Ambient Gas-Particle Partitioning of Tracers for ... - ACS Publications

Subscriber access provided by CORNELL UNIVERSITY LIBRARY

Article

Ambient gas-particle partitioning of tracers for biogenic oxidation Gabriel Isaacman-VanWertz, Lindsay D. Yee, Nathan M Kreisberg, Rebecca Wernis, Joshua A Moss, Susanne V Hering, Suzane S de Sa, Scot T. Martin, Lizabeth M. Alexander, Brett B Palm, Weiwei Hu, Pedro Campuzano-Jost, Douglas A. Day, Jose Luis Jimenez, Matthieu Riva, Jason Douglas Surratt, Juarez Viegas, Antonio Manzi, Eric S. Edgerton, Karsten Baumann, Rodrigo Souza, Paulo Artaxo, and Allen H. Goldstein Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.6b01674 • Publication Date (Web): 23 Aug 2016 Downloaded from http://pubs.acs.org on August 24, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 37

Environmental Science & Technology

ACS Paragon Plus Environment


1

Ambient gas-particle partitioning of tracers for

2

biogenic oxidation

3 4

Gabriel Isaacman-VanWertz†,‡, Lindsay D. Yee†, Nathan M. Kreisberg§, Rebecca Wernisǁ,

5

Joshua A. Moss†,‡, Susanne V. Hering§, Suzane S. de Sá%, Scot T. Martin%,∫, M. Lizabeth

6

Alexander┴, Brett B. Palm#, Weiwei Hu#, Pedro Campuzano-Jost#, Douglas A. Day#, Jose L.

7

Jimenez#, Matthieu Riva¶, Jason D. Surratt¶, Juarez Viegas▲, Antonio Manzi▲, Eric Edgerton□,

8

Karsten Baumann□, Rodrigo Souza■, Paulo Artaxo○, Allen H. Goldstein*,†, ǁ

9

†

Dept. of Environmental Science, Policy, and Management, University of California, Berkeley,

10

CA, USA, 94720

11

§

12

ǁ

13

94720

14

%

15

01451

16

∫

Aerosol Dynamics Inc., Berkeley, CA, USA, 94710

Dept. of Civil and Environmental Engineering, University of California, Berkeley, CA, USA,

School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA,

Dept. of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA, 01451


Page 2 of 37

Page 3 of 37


17

┴

18

Richland, WA, USA, 99352

19

#

20

Sciences (CIRES), University of Colorado, Boulder, CO, USA, 80309

21

¶

22

NC, USA, 27599

23

▲

24

□

Atmospheric Research & Analysis, Inc., Cary, NC, USA, 27513

25

■

Universidade do Estado do Amazonas, Manaus, AM, Brazil, 69735-000

26

○

Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory,

Dept. of Chemistry & Biochemistry and Cooperative Institute for Research in Environmental

Dept. of Environmental Sciences and Engineering, University of North Carolina, Chapel Hill,

Instituto Nacional de Pesquisas da Amazonia, Manaus, AM, Brazil, 69060-001

Universidade de São Paulo, São Paulo, SP, Brazil, 05508-020

27 28

KEYWORDS: Atmospheric chemistry, gas-particle partitioning, secondary organic aerosol

29

30

ABSTRACT: Exchange of atmospheric organic compounds between gas and particle phases is

31

important in the production and chemistry of particle-phase mass, but is poorly understood due

32

to a lack of simultaneous measurements in both phases of individual compounds. Measurements

33

of particle- and gas-phase organic compounds are reported here for the southeastern U.S. and

34

central Amazonia. Polyols formed from isoprene oxidation contribute 8% and 15% on average to

35

particle-phase organic mass at these sites, but are also observed to have substantial gas-phase



36

concentrations contrary to many models that treat these compounds as non-volatile. The results

37

of the present study show that the gas-particle partitioning of approximately 100 known and

38

newly observed oxidation products is not well explained by environmental factors (e.g.

39

temperature). Compounds having high vapor pressures have higher particle fractions than

40

expected from absorptive equilibrium partitioning models. These observations support the

41

conclusion that many commonly measured biogenic oxidation products may be bound in low-

42

volatility mass (e.g. accretion products, inorganic-organic adducts) that decomposes to individual

43

compounds on analysis. However, the nature and extent of any such bonding remains uncertain.

44

Similar conclusions are reach for both study locations, and average particle fractions for a given

45

compound are consistent within ~25% across measurement sites.

46 47

Introduction

48

Most reactive carbon emitted globally to the atmosphere is biogenic, consisting predominately

49

of isoprene, followed by monoterpenes, and other terpenoid compounds.1 Reactions of these gas-

50

phase species with atmospheric oxidants can yield lower-volatility oxygenated products that

51

partition to the particle phase, increasing the mass concentration of organic aerosol (“OA”, all

52

discussion in this work limited to submicron particles)2–4 through the formation of oxidized

53

secondary organic aerosol (SOA).5 While health and climate impacts of atmospheric organic

54

carbon are primarily associated with particle-phase mass,3 organic aerosol is a dynamic, multi-

55

phase system6 that undergoes oxidation reactions both within and between the gas and particle

56

phases that can yield both higher and lower volatility products.

57

Many compounds have been identified in ambient and laboratory studies that provide insight

58

into the production of organic aerosol mass, but observations of actual ambient gas-particle


Page 4 of 37

Page 5 of 37


59

partitioning are sparse. The volatility of organic species in the matrix of the ambient particle

60

population is consequently uncertain, though it is critical for understanding the lifecycle of

61

organic aerosol in the atmosphere.7–10 In many models, SOA is treated as semi-volatile, meaning

62

that molecular constituents partition between the gas and particle phases. Thermodynamic

63

equilibrium is often used to model the extent of absorption into the particle phase and hence the

64

organic mass concentration, which is primarily treated as a function of compound vapor pressure

65

and total OA mass.6,11 This “organic absorptive equilibrium partitioning model” is often

66

extended to individual monoterpene oxidation products (e.g. pinic acid) and for binned isoprene

67

and monoterpene oxidation products.12 In the case of isoprene photooxidation, which is a

68

dominant process over the study areas described herein, chemically explicit models often treat

69

individual identified particle-phase products and surrogate species to be essentially non-volatile,

70

formed after heterogeneous uptake of their precursor and not re-volatilized.13,14 This assumption

71

is reasonable for species having low expected vapor pressures such as organosulfates and multi-

72

component esters and ethers (dimers, hetero-oligomers, etc.), which are thought to comprise a

73

significant fraction of isoprene SOA.15–17 However, most compounds used to quantify isoprene-

74

derived particle-phase mass (primarily C5 polyols, e.g. ref.

75

high enough to have a substantial or majority gas-phase component under typical atmospheric

76

conditions and organic aerosol concentrations, supported by recent measurements of gas-phase

77

2-methyltetrols in a relatively dry environment.19 They may be sufficiently soluble in water,

78

however, to remain largely in the particle-phase in humid environments, but solubility

79

parameters are also often highly uncertain. While organic absorptive equilibrium partitioning has

80

been widely used to effectively parameterize OA mass concentrations produced in chamber

81

studies, recent ambient measurements have found that some organic compounds are well

18

) have estimated vapor pressures



82

described while others are not.20 Neither an assumption of non-volatility for compounds of

83

relatively high vapor pressures, nor an assumption of absorptive equilibrium partitioning of

84

individual compounds has been extensively validated through observation of ambient aerosols.

85

Several additional processes have also been proposed to strongly influence gas-particle

86

partitioning and in some cases better explain laboratory data, such as particle phase transport

87

limitations,21,22 influence of inorganic ions on solubility,23 and oligomerization reactions.24 In

88

particular, recent measurements of ambient particles have demonstrated the likely importance of

89

oligomers15 and low-volatility products of accretion reactions that can thermally decompose to

90

form fragments with the same formulas as many biogenic oxidation products.25 Modeling of gas-

91

particle partitioning has improved in recent years through the inclusion of kinetics and particle-

92

phase reactions,26,27 and some work has suggested that accounting for these processes reconciles

93

observed and expected partitioning.25 However, a complete model treatment of organic

94

absorption including these processes remains elusive due to uncertainties in both model

95

parameters and OA composition.

96

We present here measurements of fraction in the particle phase (FP) of several compounds

97

commonly used as “tracers” for the oxidation of biogenic volatile organic compounds (BVOC),

98

as well as many newly measured compounds. Tracers uniquely identified by structure, mass

99

spectrum, and/or chromatographic retention times are used to constrain the impact of known

100

precursors or chemical pathways on the formation and removal of SOA. Decomposition of

101

thermally labile low-volatility compounds may contribute to particle-phase concentrations of

102

tracers,28–30 but the extent to which this process impacts tracer measurements is uncertain and

103

must depend on the way the specific tracer is chemically bound in SOA. The practical

104

consequence is that a “tracer” may represent a pure unambiguously identified compound or a


Page 6 of 37

Page 7 of 37


105

pure compound plus mass from a converted analogue. This does not necessarily reduce its

106

applicability as a tracer, but can increase the observed tendency of a tracer to stay in the particle

107

phase and impact interpretation of tracer data. Ambient measurements of partitioning for tracers

108

are consequently useful to understand the processes that drive exchange between gas and particle

109

phases and to accurately include these transitions in chemically explicit models.

110

In this work, hourly concentrations of BVOC oxidation products in both the gas and particle

111

phases were measured in two forested locations: the southeastern U.S. in rural Alabama during

112

the summertime (“SOAS”) and central Amazonia (wet season: “IOP1”; dry season: “IOP2”)

113

(details in ref.

114

biogenic emissions account for a major fraction of summertime particle-phase organic carbon in

115

mid-latitude forested regions such as the southeastern U.S. (e.g ref

116

studies33 have quantified BVOC oxidation products in the tropics. Consequently, though the bulk

117

properties of the particles in the Amazon are well-studied,34,35 improved measurements of tracer

118

compounds in the Amazon are needed to better understand the sources and atmospheric

119

processing of organic compounds and study the processes controlling aerosol formation from

120

biogenic emissions.

31

). Previous measurements have shown that oxidation of isoprene and other

32

and refs therein), but few

121 122

Experimental

123

This work is the result of measurements from several complex instruments at multiple field

124

sites, compared to a number of different modeling approaches. All instruments, locations, and

125

models are therefore described only briefly here, with greater detail in the Supporting

126

Information.

127

Partitioning and concentration measurements



128

Individual organic compounds were measured using a Semi-Volatile Thermal desorption

129

Aerosol Gas chromatograph (SV-TAG) with in-situ derivatization,36,37 which provides hourly-

130

resolved concentrations and gas-particle partitioning of most common BVOC oxidation tracers

131

(e.g. those measured by refs.

132

containing a high-surface-area passivated metal-fiber filter that quantitatively collects gas- and

133

particle-phase species less volatile than tetradecane.37 Prior to collection, gas-phase mass is

134

removed from one sample with a multi-channel carbon denuder (MAST Carbon) – gas-particle

135

partitioning is calculated as the ratio of these channels. Samples are desorbed with a ramped

136

temperature (8 minutes: 30-310 oC, 35 oC/min) into helium saturated with a trimethylsilylation

137

agent (“MSTFA,” N-Methyl-N-(trimethylsilyl) trifluoroacetamide) to a pre-concentrating trap,

138

which is then back-flushed and desorbed with a ramped temperature (3 minutes: 30-310 oC, held

139

for 2 minutes) to a gas chromatography column for analysis by GC/MS (7890A/5975C, Agilent

140

Technologies). Chromatography uses a non-polar column (Rtx-5Sil MS, 20m×0.18mm×0.18µm;

141

Restek) with a 23.6

142

temperatures of high surface area components over several minutes, exposure of analytes to

143

temperatures higher than necessary is minimized.

o

32,38–40

). Air is sampled at 10 Lpm through two identical cells

C/min temperature ramp (11.9 minutes: 50-330 oC). By ramping

144

Compounds are quantified using isotopically labeled internal standards to correct for run-to-

145

run variability in instrument response, and regular multi-point calibrations of approximately 100

146

authentic standards. Details and associated uncertainties of this approach are discussed in depth

147

by in Isaacman et al.36 and provided in the Supplementary Information; uncertainty is generally

148

on the order of 20% of measurement.

149

SV-TAG measurements are validated wherever possible against known measurements and

150

previously published results. Concentrations are compared with contemporaneous measurements


Page 8 of 37

Page 9 of 37


151

based on traditional filter collection and extraction. Partitioning is compared to models for n-

152

alkanoic acids, which have been shown previously20 to be well-described by organic absorptive

153

equilibrium partitioning (within the error uncertainty of SIMPOL vapor pressure estimation,

154

approximately half an order of magnitude41).

155

Field sites

156

Measurements were collected during three periods from two locations situated in rural areas

157

where atmospheric composition is expected to be dominated by regional BVOC emissions and

158

their oxidation products, with variable anthropogenic influence (map provided in Fig. S1).

159

Data from the southeastern United States were collected from June 1 to July 15, 2013 as part of

160

the Southern Oxidant and Aerosol Study (“SOAS”) at the Centreville, Alabama ground site

161

(“CTR”; 32.90325 N, 87.24995 W; time zone: UTC-5). SV-TAG data are reported from June 4

162

to July 5, 2013, [OA]avg = 4.8 ug m-3, Tavg = 24.7 ○C, RHavg = 82%. Samples were collected for

163

50 minutes every 1.5 hours until June 18, and the first 22 minutes of each hour thereafter. The

164

site is rural, primarily forested with a mixture of deciduous and coniferous trees. These

165

measurements are referred to as “SOAS” or as “southeastern U.S.”

166

Data were collected in rural Brazil as part of the GoAmazon2014/5 field campaign,42

167

consisting of several ground sites and aircraft near Manaus, Brazil. Measurements presented in

168

this work were made during the wet season (Intensive Operating Period 1, IOP1: Feb 1 to Mar

169

31, 2014) and the dry season (IOP2: Aug 15 to Oct 15, 2014) at the rural “T3” ground site (3.213

170

S, 60.599 W; time zone: UTC-4), 60 km west (downwind) of Manaus, with periods of strong

171

anthropogenic influence from Manaus or regional biomass burning. SV-TAG data reported here

172

span IOP1: Feb 14 to March 27, 2014, [OA]avg = 1.2 ug m-3, Tavg = 26.5 ○C, RHavg = 90%, and

173

IOP2: Aug 26 to Oct 4, 2014, [OA]avg = 8.7 ug m-3, Tavg = 27.9 ○C, RHavg = 83%. Samples were



174

collected during the first 22 minutes of every hour. These measurements are hereafter referred to

175

as “GoAmazon2014/5 IOP1 or IOP2” or as “Amazon wet/dry season.”

Page 10 of 37

176

Supporting measurements and models

177

Particle-phase mass and chemical composition (organic, sulfate, nitrate, ammonium, and

178

chloride) were measured at both sites using an Aerodyne High Resolution Time-of-Flight

179

Aerosol Mass Spectrometer (“AMS”), which focuses particles through an aerodynamic lens onto

180

a 600 °C vaporizer for analysis using electron impact mass spectrometry.43–45 Temperature and

181

relative humidity were measured at both sites with commercially available instrumentation. Gas-

182

phase ammonia concentrations were measured at the SOAS CTR site, but not during

183

GoAmazon2014/5. SV-TAG measurements were validated in part by comparison to filter

184

samples collected at 1000 Lpm on quartz-fiber filters baked prior to sample collection at 550 °C.

185

Samples were stored at -20 °C until extraction in methanol, trimethylsilylation, and analysis by

186

GC/MS (5890/5971A, Hewlett Packard).

187

Total particle-phase liquid water is calculated as the sum of water associated with inorganic

188

ions and water associated with organic mass, assuming no interaction. The inorganic contribution

189

is calculated using the thermodynamic model ISORROPIA-II,46 shown previously to provide

190

reasonable values at SOAS.47,48 Temperature, relative humidity, and AMS measurements of

191

inorganic ions were used as inputs for modeling of both SOAS and IOP1. Gas-phase ammonia

192

was included for SOAS but does not substantially impact calculated water concentrations (Fig.

193

S10). Water associated with organic mass is calculated based on the approach of Guo et al.48,

194

using as inputs ambient relative humidity and the concentration and hygroscopicity of particle-

195

phase organic mass measured by the AMS (Eq. S1). Hygroscopicity is estimated from the AMS

196

oxygen-to-carbon ratios.49


Page 11 of 37


197

Observations of gas-particle partitioning are placed in the context of expected FP using an

198

organic absorptive equilibrium partitioning model that assumes instantaneous equilibrium

199

between the gas and particle phases based on a compound’s vapor pressure (calculated using

200

SIMPOL,41 a structure-activity relationship).11,50 Solubility in particle-phase liquid water is

201

explored by expanding this model to assume equilibrium absorption into a mixed liquid phase

202

composed of both organics and water, based on an empirically calculated effective partitioning

203

coefficient. Phase separation of the particle is not considered here as estimated solubility

204

parameters for the tracer compounds studied here are too uncertain to explicitly treat uptake into

205

an aqueous phase (Fig. S6 and Sect. S1.3.3). These models – absorptive equilibrium partitioning

206

into pure organics or a mixed organic-aqueous phase – provide a baseline based on partitioning

207

theory as it is often implemented in chemical transport models against which observations can be

208

placed, as well as a framework for understanding the expected impact on partitioning of changes

209

in meteorology and particle composition.

210 211

Results and Discussion

212

Prevalence and variability of isoprene oxidation products

213

Products of isoprene (2-methyl-1,3-butadiene, C5H8) are of particular interest because this

214

single species accounts for approximately half (600 TgC per year) of all non-methane

215

hydrocarbon emissions.1 Laboratory studies of isoprene oxidation pathways expected in typical

216

rural ambient environments have shown SOA formation to occur in large part through gas-phase

217

formation of isoprene epoxydiol (IEPOX), followed by uptake into the particle and epoxide-

218

opening reactions that yield polyols, oligomers, and organosulfates, amongst other possible

219

products.51–54 Five of the most commonly measured tracers are reported here: two 2-



220

methyltetrols (2-methylthreitol and 2-methylerythritol) and three isomers of methyl-trihydroxy-

221

butene (“C5 alkene triols”).55 These species are formed in the particle and, as discussed above,

222

most chemically explicit models do not include their re-volatilization, but the data presented here

223

demonstrate substantial concentrations of these tracers in both the gas- and particle-phase.

Page 12 of 37

224

Isoprene oxidation products comprise a substantial fraction of SOA at both measurement

225

locations (Fig. 1a). In the summertime in rural Alabama, these five IEPOX-derived polyols

226

account for on average 8.0% of all OA mass (with some periods up to ~20%), and approximately

227

15% of OA mass in the Amazonian wet season (up to ~35%), in reasonable agreement with

228

contemporaneous and previous bulk measurements at both sites.34,48,56,57 The close correlation (r2

229

= 0.80) of total gas plus particle phase concentrations of these two compound classes (Fig. 1b,

230

data from IOP1 shown as an example) suggests that, as expected, they are formed through

231

similar chemical pathways.

232

Given their high concentrations, the dynamics, partitioning, and variability of these polyols

233

serve as useful examples for the properties and chemistry of particle-phase organic mass. Total

234

gas-plus-particle-phase isoprene oxidation products (Fig. 1c,d) are maximum during the

235

afternoon, approximately coincident with isoprene concentrations in both regions.35,47 Gas-phase

236

concentrations increase in the afternoon, while the daily cycle in particle-phase concentrations is

237

weaker (GoAmazon2014/5) or absent (SOAS). Though the correlation between total

238

concentrations of these compound classes supports a shared or related formation mechanism (e.g.

239

particle-phase reactions of IEPOX), the presence of a large but variable gas-phase component

240

indicates that either partitioning to the gas-phase occurs following in-particle formation of these

241

compounds, or gas-phase formation of these species are more important formation pathway than


Page 13 of 37


242

typically considered in models. In either case, a large gas-phase component must be considered

243

in understanding these tracers.

244

Diurnal trends in concentrations result in an average afternoon minimum in the fraction of each

245

compound that is in the particle (Fig. 2a,b), FP. However, changes in FP of these two compound

246

classes do not have the same daily pattern, with a sharper valley in the partitioning of C5 alkene

247

triols around 1600 (local) at both sites. Furthermore, while 2-methyltetrols have a similar or

248

smaller fraction in the particle phase in the Amazon wet season than at the SOAS site, the

249

opposite is true for C5 alkene triols. In general, IOP1 has lower particle-phase organic mass and

250

higher temperatures, though SOAS exhibits stronger diurnal changes in temperature; diurnal

251

profiles of these environmental conditions are provided in Fig. S11. Observed FP is closely

252

correlated for isomers (e.g. 2-methylthreitol vs. 2-methylerythritol, r2 = 0.86, shown in Fig. S14).

253

However, despite similar average FP and qualitative diurnal trends, observed partitioning of the

254

tetrols is in fact poorly correlated with partitioning of the triols (Fig. 2c). This conclusion

255

qualitatively extends beyond isoprene products to α-pinene oxidation products, which also have

256

poorly correlated partitioning (r2 < 0.2) even for compounds with similar average FP and

257

functional groups (examples shown in Fig. S17).

258

partitioning is driven primarily by external conditions (e.g. temperature or particle composition)

259

without consideration for compound functionality or identity is therefore insufficient to describe

260

the observations. Instead, observed partitioning appears to be driven in large part by compound-

261

specific processes that are not clear in typical measurements of these species (e.g. accretion

262

reactions, association with inorganic ions, compound-dependent temperature sensitivity of vapor

263

pressure, etc.), or by physical processes that are independent of vapor pressure (e.g. in-particle

264

diffusion).

A model in which variability of tracer



Page 14 of 37

265

Factors influencing partitioning of individual tracers

266

Observations of polyol partitioning are in fact found to be poorly described by either an

267

absorptive equilibrium partitioning model, as expected from the discussion above, or by

268

assumptions of non-volatility. The histogram of observed partitioning for these tracers (2-

269

methylerythritol shown as a representative example in Fig. 2d) demonstrates that FP is highly

270

variable, ranging from all mass in the particle-phase to nearly all mass in the gas-phase; on

271

average it is approximately evenly split between phases. These data do not support an

272

assumption of non-volatility. However, FP is also substantially higher than expected based on

273

vapor pressure (Fig. 2d, black solid line) and correlation with an organic absorptive partitioning

274

model is poor (r2 = 0.14); vapor pressure is calculated for each sample at the measured ambient

275

temperature using SIMPOL. High FP is not due simply to bias in the model from overestimation

276

of vapor pressure of this species or poor prediction of partitioning coefficients; the dashed line

277

Fig. 2d shows the results of the absorptive equilibrium partitioning model in which an empirical

278

uptake parameter (e.g. vapor pressure) is optimized to minimize the average difference between

279

the observed and expected FP (i.e. the slope of the fit) with no a priori assumptions about

280

volatility or solubility. This optimized equilibrium model necessarily better estimates average FP

281

but is no better at predicting temporal variability (r2 = 0.13), whether solubility in particle-phase

282

liquid water is included (as shown in Fig. 2d, with the inclusion of solubility described in the

283

Supplemental Information), or whether only absorption into organic mass is considered as in the

284

case of the organic absorptive equilibrium partitioning model (both correlations shown in Fig.

285

S7b,c). Changing the temperature dependence of the vapor pressure also does not improve its

286

ability to capture temporal variability. While these models do exhibit some skill in predicting FP

287

of some compounds (e.g. pinic acid, shown in Fig. S16 and ref.


20

), variability of observed

Page 15 of 37


288

partitioning for many individually measured tracers is not easily explained by current widely

289

used atmospheric modeling frameworks.

290

It should be noted that many details of equilibrium absorption are still uncertain and, as such,

291

are simplified in the models tested here. Time-independent factors (e.g. biases in vapor pressure)

292

likely do not contribute substantially to the observed differences, as they would be compensated

293

for by the optimization of the uptake parameter (dashed line, Fig. 2d). However, time- (or

294

composition-) dependent changes to equilibrium partitioning parameters cannot be excluded. In

295

many

296

partitioning models assume ideal conditions (an activity coefficient of unity) despite expected

297

dependence on compound functionality, particle phase state, and organic composition.60 Some

298

compounds discussed here have chemical properties (e.g. oxygen to carbon ratio, solubility, etc.)

299

similar to the bulk organic particle composition. For such compounds strong barriers to

300

partitioning are not expected for organic absorptive equilibrium partitioning and an assumption

301

of approximate ideality is not unreasonable, but even in these cases time-dependent variation in

302

activity cannot be discounted as a factor and non-ideal aqueous uptake may impact partitioning.

303

A comprehensive implementation of absorptive equilibrium partitioning with compound-

304

dependent and temporally-variable activity coefficients may better describe the observations, but

305

the complexity of organic composition and the uncertainty in relevant parameters create barriers

306

to any such comprehensive implementation.

implementations (though not all, e.g refs.

58,59

), organic absorptive equilibrium

307

Several factors impacting ambient observations may contribute to real complexity that require

308

greater understanding of atmospheric composition and chemistry, not only improved

309

implementation of absorptive equilibrium partitioning models. These include complex bonding

310

in particle-phase organic compounds that result in poorly characterized organic mass (e.g. the



Page 16 of 37

311

presence of low-volatility compounds that decompose during analysis), physical limitations that

312

impede equilibrium on the timescales of formation and measurement, or dissolution processes

313

that are not captured by an assumption of instantaneous thermodynamic equilibrium with an

314

aqueous phase. Diffusion limitations and high viscosity within the particle are not likely as the

315

particles are expected to be liquid at most times, which has been demonstrated for

316

GoAmazon2014/5.61 Solubility of organic species in pure particle-phase liquid water may, like

317

absorption into an organic phase, occur at approximate thermodynamic equilibrium.

318

Consequently the inability of the optimized equilibrium absorption model to capture observed

319

variability in FP, as well as poorly correlated FP for chemically similar compounds, suggests that

320

unmediated solubility is unlikely to dominate partitioning for most observed species. However,

321

there are complex uptake processes (e.g. solvation mediated by inorganic ions) that cannot be

322

discounted but are not yet sufficiently understood to be comprehensively implemented in

323

models. The high fraction in the particle of these tracers might therefore be explained by recent

324

work suggesting that biogenic oxidation products exist in large fraction as low-volatility products

325

that can thermally decompose to form constituent higher volatility components.25 The tracers

326

measured in this work represent many compounds traditionally included in chemically explicit

327

models, so these data provide a crucial step toward reconciling our traditional understanding of

328

the components of biogenic SOA – most of which have relatively high vapor pressures – with

329

bulk measurements, which often find ambient SOA to have relatively low volatility.21

330

Assuming relatively minor influence of dissolution and particle viscosity on observed

331

partitioning, contribution to particle-phase tracer concentrations by low-volatility mass can be

332

roughly estimated as the difference between observed FP and that estimated from vapor pressure.

333

Error on vapor pressure is estimated at approximately half an order or magnitude, so this


Page 17 of 37


334

approach is imprecise, but it is nevertheless a useful means to explore the potential ubiquity of

335

these low-volatility products. On average half of particle-phase 2-methylerythritol mass in the

336

southeastern U.S. is unexplained by vapor pressure of the pure species, and nearly three-quarters

337

during the Amazon wet season. While this analysis provides some insight, quantitative

338

understanding of low-volatility content of these tracer species assumes that absorptive

339

equilibrium partitioning models based on vapor pressure are sufficient to describe partitioning of

340

the pure species of these tracer compounds. There is some recent evidence for this assumption,25

341

but it has not yet been tested in a diversity of environments and does not include the potentially

342

important impact of processes that may slow timescales to reach equilibrium (e.g. kinetically-

343

limited reactions), or are poorly understood modifiers on that equilibrium (e.g. formation of

344

inorganic adducts, composition-dependent activity coefficients). Furthermore, the high particle-

345

phase liquid water content during these campaigns may increase partitioning to the particle

346

phase,58,62 but quantitative estimation of this effect is impeded by large uncertainties in solubility

347

and activity parameters. Consequently, the estimate of low-volatility content provided here is

348

qualitative. However, this inferred fraction of low-volatility mass is substantially higher than

349

measurement differences between SV-TAG and co-located filter-based measurement techniques

350

(Fig. S3), suggesting potential widespread measurement of low-volatility mass as pure discrete

351

tracer species across measurement platforms.

352

The observed high FP values for these tracers, coupled with recent observations of low-

353

volatility accretion products raises significant questions about the nature of the bonds keeping

354

these tracers in the particle phase. Many classes of low-volatility compounds have been observed

355

in ambient particle samples, but most of these cannot easily explain the observed tracer FP

356

results. Thermal decomposition of organosulfates, for instance, was found in laboratory tests of



Page 18 of 37

357

SV-TAG to negligibly contribute to measured mass (

Ambient Gas-Particle Partitioning of Tracers for ... - ACS Publications

Recommend Documents