Modeling the influence of alkane molecular structure on secondary organic aerosol formation Bernard Aumont,*a Marie Camredon,a Camille Mouchel‐Vallon,a Stéphanie La,a Farida Ouzebidour,a Richard Valorso,a Julia Lee‐Taylorb and Sasha Madronichb a LISA, UMR CNRS 7583, Université Paris Est Créteil et Université Paris Diderot, France. E‐ mail:
[email protected]‐pec.fr b National Center for Atmospheric Research, Boulder, CO, USA. Abstract Secondary Organic Aerosols (SOA) production and ageing is a multigenerational oxidation process involving the formation of successive organic compounds with higher oxidation degree and lower vapor pressure. Intermediate Volatility Organic Compounds (IVOC) emitted to the atmosphere are expected to be a substantial source of SOA. These emitted IVOC constitute a complex mixture including linear, branched and cyclic alkanes. The explicit gas‐phase oxidation mechanisms are here generated for various linear and branched C10‐C22 alkanes using the GECKO‐A (Generator for Explicit Chemistry and Kinetics of Organics in the Atmosphere) and SOA formation is investigated for various homologous series. Simulation results show that both the size and the branching of the carbon skeleton are dominant factors driving the SOA yield. However, branching appears to be of secondary importance for the particle oxidation state and composition. The effect of alkane molecular structure on SOA yields appears to be consistent with recent laboratory observations. The simulated SOA composition shows however an unexpected major contribution of multifunctional organic nitrates. Most SOA contributors simulated for the oxidation of the various homologous series are far too reduced to be categorized as highly oxygenated organic aerosols (OOA). On a carbon basis, the OOA yields never exceeded 10% regardless of carbon chain length, molecular structure or ageing time. This version of the model appears clearly unable to explain a large production of OOA from alkane precursors.
1. Introduction Long carbon chain hydrocarbons (C>10) are emitted in the atmosphere from biomass and as unburnt byproducts of biomass and fossil fuel combustion. These emissions form a complex mixture, including a large fraction of linear, cyclic and branched alkanes.1–4 These organic compounds of intermediate volatility (IVOC) are emitted largely in the condensed phase and then volatized by dilution5 in the atmosphere. Gas phase oxidation of IVOC leads to the production of organic species of enough low volatility to condense and contribute to secondary organic aerosols (SOA) production.6 Laboratory studies have shown that IVOC alkanes are potential SOA contributors.7–12 Modeling studies have also identified IVOC as substantial SOA precursors at the continental scale and in the plumes of megacities.13–17 SOA production in the Mexico City plume was explored using a detailed chemical scheme: the Generator of Explicit Chemistry and Kinetics of Organics in the Atmosphere – GECKO‐A.17 In these simulations, n‐alkanes were used as surrogate species of IVOC emissions. The model was found to be able to explain typical SOA levels but the prediction of the atomic O/C ratio fell short.17,18 SOA formation from the gas phase oxidation of intermediate volatility n‐alkanes was recently examined using the GECKO‐A tool.19 For the n‐alkane series, most SOA contributors were found to be reduced enough to be categorized as hydrocarbon‐like organic aerosols (HOA), although of secondary origin.19 Simulated results with GECKO‐A suggest that gas phase
n‐alkane oxidation cannot explain the large fraction of highly oxygenated organic aerosols (OOA) observed in the atmosphere. Branched alkanes are a major fraction of the emitted hydrocarbons in engine exhausts.4 These branched alkanes are more prone to fragment in the early stage of the oxidation than their corresponding linear analogues.7,9,11 This enhanced fragmentation is expected to alter both the SOA yields and the mean SOA oxidation state. The aim of this study is to examine the influence of the parent alkane molecular structure on SOA yield and oxidation states. The study focuses on the effect of the degree of branching of the carbon backbone for various C10‐C22 alkane series. The mechanism self‐generator GECKO‐A20 is used to generate highly detailed oxidation schemes, and box model simulations are performed to explore the formation and ageing of the particle phase subjected to continuous gas phase oxidation of the organic matter. Section 2 briefly presents the GECKO‐A modeling and results are discussed in Sect. 3.
2. Model description 2.1 The GECKO‐A modeling tool The number of species involved in the multigenerational gas phase oxidation of hydrocarbons grows exponentially with the size of the carbon skeleton.20,21 For a C8 hydrocarbon, the explicit description of the gas phase oxidation processes up to the final production of CO2 involves more than one million species, far exceeding the size of chemical mechanisms that can be managed manually. The chemical mechanism generator GECKO‐A is a computer program designed to overcome this difficulty. This tool is used here to develop oxidation scheme for alkane series with various degrees of branching. Reactions pathways and rate constants are assigned during the chemical mechanism generation on the basis of experimental data and Structure Activity Relationships (SAR). The protocol implemented in GECKO‐A to self generate the mechanisms is described by Aumont et al.,20 with chemistry updates described by Aumont et al.19 The relative contribution of fragmentation vs. functionalization oxidation routes is a key ratio in the context of SOA formation.22 Evolution along the fragmentation or functionalization routes is mostly dictated by fate of the alkoxy radicals produced as intermediates during the gas phase oxidation.7 In the previous version of GECKO‐A,19 the alkoxy radical chemistry was based on estimations provided by the SAR developed by Atkinson.23 Vereecken and Peeters24 have recently reported a SAR based on quantum chemical calculations to estimate the barrier heights for alkoxy radical decomposition for a large set of organic moieties. In particular, the SAR includes rates for leaving groups such as the nitrate moieties, which are prevalent in the oxidation mechanisms of long‐chain alkanes. The recent SAR provided by Vereecken and Peeters24 for alkoxy radical decomposition has therefore been implemented in GECKO‐A. Phase partitioning is described in the model assuming that the condensed phase behaves as an ideal well mixed homogeneous liquid phase. Gas/particle phase equilibrium is described by Raoult’s law:25 Pi=xiPivap (1) vap where Pi is the vapor pressure of the species i, xi its mole fraction in the condensed phase and Pi its equilibrium partial pressure. At phase equilibrium, the ratio ξi of the species in the condensed phase is given by:26
ξ
aer i
=
N i ,aer N i ,aer + N i , gas
−1
⎞ ⎛ M OA Pi vap ⎛ C* ⎞ = ⎜1 + × 106 ⎟ = ⎜⎜1 + i ⎟⎟ ⎟ ⎜ COA RT ⎝ COA ⎠ ⎠ ⎝
−1
(2)
where Ni,j is the number concentration in phase j (molecule of species i per cm3 of air), T is the temperature (K), R the ideal gas constant (atm m3 K‐1 mol‐1), COA the aerosol mass concentration (μg m‐3 of air), MOA the mean organic molar mass in the aerosol (g mol‐1) and C*i is an effective saturation mass concentration (μg m‐3 of air). The Nannoolal et al.27 group contribution method was used to estimate the vapor pressure of each non radical species included in the mechanism, as described by Valorso et al.28 This method was used in conjunction with the Nannoolal et al.29 method to estimate boiling points. Contributions for functional groups not provided by Nannoolal et al.27,29 were taken from Compernolle et al.30 The explicit description of n‐decane oxidation is expected to lead to a mechanism including about 108 species,20 one order of magnitude above the size of chemical schemes that can be both generated and solved. Simplifications are therefore required to reduce the schemes down to a manageable size. Usually, simplifications in GECKO‐A are performed based on the lumping of structural isomers.17,28,31 The description of oxidative trajectories along fragmentation or functionalization routes is a key aspect of the modeling study performed here and requires a detailed representation of alkoxy radical chemistry. The chemical fate of the alkoxy radical (O2 reaction, H‐shift isomerisation, or C‐C bond breaking) depends on the chemical structure in the vicinity of the alkoxy moiety.23,24 No lumping is therefore performed to keep the molecular structures of the successive generations of secondary organic species. Mechanism reduction has been performed as follows. Species produced with a maximum yield below 5×10‐6 are not treated further in the mechanism and the chemical removal of these species is considered as a final sink. This approximation does not significantly alter the mass budget, the cumulative loss of carbon atoms at the end of the simulation being less than 1 % of the initial organic carbon. Furthermore, gas phase oxidation becomes negligible for non‐ volatile species, where most of the species mass occurs in the condensed phase. According to Eq. 2 and for COA representative of polluted conditions (10 μg m‐3), a species is almost exclusively in the condensed phase at thermodynamic equilibrium when Pvap is below 10‐13 atm. The gas phase chemistry is therefore omitted for such low volatility species. Finally, to reduce the mechanisms to a manageable size, we assume high NOx conditions, i.e. the peroxy+peroxy reactions are ignored. This approximation is appropriate under most polluted conditions. Nevertheless, this high NOx condition remains hardly representative of the atmospheric oxidation on timescales exceeding one day. Restricting the study to high NOx conditions is therefore a severe simplification.19 Simulations performed here are clearly exploratory and intended to examine some, though not all, aspects of the alkane molecular structure on SOA formation. The numbers of species ultimately considered in the mechanisms are listed in Table 1 for various alkane series. Time integration of the set of ordinary differential equations associated with gas phase oxidation is solved using the “2‐step” solver.32,33 Phase equilibrium is enforced at each time step, as described by Camredon et al.34 Reactions in the condensed phase are not considered in this model configuration. Ageing of the particles is driven by gas phase chemistry that progressively shifts the various gas/particles equilibria as oxidation progresses. 2.2 Simulation conditions Simulations were conducted under conditions similar to those described by Aumont et al.19 The simulations are run for constant environmental conditions. The temperature is set to 298 K. Photolysis frequencies are calculated using the TUV model35 for mid latitude conditions and for a zenith angle of 45°. A constant OH source of 2×107 radicals cm‐3 s‐1 is included to initiate oxidation in the model. The NOx mixing ratio is set to 1 ppb. Note that the simulated SOA formation is almost insensitive to the prescribed NOx value, the fate of the organic peroxy radical being exclusively the reaction with NO as described above (i.e. high NOx conditions). In
this study, COA is set to a constant value of 10 μg m‐3, representative of polluted tropospheric conditions. The composition is assumed to be non‐volatile organic matter with a mean molar mass MOA of 250 g mol‐1, within the expected values for atmospheric organic aerosol.25,36,37 Oxidation of the parent hydrocarbon may obviously lead to the formation of SOA and therefore alters the prescribed aerosol composition and mass concentration. The hydrocarbon initial mixing ratio is set to an arbitrary value of 10 ppt carbon (ca. C0=6.5×10‐3 μg m‐3), a value low enough to not modify the prescribed aerosol properties. For these conditions (C0 3
< parent
Figure 4. Contribution of the species from a given generation to the particle composition (color histograms) and fraction of the species in the particle phase having the same carbon skeleton as the parent compound (grey histograms). Four homologous series are considered: linear (1st column), 2‐methyl series (2nd column), 2 and 3 methyl vicinal methyl groups in the middle of the backbone (3rd and 4th column) and for C10 (first row) to C22 alkanes (4th row). Results are given for Nτ=5. Budget is performed on a carbon atom basis.
H R
R H2O
OH
O2 OO.
ONO2
NO
R
R
R
(N)
NO2 O.
O2
R
OH
O
HO2 R
R
(K)
R
R
OO.
R
ONO2 (NO)
NO2 O2
OH R
HO2
R
O R
R
O.
OH OO.
OH R
R
R
OH R
R OH
NO2 O.
O2 R
(KO)
ONO2
OH
NO
OH
R
OH
NO
R
R
HO2
(NOO) OH
OH R
R
(KOO)
Figure 5. The main reactions involved in the functionalization trajectories between 2 successive generations. OH
O
NNOO
NNKO
NNKK
NNO
NNOO
NNN
NNKO
3 vicinal methyls
NKKO
NKOO
NNN
NKKO NNOO
NKOO
NNN
NNO
0.1
NNO
NNN
0.2
2 vicinal methyls
NNKK
NNKO NKOO NKKO NNKK NNOO
0.3
NNO
carbon atom ratio
C10 alkane
0.4
2-methyl
NNKO
linear
NNKO NKOO
other
KOO
NNO NOO KKOO NKOO NNKO NKKO
NN NO
NNKK
NNN NKO NN NKO
other
NNOO
NKKO
other
NNO
other NKOO NNKO NKKO NNKK NNOO
NNO NNN
NKO
NNO NOO KKOO NKOO NNKO
NKO KOO
NO
NN
KKOO NKOO NNKO NKKO NNOO
NNO
NN KOO
NO
KOO
other
KKOO NKOO NNKO NKKO
NNO
NO NKO NN
N
other
KKOO NKOO NNKO NKKO
NO NN NKO KOO
other N
NNO NOO KKOO NKOO
KO
KO
KO NO NKO KOO NN
0.1
NKOO
0.2
KKOO
N
other N
0.3
NO NN NKO KOO NNO
carbon atom ratio
0 0.4
C22 alkane
NKO
other NNO
KKOO NKOO NNKO NKKO
NKO KOO
0.1
N NO
0.2
NN
0.3
KO
carbon atom ratio
C18 alkane
0 0.4
other
NN
NOO
NN
0.1
NN
other NNN KKOO NKOO NNKO NKKO NNKK NNOO
NKO
0.2
NNO
carbon atom ratio
C14 alkane
0.3
NKO NNO NNN NOO KKOO NKOO NNKO NKKO NNKK NNOO other
0 0.4
0
N = nitrate group
K = ketone group
O = hydroxy group
Figure 6. Major contributors to the SOA composition. Position isomers are lumped into the same histogram. Colors discriminate contributors by number of functional groups borne by the species: dark blue = monofunctional species, light blue = difunctional species, orange = trifunctional species and red = species with more than 3 functional groups. N, K, O denotes species bearing the nitrate, ketone and hydroxy group, respectively. Four homologous series are considered: linear (1st column), 2‐methyl series (2nd column), 2 and 3 methyl vicinal methyl groups in the middle of the backbone (3rd and 4th column) and for C10 (first row) to C22 alkanes (4th row). Results are given for Nτ=5. Budget is performed on a carbon atom basis.