Secondary organic aerosol from limona ketone: Insights into terpene ozonolysis via synthesis of key intermediates

Article abstract of DOI:10.1039/b701333g

Limona ketone was synthesized to explore the secondary organic aerosol (SOA) formation mechanism from limonene ozonolysis and also to test group-additivity concepts describing the volatility distribution of ozonolysis products from similar precursors. Limona ketone SOA production is indistinguishable from α-pinene, confirming the expected similarity. However, limona ketone SOA production is significantly less intense than limonene SOA production. The very low vapor pressure of limonene ozonolysis products is consistent with full oxidation of both double bonds in limonene and furthermore with production of products other than ketones after oxidation of the exo double bond in limonene. Mass-balance constraints confirm that ketone products from exo double-bond ozonolysis have a minimal contribution to the ultimate product yield. These results serve as the foundation for an emerging framework to describe the effect on volatility of successive generations of organic compounds in the atmosphere. the Owner Societies.

Full text of DOI:10.1039/b701333g

View Article Online / Journal Homepage / Table of Contents for this issue
PAPER
www.rsc.org/pccp | Physical Chemistry Chemical Physics
Secondary organic aerosol from limona ketone: insights into terpene
ozonolysis via synthesis of key intermediateswz
Neil M. Donahue,* Joshua E. Tischuk,y Bryce J. Marquisz and
Kara E. Huff Hartz8
Received 29th January 2007, Accepted 13th March 2007
First published as an Advance Article on the web 17th April 2007
DOI: 10.1039/b701333g
Limona ketone was synthesized to explore the secondary organic aerosol (SOA) formation
mechanism from limonene ozonolysis and also to test group-additivity concepts describing the
volatility distribution of ozonolysis products from similar precursors. Limona ketone SOA
production is indistinguishable from a-pinene, confirming the expected similarity. However,
limona ketone SOA production is significantly less intense than limonene SOA production. The
very low vapor pressure of limonene ozonolysis products is consistent with full oxidation of both
double bonds in limonene and furthermore with production of products other than ketones after
oxidation of the exo double bond in limonene. Mass-balance constraints confirm that ketone
products from exo double-bond ozonolysis have a minimal contribution to the ultimate product
yield. These results serve as the foundation for an emerging framework to describe the effect on
volatility of successive generations of organic compounds in the atmosphere.
Roughly half of the PM2.5 mass consists of a vast mixture of
1
. Introduction
8,9
organic compounds, known as organic aerosol (OA). The rest
are reasonably well-understood inorganic salts (principally
ammonium sulfate and ammonium nitrate). The particles are
generally either acidic or neutral. The organic fraction is of great
interest because it is poorly understood, because organic composi-
tion strongly influences the properties described above, and be-
cause organics exhibit an extremely rich combination of chemical
behaviors, including both gas- and condensed-phase oxidation
Limona ketone (4-acetyl-1-methylcyclohexene) is a potentially
1
important intermediate in limonene ozonolysis (Scheme 1).
The system is both intrinsically important as a major atmo-
spheric source of secondary organic aerosol (SOA) and che-
mically interesting as a model system and a tool to deconvolve
multiple generations of oxidation in SOA formation.
Fine particulate matter plays a pivotal role in the atmo-
sphere. It is typically defined as aerosol particles less than
10
chemistry and condensed-phase macromolecular chemistry.
2
.5 microns in diameter (PM2.5). Fine particles strongly influ-
OA is traditionally divided into two categories with simple
definitions: primary OA (POA), which are particles directly
emitted into the atmosphere, most often from combustion
sources, and the SOA mentioned above, which is condensed-
phase organic material formed during gas-phase chemical
ence climate, most notably through the indirect effect, where
their role as cloud condensation nuclei (CCN) controls the size
distribution and total surface area of cloud droplets and thus
2
,3
cloud lifetimes and reflectivity. Feedbacks associated with
the indirect effect are one of the largest uncertainties in the
4
climate system. Fine particles also have severe negative health
consequences, causing roughly 50 000 premature deaths each
5
year in the United States alone along with a host of chronic
health effects. The mechanism responsible for these effects
appears to be deposition in the deep lung followed by cardi-
6
opulmonary distress, though other mechanisms including
direct passage up olfactory nerves to the brain have been
7
reported.
Center for Atmospheric Particle Studies, Carnegie Mellon University,
USA
w The HTML version of this article has been enhanced with colour
images.
z Electronic supplementary information (ESI) available: NMR of
limona ketone synthesis. See DOI: 10.1039/b701333g
y Present address: J. E. T. now at Agilent Technologies, Cambridge
MA, 15237.
z Present address: B. J. M. now at University of Minnesota,
Chemistry Department, Minneapolis, MN, 55455.
8 Present address: K. E. H. H. now at Department of Chemistry and
Biochemistry, Southern Illinois University, Carbondale, IL 62901.
Scheme 1 Limonene oxidation pathways.
This journal is ꢀc the Owner Societies 2007
Phys. Chem. Chem. Phys., 2007, 9, 2991–2998 | 2991

View Article Online
reactions in the atmosphere. SOA is known to be semi-
pounds. For several reasons we follow the total mass concen-
ꢁ3
1
volatile, and so SOA formation experiments typically seek
1
tration of organic aerosol particles, COA, typically in mg m
.
ꢁ
3
to parameterize the vapor pressures of a large set of (typically
Atmospheric levels range from 1 mg m in the remote atmo-
1
unidentified) reaction products. Recent experiments have
2
24
ꢁ3
25,26
sphere to 100 mg m in very polluted urban settings.
The corresponding total mass concentration of a compound
demonstrated that a large fraction of POA is also semi-
1
3–15
volatile.
We thus now view all organic aerosol as a highly
(in all phases) is C . The vapor pressure can be converted into
i
*
i
dynamic, evolving mixture subject to multiple generations of
0,15–17
oxidation in both the vapor and condensed phases.
an effective saturation concentration, C
. This subsumes non-
1
ideality such as activity coefficients; derivations and extensive
1
1,12,15
Emissions, dilution, and chemistry will all alter the ambient
volatility distribution over time. Now we must understand and
describe multiple generations of oxidation of semi-volatile
compounds, with reactions that transform the volatility
distribution in either direction in a mixture containing hun-
dreds of important compounds.
discussions are presented elsewhere,
cases C* is presented as a partitioning constant, K = 1/C*.
Given these values, one can then find a partitioning coeffi-
cient, x . This is the fraction of compound i found in the
though in some
1
1,12
i
condensed phase at equilibrium
1
xi ¼
ð1Þ
These issues pose a dilemma. On the one hand, the complex-
ity of the system demands a simplified representation that still
describes the range of behaviors and sensitivities occurring in
the atmosphere. On the other hand, we must understand at least
a subset of the system in detail in order to be able to frame and
test that simplified representation. Good model systems are
both intrinsically important to the atmosphere and suitable
archetypes for broader behavior. Ozonolysis of limonene is one
such system. Limonene has a significant flux into the atmo-
sphere, and monoterpene ozonolysis is a major recognized SOA
source. Because limonene is doubly unsaturated, it is also an
excellent model system for multi-generational evolution of the
vapor-pressure distribution. Ozonolysis of terpenoids is a major
ꢂ
1
þ C =COA
i
Eqn (1) is an expression of Raoult’s law. It is a simple
saturation curve, of the same form as a Langmuir–Hinshel-
wood isotherm or a Lindemann–Hinshelwood pressure falloff
curve in kinetics. The interpretation is simple as well: when the
mass concentration of organic aerosol is equal to the satura-
tion concentration of some compound, that compound will be
i
evenly divided between the condensed and vapor phases (x =
*
0.05). However, if COA greatly exceeds C
that compound will be in the condensed phase (x
the other extreme almost none of that compound will be in the
condensed phase (x C 0). The total aerosol concentration is in
i
, then almost all of
i
C 1), and at
i
1
8,19
source of SOA,
and the selectivity of ozone to double
fact derived iteratively from all of the individual concentra-
bonds—and the wide range of ozone–alkene rate con-
stants—makes the chemical mechanism vastly simpler than
oxidation of large organic compounds by OH radicals.
tions and partitioning coefficients
X
It is important to note that coexistence demands that if there is a
condensed phase, some amount of any compound will be in it,
and of course there will always be some vapor pressure even for
COA ¼
xiCi:
ð2Þ
i
Recent experiments on the photoxidation of isoprene have
shown that SOA can be formed as a second-generation
product from oxidation of first-generation methacrolein
2
0
(
MAK). Though only a very small fraction of the mass of
‘
non-volatile’ compounds. This can be important in either direc-
oxidized isoprene winds up as SOA, this may well be impor-
17
2
1
tion, with gas-phase oxidation of low-volatility material and
20,27
condensed-phase processing of volatile monomers
tant due to the enormous biogenic isoprene flux. However,
monoterpenes have twice as many carbons as isoprene, and
consequently tend to yield products with much lower vapor
potentially
playing an important role in the atmosphere by altering volatility
and mass distributions of semi-volatile organics.
2
2
pressures when they are oxidized. Oxidation of terpenes is
thus a major, well-known source of SOA, but in almost all
cases only a single generation of oxidation has been considered
in chamber experiments. In one recent study, though, multi-
ply-unsaturated terpenes were shown to exhibit more compli-
cated SOA-formation behavior than singly-unsaturated
terpenes, with continued SOA formation even after complete
We have recently proposed that organic aerosol can be
*
described by a ‘semi-volatile basis set’, {C }. Individual com-
i
pounds are lumped into volatility bins and each bin is sepa-
rated by an order of magnitude in saturation concentration
ꢂ
4
5
6
ꢁ3
C ¼ f0:01; 0:1; 1; 10; 100; 1000; 10 ; 10 ; 10 g mg m
i
ð3Þ
2
3
terpene consumption. This behavior was observed in both
photooxidation’ experiments, in which both ozone and OH
ꢁ
3
ꢁ7
‘
For reference, 1 mg m is roughly 10 Torr. We then reduce
a complex reaction, such as ozonolysis, to
radicals are formed during photochemical oxidation of hydro-
carbons in an illuminated chamber, and pure ozonolysis
Precursor þ O3 ! a1p1 þ a2p2 þ :::
ð4Þ
is the mass yield of products p with saturation
experiments, in which a ‘HO
x
conditioner’ (i.e., 2-butanol)
was added to an ozone + terpene mixture in order to convert
2
OH radicals generated in the ozonolysis reaction.
where a
concentration C
1
1
3
*
ꢁ1
1
= 0.01 mg m , etc. Our objective is to
identify the volatility distribution of the products {a}, where
Sa 4 1 because the oxidation reactions typically add mass to
i
2
. Background
the products. This greatly simplifies representation of the
aerosol and of SOA formation chemistry when dealing with
multiple reactions and multiple generations of oxidation
We need to understand how organic reaction products parti-
tion between the vapor and a condensed phase consisting of a
complex mixture of thousands of individual organic com-
1
7
chemistry.
2
992 | Phys. Chem. Chem. Phys., 2007, 9, 2991–2998
This journal is ꢀc the Owner Societies 2007

View Article Online
limonene is approximately 30 times faster than the exo double
3
bond, but by studying limona ketone we can effectively ‘pre-
1
select’ one of the two possible pathways for the exo ozonolysis
(
pathway 2b in Scheme 1)—the only pathway in which the
large reaction fragment is a stable molecule rather than a
Criegee intermediate (CI).
3. Limona ketone synthesis
Limona ketone was synthesized using a Diels–Alder reaction
3
of isoprene and methyl vinyl ketone on dry silica gel. SiO
2
2
was dried in an oven overnight. Equimolar amounts of
isoprene and methyl vinyl ketone were added on top of the
dried SiO in a round bottom flask and stirred overnight in a
2
water bath at 55 1C. The product was extracted from the
reaction flask with methanol. After washing the organic phase
4
with water, it was dried over MgSO . A yellow-colored oil was
Fig. 1 The fraction of material in the condensed phase for different
collected after thorough rotary evaporation down to 50 mbar.
We analyzed the collected product with H- and C-NMR.
The spectra are shown in the supporting information, along
with simulations. H-NMR analysis of the product shows the
requisite unsaturated proton at 5.3 ppm and two methyl
groups at 1.65 and 2.15 ppm in both the experimental and
simulated spectra. Isomerization of the product with one
major and one minor compound appears as splitting at 1.65
and 2.16 ppm, revealing the presence of a 4-acetyl-2-methyl-
cyclohexene Diels–Alder isomer. In the C-NMR spectra, the
major isomer’s unsaturated carbons appear at 119 and
134 ppm, with the ketone carbon found at 210 ppm. The ring
carbons and methyl groups are found between 23 and 50 ppm,
with the experimental data matching the simulation nicely.
The major and minor isomers were structurally elucidated
by heteronuclear multiple-bond coupling (HMBC) long-range
coupling experiments. The ratio of limona ketone to the
4-acetyl-2-methylcyclohexene isomer was determined to be
approximately 3 : 1 by integration of the methyl proton peaks.
HMBC analysis initially shows coupling of the 1-methyl
protons at 1.64 ppm in the 1-D H-NMR with both isomers’
vinylic tertiary C-NMR carbons at 119 and 121 ppm. The
major isomer’s methyl protons couple with the major vinylic
tertiary carbon peak seen in the spectrum at 119 ppm while the
minor proton peak couples with the minor vinylic C-NMR
peak at 121 ppm. After determining which tertiary vinylic
carbon was associated with the major compound, long-range
coupling shows that the 119 ppm major tertiary vinylic carbon
couples with the proton on the acetyl bound carbon. The
minor vinylic carbon does not show this association as it is out
of range, three carbons away. The minor isomer’s proton,
found adjacent to the acetyl substituent, shows a slight
coupling with the 134 ppm quaternary vinylic carbon. This
proves that the minor isomer has the acetyl substituent
2 carbons away from a quaternary unsaturated carbon in the
meta position, while the major product has a para substituted
acetyl group to C-1. From this analysis, we conclude that the
major product is limona ketone while the minor product is the
meta substituted isomer, 4-acetyl-2-methylcyclohexene.
ꢁ3
bins in the semi-volatile basis set for COA = 1 mg m (arrow). Lower-
volatility material is almost completely in the condensed phase
(
x = 1), while higher-volatility material is almost completely vapor-
ized (x = 0). If the aerosol loading were higher, for example COA
=
ꢁ3
1
0 mg m , the curve would be shifted 1 decade to the left, with the
ꢁ3
C* = 10 mg m bin showing 50–50 partitioning.
ꢁ
3
The bins at 1000 mg m and above represent ‘intermediate
volatility organic compounds’ (IVOC), which are not signifi-
cantly partitioned into the condensed phase in the atmosphere
but which have a large mass concentration compared with
1
4
typical COA. These derive from combustion emissions as well
as first-generation oxidation of ‘traditional SOA’ precursors.
1
7
The IVOC are a potentially potent source of SOA. Tradi-
tional SOA precursors such as monoterpenes have C*s of
8
ꢁ3
1
0 mg m and above.
The great value of the basis set is that it allows us to describe
partitioning without identifying all of the compounds forming
the organic aerosol solution, and this lets us define chemical
transformations that alter saturation concentrations and par-
titioning without either identifying all of the material or
writing an essentially unknowable complete chemical mecha-
1
5
nism. For example, Fig. 1 shows how material will partition
ꢁ
3
for COA = 1 mg m (indicated on the figure with an arrow).
The basis set also underscores the importance of designing
experiments spanning the atmospheric range of aerosol con-
centrations. To enhance signal to noise, many experiments
have been conducted at elevated aerosol levels (often
ꢁ
3
4
1000 mg m ), where partitioning will be completely differ-
ent than in the atmosphere. This is a serious mistake.
2
8–30
It is very easy to fit data from both SOA experiments
1
4
and primary-emissions dilution experiments using eqn (1)
and (2). One obtains meaningful parameters spanning the
basis set (for example product yields in SOA experiments
and a flux distribution in emissions experiments) though
parameters are only constrained if the actual COA range of
the experiment falls near the C* for a given bin. In this paper
we shall use basis-set fits of SOA production from limona
ketone + ozone as well as limonene + ozone to constrain the
atmospheric mechanism for oxidation of the exo double bond
in limonene. Gas-phase ozonolysis of the endo double bond in
From GC/MS analysis, the strongest elution shows m/z of
138 and a retention time consistent with limona ketone. This
peak contains approximately 98% of the eluted material based
This journal is ꢀc the Owner Societies 2007
Phys. Chem. Chem. Phys., 2007, 9, 2991–2998 | 2993

View Article Online
on GC peak-height integration. All in all, the calculated
limona ketone yield was approximately 70%, with 25%
different scavengers have been observed, but for endocyclic
alkenes such as limona ketone the primary effect appears to be
associated with HO2 production from the scavenger, with
4
-acetyl-2-methylcyclohexene secondary product. The hypoth-
esis under consideration here is that mono-unsaturated mono-
terpenes with a 1-methylcyclohexenyl functionality should all
give similar SOA product yields, with respect to volatility.
Specifically, we wish to determine whether limona ketone and
a-pinene ozonolysis products show similar volatility. Confir-
mation of this hypothesis is not impeded by the mixture of
isomers we obtained and so we continued with SOA experi-
ments using the product as synthesized.
2-butanol showing relatively high HO formation (which is
2
3
6
desirable). We use 2-butanol for consistency in all of our
experiments. Also, in roughly 2/3 of the experiments we use a
small mass of ammonium sulfate seeds to promote condensa-
3
0
tion of organic vapors. We see no systematic dependence on
the presence of inorganic seed aerosol. In all other respects it is
assumed that the target chemistry remains constant as we vary
both limona ketone and ozone concentrations.
4
. Experimental
5. Results
We measure a mass partitioning coefficient by oxidizing a
The results of twelve individual experiments are reported in
Table 1 and plotted in Fig. 2, showing the aerosol mass
known mass of a precursor (limona ketone, DC ) and obser-
lk
0
ving the mass of organic aerosol produced (DCOA). The
resulting aerosol mass fraction is
fraction, x , versus the mass of aerosol produced, C . This
OA
1
2
form of presentation was first proposed by Odum et al. to
emphasize how the partitioning of semi-volatile compounds
depends on the total condensed-phase mass available for an
ideal solution. This form also enables direct fitting under the
DCOA
DClk
x ¼
ð5Þ
1
5,28
We measure DClk using both proton transfer reaction mass
spectrometry (PTRMS) and gas chromatography, and we
measure aerosol volume using ion mobility with a scanning
mobility particle sizer (SMPS). The volume is converted to a
volatility basis set.
For example, we see here that the
condensed-phase products amount to around 5% by mass of
ꢁ3
the consumed limona ketone at 1 mg m total aerosol con-
ꢁ
3
centration, but upwards of 20% by mass at 100 mg m . The
cumulative mass yield of products with saturation vapor
ꢁ3
normalized mass (DCOA) with a nominal density of 1 g cm
;
ꢁ3
to indicate this we represent the normalized aerosol mass
fraction as x. These normalized values can be converted easily
to absolute masses after later experiments determine the SOA
density.
pressures less than or equal to 1 mg m is thus no more than
0.1 or so, while the cumulative mass yield of products with
ꢁ
3
saturation vapor pressures less than or equal to 100 mg m is
roughly 0.3. In fact, a basis-set fit to the data using the basis set
ꢁ3
There are numerous experimental details vital to obtaining
accurate mass fractions, all of which have been described in a
succession of papers treating our own experimental cham-
shown in eqn (3) (0.01–1000 mg m ) gives mass yields of
ai ¼ f0:004; 0:005; 0:08; 0:05; 0:09; 0:18g ð6Þ
2
8,31,33,34
ꢁ3
ber.
The salient features are that we perform experi-
and indeed we see a cumulative yield of 0.09 at 1 mg m and
3
ꢁ3
ments in a cleaned 10 m Teflon reactor enclosed in a
temperature-controlled room and filled with dried, filtered
air at relative humidities of 5–15%. Cleaning includes over-
night purging with filtered 40 1C air containing 41 ppm ozone
under UV illumination.
0.23 at 100 mg m . Details of the fitting procedure are
2
8
described in Presto and Donahue. A 63% confidence interval
is shown with dashed lines in Fig. 2—it is meaningless to quote
uncertainties for any single yield because of covariance
amongst the fit parameters, but the overall confidence of the
AMF values over the range covered by the data is of order
20%.
During these experiments, with ozone between 0.5 and
2
ppm, the limona ketone was completely removed (3 e-folds)
in 3–15 min. Suspended particles are lost via deposition to the
chamber walls with a timescale of B2 hr. When the deposition
timescale is sufficiently longer than the precursor oxidation
timescale (as it is here), the suspended aerosol volume can be
adjusted for wall losses with a simple first-order correction,
which typically adds 10–20% to the maximum observed
During one experiment with relatively low ozone concen-
trations (85 ppb) we measured limona ketone decay using the
PTRMS, following 2.5 e-folds of log–linear decay over 400 s to
Table 1 Aerosol production results
3
3,35
33
aerosol mass.
As with a-pinene, the SOA shows no sign
ꢁ3
ꢁ3
ꢁ3
w
lk (ppb) DClk/mg m DCOA/mg m
wO3 (ppb) T/1C Cseed/mg m
of aging, instead settling down to a constant modal size after
the aerosol growth period. Repeatability is of order ꢃ10%,
with an added absolute uncertainty of approximately ꢃ0.01 mg
112.13
9.07
23.60
1.80
.18
176.53
636.0
333.0
133.0
66.7
109.00
42.00
10.04
10.74
0.32
168.90
184.50
52.49
63.04
61.29
58.62
130.42
0.50
0.98
1.00
1.50
1.70
1.50
1.50
0.70
0.83
1.20
0.85
0.91
25
22
24
25
25
25
25
24
24
24
23
24
2.00
15.20
2.30
1.50
2.68
2.18
2.94
2.00
0.00
0.00
0.00
0.00
5
ꢁ
3
m
, while comparison with the literature suggests that an
1
1
uncertainty of approximately ꢃ15% is appropriate for similar
experiments.
6.7
1000.0
800.0
200.0
200.0
200.0
200.0
700.0
142.00
35.50
We routinely add 10 ppm of 2-butanol as an OH radical
3
3,36
scavenger.
Ozone–alkene reactions are a source of OH
3
5.50
radicals in the gas phase, and these OH radicals would
otherwise react with the limona ketone and interfere with
the target chemistry. Differences in SOA formation with
35.50
35.50
24.00
1
2
994 | Phys. Chem. Chem. Phys., 2007, 9, 2991–2998
This journal is ꢀc the Owner Societies 2007

View Article Online
stantially reduced vapor pressures; this means that ketone
production from exo ozonolysis is at best a minor pathway.
We shall discuss these conclusions in turn.
Endo oxidation
There is every reason to believe that the basic mechanism for
compounds with a 1-methylcyclohexenyl moiety does not
change much with distant substitutions, including the 4–6
bridge of a-pinene. A nominal mechanism is shown in Scheme
2
.
This includes pathways 1a and 1b in Scheme 1 for limonene.
The mechanism applies to gas-phase ozonolysis, where the CI
is formed with an enormous amount of internal vibrational
ꢁ
1 38
energy (B200 kJ mol ). This is a representative pathway
for low-NO conditions; we use it only to place a reasonable
x
constraint on the expected product mass. For the sake of
simplicity we show a single reaction pathway, with several
steps subsumed in the second arrow including isomerization
and prompt dissociation of the ‘hot’ CI, subsequent therma-
lization of the resulting organic radical followed by addition of
molecular oxygen to form an organo-peroxy radical (RO2),
and finally reaction of that RO2 with HO2 to form the
peroxide shown.
0
Fig. 2 Aerosol mass fractions x = DCOA/DClk for the limona ketone
+ ozone reaction, plotted vs. the organic aerosol mass concentration
(
C
OA). SOA formation from limona ketone is similar to that from
a-pinene (lower dashed curve), but a factor of four less intense than
from limonene (upper dashed curve).
give a second-order rate constant at 298K for the ozone +
ꢁ1 ꢁ1
The actual endo mechanism almost certainly involves many
reaction products—the wide range in C* values needed for the
ꢁ
16
3
cm molec
limona ketone reaction of 2.7 ꢄ 10
s , in
good agreement with other compounds containing the
basis-set fit suggest this, and several analytical studies on
39–41
3
-methylcyclohexenyl moiety. There are numerous potential
7
1
terpene oxidation products confirm it.
This general path-
sources of systematic error in rate constant determinations
based on slow decay in a smog chamber; our best estimate
based on experience for alkenes with well-known rate con-
stants is an accuracy of order 30%.
way is also consistent with the large observed OH yields from
4
2,43
cyclohexene, a-pinene and limonene.
However, the details
of the gas-phase mechanism remain uncertain; this is one of
several reasons why we use the volatility basis-set formalism in
order to represent the vapor-pressure distribution in spite of
ill-defined reaction products.
The bottom line for these experiments is that limona ketone
behaves kinetically and mechanistically very much like
a-pinene, consistent with our hypothesis.
Finally, group additivity relationships developed to describe
vapor-pressure changes with substitution suggest that repla-
cing the vinyl CH
2
in limonene with a carbonyl in limona
6
. Discussion
ketone should have at most a modest influence on the vapor
2
2
The most obvious feature of the data shown in Fig. 2 is the
comparison with AMF values for limonene and a-pinene
shown as dashed curves. Limona ketone generates SOA with
an efficiency that is statistically indistinguishable from
pressure. All of the available indicators thus suggest a similar
vapor-pressure distribution for the ozonolysis products of
limona ketone and a-pinene, as well as many C10 compounds
containing a 1-methyl-cyclohexene moiety. This is consistent
with our observations.
2
8
i
a-pinene (where a = {0.004, 0, 0.05, 0.09, 0.12, 0.18} ), but
is much lower than limonene.
Exo oxidation
We can draw two general conclusions from this. First, the
common features for limona ketone and a-pinene are consis-
tent with a common mechanism for the endocyclic ozonoly-
sis—this is presumably shared by the endocyclic double bond
in limonene as well. Second, the greatly enhanced SOA for-
mation for limonene indicates that additional oxidation of the
exo double bond in limonene generates products with sub-
The second issue is the branching of the reaction between
ozone and the exo double bond (pathway 2a vs. 2b in
Scheme 1). If the reaction occurs in the gas phase, this
branching ratio should not depend strongly on whether path-
3
1
way 1 or pathway 2 occurs first. In fact, k I 30 k , so
1
2
pathway 1 predominates in limonene ozonolysis. However, we
Scheme 2 Endocyclic gas-phase mechanisms without NO
This journal is ꢀc the Owner Societies 2007
x
.
Phys. Chem. Chem. Phys., 2007, 9, 2991–2998 | 2995

View Article Online
can use limona ketone ozonolysis as a surrogate to probe half
of the possible second-generation products following pathway
1
a and 1b (in essence simply substituting the terminal CH2
with an O).
Because ozonolysis of limonene yields significantly less-
volatile products than either limona ketone or a-pinene, we
can quickly conclude that limona ketone itself is a minor
product in limonene ozonolysis.
We can develop stronger constraints with a mass balance.
There is strong evidence that for SOA from limonene ozono-
2
9
x
lysis (at low NO ), both double bonds have been oxidized.
Our basis-set fits from limona ketone and limonene reveal the
mass in each volatility bin in each case. For example, the mass
yields of products from complete limonene oxidation (plotted
Lim
as a dashed grey line in Fig. 2) are a
= {0, 0.03, 0.29, 0.31,
i
2
9
Lim
Lim
0
.30, 0.60}, so a
consistent with complete oxidation of C limonene leading
= Siai
= 1.53. This is broadly
1
0
Lim
Fig. 3 Constraints on the limona ketone formation branching ratio
from ozonolysis of the exo double bond in limonene (b = 1 is
complete branching toward the carbonyl oxide on the limonene
residue).
to a C C 1.65 for all products containing
more than 1 carbon) where almost all of this mass is accounted
9
O
7
product (a
ꢁ3
for at C* r 1000 mg m
.
If the ozonolysis of limonene consists of pathway 1 and
pathway 2 in Scheme 1 in succession, if the branching ratio of
each pathway does not depend on the order, and if the
reactions occur in the gas phase, we can use the limona ketone
to close the mass balance. Specifically, we shall assume that the
volatility distribution of the ultimate products does not de-
pend on whether the endo or exo double bond is oxidized first.
Furthermore, because the endo ozonolysis products are teth-
ered, we shall assume that the ultimate volatility distribution is
double bond would lead to rapid secondary ozonide forma-
7
tion. The end result might be a C10O substituted secondary
ozonide, rather than limona ketone products. This would have
a similar mass to our ‘nominal product’ as well as a very low
vapor pressure.
In the end we can determine what fraction of the exo double
bond does not resemble limona ketone oxidation, either
because the primary ozonide of the terminal double bond
determined by the products of pathway 2 in Scheme 1.
LCI
2
favors pathway 2a, or because any CH OO is scavenged
Products from pathway 2a we label a
and products from
within the aerosol to form a secondary ozonide. In all like-
lihood a combination of these factors is responsible for what
we observe. Either way, there is very little gas-phase produc-
tion of ketones from the terminal double bond in limonene
ozonolysis.
LK
pathway 2b we label a . From our fit to the limona ketone
LK
ꢁ3
data, a
= 0.41 for C* r 1000 mg m , with a ‘nominal
9 5
product’ formula (neglecting H) of C O . The branching ratio
toward pathway 2a is b. In general, branched terminal alkenes
have been observed to slightly favor the more substituted
LCI
44,45
carbonyl oxide (a ), with b C 0.65–0.80.
With our mass-balance constraints, the total mass yield
7. Conclusions
Lim
from limonene ozonolysis (a ) is
The work presented here establishes constraints on the branch-
ing of the second oxidation step in limonene ozonolysis. This is
one piece of a challenging analytical problem convolving the
difficulties of sampling and measuring very low levels of
Lim
LCI
LK
a
¼ ba
þ ð1 ꢁ bÞa
ð7Þ
LCI
where a
and b are the unknowns. Eqn (6) can be trivially
LCI
ꢁ3
solved for a
as a function of b
products (o10 mg m ) with the complex and poorly con-
strained gas-phase Criegee mechanism. The findings lead to
two important insights into the oxidation behavior of
terpenes. First, the volatility distributions following oxidation
of similar precursors are likewise similar. Specifically in this
case, limona ketone and a-pinene share a 1-methylcyclohexene
functionality and their ozonolysis products show very similar
aerosol volatility distributions. Second, multiple oxidiation
steps of poly-unsaturated terpenes can generate high yields
of very low vapor pressure products. This is not prima facie
obvious, as complete oxidation will ultimately lead to volatile
1
1 ꢁ b
LCI
Lim
LK
a
¼
a
ꢁ
a
ð8Þ
b
b
so the observed SOA from limonene could be explained by a
LCI
number of different combinations of a
Fig. 3. However, it is difficult to imagine a mass yield ratio
and b, as shown in
9
much greater than 1.7—a C O : C of nearly 1 : 1, meaning that
b Z 0.85.
One other significant possibility exists—we have shown that
the exo double bond in limonene is oxidized by rapid hetero-
geneous uptake of ozone onto fresh, unsaturated SOA parti-
1
6
2
CO .
cles, at least under low-NO
3
x
conditions such as those
Two possibilities remain viable. First, the regio chemistry of
the large terminal alkenes in the gas phase may favor leaving
the Criegee moiety on the larger, more substituted fragment.
This would also explain, for example, why ozonolysis of
b-pinene gives only slightly less SOA than a-pinene. Alternately,
1
employed in this study. In a solution where essentially all
of the organic compounds contain at least one carbonyl
moiety (the fresh SOA), it is very likely that any CH OO
2
produced following condensed-phase ozonolysis of a terminal
2
996 | Phys. Chem. Chem. Phys., 2007, 9, 2991–2998
This journal is ꢀc the Owner Societies 2007

View Article Online
if the terminal double bond is in fact oxidized in the condensed
phase the ketone could still be an important intermediate
product, only to be lost to SOZ formation. These are not
exclusive; what we know with certainty is that ketone produc-
tion is very small.
11 J. F. Pankow, An absorption model of the gas/aerosol partitioning
involved in the formation of secondary organic aerosol, Atmos.
Environ., 1994, 28, 189.
1
2 J. Odum, T. Hoffmann, F. Bowman, D. Collins, R. Flagan and J.
Seinfeld, Gas/particle partitioning and secondary organic aerosol
yields, Environ. Sci. Technol., 1996, 30, 2580.
1
3 E. M. Lipsky and A. L. Robinson, Effects of dilution on fine
particle mass and partitioning of semivolatile organics in diesel
exhaust and wood smoke, Environ. Sci. Technol., 2006, 40,
155–162.
This work represents a step toward solution of the ‘genera-
1
6
tion problem’ in atmospheric organic aerosol chemistry. The
problem is how to represent the complex, multi-generation
chemistry associated with the complex mixture that comprises
atmospheric organic aerosols. This work reveals clearly that
ozonolysis, at least, can lead to highly oxidized products from
terpenes with a large carbon number and an O : C ratio
approaching unity, and that the effect of the oxidation chem-
istry on similar precursors shares common features. The next
question is whether continued oxidation of these products
1
4 M. K. Shrivastava, E. M. Lipsky, C. O. Stanier and A. L.
Robinson, Modeling semi-volatile organic aerosol mass emissions
from combustion systems, Environ. Sci. Technol., 2006, 40,
2
671–2677.
15 N. M. Donahue, A. L. Robinson, C. O. Stanier and S. N. Pandis,
Coupled partitioning, dilution, and chemical aging of semivolatile
organics, Environ. Sci. Technol., 2006, 40, 2635–2643.
1
6 N. M. Donahue, K. E. Huff Hartz, B. Chuong, A. A. Presto, C. O.
Stanier, T. Rosenørn, A. L. Robinson and S. N. Pandis, Critical
factors determining the variation in SOA yields from terpene
ozonolysis: A combined experimental and computational study,
Faraday Discuss., 2005, 130, 295–309.
(
presumably by OH radical attack) will continue to reduce the
vapor pressures of the products, or whether fragmentation will
drive the vapor pressures back up as the products head toward
complete oxidation.
1
7 A. L. Robinson, N. M. Donahue, M. Shrivastava, A. M. Sage, E.
A. Weitkamp, A. Greishop, T. E. Lane and S. N. Pandis, Rethink-
ing organic aerosols: Semivolatile emissions and photochemical
aging, Science, 2007, 315, 1259–1263.
1
8 S. Pandis, S. Paulson, J. Seinfeld and R. Flagan, Aerosol forma-
tion in the photooxidation of isoprene and b-pinene, Atmos.
Environ., 1991, 26, 997.
Acknowledgements
This work was supported by grants RD-83108101 from the US
EPA and ATM-0446495 from NSF. Although the research
described in this manuscript has been funded in part by the
United States Environmental Protection Agency, it has not
been subjected to the Agency’s required peer and policy review
and therefore does not necessarily reflect the views of the
Agency and no official endorsement should be inferred.
19 R. Griffin, D. Cocker, R. Flagan and J. Seinfeld, Organic aerosol
formation from the oxidation of biogenic hydrocarbons, J. Geo-
phys. Res., 1999, 104, 3555.
2
0 J. H. Kroll, N. L. Ng, S. M. Murphy, R. C. Flagan and J. H.
Seinfeld, Secondary organic aerosol formation from isoprene
photooxidation, Environ. Sci. Technol., 2006, 40, 1869–1877.
1 M. Claeys, B. Graham, G. Vas, W. Wang, R. Vermeylan, V.
Pashynska, J. Cafmeyer, P. Guyon, M. Andreae, P. Artaxo and
W. Maenhaut, Formation of secondary organic aerosols through
photooxidation of isoprene, Science, 2004, 303, 1173.
2
2
2 P. Ziemann, Evidence for low-volatility diacyl peroxides as a
nucleating agent and major component of aerosol formed from
References
3
reactions of O with cyclohexene and homologous compounds,
1
2
3
4
5
6
7
8
9
D. Grosjean, E. L. Williams and J. H. Seinfeld, Atmospheric
oxidation of selected terpenes and related carbonyls: Gas-phase
carbonyl products, Environ. Sci. Technol., 1992, 26, 1526–1533.
P. J. Adams and J. H. Seinfeld, Disproportionate impact of
particulate emissions on global cloud condensation nuclei concen-
trations, Geophys. Res. Lett., 2002, 30, 1239.
Intergovernmental Panel on Climate Change, Climate Change: The
Physical Science Basis, IPCC Secretariat, 7 bis, Avenue de la Paix,
C. P. N1 2300, 1211 Geneva 2, Switzerland, 2007.
R. J. Charlson, J. E. Lovelock, M. O. Andreae and S. G. Warren,
Ocean phytoplankton, atmospheric sulfur, cloud albedo, and
climate, Nature, 1987, 326, 655–661.
C. I. Davidson, R. F. Phalen and P. A. Solomon, Airborne
particulate matter and human health: A review, Aerosol Sci.
Technol., 2005, 39, 737–749.
L. Clancey, P. Goodman, H. Sinclain and D. W. Dockery, Effects
of air pollution control on death rates in Dublin, Ireland: An
intervention study, Lancet, 2002, 360, 1210–1214.
G. Oberdorster, Z. Sharp, V. Atudorei, A. Elder, R. Gelein, W.
Kreyling and C. Cox, Translocation of inhaled ultrafine particles
to the brain, Inhalation Toxicol., 2004, 16, 437–445.
A. E. Wittig, N. Anderson, A. Y. Khlystov, P. S. N., C. Davidson
and A. L. Robinson, Pittsburgh Air Quality Study overview,
Atmos. Environ., 2004, 38, 3107–3125.
P. Formenti, W. Elbert, W. Maenhaut, J. Haywood, S. Osborne
and M. O. Andreae, Inorganic and carbonaceous aerosols during
the Southern African Regional Science Initiative (SAFARI 2000)
experiment: Chemical characteristics, physical properties and emis-
sion data for smoke from African biomass burning, J. Geophys.
Res. A, 2003, 108, 8488.
J. Phys. Chem. A, 2002, 106, 4390.
23 N. L. Ng, J. H. Kroll, M. D. Keywood, R. Bahreini, V. Var-
utbangkul, R. C. Flagan, J. H. Seinfeld, A. Lee and A. H. Gold-
stein, Contribution of first- versus second-generation products to
secondary organic aerosols formed in the oxidation of biogenic
hydrocarbons, Environ. Sci. Technol., 2006, 40, 2283–2297.
24 S. H. Chung and J. H. Seinfeld, Global distribution and climate
forcing of carbonaceous aerosols, J. Geophys. Res., 2002,
107, 4407.
25 Q. Zhang, D. R. Worsnop, M. R. Canagaratna and J.-L. Jimenez,
Hydrocarbon-like and oxygenated organic aerosols in Pittsburgh:
insights into sources and processes of organic aerosols, Atmos.
Chem. Phys., 2005, 5, 3289.
26 R. Volkamer, J. L. Jimenez, F. San Martini, K. Dzepina, Q.
Zhang, D. Salcedo, L. T. Molina, D. R. Worsnop and M. J.
Molina, Secondary organic aerosol formation from anthropogenic
air pollution: Rapid and higher than expected, Geophys. Res. Lett.,
2006, 33, L17811.
27 M. Kalberer, D. Paulsen, M. Sax, M. Steinbacher, J. Dommen, A.
S. H. Prevot, R. Fisseha, E. Weingartner, V. Frankevic, R. Zenobi
and U. Baltensperger, Identification of polymers as major compo-
nents of atmospheric organic aerosols, Science, 2004, 303, 1659.
28 A. A. Presto and N. M. Donahue, Investigation of a-pinene +
ozone secondary organic aerosol formation at low total aerosol
mass, Environ. Sci. Technol., 2006, 40, 3536–3543.
29 J. Zhang, K. E. Huff Hartz, R. K. Pathak, S. N. Pandis and N. M.
Donahue, Secondary organic aerosol formation from limonene
ozonolysis: Homogeneous and heterogeneous influences as a func-
x
tion of NO , J. Phys. Chem. A, 2006, 110, 11053–11063.
30 R. K. Pathak, C. O. Stanier, N. M. Donahue and S. N. Pandis,
Ozonolysis of a-pinene at atmospherically relevant concentrations:
Temperature dependence of aerosol mass fractions (yields),
J. Geophys. Res., 2007, 112, D03201, DOI: 10.1029/2006JD007436.
1
0 Y. Rudich, N. M. Donahue and T. F. Mentel, Aging of organic
aerosol: bridging the gap between laboratory and field studies,
Annu. Rev. Phys. Chem., 2007, 58, 321–352.
This journal is ꢀc the Owner Societies 2007
Phys. Chem. Chem. Phys., 2007, 9, 2991–2998 | 2997

View Article Online
3
1 J. Zhang and N. M. Donahue, Constraining the mechanism and
kinetics of OH + NO and HO + NO using the multiple-well
38 B. Chuong, J. Zhang and N. M. Donahue, Cycloalkene ozonolysis:
Collisionally mediated mechanistic branching, J. Am. Chem. Soc.,
2004, 126, 12363–12373.
2
2
master equation, J. Phys. Chem. A, 2006, 110, 6898–6911.
2 V. V. Veselovsky, A. S. Gybin, A. V. Lozanova, M. M. Moiseen-
kov, W. A. Smit and R. Caple, Dramatic acceleration of the
Diels–Alder reaction by adsorption on chromatography adsor-
bents, Tetrahedron Lett., 1988, 29, 175–178.
3 A. A. Presto, K. E. Huff Hartz and N. M. Donahue, Secondary
organic aerosol production from terpene ozonolysis: 1. Effect of
UV radiation, Environ. Sci. Technol., 2005, 39, 7036–7045.
4 A. A. Presto, K. E. Huff Hartz and N. M. Donahue, Secondary
organic aerosol production from terpene ozonolysis: 2. Effect
3
39 R. Atkinson and J. Arey, Atmospheric chemistry of biogenic
organic compounds, Acc. Chem. Res., 1998, 31, 574–583.
40 R. Kamens, M. Jang, C. Chien and K. Leach, Aerosol formation
from the reaction of a-pinene and ozone using a gas-phase kinetics-
aerosol partitioning model, Environ. Sci. Technol., 1999, 1430.
41 S. Leungsakul, M. Jaoui and R. M. Kamens, Kinetic mechanism for
predicting secondary organic aerosol formation from the reaction of
d-limonene with ozone, Environ. Sci. Technol., 2005, 39, 9583.
42 J. D. Fenske, K. Kuwata, K. Houk and S. E. Paulson, OH radical
yields from the ozone reaction with cycloalkenes, J. Phys. Chem. A,
2000, 7246.
3
3
of NO
046–7054.
x
concentration, Environ. Sci. Technol., 2005, 36,
7
3
3
5 C. O. Stanier, R. K. Pathak and S. N. Pandis, Measurements of the
volatility of aerosols from a-pinene ozonolysis, Environ. Sci.
Technol., 2007, in press.
6 M. D. Keywood, J. H. Kroll, V. Varutbangkul, R. Bahreini, R. C.
Flagan and J. H. Seinfeld, Secondary organic aerosol formation
from cyclohexene ozonolysis: Effect of oh scavenger and the role of
radical chemistry, Environ. Sci. Technol., 2004, 38, 3343.
7 R. Atkinson, Gas phase tropospheric chemistry of organic com-
pounds, J. Phys. Chem. Ref. Data, 1994, Monograph 2, 216 pp.
43 A. A. Presto and N. M. Donahue, Ozonolysis fragment quenching
by nitrate formation: The pressure dependence of prompt OH
radical formation, J. Phys. Chem. A, 2004, 108, 9096–9104.
44 R. Atkinson, Gas phase tropospheric chemistry of organic com-
pounds, J. Phys. Chem. Ref. Data, 1997, 26, 215–290.
45 S. E. Paulson, M. Y. Chung and A. S. Hassan, OH radical
formation from the gas-phase reaction of ozone with terminal
alkenes and the relationship between structure and mechanism,
J. Phys. Chem. A, 1999, 103, 8125.
3
2
998 | Phys. Chem. Chem. Phys., 2007, 9, 2991–2998
This journal is ꢀc the Owner Societies 2007