot of concentrations of each compound versus reacted precursor (toluene determined by FTIR) was evaluated for slope following a least-squares linear regression.The slope in this analysis was used to determine yield (Gomez Alvarez et al., 2007)  3 Results and discussion

SOA production
Figure 1 displays a 2006 diesel in dark baseline experiment (D-1, 05/31/06) with both collected and wall loss corrected SMPS data (as discussed above).After over 2 h of aging, the corrected value is within 0.1% of the initial aerosol mass following injection.This indicates that the wall loss correction procedure does account for the main loss mechanism in a dark baseline experiment (values are included in Table 2).In addition, examination of corrected particle profiles at different stages of aging (e.g. 1 h, 2 h) also indicates an appropriate accounting for the main aerosol loss mechanism with profiles that appear to be uniformly conserved and centered in the accumulation mode (see Fig. S1 in supplement: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf).SOA mass production values from the 2006 campaign can be viewed in Table 2.The %SOA of the final inchamber mass concentration following aging was calculated using the mass difference (i.e.mass production) between the post-injection and pre-sampling masses corrected for wall losses.SOA % production increased with OH radical (from HCHO photolysis) experiments (L-2), versus diesel in light (L-1) experiments.The greatest % production values are from the diesel in light with toluene addition (L-3) experiments.The OH radical concentrations for the L-3a and L-3b (toluene addition) experiments were 5.4×10 6 and 3.3×10 6 molecules cm −3 , respectively, indicating a similar in-chamber atmosphere reactivity (values calculated using rate constant from Atkinson, 1994).See the toluene addition experiment Sect.3.3 for more discussion.
Figure 2 displays the L-2b experiment (DE+OH radical +VOC mix, Table 2) with aerosol mass concentrations (µg m −3 ) and VOC (ppb) data through out the DE injection, aging, and the beginning of sample collection.This figure includes both non-corrected and size-resolved wall loss corrected SMPS data.Non-corrected particle profiles observed during DE aging in the L-2b experiment are provided in supplemental Fig. S2: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf.Nearly a quarter of the resulting aerosol mass (L-2b in Table 2) at the beginning of sampling is newly formed SOA.Decomposition of shorter life time VOC (o-xylene, isobutyl benzene) is observed, while less reactive VOC (benzene) displays very little decomposition for the 4.75 h sunlight exposure time (Finlayson-Pitts and Pitts, 2000).GC/MS analysis indicated a similar decomposition for all substituted benzene derivatives (listed in Table S1: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf), and naphthalene concentrations decreased below the detection limit prior to sample collection (see discussion below).The significant production of o-toluic acid in this experiment is notable with approximately 3 times the concentration found in the particle phase, relative to the L-2a (DE+OH) experiment without VOC addition.See Sect.3.2 for more discussion on the o-toluic acid concentrations and production pathway.

SOA and VOC
Following each chamber experiment, DPM were collected on quartz fiber filters and analyzed for OC/EC by thermal/optical reflectance method (see Sect. 2.2).The OC/EC ratios for 2006 EUPHORE experiments are shown in Fig. 3.With the exception of the L-1a (DE only) experiment, all light exposures exhibit an increase in OC/EC relative to dark experiments.The L-1a atmosphere was not photochemically active (as discussed above) due to the very low concentrations of VOC.Several experiments exhibit a percent increase in OC/EC (relative to dark experiments) that exceeds 100% (L-1b, L-2b, L-3a), with the L-1b experiment displaying the highest increase at 204%.The L-3 (toluene addition) and VOC addition experiment in sunlight L-1b, show significant contribution of pyrolized organic carbon, which indicates the presences of highly polar and/or oligomeric organic compounds formed as SOA (Chow et al., 2004).See Zielinska et al. (2009) for more detail on OC/EC, and pyrolized organic carbon results.
These high increases of OC/EC ratio are not always consistent with the lack of mass production calculated using the SMPS method discussed above.For example, the L-1b (DE+VOC) experiment displays a significant increase of OC/EC while no mass production is observed (Table 2).There may be several reasons explaining these apparent inconsistencies: (1) positive quartz filter artifact (i.e.adsorption of VOC on the quartz filter) which would result in artificially high OC contents; (2) uncertainties of OC/EC measurements (Chow et al., 2004), amplified by a very low mass of particles collected on quartz filters (Zielinska et al., 2009); and (3) assumption of spheroid particles for SMPS data.If fresh emissions evolve from initially non-spherical to more spherical particles during the SOA production process, then the SMPS method of mass determination would potentially include an associated uncertainty of measurement.
In contrast to recent work (Weitkamp et al., 2007;Robinson et al., 2007), the modern DE and the experimental setup used at EUPHORE suggests that no significant SOA production results from aging of DE in sunlight without addition of an exogenous radical source (HCHO) and/or addition of VOC (including toluene addition experiments).For example, the L-1a experiment (DE in light with no additions) actually exihibits a lower OC/EC value relative to D-1 (DE in dark).Several factors may explain the difference between the EUPHORE and the Weitkamp et al. ( 2007

POC concentrations
The analysis for a limited number of POC (54 listed in Table S2: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf) was conducted to assess the composition of modern DE and the formation of photochemical transformation products.Understanding the relationship between SOA production and changes in POC composition in light versus dark chamber experiments may help to delineate the compound specific contribution to overall SOA mass production.Gas and particle phase concentrations of selected POC are shown in Table 3.Note that all organic compound concentrations are expressed in relation to elemental carbon (i.e.microgram of species per milligram of EC).The inherent assumption in this normalization procedure is that EC mass will remain unchanged while atmospheric transformations of speciated organic compounds take place within the chamber atmosphere.Supplemental Tables S4-S7: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdfinclude all measured POC concentrations.
Gas phase POC concentrations are generally more dominated by alkanoic and aromatic acids, while particle phase alkanedioic (diacids) values are relatively higher , as would be expected from partitioning theory (Pankow, 1994;Seinfeld and Pandis;1998)  be due to the reduced solar radiation input, due to overcast conditions on 06/09/06 (see Sect. 3.3 for further discussion).
A closer look at the light L-1a (DE) and L-1b (DE+VOC) experiments in Figs. 4 and 5, reveals the significant changes to the overall POC concentrations and profile resulting from the addition of VOC.While other experimental conditions are similar (e.g.post-injection mass, aging time) and the two experiments exhibit no significant SOA production, the changes in chemical composition do suggest a significant chemical transformation has taken place in both gas and particle phases.The L-2a (DE+OH) and L-2b (DE+OH+VOC) experiments do not display such drastic changes in measured total POC concentrations, but do exhibit a significant difference in SOA production (Table 2).Although post-injection mass is lower in the L-2b experiment, leading to a reduced total aerosol surface area (i.e.particle size profiles are similar), the % SOA production is over 2.6 times greater than the L-2a experiment.However, only targeted POC were quantified for these exposures by the derivitization followed by GC/MS analysis method.This technique does not include analysis for multifunctional, highly oxygenated and/or nitrogenated, oligomeric compounds that may have been produced in the L-2a and L-2b experiments and may explain the differences in the observed % SOA production (Zielinska et al., 2009).Further discussion of POC composition is provided in the supplemental: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf.
Figure 6 displays the particle phase concentrations (µg mg −1 EC) of 1,2-benzenedicarboxylic acid (phthalic acid) and phenylacetic acid (only detected in L-1b, L-2b w/VOC experiments).These two compounds were not detected in the dark baseline (D-1) experiments, and the highest concentrations are associated with experiments that included VOC (w/naphthalene) additions (D-2, L-2b, L-1b).Phthalic acid has been shown to be the only major product from the photolysis of naphthalene on SiO 2 and Al 2 O 3 surfaces in air (Barbas et al., 1993), and more recent chamber experiments have identified phthalic anhydride and phthaldialdehyde as a transformation product of the OH radical initiated reactions of naphthalene (Wang et al., 2007).
Phthalic acid has been proposed as an indicator for photochemical SOA formation, and as a surrogate for SOA contribution to ambient samples in past studies (Fine et al., 2004;Fraser et al., 2003).However, the determination of % SOA from a single compound may produce significant errors in estimations as is evident from Table 2.For example, the L-1b (DE+VOC) experiment displays the highest relative phthalic acid concentration but does not show any significant   SOA production.Possible precursors of phthalic acid are discussed in the supplemental: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf.
The D-2 (NO 3 radicals from N 2 O 5 decomposition) experiment provides an example of relatively high phthalic acid production in the absence of light.NO 3 radicals are very reactive and are observed during late evening and nighttime hours in the atmosphere, followed by rapid removal via Atmos.Chem.Phys., 10, 609-625, 2010 www.atmos-chem-phys.net/10/609/2010/photolysis and reaction with NO during the day.NO 3 radical reactions with naphthalene proceed by initial addition to the aromatic rings to form nitrooxycyclohexadienyl-type radicals, which can react with O 2 or NO 2 to yield various products (Pitts et al., 1985;Atkinson et al., 1987;Sasaki et al., 1997).While the ring retaining reaction pathways, which result in the formation of nitronaphthalenes and hydroxynitronaphthalenes represents the major product formation yields (Atkinson and Arey, 2007), the ring cleavage pathways may contribute to the production of carboxylic aromatic acids.Decomposition followed by oxidation of the substituted (≥C 2 ) benzene derivatives and/or PAH, is the likely production pathway for the observed phenylacetic acid.For example, the ethylbenzene-OH pathway proposed by Forstner et al. (1997) is a possible production route for phenylacetic acid in the L-2b (DE+OH+VOC) and L-1b (DE+VOC) experiments.A change in chamber atmosphere reactivity due to VOC addition is apparent, with the L-2b having the maximum ozone production of all experiments (251 ppb), followed by the L-1b experiment (188 ppb) (see Table S8: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf for all values).This may have promoted the production of a more complex set of transformation products.
Several dicarboxylic acids listed in Table 3, including: succinic, methylsuccinic, glutaric, suberic, and sebacic were identified as major components in atmospheric PM during photochemical smog episodes (Grosjean et al., 1978).Succinic acid has been shown to be present at relatively high concentrations in many global urban sites (Ho et al., 2006 and references therein).Possible precursors (in DE) suggested in a past study (Hatakeyama et al., 1987) include cyclopentene for glutaric or succinic acids, and 3-methylcyclopentene for methylsuccinic.However, the modern DE used in this study had very low concentrations of these potential precursors, and the contribution from other precursors (e.g.aromatic compounds) can not be ruled out.
To evaluate whether diacids are relatively more abundant than monoacids in the light exposure experiments, a ratio of the quantified (Table S2: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf)alkanedioic to alkanoic acids is shown in Fig. 8.Even though our analytical method determined the concentrations of a limited number of compounds, it is apparent that both the gas and particle phases display a higher ratio for the light versus dark experiments.The high particle phase ratio in the L-1b (DE+VOC) experiment can be explained in part by the relatively high C 4 and C 5 diacids production (Table 3), which may be the result of a combination of factors including: higher actinic flux during L-1b (J(NO 2 )=8.7±0.7×10−3 ) versus L-1a (J(NO 2 )=6.4±1.2×10−3 ), addition of VOC in L-1b resulting in a more reactive mixture with increased concentrations of possible precursors and a greater ozone concentration (190 ppb maximum in L-1b versus 20 ppb in L-1a), and the highest post-injection mass concentration (49.4 µg m −3 ; Table 2) leading to increased particle phase partitioning (Donahue et al., 2006).

Toluene addition experiments
To determine if SOA production (% yield) from toluene in the presence of a complex mixture (modern DE), is significantly different from SOA % yield in simplified experimental mixtures (Johnson et al., 2004), the four L-3 (DE+tol) experiments were further evaluated.Table 4 displays the % yield [(SOA(µg m −3 )/ tol(µg m −3 ))×100] for all L-3 experiments, with the initial toluene concentration ([tol] o ) and [tol] o /[NO x ] o ratio.Generally, the % yield values (average =5.3±1.6) are within the range of "low-NO x " ([tol] o /[NO x ] o range 10.9-17.8)experiments conducted by Johnson et al. (2004) and Hurley et al. (2001).However, these experiments were conducted without seed aerosol and the increase in yield with decreasing [tol]     Atkinson (1994).c Maximum ozone production during aging.d % yield (relative to toluene) calculated following size-resolved wall-loss correction, assuming a 20% analytical error.e corrected for photolysis loss using Koltz et al. (2001); correct for OH loss using Meller et al. (1991); 30% error applied see Volkamer et al. (2005) for detailed discussion on potential sources of error.f 30% Error applied.g Data not available.observed in our experiments is in more agreement with trends observed by Odum et al. (1997), which included seed aerosol and a more complex mixture of VOC.
SOA formation from a variety of precursors has been shown to be significantly impacted by NO x concentrations (Song et al., 2005;Kroll et al., 2006;Pathak et al., 2007) possibly due to the formation of different condensable products under different NO x conditions.Johnson et al. (2004) displayed the need to increase model species-independent partitioning coefficient scaling factors as NO x concentrations increased in past toluene photooxidation studies at EUPHORE, indicating a systematic impact on the SOA production process due to changes in the [tol] o /[NO x ] o ratio.Since most atmospheric oxidation reactions of organics produce alkyl radicals, multifunctional organic nitrates may also be produced at significant concentrations as observed by Lim and Ziemann (2005) and Forstner et al. (1997).
Although our yield values are generally lower than those reported by Odum, our highest yield experiments, L-3b (6.8±1.4) and L-3c (6.4±1.3), are in good agreement with values (within 20% analytical error) for the highest [tol] o /[NO x ] o (5.47) experiment conducted by Odum and coworkers on 07/10/96 (yield =7.4%).It is important to note that Odum did enhance the reactivity of their mixtures by adding propene (i.e.production of radicals), and this may explain the generally higher yield trend.Upon further evaluation, Odum et al. (1997) classify toluene (along with other single substituted methyl/ethyl benzene derivatives) as a "high-yield aromatic" and using a two-product model (Pankow, 1994), produce yield curves where the amount of aerosol produced ( M o ) is in the range of our experiments (22.7-64.1 µg m −3 ).Our observed % yields (3.3±0.7 to 6.8±1.4)fall within the proposed range of 4-7% (Odum et al., 1997), and are in agreement with more recently reported observations (Stroud et al., 2004).
Formation of glyoxal (CHOCHO) in our L-3 experiments was observed, and the % yield (relative to reacted toluene) is reported in Table 4. Glyoxal has been identified as a major primary product from the OH-initiated oxidation of toluene (Volkamer et al., 2001), and more recent work at EUPHORE has further confirmed the dicarbonyl route proposed by Darnall et al. (1979) (see: Atkinson et al., 1980 andGomez Alvarez et al., 2007).Our observations of nearly immediate (with 5 min time resolution) glyoxal formation following toluene addition, supports the proposed ring-cleavage pathways involving the bicycloalkyl radical (Volkamer et al., 2001), and the L-3b experiment % yield (37.6) is in agreement with values reported recently (37±2) by Gomez Alvarez et al. (2007).The significantly lower actinic flux (approximately 1/2) in the L-3b experiment may explain the higher yield.Glyoxal loss due to photolysis is addressed in Volkamer et al. (2005), and the reassessment of % yields from previous work (Volkamer et al., 2001) resulted in a reported value of 30.6±6% (previously 39±10.2%),which is in agreement (within error) with our higher yield toluene addition experiments (L-3b, L-3c, L-3d).The J(glyoxal) used for loss correction in these experiments ranged from 4.5-9.6×10−5 (s-1), and are typical values for natural sunlight at mid-latitude locations (Volkamer et al., 2005).Interestingly, the lowest % yield (L-3a) experiment was also the experiment with the lowest SOA yield, and the highest of the dicarbonylic route in this system plays a role in SOA production efficiency (Gomez Alvarez et al., 2007).Formation of glyoxal oligomer species in relation to SOA formation processes has been discussed in previous studies (Kalberer et al., 2004;Hastings et al., 2005).While some correlation between glyoxal and SOA yield is observed here (R 2 =0.61), more experimentation with atmospherically relevant mixtures is needed to sufficiently parameterize these processes.Methylglyoxal yields are included in Table 4 for the 2006 experiments.The significant production observed in the L-3b experiment is similar to % yield values recently reported (37±2) (Gomez Alvarez et al., 2007).When combined with the glyoxal yield, a significant fraction of the reacted toluene can be account for (L-3a=28%, L-3b=75%).Formation of formaldehyde (HCHO) from decomposition of more complex organic compounds can be the result of reactions involving multi-generational products (Finlayson-Pitts and Pitts, 2000).In this study HCHO % yields relative to decomposed toluene (following correction for photolysis and OH radical loss) ranged from 18-23%, and showed no significant variability between the four L-3 experiments (assuming a 30% error).HCHO can originate from multiple precursors, for example, from the photolysis of glyoxal or methylglyoxal (Koch and Moortgat, 1998).Therefore, HCHO yield values reported here should be seen as a reaction system characteristic and not as an indication of a specific production pathway.The low actinic flux experiment (L-3b) did result in the lowest yield value (18%), and observed formic acid production generally coincided with HCHO production (Fig. S3: http://www.atmos-chem-phys.net/10/609/2010/acp-10-609-2010-supplement.pdf).

Conclusions
Atmospheric transformations of emissions from a modern light-duty diesel engine equipped with oxidation catalyst have been studied in the EUPHORE outdoor simulation chamber in Valencia, Spain.It was observed that SOA production in sunlight experiments was negligible unless VOC or OH radicals were added to the chamber to enhance the reactivity of the mixtures.Maximal SOA production (in the range of 40%) was observed when toluene was added to the chamber in relatively high concentrations (400-500 ppb).
Despite the relatively low yield of SOA, significant changes were observed in the concentrations of quantified polar organic compounds (POC) at the end of each 4-6 h experiment.In particular, aromatic and aliphatic diacid concentrations were enhanced in light as compared with dark exposures.Phthalic acid, a postulated photooxidation product of naphthalene, showed significant increases in concentration when naphthalene was added to the chamber as one of the VOC mixture components.Yet a significant correlation between SOA production and phthalic acid concentrations was not observed in this study.Our data suggests that addition of VOC, including naphthalene, results in a relatively higher concentration of potential SOA tracers in a complex mixture (DE).However, the lack of correlation between %SOA and single compound concentrations (e.g.phthalic acid) suggests that more extensive chemical characterization data near sources and following aging/transport need to be considered for more accurate estimations in the ambient atmosphere.
The toluene addition experiments show consistency with previously conducted toluene oxidation studies at EU-PHORE (Johnson et al., 2004;Gomez Alvarez et al., 2007) and with trends observed by Odum et al. (1997).As suggested by earlier work (Volkamer et al., 2001), a significant fraction of the reacted toluene can be accounted for when glyoxal and methylglyoxal yields are considered.The observed yield correlation between SOA, glyoxal, and other decomposition products (such as HCHO) of more complex organic compounds, suggests that reactions involving multigenerational products were occurring within the chamber atmospheres.
Figure 5. Particle-phase POC concentrations for three compound classes.

Table 1 .
Dark (D)and Light (L) Exposure Test Matrix.

Table 2 .
%SOA of the final in-chamber mass concentration prior to sample collection."Post-injection mass" indicates in-chamber DPM immediately following injection."Pre-sampling mass" indicates DPM concentration following aging and has been correct for wall-loss.

Table 3 .
Gas and particle phase concentrations (µg mg −1 EC) for selected organic acids.
a "C#" indicates carbon number, while "d" signifies whether the compound is a diacid.b Both particle and gas phases included as sum for nonanoic acid.c ND, not detected.d

Table 4 .
L-3, toluene addition experiments conducted at EUPHORE.Aging time is the total time DE aged following chamber injection and prior to beginning of sample collection.b Decay of toluene was used to estimate the concentration of OH radicals using rate constant from a