The gas-phase ozonolysis of α-humulene

Article abstract of DOI:10.1039/c0cp02379e

α-Humulene contains three double bonds (DB), and after ozonolysis of the first DB the first-generation products are still reactive towards O ₃ and produce second- and third-generation products. The primary aim of this study consisted of identifying the products of the three generations, focusing on the carboxylic acids, which are known to have a high aerosol formation potential. The experiments were performed in a 570 litre spherical glass reactor at 295 K and 730 Torr. Initial mixing ratios were 260-2090 ppb for O₃ and 250-600 ppb for α-humulene in synthetic air. Reactants and gas-phase products were measured by in situ FTIR spectroscopy. Particulate products were sampled on Teflon filters, extracted with methanol and analyzed by LC-MS/MS-TOF. Using cyclohexane (10-100 ppm) as an OH-radical scavenger and by monitoring the yield of cyclohexanone by PTR-MS, an OH-yield of (10.5 ± 0.7)% was determined for the ozonolysis of the first DB, and (12.9 ± 0.7)% of the first-generation products. The rate constant of the reaction of O₃ with α-humulene is known as k₀ = 1.17 × 10^-14 cm³ molecule^-1 s^-1 [Y. Shu and R. Atkinson, Int. J. Chem. Kinet., 1994, 26, 1193-1205]. The reaction rate constants of O₃ with the first-generation products and the second-generation products were, respectively, determined as k₁ = (3.6 ± 0.9) × 10^-16 and k₂ = (3.0 ± 0.7) × 10^-17 cm³ molecule^-1 s ^-1 by Facsimile-simulation of the observed ozone decay by FTIR. A total of 37 compounds in the aerosol phase and 5 products in the gas phase were tentatively identified: 25 compounds of the first-generation products contained C13-C15 species, 9 compounds of the second-generation products contained C8-C11 species, whereas 8 compounds of the third-generation products contained C4-C6 species. The products of all three generations consisted of a variety of dicarboxylic-, hydroxy-oxocarboxylic- and oxo-carboxylic acids. The formation mechanisms of some of the products are discussed. The residual FTIR spectra indicate the formation of secondary ozonides (SOZ) in the gas phase, which are formed by the intramolecular reaction of the Criegee moiety with the carbonyl endgroup. These SOZ revealed to be stable over several hours and its formation was shown not to be affected by the addition of Criegee-radical scavengers such as HCOOH or H₂O. This suggests that in the ozonolysis of α-humulene at atmospheric pressures the POZ will decompose rapidly, and that a large fraction of the formed exited Criegee Intermediate will be stabilized to form stable SOZ, while the formation of OH-radicals via the hydroperoxide channel will be a minor process. the Owner Societies.

Full text of DOI:10.1039/c0cp02379e

View Online / Journal Homepage / Table of Contents for this issue
Dynamic Article Links
PCCP
Cite this: Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
www.rsc.org/pccp
PAPER
The gas-phase ozonolysis of a-humulenew
a
M. Beck,z R. Winterhalter,ab F. Herrmanna and G. K. Moortgat*a
Received 3rd November 2010, Accepted 3rd March 2011
DOI: 10.1039/c0cp02379e
a-Humulene contains three double bonds (DB), and after ozonolysis of the first DB the
first-generation products are still reactive towards O3 and produce second- and third-generation
products. The primary aim of this study consisted of identifying the products of the three
generations, focusing on the carboxylic acids, which are known to have a high aerosol formation
potential. The experiments were performed in a 570 litre spherical glass reactor at 295 K and
730 Torr. Initial mixing ratios were 260–2090 ppb for O3 and 250–600 ppb for a-humulene in
synthetic air. Reactants and gas-phase products were measured by in situ FTIR spectroscopy.
Particulate products were sampled on Teflon filters, extracted with methanol and analyzed by
LC-MS/MS-TOF. Using cyclohexane (10–100 ppm) as an OH-radical scavenger and by monitoring
the yield of cyclohexanone by PTR-MS, an OH-yield of (10.5 ꢀ 0.7)% was determined for the
ozonolysis of the first DB, and (12.9 ꢀ 0.7)% of the first-generation products. The rate constant
of the reaction of O3 with a-humulene is known as k0 = 1.17 ꢁ 10ꢂ14 cm3 moleculeꢂ1 sꢂ1
[Y. Shu and R. Atkinson, Int. J. Chem. Kinet., 1994, 26, 1193–1205]. The reaction rate constants
of O3 with the first-generation products and the second-generation products were, respectively,
determined as k1 = (3.6 ꢀ 0.9) ꢁ 10ꢂ16 and k2 = (3.0 ꢀ 0.7) ꢁ 10ꢂ17 cm3 moleculeꢂ1 sꢂ1 by
Facsimile-simulation of the observed ozone decay by FTIR. A total of 37 compounds in the
aerosol phase and 5 products in the gas phase were tentatively identified: 25 compounds of the
first-generation products contained C13–C15 species, 9 compounds of the second-generation
products contained C8–C11 species, whereas 8 compounds of the third-generation products
contained C4–C6 species. The products of all three generations consisted of a variety
of dicarboxylic-, hydroxy-oxocarboxylic- and oxo-carboxylic acids. The formation mechanisms
of some of the products are discussed. The residual FTIR spectra indicate the formation
of secondary ozonides (SOZ) in the gas phase, which are formed by the intramolecular reaction
of the Criegee moiety with the carbonyl endgroup. These SOZ revealed to be stable over several
hours and its formation was shown not to be affected by the addition of Criegee-radical scavengers
such as HCOOH or H2O. This suggests that in the ozonolysis of a-humulene at atmospheric
pressures the POZ will decompose rapidly, and that a large fraction of the formed exited Criegee
Intermediate will be stabilized to form stable SOZ, while the formation of OH-radicals via the
hydroperoxide channel will be a minor process.
which play an important role in atmospheric chemistry. The
1. Introduction
emissions of BVOCs largely exceed those of anthropogenic
origin.2 Most field and laboratory studies aimed at investigating
the chemistry of BVOCs have been focussed on isoprene and
monoterpenes (C10H16).2–4 On the other hand sesquiterpenes
(C15H24), especially b-caryophyllene and a-humulene, play a
very important role through their high reactivity in the atmo-
sphere and their much higher aerosol formation potential.5
The estimated global BVOC emission rates are very uncertain.
Kanakidou et al.6 reported that the emission of sesquiterpenes
and other terpenes can be up to 5 times higher than estimated.
Bonn et al.7 measured sesquiterpene mixing ratios of more
than 100 ppt in ambient air in the summer and occasionally
less than 1 ppt sesquiterpene in the winter. Also Kim et al.8
The photooxidation of biogenic volatile organic compounds
(BVOC) are primary sources of secondary organic aerosols (SOA),
a Max Planck Institute for Chemistry, Atmospheric Chemistry
Department, P.O. Box 3060, D-55020 Mainz, Germany.
E-mail: geert.moortgat@mpic.de
b Bavarian Health and Food Safety Authority, Pfarrstrasse 3,
80538 Munchen, Germany
¨
w This article was submitted as part of a collection following the 21st
International Symposium on Gas Kinetics, held in Leuven in July
2010.
z Current address: Johannes Gutenberg-University, Institute for
Inorganic and Analytical Chemistry, Duesbergweg 10–14, D-55128 Mainz,
Germany.
c
10970 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
used proton transfer reaction mass spectrometry (PTR-MS) to
measure total sesquiterpene mixing ratios of 20 ppt in the
summer in a deciduous forest. Some studies were performed to
take into account the contribution of the type of plants,
chemical speciation, light and the temperature dependence
of the emissions.9 The sesquiterpenes emission of a variety of
pine tree species was investigated by Helmig et al.10 They
found that these can be up to 29% of the monoterpene
emission, depending on species and temperature. In their study
a-humulene was one of the ten most frequently emitted
sesquiterpenes. Depending on the species of pine tree its fraction
on the sesquiterpene emission can be higher than 10%.
Besides pine trees10 a large variety of plants emit sesquiterpenes
and in particular a-humulene.11 Different plants such as
thale cress,12 orange orchards,13 marsh elders,14 tobacco15
and sunflowers16 are proven to be a-humulene emitters. In
hop, Humulus lupulus, a-humulene is a main constituent of the
essential oil, and depending on the hop species it can consist
up to 40% of the oil.17
with 3.1 groups per average product molecule. They also
determined 1.3 alcohol and 0.31 carboxylic acid groups per
average product molecule.
a-Humulene contains three double bonds. After oxidation
of the first double bond the first-generation products are still
reactive. All the above-mentioned studies were focussed on
the oxidation of one double bond. The further oxidation of the
first-generation products has not been studied so far. Also the
oxidation of the second-generation products, which still con-
tain one double bond, has not previously been investigated.
In this study the ozonolysis of a-humulene with various
ozone–a-humulene ratios was examined in a glass reactor.
The primary aim consisted of analysing the products of the
ozonolysis of a-humulene. Hereby the focus was laid on the
carboxylic acids, which are supposed to have a high aerosol
formation potential. Products of all three generations have
been tentatively identified. Furthermore the reaction rate
coefficients of O3 with the first- and second-generation products
have been determined. In addition, the OH yield of the
ozonolysis of the first- and second double bond was measured.
Finally the SOA yield was determined.
The reaction rate coefficient of a-humulene with ozone is very
high (1.17 ꢁ 10ꢂ14 cm3 moleculeꢂ1
s
ꢂ1) as compared to most
monoterpenes such as a-pinene (8.66 ꢁ 10ꢂ17 cm3 moleculeꢂ1 sꢂ1
)
and its lifetime under atmospheric conditions is estimated
as 1 to 2 minutes.1 Shu and Atkinson1 also measured the
rate constants for the reaction with OH-radicals (2.9 ꢁ
10ꢂ10 cm3 moleculeꢂ1
s
ꢂ1) and with NO3 radicals (3.5 ꢁ
10ꢂ14 cm3 moleculeꢂ1 ꢂ1).
s
Griffin et al.5 showed that the aerosol formation potential of
a-humulene photooxidation products is much higher than that
of monoterpenes. Huff Hartz et al.18 performed a study of
mono- and sesquiterpene ozonolysis in order to determine the
CCN-activity of the products. They reported that products of
sesquiterpenes ozonolysis are less CCN-active than those of
monoterpenes ozonolysis. Moreover the results of a-humulene
photooxidation studies performed in chamber experiments19,20
confirmed the important role of a-humulene in the formation
of secondary organic aerosol (SOA). Griffin et al.5 reported a
SOA yield of 32 to 85%, whereas Lee et al.20 measured a SOA
yield of 65 ꢀ 1% with 53% relative humidity. Jaoui and
Kamens19 measured particle size distribution and number
concentration of the formed organic aerosols. These authors
tentatively identified about 15 products besides some small
molecules such as acetone and formaldehyde. The identified
products were mostly formed by the oxidation of only one
double bond (first-generation products). They used GC-MS
for the identification of the major carbonyl products, whereas
small products were identified by HPLC-MS using DNPH for
derivatisation. Lee et al.20 used PTR-MS to study the gas-phase
products. Their results supported some products identified by
Jaoui and Kamens.19 Lee et al.20 observed acetaldehyde,
formic acid, acetone and acetic acid but not formaldehyde,
which was identified by Jaoui and Kamens.19
2. Experimental methods
An evacuable 570 l spherical glass reactor was used for
the experimental investigations. The reactor is coupled
with different analytical instruments and other devices. A
FTIR-spectrometer is permanently attached (described in
detail later), as well as a gas mixing device and an ozone
generating system. SMPS (scanning mobility particle sizer)
and PTR-MS (proton-transfer-reaction-mass spectrometer)
are also coupled to the reactor. Various pumps enable us to
reduce the pressure in the reactor to 6 ꢁ 10ꢂ6 Torr. The
temperature was kept at 297 ꢀ 2 K by the air conditioning of
the room, whereas the pressure of the purified air in the reactor
was maintained between 720 and 750 Torr. In some cases
water vapour was added by passing air through a bubbler
filled with 18 MO water (Elgastat) and quantified with a
humidity sensor (Panametrics). In some experiments cyclo-
hexane was added as an OH-radical scavenger. Formic and/or
acetic acid were occasionally added as scavengers of the
stabilized Criegee Intermediate (CI). These scavengers were
used to elucidate the mechanism of the ozonolysis. The
pressurised mixture of a-humulene in N2, as well as the used
scavengers, was flushed via a transfer cylinder of 1.38 l into the
reactor by the carrier gas N2. Because of the low vapour
pressure of a-humulene it was necessary to repeat the flushing
several times in order to introduce enough substance in the
reactor. Ozone was generated externally by leading pure
oxygen along an Hg Pen-Ray lamp. The mixing ratios of
ozone, a-humulene and scavenger (except water, see above)
were monitored by FTIR-spectroscopy. Three Teflon stirrers
Dekermenjian et al.21 performed the ozonolysis of a-humulene
to examine the chemical composition of the formed aerosol in
a Teflon chamber using an excess of a-humulene and cyclo-
hexane as an OH scavenger. They used a low pressure
impactor (LPI) and FTIR spectroscopy. Particles were
observed in all fractions between 0.05 to 1 mm. Ketones and
aldehydes were found to be the dominant functional groups,
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10971

View Online
the ambient air flow of 15 ml minꢂ1 was drawn from a faster
Table 1 Number of experiments at various selected conditions in the
ozonolysis of a-humulene
stream (300 ml minꢂ1). Only compounds with a greater proton
affinity than water can undergo efficient proton-transfer reac-
tions with the H3O+ ions to produce protonated product ions
in the drift tube. This means that a large section of OVOCs,
aromatics and alkenes was protonated, while compounds with a
lower proton affinity would not be protonated, especially such
main constituents of the air, namely N2, O2 and Argon.
A quadrupole mass filter differentiated the ions based on
their mass to charge ratio. An electron multiplier detected
the selected ions. The concentration of a compound can be
calculated by the known kinetics of the ion (H3O+)-molecule
reaction or by independent calibration.
Ratio ozone/a-humulene
Added scavenger
1–1.5
1.5–2.5
2.5 and higher
None
Acetic acid
Formic acid
Cyclohexane
Cyclohexane + formic acid
Formic acid + acetic acid
Water
4
1
2
3
2
1
3
3
2
2
3
2
2
1
4
4
2
2
3
1
1
1
4
5
Water + cyclohexane
in the reactor provided a rapid mixing of the reactants. A more
detailed description of the experimental setup was given by the
literature.22,23
In this study only calibrations for cyclohexanone were
performed in order to determine the OH-yield of the reaction
(details will be given later). Several of these calibrations were
performed during the experiments. The calibrations showed
that the detector response was linear within the measured
concentration range.
Initial mixing ratios of the reactants were varied between
250 and 600 ppb (1 ppb E 2.3 ꢁ 1013 molecule cmꢂ3 at ambient
temperature and pressure) for a-humulene and between 260
and 2090 ppb for ozone. The number of experiments at the
various selected conditions of added scavengers is summarized
in Table 1. The mixing ratios of the CI-scavengers formic and
acetic acid were about 3 ppm, whereas those of H2O varied
between 6000 and 15 000 ppm (equivalent to 20% to 55% relative
humidity at the above conditions). Generally cyclohexane was
added at a mixing ratio of about 300 ppm, sufficient to scavenge
more than 97% of the formed OH-radicals.
The ion source produces besides the needed H3O+ also
other unwanted ions such as O2 and NO+. Under these
+
conditions cyclohexane, which is used in relatively high mixing
ratios, is oxidised. Thus there is a permanent background at
the mass of the protonated cyclohexanone. This background is
very stable over time, and could easily be subtracted from our
measurements. A poor detection limit for cyclohexane (5 ppb)
is the result of the difference between this background signal
and that obtained by sampling zero air. Because of the
stability of the background, changes in the cyclohexanone
signal could be readily and quickly registered, even at a level
of the background plus three times its standard deviation
(background: 5 ppb; standard deviation: 0.3 ppb).27
2.1 FTIR spectroscopy
Long-path FTIR spectroscopy (Bruker IFS 28, 0.5 cmꢂ1
resolution, 43.4 m path-length) was used to determine the
mixing ratios of reactants and products. This spectrometer
contained two detectors: a HgCdTe detector used for the
spectral range 700–2000 cmꢂ1 and an InSb detector for the
range 2000–4000 cmꢂ1. In most experiments only the HgCdTe
detector was used, because the main interesting features were
obtained in its spectral range. For each spectrum 128 scans
were averaged resulting in a time resolution of about 3 min for
the acquisition of the data. For the kinetic experiments only
16 scans were averaged at the very beginning in order to realize
a better time resolution. The calibration of HCOOH was
described by Finkbeiner et al.24 All other compounds have
been calibrated by standard volumetric methods.
2.3 Scanning mobility particle sizer (SMPS)
A scanning mobility particle sizer (SMPS, TSI 3936) monitored
the aerosol concentration and its size distribution. The SMPS
consists of an electrostatic classifier (TSI 3080) with a long
differential mobility analyzer (LDMA; TSI 3081) and an ultra
fine condensation particle counter (CPC; TSI 3025A) as a
detector.
2.4 HPLC-MS
The generated aerosol particles were collected during 30–60 min
on Teflon (PTFE) filters (45 mm diameter, 0.45 mm pore size).
The filters were enclosed in a 7 cm3 glass flask with 4 ml pure
methanol added for extraction. After conditioning for at least
two days, 3 ml of the solution was taken and concentrated to
about 0.75 ml. The cooled sample solutions were concentrated
by removing the methanol with a nitrogen stream. Small
volatile compounds such as acetone could also be removed,
but these compounds were not detectable with the used
method. The exact volume was determined gravimetrically.
A same volume of pure water was added, since pure methanol
solutions cause a fronting problem in the used liquid chromato-
graphy. 100 ml of the solution was directly injected into the
HPLC-MS system.
2.2 Proton-transfer-reaction mass spectrometry (PTR-MS)
A proton transfer reaction mass spectrometer (PTR-MS,
Ionicon Analytik, Austria) was used for online measurements
of volatile organic species. Lindinger et al.25 for example
described in detail the general principle of operation of the
PTR-MS. The specific mass spectrometer used in this study
has been successfully employed in numerous previous field26
and laboratory27 studies. For this study practically the same
setup was used such as described in Herrmann et al.27
A water-fed hollow cathode discharge source produced
high concentrations of H3O+ ions (circa 5 ꢁ 106 ions cmꢂ3).
These are introduced by a stable, regulated flow of ambient
air (15 ml minꢂ1) into an ion/molecule reaction drift tube
maintained at 50 1C and 2.2 mbar pressure. In order to increase
time resolution and to minimise sample line residence time,
The HPLC system consisted of an autosampler (Series 200,
Perkin Elmer, Norwalk, Connecticut, USA), a degasser and
c
10972 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
a quaternary pump (Series 1100, Agilent Technologies,
Waldbronn, Germany). The analytical column was
2.5 Chemicals
a
Sources of the chemicals (stated purities in parentheses) were
as follows: cyclohexane (99.9%), formic acid (98%), acetic
acid (98%), a-humulene (98%) from Sigma-Aldrich, Steinheim,
Germany. All chemicals were of highest purity, commercially
available and used without further purification.
ReproSil-Pur C18-AQ (250 mm ꢁ 2 mm id, 5 mm particle size)
in a PEEK (poly-ether-ether-ketone) cartridge (Dr Maisch
GmbH, Ammerbuch, Germany). The flow rate of the gradient
elution was 400 ml minꢂ1. An eluent solution of 0.1%
(v/v) formic acid in deionised water (eluent A) and aceto-
nitrile (eluent B) was used. The gradient started with 100% A
for 30 s, followed by a gradient from 100% A to 95% B during
20 min; thereafter the gradient turned back to 100% A in
3 min. The electrospray ionization in the negative mode
(ESIꢂ) was used as an ionization technique at 400 1C and an
ionization voltage of ꢂ4 kV. This system was coupled to a
hybrid mass spectrometer (LC-Triple Quad-MS/MS-TOF
QSTAR, Applied Biosystems MDS-SCIEX, Toronto, Canada).
This instrument combines tandem mass spectrometry
(MS/MS) with the high mass resolution of a time-of-flight
detector (TOF). The tandem mass spectrometry (MS/MS)
selects ions with different mass-to-charge ratios and through
collision with neutral gas molecules (here N2) the ions
dissociate in various fragments (also called CID, Collision-
Induced Dissociation). The MS-system and its software can
display the Total Ion Chromatogram (TIC) or a chromato-
gram of selected mass-to-charge ratios (XIC, eXtracted Ion
Chromatogram).
3. Results and discussion
The main focus of this study was laid on the product identifi-
cation and the determination of the reaction mechanism
of the ozonolysis of a-humulene. As mentioned above
FTIR-spectroscopy, PTR-MS and HPLC-MS were used for
the detection of various products. Moreover the OH-radical
yield was determined by PTR-MS. The use of a scanning
mobility particle sizer (SMPS) allowed the quantification of
the aerosol yield under various conditions.
Since a-humulene contains three endocyclic double bonds,
the products can be classified into three generations. The
ozonolysis of the first double bond (DB) leads to first-generation
products, whereas the consecutive ozonolysis of the remaining
DB of the first-generation products leads to second-generation
products and so on. To produce a specific generation of
products, different ozone/a-humulene-ratios were used.
Further mechanistic insight can be obtained by using
OH-radical- and CI-scavengers. Formic acid, acetic acid and
water represented the CI-scavengers, whereas cyclohexane the
OH-scavenger. A total of 58 experiments were conducted, as
summarized in Table 1.
The combination of the used gradient elution and the
ionisation method has been optimised for the separation of
carboxylic acids.28 Thus organic compounds without any
carboxylic group are nearly undetectable in the present concen-
trations used in this study.
Most products of the particle phase were only tentatively
identified because the needed standards were not available.
Due to the high resolution of the used mass spectrometer,
mostly one elemental formula was obtained, which was
conform to the ozone/a-humulene reaction. The functional
groups could be identified by CID: carboxylic acids usually
show the loss of CO2 whereas aldehydes the loss of CO. For
example the loss of CO2 and water indicates the presence
of a dicarboxylic acid or an oxo-carboxylic acid. With the
knowledge of the elemental formula and the CID fragmenta-
tion pattern it was thus possible to identify the type of the
compound observed. The resulting structures of the observed
compounds were further deduced from the postulated reaction
pathways of the a-humulene ozonolysis described later in
detail. In addition the earlier described experimental setup of
the dicarboxylic acids showed also the formation of sodium-
adducts. This has been observed in other studies.29 The
sodium-adducts helped to differentiate between dicarboxylic
acids and hydroxy-oxocarboxylic acids with the same elemental
formula.
3.1 Reaction kinetics of the a-humulene + ozone reaction
Two of the three DB’s of a-humulene (Scheme 1) have three
alkyl substituents (DB 1 and 2) and are expected to have
similar reactivities, whereas DB 3 with only two alkyl substi-
tuents has probably a different reactivity. By analogy to the
reaction of ozone with small alkenes, the reaction rate constant
of ozone with trans-2-butene (analogous to DB 3) is 1.9 ꢁ
10ꢂ16 cm3 moleculeꢂ1 ꢂ1, whereas the reaction rate constant
s
of ozone with 2-methyl-2-butene (analogous of DB 1 and DB 2)
is 4.0 ꢁ 10ꢂ16 cm3 moleculeꢂ1 sꢂ1 30
.
Thus it can be assumed
that DB 3 is the slowest reacting double bond. But the
difference of the reaction rate constants is too small to establish
which double bond will be preferentially attacked by ozone.
To determine the rate constants the following assumptions
were made:
Here k0 is the rate constant of the reaction of O3 with the
first double bond (DB 1) of a-humulene. This reaction leads to
various products of the first generation. These first-generation
products can further react with ozone and k1 is the averaged
Scheme 1
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10973

View Online
The rate constants derived for the second and third double
bonds are k1 = (3.6 ꢀ 0.9) ꢁ 10ꢂ16 and k2 = (3.0 ꢀ 0.7) ꢁ
10ꢂ17 cm3 moleculeꢂ1 ꢂ1, respectively. Under the conditions
s
of excess O3 over a-humulene, the initial attack reaction of
O3 with a-humulene with the very fast rate constant
(1.17 ꢁ 10ꢂ14 cm3 moleculeꢂ1 sꢂ1) will cause an instantaneous
drop of the O3 concentration, producing first-generation
products. The rate of removal of these first-generation
products is governed by the magnitude of k1, and can be
observed by the curvature during the first 1000 s. The optimal
simulations of the three data sets were obtained with
an uncertainty of 25%. The slope of the tail of the curve
(t 4 1000 s) is controlled by the reaction rate constant k2. The
overall uncertainty of k2 is also estimated to be 25%. This was
based on sensitivity runs where the magnitude of both k1 and
k2 was varied independently by 25 and 50%.
Fig. 1 Consumed ozone mixing ratio during the reaction of O3 with
a-humulene (aH), measured by FTIR; initial concentrations: Exp. 1
[O3]1 = 2181 ppb; [aH]3 = 480 ppb; Exp. 2 [O3]2 = 1085 ppb; [aH]2 =
255 ppb; Exp. 3 [O3]3 = 1656 ppb; [aH]1 = 325 ppb. For the
The magnitude of the k1
=
(3.6
ꢀ
0.9)
ꢁ
10ꢂ16 cm3 moleculeꢂ1 sꢂ1 reflects the reaction rate constant
of ozone with alkenes with increasing substitution around
the –CHQCz double bond (such as 2-methyl-2-butene
4.0 ꢁ 10ꢂ16, cis-3-methyl-2-pentene 4.5 ꢁ 10ꢂ16, trans-3-methyl-
2-pentene 5.6 ꢁ 10ꢂ16, cis + trans-3,4-dimethyl-3-hexene 4.5 ꢁ
simulations by Facsimile (full lines) the values k0 = 1.17 ꢁ 10ꢂ14
,
k1 = 3.6 ꢁ 10ꢂ16 and k2 = 3 ꢁ 10ꢂ17 cm3 moleculeꢂ1 sꢂ1 were used.
10ꢂ16, all in units cm3 moleculeꢂ1 ꢂ1). The magnitude of
s
rate constant of all first-generation products. Analogously,
this can be applied for the second-generation products with a
rate constant k2. The formed OH-radicals may also attack
a-humulene but this influence was neglected, since the OH-rate
(= kOH ꢁ [humulene]) is much smaller than the O3-rate
(= kO3 ꢁ [humulene]). As described later in detail the OH yield
of the ozonolysis is about 10.5% for a-humulene and about
12.9% for the first-generation products. The ozone concentration
was always much higher than the OH concentration.
k2 = (3.0 ꢀ 0.7) ꢁ 10ꢂ17 cm3 moleculeꢂ1 sꢂ1 is somewhat larger
than some typical reactions of O3 with substituted linear alkenes
CH2QCH– (such as 3-methyl-1-butene 0.95 ꢁ 10ꢂ17, 4-methyl-
1-pentene 1.0 ꢁ 10ꢂ17, 3,3-dimethyl-1-butene 0.39 ꢁ 10ꢂ17, all in
units cm3 moleculeꢂ1 ꢂ1).30
s
The rate constant k0 is anomalously high, which cannot be
explained by the alkyl substituents at the DB, since at least one
type of DB is still present in the first generation products with
a rather normal (expected) rate constant. A possible reason
for the high rate constant of a-humulene and other cyclic
sesquiterpenes might be the release of ring strain upon ozone
attack on the DB.
The consumed ozone mixing ratios were plotted against
reaction time for three experiments with different initial reactant
mixing ratios (see Fig. 1). The amount of ozone was detected
by FTIR. At the beginning of the reaction 16 scans were
averaged for the generation of one spectrum, which resulted
in a time resolution of less than 30 s. Later it was changed
back to 128 scans per spectrum. In Table 2 the results of the
three experiments and their initial mixing ratios are shown.
The first constant k0 was reported by Shu and Atkinson1 as
At typical atmospheric ozone mixing ratios of about 30 ppb
a-humulene has a lifetime of 2 min, whereas the averaged
first-generation products have a lifetime of about 1 h and the
second-generation products of about 12.5 h.
3.2 SOA-yield in the ozonolysis of a-humulene
1.17 ꢁ 10ꢂ14 cm3 moleculeꢂ1
s
ꢂ1. The program Facsimile31
was used to determine the rate constants of ozone with the
first- and second-generation products using the known k0 and
varying k1 and k2 to obtain the best fit of the simulated ozone
decay with the experimentally determined ozone time profile.
The rate constants k1 and k2 were optimized so that the
simulations of the three experimental ozone profiles gave the
best fit with identical values for k1 and k2. These simultaneous
simulations of the three data set are shown in Fig. 1.
Two groups of experiments were performed in order to
determine the SOA-yield of the ozonolysis of a-humulene.
The earlier described SMPS was used for this issue. No OH- or
CI-scavengers were added and the experiments were performed
under dry conditions. Also no seed aerosol was used. Three
experiments (Exp. 1 to 3) with a-humulene in excess and two
experiments (Exp. 4 and 5) with 4-fold ozone excess were
performed.
Table 2 Determined rate constants (in cm3 moleculeꢂ1 ꢂ1) of a-humulene and initial mixing ratios
s
AU1409
AU1809
AU1909
a-Humulene
Ozone
k0
k1
k2
480 ppb
2182 ppb
a-Humulene
Ozone
k0
k1
k2
255 ppb
1085 ppb
a-Humulene
Ozone
k0
k1
k2
325 ppb
1656 ppb
1.17 ꢁ 10ꢂ14
3.6 ꢁ 10ꢂ16
3.0 ꢁ 10ꢂ17
1.17 ꢁ 10ꢂ14
3.6 ꢁ 10ꢂ16
3.0 ꢁ 10ꢂ17
1.17 ꢁ 10ꢂ14
3.6 ꢁ 10ꢂ16
3.0 ꢁ 10ꢂ17
c
10974 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
During the experiment with excess a-humulene, the aerosol
number concentration reached a maximum within the first
10 min, with a particle diameter from 50 to 100 nm. At longer
reaction times the particles grow in size, but their number
gradually decreased due to coagulation and/or loss by
deposition on the wall. Wall deposition rates depend on
particle size, geometry of the reactor, turbulence in the chamber
and charge distribution.32–34
factors could affect the yield. A part of the increasing yield is
caused by the higher oxygen atom to carbon atom ratio
(for example C15-dicarboxylic acid III has a ratio of 4 to 15,
whereas the ratio for C6-dicarboxylic acid XIII is 4 to 6). This
oxygenation increased the absolute yield. Also the compounds
are more polar and thus less volatile. But on the other hand
carbon–carbon double bonds undergo ozonolysis and are
therefore destroyed. This lowered the molecular weight and
made the products more volatile. Donahue et al.37 also
mentioned the potential difference between gas phase and
condensed phase oxidation subsequent to OH-attack, especially
for cross-linking and oligomerization. Especially oligomeriza-
tion could play an important role for the observed effect of
higher SOA-yield for second- and third-generation products.
Sadezky et al.29 reported that main compounds of SOA
originating from small alkenes–ozone reactions are oligomers.
Pun and Seigneur38 mentioned that the effect of oligomeriza-
tion for the products of a-humulene and ozone is small
because the monomers were already in the particle phase due
to their low vapour pressure. As Kroll and Seinfeld39 pointed
out the volatility of products is very much influenced by the
reactions of the formed peroxy- and oxy-radicals.
The wall loss in the glass reactor has been determined in a
separate a-humulene experiment using excess ozone (typically
100 ppb a-humulene, 600 ppb ozone and 270 ppm cyclohexane)
where the aerosol volume concentration was monitored at the
end of the reaction (t = 25 min) for at least 30 minutes. The
typical loss rate amounted to 5.6 ꢁ 10ꢂ5 sꢂ1 = 0.20 hꢂ1
,
however with an uncertainty of 50%.35,36 In the present study
it was assumed that the wall loss rate was constant during and
between the performed experiments.
SOA yields were calculated from the aerosol mass occurring
at the maximum number concentration, using a density of the
aerosol particles of 1.0 g cmꢂ3. For the experiments using
excess a-humulene, a 5% wall loss correction was applied,
assuming a fast consumption of the sesquiterpene. For the
experiments using excess ozone, wall loss corrections of 15%
were applied, taking into account that the aerosol number
concentration was monitored at about 30 minutes reaction
time, when the sesquiterpene was almost completely converted.
As seen in Table 3 the corrected SOA-yield in the experi-
ments with a-humulene excess varied between 14 and 17%. A
high excess of ozone increased the yield to almost 45%. From
the identification of the products by HPLC-MS (as described
in detail below in Section 3.6), the different SOA yields can be
explained by the degree of ozonolysis of the different double
bonds. The SOA of the experiments with excess a-humulene
should predominantly contain first-generation products. It
could be assumed that these products have a relatively high
molecular weight, mostly possessing two or more functional
groups (see Section 3.6 in addition to an unknown number of
unidentified products). All these products are expected to
have a relatively low volatility and therefore have high SOA
formation potential. On the contrary, the SOA of the experi-
ments with ozone in excess contained mainly products of the
second- and third-generation (second-generation products
have a relatively long life-time, see above). The latter compounds
have a relatively low molecular weight, but because of their
functional groups they also possess a relatively low volatility.
The reason for the higher SOA-yield of the second- and
third-generation cannot readily be explained. Donahue et al.37
presented a study discussing the critical factors for SOA-yield
from terpene ozonolysis. They showed that some competing
In the literature only values for the SOA-yield of a-humulene
photooxidation5,20 are available. They reported SOA-yields
between 32 and 85%. The photooxidation leads to a totally
different chemistry. Therefore a comparison between those
data and this study is very difficult. Winterhalter et al.23
reported under compatible conditions SOA-yields for the
ozonolysis of b-caryophyllene between 6 and 24% using excess
sesquiterpene and no added scavenger. Thus it can be assumed
that the SOA formation potential for the first-generation
products of these two sesquiterpenes is similar.
3.3 Reaction mechanism of the a-humulene + ozone reaction
Ozone addition at DB 1 and DB 2 leads to the two primary
ozonides (POZ 1 and POZ 2), whereas the less reactive DB
yields POZ 3. Each POZ decomposes to two excited Criegee
Intermediates CI 1a* and CI 1b*, CI 2a* and CI 2b*, CI 3a*
and CI 3b*, respectively (with syn and anti conformations a
total of 12 CI). This is shown in Scheme 2.
The excited CI has different reaction possibilities (see Scheme 3):
it can react via the ‘‘hot acid channel’’, also referred as ‘‘ester
channel’’ or ‘‘dioxirane channel’’,40 or via the ‘‘hydroperoxide
channel’’, also referred as ‘‘vinyl hydroperoxide channel’’. The
‘‘O-atom channel’’ is not important in most cases and was not
considered.40 The excited CI can also thermalize through
collision with the carrier gas air, to form a stabilized CI (SCI).
In this paper the Criegee Intermediate (carbonyl O-oxide) is
Table 3 SOA yield of a-humulene with various initial mixing ratios
Reacted a-humulene
in ppb
Ozone/reacted
a-humulene ratio
SOA in mg mꢂ3
uncorrected
SOA in mg mꢂ3
corrected
Experiment
SOA yield in %
Exp. 1
Exp. 2
Exp. 3
Exp. 4
Exp. 5
150
150
150
270
410
1.0
1.0
1.0
4.0
4.0
168
200
179
880
176
210
188
1012
1495
14
17
15
45
44
1300
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10975

View Online
Scheme 2
Scheme 3
represented as a biradical, but some theoretical calculation
have collisionally been stabilized. These authors showed through
suggested a zwitterionic structure (e.g. Cremer et al.).41 Because
of their relatively long lifetime, SCIs can bimolecularly react
with scavenger molecules such as aldehydes, ketones, organic
acids (e.g. formic and acetic acid) and H2O.42
theoretical calculations that the formation of SOZ will be the
dominant process for C15 alkyl-substituted cyclohexenes, which
served as surrogate for sesquiterpenes. The SOZ formation and
the pathways of both the hot acid and the hydroperoxide species
will be described later in detail.
As shown as an example in Scheme 4, the SCI CI 1a can
react with water as CI-scavenger to a dicarbonyl-compound
(DC1) or to an oxocarboxylic acid (IVa, tentatively identified).
Acids (RCOOH) as scavengers lead to the corresponding acid
anhydrides.
3.4 Mechanism of OH-formation and measurement
of OH-yields
Another possibility is the formation of internal secondary
ozonides (SOZ). This intramolecular reaction is known to
occur in the ozonolysis of cycloalkenes and was postulated by
Chuong et al.43 to occur if the excited Criegee Intermediates
Every CI reacting through the hydroperoxide channel yields
an OH-radical. The fraction of molecules, which entered this
channel, depends on the molecule structure and needs to be
experimentally determined for every different alkene reacting
c
10976 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Scheme 4
with ozone. The formation of OH is shown in Scheme 5
Since the results of Berndt et al.45 and Atkinson et al.44 are
starting from a CI of a simple alkene (CIA1). The excited CI
forms a hydroperoxide (HPA1) via transfer of an H atom in
the a-position (1,4-hydrogen shift). This hydroperoxide can
split off an OH-radical with the formation of a new radical
(RA1).40 The hydroperoxide channel and its products will be
discussed later for the particular case of a-humulene.
A common method to determine OH-radical yields is the
use of cyclohexane as an OH-scavenger. The OH-scavenging
effect of cyclohexane is well known and explained in detail in
the literature.44 The reaction of cyclohexane with OH-radicals
mainly leads to cyclohexanol and cyclohexanone,44,45 as seen
in Scheme 6.
very similar and since the latter measured a series of terpenes,
the value of the latter (55%) was used for calculation of
the total (cyclohexanol + cyclohexanone) yield. The ratio of
cyclohexanone/cyclohexanol for the calculation was taken
from Shu and Atkinson.1 Although these values are based
only on the ozonolysis of the first double bond of a molecule,
they were also used for the calculation of the OH-yield of the
first-generation products.
The OH-yield of 22% from ozonolysis of the first DB was
previously measured.1 However, there are no data available
for the OH-yield of the second DB. The used experimental and
calculation methods to determine the yield of first and second
DB were developed by Herrmann et al.27 The initial mixing
ratios of cyclohexane were about 10 ppm in order to scavenge
about 80% of the formed OH-radicals. For the calculation of
the reaction ratios, the OH reaction rate constants of Atkinson47
for cyclohexane and Shu and Atkinson48 for a-humulene were
used.
As described earlier, only cyclohexanone could be measured
by PTR-MS. Therefore the relative yield of cyclohexanone has
to be known. Unfortunately the yield of cyclohexanone depends
on the initial compound undergoing ozonolysis. Atkinson
et al.44 reported a combined (cyclohexanol + cyclohexanone)
yield of 55% for a series of monoterpenes, with a ratio of
cyclohexanone/cyclohexanol between 0.8 and 1.4, depending
on the terpene. For simple alkenes this combined yield is
50%.46 Berndt et al.45 measured for a-pinene a total yield of
60% and for (cyclohexanol + cyclohexanone) with a ratio of
1.0. For sesquiterpenes in general and a-humulene in particular
there are no specific data in the literature about the combined
yield of (cyclohexanone + cyclohexanol). Shu and Atkinson1
reported only a cyclohexanone/cyclohexanol ratio of 2.02 for
the reaction of the first double bond of a-humulene.
The measured cyclohexanone mixing ratios have been corrected
to 100% complete scavenging. Because of the excess of ozone,
the slower second and the third DB were also attacked by
ozone. As seen in Fig. 2 the OH-reaction with cyclohexane
started very fast, causing initially a sudden formation of cyclo-
hexanone. During the further progress of the reaction the
formation of cyclohexanone slowed down. The time was calculated
at which 95% of the first DB of a-humulene had reacted with
ozone (marked in Fig. 2 with the left vertical black line).
Unfortunately, in this relevant time period the PTR-MS was
measuring the background, thus consequently the cyclohexanone
yield was calculated by fitting the cyclohexanone signal before
and after the background measurement. As can be seen in
Fig. 2, even small uncertainties in the reaction time can affect
the calculated yield. The same procedure was done to calculate
the OH-yield of the second DB. It should be mentioned that
this corresponds to the OH-yield of the first-generation products.
By calculating the time at which 95% of the first-generation
products had reacted, the initial ozone mixing ratios were
corrected for the previously reacted ozone. The yield of the
first DB was subtracted from the total measured yield at the
Scheme 5
Scheme 6
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10977

View Online
bond is less than half the yield reported by Shu and Atkinson1
(22%, using a small excess of a-humulene).
It is not obvious to explain the observed difference in OH
yield. It is interesting to note that the sum of the OH yield
of the first and second double bonds obtained in this study,
(10.5 ꢀ 0.7) + (12.9 ꢀ 0.7) = 23.5 ꢀ 0.7%, is in good
agreement with 22% yield obtained by Shu and Atkinson.1
The latter authors used a different combined yield of cyclo-
hexanone and cyclohexanol (50% versus 55% in this study),
but the same cyclohexanone/cyclohexanol ratio for the
OH + cyclohexane reaction and another experimental method
for detecting cyclohexanone. Also, in their experiments the
ozone addition was made in 5 successive steps. As mentioned
above the average reaction rate constant of O3 with the
first-generation products is relatively fast (k1 = 3.6 ꢁ
10ꢂ16 cm3 moleculeꢂ1 ꢂ1), so that it cannot be excluded that
s
Fig. 2 Cyclohexanone yield used for calculation of the OH-yield. The
dotted line marks the yield of the first double bond, the dashed line of
the second double bond, whereas the vertical lines mark the reaction
time when 95% of the respective double bond has reacted.
first-generation products have accumulated in the reactor between
successive ozone additions, so that ozonolysis of these first-
generation products took place, enhancing thus the OH yield.
calculated reaction time.27 It was assumed that the rate con-
stant for the OH/first-generation products-reaction is in the
same range as the OH/a-humulene-reaction, so that the same
amount of OH is scavenged by cyclohexane. Two experiments
of this kind were performed for a-humulene (Table 4).
The OH-yield for the first double bond was 9.8 and 11.2%
(average 10.5 ꢀ 0.7%) with an estimated uncertainty of
25%. The second double bond (average of all first-generation
products) has a measured OH-yield of 12.3 and 13.6%
(average 12.9 ꢀ 0.7%) with an estimated uncertainty of
30%. The measured OH-yield of 10.5% for the first double
3.5 Secondary ozonides
Secondary ozonides originate from the cyclisation of the
stabilized Criegee Intermediates. The carbonyloxide-moiety
of the CI-species reacts intramolecularly with the carbonyl-
end to form an internal SOZ-group. Three different internal
SOZ structures are possible (without considering any stereo-
chemistry, see Scheme 7).
The formation of the internal secondary ozonides was
detected in the gas phase by FTIR-spectroscopy. The FTIR
spectra from experiments with less than a two-fold excess of
ozone showed the characteristic absorptions of SOZ between
1100 and 1200 cmꢂ1 (Fig. 3), which represent the range of the
C–O stretching vibration. It is most probable that all three
SOZ are formed. It is however suggested that the formation of
SOZ 1 and SOZ 2 is favoured, because the three maxima in
the relevant range of the FTIR-spectra are similar to those of the
propene–SOZ49 (having analogue DB 1 and 2). The analogue
2-butene–SOZ to SOZ 3 has only one sharp maximum in this
range.50
Table 4 OH-yield of first (1 DB) and second (2 DB) double bond
OH-yield
1 DB in %
OH-yield
2 DB in %
Literature OH-yield
1 DB in %
Experiment
Exp. 1
Exp. 2
Average
9.8
11.2
10.5 ꢀ 0.7
12.3
13.6
12.9 ꢀ 0.7
22a
22a
As shown in Fig. 3 the influence of the CI-scavengers
HCOOH, acetic acid and water is not significantly affecting
a
From ref. 48.
Scheme 7
c
This journal is the Owner Societies 2011
10978 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001

View Online
Fig. 3 FTIR-spectra of the a-humulene–SOZ under various experi-
mental conditions (direct after ozone addition, similar initial a-humulene
and ozone mixing ratios but different scavengers).
Fig. 4 FTIR-spectra of the formed a-humulene–SOZ directly after
ozone addition and after 90 min, indicating the complete decay of
the SOZ.
For example, the SOZ 1 can react with water forming a
molecule possessing an aldehyde (or a keto) group and a
hydroxy-hydroperoxide group, as shown in Scheme 8. The
latter compound can lose H2O2 and form an aldehyde with an
additional oxo-group.53,54 Further oxidation of the aldehyde
leads to the C15-oxocarboxylic acid IVa. Another possibility
is that the hydroxy-hydroperoxide group loses water and IVa
is directly formed.
the magnitude of the spectra. This is an interesting finding
because the suppression of SOZ-formation in the presence of
CI-scavengers was shown in earlier studies, e.g. for b-pinene.51
Also the internal SOZ-formation of b-caryophyllene is partially
suppressed by water and formic acid.23 The reason for this
non-suppressing behaviour is probably caused by the very fast
intramolecular reaction of the CI-end with the aldehyde-end of
the Criegee Intermediates.
The oxocarboxylic acids IVb can be formed via similar
pathways from the SOZ 2, and IVc and IVd from SOZ 3.
The SCI formed by the ozonolysis of endocyclic alkenes was
examined by Chuong et al.43 for cyclohexene and a series of
alkyl-substituted cyclohexenes. In this case the CI ozonolysis
products are tethered and thus form a single disubstituted
product, as in a-humulene. The authors disclosed that production
of SOZ depends on the degree of collisional stabilization and
the size of the Criegee Intermediates. Their theoretical calcula-
tions predicted that the stabilization of the CI leads to a
dramatic transformation in the dominant oxidation pathway
from an OH-radical-forming process at low carbon number to
a SOZ-forming process at high carbon number. They concluded
that at some carbon number between 8 and 15 (or something
slightly smaller than monoterpenes and sesquiterpenes) SOZ
formation dominated at all pressures. They also expected that
any SCI formed from endocyclic sesquiterpenes would quickly
cyclise to form SOZ, and therefore will not undergo bimolecular
reactions with H2O and other CI scavengers, as was observed in
the case of a-humulene.
Another interesting aspect of the SOZs of a-humulene is
their relative long lifetime. Under wet conditions about 40%
of the SOZs remained after 140 minutes, while under dry con-
ditions about 80% remained after 90 minutes. An excess of
ozone decreased the yield of SOZ. More than a twofold excess of
ozone destroyed all SOZs (Fig. 4). This means that no molecule
with two ozonide-groups is formed or it is too unstable to be
detected. The ozonolysis of first- and second-generation products
did not seem to form SOZs in detectable amounts.
3.6 Tentatively identified products by HPLC-MS and CID
A summary of all tentatively identified oxidation products in
the aerosol samples of a-humulene ozonolysis is given in
Table 5. Some mentioned third-generation products were only
detected in the gas-phase. Table 5 also presents the systematic
IUPAC names (obtained with the software ACD/ChemSketch).
3.6.1 First-generation products. The six possible CIs lead to
a multitude of theoretical products. Besides the three SOZs
and their decomposition products mentioned above, the hot
acid and the hydroperoxide channel can produce many other
products. It should be noted that for most of the identified
products more than one reaction pathway is possible, which
could occasionally be proven by the different effects of the
added scavengers. In total 5 products are tentatively identified,
which are formed via the hot acid channel or also through the
decomposition of the SOZs, while 24 products are obtained
via the hydroperoxide channel. Fig. 5 to 7 show the possible
products originating from the three POZs and a very rough
estimation of the relative yield according to an excess a-humulene
experiment without any scavengers. The relative yield is estimated
assuming a similar sensitivity for the detection of the carboxylic
acids and is indicated by the symbols L (large), M (medium) and
S (small).
Some tentatively identified products can originate from the
SOZs. The decomposition of the SOZs can be initiated by a
hydrolysis or an ozonolysis in the reaction chamber or during
the processing of the samples. Photolysis with UV-light has
also recently been observed.52
Hydroperoxide channel. The reaction pathway via the hydro-
peroxide channel leads to more theoretically and actually
tentatively identified products than the hot acid channel. The
hydroperoxide channel is the source of the OH-radicals
that are formed during the ozonolysis of a-humulene. In the
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10979

View Online
Scheme 8
present study the yield of OH-radicals was determined as
addition of O2 leads via a C15-peroxy radical to a C15-oxy
10.5% for the first double bond and 12.9% for the second
double bond. The products IIa, IIb, IIIa–IIIh and Va–VIIb
(Table 5) are (most probably) formed through the hydroperoxide
channel. The formation of the OH-radicals is shown below in
Scheme 9 for CI 1a and CI 1b (with the same analogy for the
other CIs). The CI forms an unsaturated hydroperoxide via a
1,4-hydrogen shift, which loses an OH-radical. New radical-
intermediate species are thus formed from the different CIs: CI
1a H1 from CI 1a, and CI 1b H1 and CI 1b H2 from CI 1b.
The C15-oxodicarboxylic acid IIa originates from the
C15-radical CI 1b H2 (see Scheme 10) and its analogue IIb
was obtained from the corresponding CI 2b. An O2 molecule
adds to the radical CI 1b H2 to form a peroxy radical. Two
peroxy radicals can undergo a disproportionation reaction to
form a molecule with an aldehyde group (or keto group,
respectively) and one with an alcohol group. One of the newly
formed molecules contains two aldehyde groups which can be
oxidised to two acid groups. The formation pathway of product
IIa is shown in Scheme 10. Analogous reactions occur for IIb.
Theoretically the third C15-oxodicarboxylic acid IIc should
be also formed. But the extracted ion chromatograms XIC of
m/z 281 (M ꢂ Hꢂ) and m/z 303 (M + Na ꢂ 2Hꢂ) only showed
two different signals, which could be assigned to the possible
structures. A significant loss of water was seen in both mass
spectra using collision induced dissociation (CID). Thus it is
assumed that only IIa and IIb were formed because IIc
contains no H-atom in the b-position to a functional group,
which would be necessary for the easy loss of water. For
steric reasons it was assumed that IIa elutes earlier than IIb
(both medium yield M).
radical, which loses a formaldehyde molecule. The formed
C14-radical adds O2 forming a C14-acylperoxy radical, which
can react with HO2 either to form an acid (VIb) or a
peroxoacid. The peroxoacid can undergo a Baeyer–Villiger-
rearrangement yielding the dicarboxylic acid Va. The aldehyde
group of the acid can be oxidised to the dicarboxylic acid Va.
This postulated pathway is supported by the observed decreasing
yield after the addition of the cyclohexane, because it destroyed
one oxidizing species (OH-radicals). Therefore the OH-induced
oxidation of the acid (VIb) is also a plausible formation pathway
of the dicarboxylic acid Va.
The C14-oxocarboxylic acids VIa–VIe were also detected
(all small yield S). A possible pathway for the formation of
VIb was shown in Scheme 11. Analogue products are implied
for the other oxocarboxylic acids. In the XIC of m/z 237
(M ꢂ Hꢂ) various peaks are observed (Fig. 9). Three not-
completely-resolved peaks (marked in the figure) seemed to be
analogous to those of the corresponding C14-dicarboxylic
acids. Thus the three peaks are supposed to belong to the
C14-oxocarboxylic acids VIa–VIe. The CID spectra of all
three oxocarboxylic acids mainly exhibit the loss of CO2.
Therefore, because of steric reasons the following elution-
order can be estimated: first elute VIa and VIb, then VIc and
VId and finally VIe.
Also both C13-dicarboxylic acids VIIa and VIIb were
tentatively identified (both small yield S). In the XIC of m/z
239 (M ꢂ Hꢂ) and m/z 261 (M + Na ꢂ 2Hꢂ) one correlating
signal was observed. The CID of m/z 239 showed the typical
loss of CO2, H2O and a combination of both plus the loss
of C7H10O2. The structures of both dicarboxylic acids are
consistent with this fragmentation pattern. In combination
with the similar reactivity of the double bonds and the similar
structure of the products (no additional methyl groups in the
a-position), it can be assumed that both co-elute. The forma-
tion pathway is very similar as to the C14-dicarboxylic
acids Va–Vc (all medium yield M). The starting point here is
the C15-radical CI 1b H1 or CI 2b H1, respectively. Instead of
formaldehyde an acetyl radical is split. The intermediate
radical W1 is formed by H-atom abstraction from an aldehyde
group through e.g. an OH-radical, see Scheme 12. Alternatively
the two aldehyde groups can be oxidised by OH-radicals (via W2).
These assumptions were proposed because a decreasing yield
First-generation products that contain only 14 or 13 carbon
atoms are also formed via the hydroperoxide channel. All
three possible C14-dicarboxylic acids (Va–Vc) were tentatively
identified. As shown in Fig. 8 three not-completely-resolved
signals in the extracted ion chromatogram (XIC) of m/z 253
(M ꢂ Hꢂ) and m/z 275 (M + Na ꢂ 2Hꢂ) display a correla-
tion. The data of the CID experiments showed only loss of
H2O, CO2 and the combination of the two. Thus only through
steric reasons it can be assumed that the C14-dicarboxylic
acids elute in the order Va, Vb and Vc (all medium yield M).
As an example the formation pathway of the dicarboxylic
acid Va is shown in Scheme 11. Starting from CI 1b H2 the
c
10980 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Table 5 Tentatively identified products of a-humulene ozonolysis
Product ID used in
text
Molecular Systematic
weight
Structure
abbreviationb IUPAC nomenclature [terpene nomenclaturea]
(1E,4E,8E)-2,6,6,9-Tetramethylcycloundeca-1,4,8-triene
[a-humulene]
I
204
AU
IIa
282
C15KDAc1 (4E,8E)-6,6,9-Trimethyl-2-oxododeca-4,8-dienedioic acid
(M)
IIb
282
C15KDAc2 (4E,7E)-3,3,7-Trimethyl-11-oxododeca-4,7-dienedioic acid
(M)
IIcc
282
C15KDAc3 (3E,7E)-3,7,10,10-Tetramethyl-2-oxoundeca-3,7-dienedioic acid
III
(3E,7E)-3,7,10,10-Tetramethylundeca-3,7-dienedioic acid
[a-humuladionic acid 10-acid]
268
C15DAc
(L)
IIIa
268
C15AcKH1a1 (4E,8E)-2-Hydroxy-4,7,7-trimethyl-11-oxododeca-4,8-dienoic acid
(L)
IIIb
(4E,8E)-10-Hydroxy-4,7,7-trimethyl-11-oxododeca-4,8-
268
C15AcKH1b1
dienoic acid
(L)
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10981

View Online
Table 5 (continued )
Product ID used in
text
Molecular Systematic
weight
abbreviationb IUPAC nomenclature [terpene nomenclaturea]
Structure
IIIc
268
C15AcKH1b2 (4E,8E)-12-Hydroxy-4,7,7-trimethyl-11-oxododeca-4,8-dienoic acid
(L)
IIId
(L)
268
C15AcKH2a1 (4E,7E)-2-Hydroxy-3,3,7-trimethyl-11-oxododeca-4,7-dienoic acid
IIIe
(L)
IIIf
(L)
268
C15AcKH2b1 (4E,7E)-10-Hydroxy-3,3,7-trimethyl-11-oxododeca-4,7-dienoic acid
268
C15AcKH2b2 (4E,7E)-12-Hydroxy-3,3,7-trimethyl-11-oxododeca-4,7-dienoic acid
IIIg
(L)
(3E,7E)-2-Hydroxy-3,7,10,10-tetramethyl-11-oxoundeca-3,7-dienoic
268
C15AcKH3a1
acid
IIIh
(L)
(4E,8E)-10-Hydroxy-2,2,5,9-tetramethyl-11-oxoundeca-4,8-dienoic
268
C15AcKH3a2
acid
IVa
(4E,8E)-4,7,7-Trimethyl-11-oxododeca-4,8-dienoic acid
[3-seco-a-humulaic-7-acid]
252
C15KAc1
(S–M)
c
10982 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Table 5 (continued )
Product ID used in
text
Molecular Systematic
weight
abbreviationb IUPAC nomenclature [terpene nomenclaturea]
Structure
IVb
(4E,7E)-3,3,7-Trimethyl-11-oxododeca-4,7-dienoic acid
[3-seco-a-humulaic-3-acid]
252
C15KAc2
(S–M)
IVc
(S–M)
(3E,7E)-3,7,10,10-Tetramethyl-11-oxoundeca-3,7-dienoic acid
[a-humulalic 10-acid]
252
C15KAc3a
IVd
(4E,8E)-2,2,5,9-Tetramethyl-11-oxoundeca-4,8-dienoic acid
[a-humulalic 11-acid]
252
C15KAc3b
(S–M)
Va
(M)
254
C14DAc1
(3E,7E)-5,5,8-Trimethylundeca-3,7-dienedioic acid
Vb
254
C14DAc2
(4E,7E)-3,3,7-Trimethylundeca-4,7-dienedioic acid
(M)
Vc
254
C14DAc3
(2E,6E)-2,6,9,9-Tetramethyldeca-2,6-dienedioic acid
(M)
VIa
(S)
238
C14KAc1a
(3E,7E)-3,6,6-Trimethyl-10-oxoundeca-3,7-dienoic acid
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10983

View Online
Table 5 (continued )
Product ID used in
text
Molecular Systematic
weight
abbreviationb IUPAC nomenclature [terpene nomenclaturea]
Structure
VIb
238
C14KAc1b
(3E,7E)-5,5,8-Trimethyl-11-oxoundeca-3,7-dienoic acid
(S)
VIc
(S)
238
238
238
C14KAc2a
C14KAc2b
C14KAc3
(3E,6E)-2,2,6-Trimethyl-10-oxoundeca-3,6-dienoic acid
(4E,7E)-5,9,9-Trimethyl-11-oxoundeca-4,7-dienoic acid
(2E,6E)-2,6,9,9-Tetramethyl-10-oxodeca-2,6-dienoic acid
VId
(S)
VIe
(S)
VIIa
(S)
240
240
214
C13DAc1
C13DAc2
C11DAc
(2E,6E)-4,4,7-Trimethyldeca-2,6-dienedioic acid
(3E,6E)-4,8,8-Trimethyldeca-3,6-dienedioic acid
(4E)-2,2,5-Trimethyloct-4-enedioic acid
VIIb
(S)
VIII
VIIIa
198
C11KAc2a
(4E)-4,7,7-Trimethyl-8-oxooct-4-enoic acid
c
10984 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Table 5 (continued )
Product ID used in
text
Molecular Systematic
weight
abbreviationb IUPAC nomenclature [terpene nomenclaturea]
Structure
VIIIb
198
C11KAc2b
(4E)-2,2,5-Trimethyl-8-oxooct-4-enoic acid
IXa
184
C10KAc2a
(3E)-3,6,6-Trimethyl-7-oxohept-3-enoic acid
IXb
184
C10KAc3
(4E)-3,3-Dimethyl-7-oxooct-4-enoic acid
X
186
C9DAc
(3E)-5,5-Dimethylhept-3-enedioic acid
XI
186
172
C9AcKH1
C8AcKH1
(3E)-7-Hydroxy-3-methyl-8-oxooct-3-enoic acid
(4E)-6-Hydroxy-5-methyl-7-oxohept-4-enoic acid
XIIa
XIIb
XIII
172
146
130
C8AcKH2
(2E)-6-Hydroxy-2-methyl-7-oxohept-2-enoic acid
C6DAc
2,2-Dimethylbutanedioic acid
XIVa
C6KAc1
3,3-Dimethyl-4-oxobutanoic acid
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10985

View Online
Table 5 (continued )
Product ID used in
text
Molecular Systematic
weight
Structure
abbreviationb IUPAC nomenclature [terpene nomenclaturea]
XIVb
130
C6KAc2
C6DK
2,2-Dimethyl-4-oxobutanoic acid
XV
114
2,2-Dimethylbutanedial
XVI
116
100
C5KAc
C5DK
4-Oxopentanoic acid
4-Oxopentanal
XVII
XVIII
102
C4KAc
C4DK
3-Oxobutanoic acid
XIX
86
3-Oxobutanal
a
b
By ref. 19. Cx = number of carbon atoms; K = keto-, oxo- or aldehyde-group; H = hydroxy-group; Ac = carboxylic acid-group; DX = two
c
functional groups of X. Substance was not detected. The letters L, M and S indicate the estimated relative yields of the first-generation products.
was observed caused upon addition of cyclohexane. The suggested
reactions are displayed in Scheme 12.
Also the analogous products IVb–IVd (shown in Table 5) were
tentatively identified. Another first-generation product, which
possibly originated mainly from the hot acid channel, is the
C15-dicarboxylic acid III (a-humuladionic acid 10-acid). This
product could be formed through the oxidation of IVc and/or
IVd, as shown in Scheme 15.
The C15-hydroxy-oxocarboxylic acids can be also formed
via the hydroperoxide channel. For the simplest possible
reaction pathways eight possible compounds (IIIa to IIIh, all
large yield L) could be formed. The formation of IIIa is shown
as an example in Scheme 13. In the corresponding XIC of
m/z 267 eight peaks were observed besides the peak of III
(C15-dicarboxylic acid), see Fig. 10. The MS/MS experiments
could not clarify the assignment of the peaks to the corres-
ponding compounds; therefore no further assignment is made
for these compounds.
The C15-dicarboxylic acid III could be tentatively assigned
(large yield L). One corresponding signal in the XIC of m/z 267
(M ꢂ Hꢂ) and m/z 299 (M + Na ꢂ 2Hꢂ) was observed. The
addition of the CI-scavengers formic and acetic acids caused a
large increase of the yield (up to factor 4, see Fig. 11). The
addition of the OH-scavenger cyclohexane decreased the yield
significantly, whereas the addition of water had hardly any effect
on the yield (Fig. 11). The increasing yield of III can be explained
by an additional pathway. The scavenging acids can react with
the stabilized CI 3a and CI 3b (C15 CIs) to form an anhydride, see
for example CI 3a in Scheme 16. The resulting anhydride hydro-
lysed either in the chamber or more likely during the sample
preparation. The decreasing yield caused by cyclohexane was
attributed to the loss of one oxidative species, the OH-radical.
Hot acid channel. The reaction pathway via the hot acid
channel for CI 1a and CI 1b is shown in Scheme 14. CI 2b reacts
analogously to CI 1b, all the other CIs analogously to CI 1a.
Only the C15-oxocarboxylic acid (IVa) was observed
among these 5 possible products (Scheme 14). The detection
of the other products shown in Scheme 14 was limited by the
analytical technique, which only could detect carboxylic acids.
c
10986 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Fig. 5 Tentatively assigned products which originate from POZ 1 (for symbols see Table 5). A very rough estimation of the relative yields is given
by S for a small yield, M for a medium yield and L for a large yield.
Fig. 6 Tentatively assigned products which originate from POZ 2 (for symbols see Table 5 and Fig. 5).
The C15-oxocarboxylic acids IVa–IVd (all small to medium
yield S–M) were observed in the XIC of m/z 251 (M ꢂ Hꢂ), see
Fig. 12, but the four signals are poorly resolved. As shown in
Fig. 12 the effect of the addition of acetic acid as CI-scavenger
is not the same for each of the oxocarboxylic acids. Therefore
a relative quantification was not possible. It was also observed
that the yield of the oxocarboxylic acids increased upon the
addition of formic acid. The possible reason is explained
analogously in Scheme 16. The formation of IVa is shown in
Scheme 14 as an example. Even a slightly higher excess of
ozone decreased significantly the yield. This extreme sensitivity
to the ozone–a-humulene ratio made it impossible to quantify
the effect of cyclohexane. The eluting order of these compounds
is assumed to be in the sequence IVa to IVd, because of steric
effects. Furthermore, all five products can also originate from
the decomposition of the SOZ as shown above.
It is interesting to note that 2/3 of the identified products
of the first generation maintained their C15-skeleton.
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10987

View Online
Fig. 7 Tentatively assigned products which originate from POZ 3 (for symbols see Table 5 and Fig. 5).
Scheme 9
Scheme 10
The C15-hydroxy-oxocarboxylic acids (IIIa to IIIh) were
C15-dicarboxylic acid III was also formed with large yield
observed with the largest yield, whereas the C15-oxodicarboxylic
acids IIa and IIb were formed with an intermediate yield,
all arising via the hydroperoxide channel. The identified
via the hot acid channel, as well as the remaining C15-oxo-
carboxylic acids (IVa to IVd) however with a rather smaller
yield (S–M). The remaining C14- and C13-products were
c
10988 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Fig. 8 Extracted ion chromatogram of m/z 253 and m/z 275 (C14-dicarboxylic acids Va, Vb and Vc).
Scheme 11
formed via the hydroperoxy channel with an intermediate (the C14
second generation. The ozonolysis of two double bonds is
dicarboxylic acids Va to Vc) or small yield (the 14-oxo-
carboxylic acids VIa to VIe, and the C13 dicarboxylic acids
VIIa and VIIb). It is also observed that the relative yield of a
specific type product is independent of its formation pathway.
For example all C15-type III products (IIIa to IIIh) with large
yield originate from all three POZs and do not seem to be
affected by the initial attack of O3 on the tri- or bisubstituted
double bonds of a-humulene. This is also true for the C14-type
VI (VIa to VIe) products with small yield.
necessary to initiate their formation. Consequently, all the
second-generation products possess one remaining double
bond with 8 to 11 carbon atoms. The products of the second
generation can originate from different possible pathways.
For example, as demonstrated in Fig. 13, the C9-dicarboxylic
acid X (C9DAc) can originate from 9 tentatively identified
products. The relative yields of the second-generation products
depend very much on a-humulene–ozone ratio, therefore even a
rough estimation of the relative yield was not attempted.
In the XIC chromatogram of m/z 213 one peak was tentatively
identified as the quasi-molecular ion of the C11-dicarboxylic acid
VIII. It was also observed as the sodium-adduct at m/z 235.
3.6.2 Second-generation products. The molecules VIII–XIIb
(Table 5) represent these tentatively identified products of the
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10989

View Online
Fig. 9 Extracted ion chromatogram of m/z 237 (C14-oxocarboxylic acids VIa to VIe).
Scheme 12
Scheme 13
The CID experiment with 25 eV collision energy showed the loss
C11 CI1 can be stabilized and react with a CI-scavenger. Another
of CO2, H2O and a combination of both. Different pathways are
thinkable to form VIII (Scheme 17). The simplest one is the
first-generation CIs reacting through the hot acid channel to
the C15-oxocarboxylic acids IVa or IVd. Upon ozonolysis of
the corresponding second double bond resulting in the
formation of CI (C11 CI1), which can further react through
the hot acid channel to form the C11-dicarboxylic acid VIII, the
pathway is the oxidation of the C11-oxocarboxylic acids VIIIa
and VIIIb. The last two possibilities are supported by the
decreasing yield of VIII in the case of cyclohexane addition
and the increasing yield in the case of formic acid or water
addition (see Fig. 14). Thus it can be concluded that the C11-
dicarboxylic acid VIII is formed by the different explained
pathways, as is demonstrated in Scheme 17.
c
10990 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Fig. 10 XIC of m/z 267. Peaks A–H representing the assumed C15-hydroxy-oxocarboxylic acids IIIa to IIIh. (The labelled letters have no
correlation with the sequence of the products assigned in Table 5.)
Scheme 14
Scheme 15
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10991

View Online
spectra of m/z 197 show the loss of CO2, followed by the loss
of a methyl group, H2O, CO and acetaldehyde. The products
VIIIa and VIIIb can easily lose CO2 and CO, but the elimina-
tion of the CH3 radical and acetaldehyde is more favourable
for VIIIa than for VIIIb, whereas the loss of H2O is only
possible for VIIIb (Fig. 16). The CI-scavengers formic and
acetic acids increased the yield, whereas cyclohexane decreased
it. Also the assertions for these effects are identical as for VIII.
The retention time of the C11-dicarboxylic acid VIII is about
2 min longer than for the C11-oxocarboxylic acids (VIIIa+b).
Under the used HPLC conditions it would be expected that
an aldehyde eluted later than the corresponding acid, but the
observed trend (aldehyde elutes earlier than acid) has been
mentioned in the literature under various conditions for
different similar ‘‘couples’’.55
Fig. 11 Relative yields of III (C15-dicarboxylic acid) under various
Two peaks were observed in the XIC chromatogram of
m/z 183. The exact mass–charge-ratio of both peaks in com-
bination with the structure of a-humulene indicated that they
correlated with C10-ketocarboxylic acids. The CID spectra of
the peak with the shorter retention time, as seen in Fig. 17,
showed besides the typical M–H–CO2ꢂ-fragment also the loss
of CH3COOH and the combined loss of CO2 and CO. There-
fore the most probable structure is the C10-oxocarboxylic acid
IXa. The other possibility for the molecular formula is the
conditions.
Also evidence for the existence of the corresponding
C11-oxocarboxylic acids VIIIa and VIIIb was seen in the XIC
chromatogram of m/z 197. The exact mass to charge ratio
correlated to the correct molecular formula. All fragments of
the CID experiment can only be explained if VIIIa and VIIIb
are assumed to co-elute (see Fig. 15 and 16). The CID mass
Scheme 16
Fig. 12 XIC of m/z 251 (C15-oxocarboxylic acids IVa to IVd) (a) without scavenger, (b) with CI-scavenger acetic acid.
c
10992 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Fig. 13 Possible precursor pathways of X (C9-dicarboxylic acid).
Scheme 17
C10-oxocarboxylic acid IXb, but this molecule cannot easily
yield of the product remained almost constant. The reason of this
lose CO because it is not an aldehyde. The CID spectra of the
peak with the longer retention time can be explained without
any problems with the structure of IXb. Both CID spectra are
shown in Fig. 17.
observation is unknown. The effect of formic and acetic acids on
the yield was not unambiguous. Cyclohexane lowered the yield,
may be because of the loss of one oxidative species (OH-radicals).
In Scheme 18 two possible pathways to form the C9-dicarboxylic
acid X from C14-dicarboxylic acid VId are shown. Pathway W2
includes one oxidation reaction to form X, so the scavenging of
the OH radical should reduce this oxidation channel.
The C9-dicarboxylic acid X was tentatively identified. One
correlating peak was observed in the XICs of m/z 185
(M ꢂ Hꢂ) and m/z 207 (M + Na ꢂ 2Hꢂ). The a-humulene
structure limited the number of possible products. Therefore it
is suggested that X is the most probable C9-dicarboxylic acid.
An interesting effect on the yield was observed upon the
addition of water. It doubled the maximum yield and also
seemed to stabilise the yield of the product, as can be seen in
Fig. 18. Even with more than a twofold excess of ozone the
Also in the XIC of m/z 185 a peak belonging to the
C9-hydroxy-oxo-carboxylic acid XI was observed. The CID
spectra showed the loss of water, CO, CO2 and some combina-
tions of those (see Fig. 19). The formation mechanism is
displayed in reaction, Scheme 19, starting from the ozonolysis
of the C15 oxo-carboxylic acid III.
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10993

View Online
For these reasons the early peak should belong to XIIa and
the late one to XIIb. The latter can be formed from the
C14-dicarboxylic acid Vc (analogous Scheme 19). Starting
from Vb analogous reactions led to the product XIIa.
3.6.3 Third-generation products. Only three C6-compounds
(XIII, XIVa and XIVb) of the third-generation products could
tentatively be identified by HPLC-MS. The C6-dicarboxylic
acid XIII, with trivial name 2,2-dimethyl succinic acid, could
positively be identified by comparison with reference material.
The retention time of the signal from both the samples and the
reference substance fully agreed, also the spectra of the CID
experiments (loss of CO2, H2O and combination of both
species). Both sodium-adducts (M + Na ꢂ 2Hꢂ at m/z 167
and 2M + Na ꢂ 2Hꢂ at m/z 313) were also observed in the
sample and the reference XID, showing identical fragmenta-
tion patterns. Incidentally, 2,2-dimethyl succinic acid is also
known as a main product of the oxidative degradation of
a-humulene in the liquid phase.56 It is also a major product of
the a-humulene gas phase ozonolysis.
Only one relative weak peak was observed in the XIC of m/z
129 (M ꢂ Hꢂ) for the C6-oxocarboxylic acids XIVa and XIVb.
It is very probable that a relatively large amount of this
compound exists in the gas phase. The retention time was
surprisingly shorter than that of the C6-dicarboxylic acid XIII,
but it was longer than that of adipic acid. CID-experiments
showed the loss of CO2, CO (which is typical for aldehydes)
and H2O. It is suggested that both C6-oxocarboxylic acids are
formed, but probably not with the same yield. These three
third-generation products could originate from various known
first- and second-generation products, as well from an unknown
number of unidentified products. Three general possible path-
ways leading to the formation of XIII, XIVa and XIVb are
shown in reaction in Scheme 20.
Fig. 14 Relative yield of VIII (C11-dicarboxylic acid) under various
conditions.
The smallest molecules of the second generation that tentatively
could be identified were the C8-hydroxy-oxocarboxylic acids
XIIa and XIIb. The corresponding XIC of m/z 171 (M ꢂ Hꢂ)
showed two relevant signals at around 12.5 and 14 min
(Fig. 20). The CID spectra of both peaks showed nearly the
same fragment, showing the loss of CO, CO2, water, but with
different relative intensities of each fragment. Both products
XIIa and XIIb can form each of these fragments, as is shown in
Fig. 21. Because of the keto–enol-tautomerism the hydroxy-
and the carbonyl-group can change their position in both
molecules, what probably occurred in solution. In the case
of XIIa the keto-form should be more preferred than in the
case of XIIb, because of the conjugated p-system that is
consequently formed. Therefore the intensity of the signals
caused by the CO-loss should be lower in the CID spectra of
XIIa. Also the fragment m/z 109 of XIIa, after the loss of
H2O and CO2, should be less stable than that of XIIb (Fig. 21).
The postulated oxidation of the aldehydes is supported by
the decreasing yield of XIII upon the addition of cyclohexane.
Fig. 15 CID mass spectra of m/z 197 at 14.7 min (C11-oxocarboxylic acids VIIIa and VIIIb).
c
This journal is the Owner Societies 2011
10994 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001

View Online
Fig. 16 CID fragmentation patterns of VIIIa and VIIIb (C11-oxocarboxylic acids).
Fig. 17 CID of m/z 183 (C10-oxocarboxylic acid); left frame: late peak (IXb); right frame: early peak (IXa).
In addition, an interesting effect of the addition of CI-scavengers
was also observed for the C6-dicarboxylic acid XIII. The
addition of H2O caused the yield slightly to decrease, whereas
the formic and acetic acid caused the yield to increase. The
signal of XIVa and XIVb was too weak to estimate the effects
of the different scavengers.
possible and probable. The same issues are even more valid
for the oxocarboxylic acids XVI and XVIII. The products of
the third generation are shown in Table 6, including the
probability of their identification. The relative yields of the
dicarbonyl products XIX (m/z 87, full circles), XVII (m/z 101,
dotted circles) and XV (m/z 115, triangles) are shown in
Fig. 22. Only the relative yields are shown because no calibra-
tions for these compounds were performed. It is not excluded
that isomeric compounds were formed. It is also possible that
other formed compounds have fragments with the same
mass to charge ratio. For the interpretation of the signals it
was assumed that in the considered concentration range the
correlation of the signal intensity and the concentration
is almost linear. As seen in Fig. 22 the yield of the three
No other small compounds could be identified with the
HPLC-MS, but some indications of the presence of smaller
molecules were seen with the PTR-MS technique. Without any
chromatographic separation and without the possibility of
performing MS/MS experiments, the clear identification of
the compounds is very difficult. Scheme 21 is showing the most
probable observed dicarbonyl products (XV, XVII and XIX).
Other compounds producing equal mass–charge-ratios are
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10995

View Online
were comparable). An excess of a-humulene led to a relatively
low yield of the three dicarbonyl compounds. The 1.7-fold
excess of ozone increased the signal intensity of all compounds,
especially the one of C5 compound XVII. High excess of ozone
led to a much higher yield of XV (C6) and also XIX (C4), but
only to a slightly increased yield of XVII (C5), compared to
the yields of 1.7-fold excess ozone. Because of the different
ionization efficiencies of the substances the absolute signal
intensities do not indicate which compound has the highest or
lowest mixing ratio.
4. Conclusions
Various aspects of ozonolysis of a-humulene have been investi-
gated, including the kinetics of ozone with the three double
bonds, the determination of the yield of OH radicals, the
formation of secondary ozonides, the production of secondary
aerosols and the tentative identification of a variety of the
three generation products. The gas phase ozonolysis was
studied under various experimental conditions, including the
Fig. 18 Relative yield of X (C9-dicarboxylic acid) under various
conditions.
dicarbonyl compounds increased generally with higher ozone
excess (the a-humulene mixing ratios of the experiments
Scheme 18
Fig. 19 CID spectra of XI (C9-hydroxy-oxocarboxylic acid).
c
This journal is the Owner Societies 2011
10996 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001

View Online
Scheme 19
Fig. 20 XIC of m/z 171 (C8-hydroxy-oxocarboxylic acids XIIa and XIIb) and CID of correlated peaks.
Fig. 21 Fragmentation pattern of XIIa and XIIb (C8-hydroxy-oxocarboxylic acids).
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10997

View Online
Scheme 20
Scheme 21
addition of cyclohexane as an OH-radical scavenger and stabilized and react intramolecularly to form stable SOZ,
HCOOH, CH3COOH and H2O as CI-scavengers, and the
variation of the a-humulene to ozone ratio.
while the formation of OH-radicals via the hydroperoxide
channel will be a minor process. The results obtained in this
study support the theoretical calculations by Chuong et al.,43
who predict that the ozonolysis of C15 endocyclic alkenes will
dominantly produce secondary ozonides.
A striking result was the detection by FTIR of secondary
ozonides (SOZ) in the gas phase. More important, the addition
of CI-scavengers, such as HCOOH and H2O, had not a
significant effect on their formation. In a nearly ozone-free
environment the SOZs remained relatively stable. Only about
20% of the SOZs was degraded after 90 min with a dew point
below 193 K (ꢂ80 1C). However, 40% of the SOZs remained
after more than two hours with a humidity of about 30%.
Evidence was found that ozone destroyed the SOZs. In experi-
ments with more than twofold excess of ozone, the SOZ was
degraded very fast; in most cases no SOZ was detectable after
few minutes.
Kinetic experiments provided an average rate coefficient for
the first-generation products (second DB) of k1 = (3.6 ꢀ 0.9) ꢁ
10ꢂ16 cm3 moleculeꢂ1 ꢂ1; this means more than 30 times
s
slower than for the reaction of ozone with the first DB
(k1 = 1.17 ꢁ 10ꢂ14 cm3 moleculeꢂ1 sꢂ1). The second-generation
products (third DB) have an average rate coefficient of
k2 = (3.0 ꢀ 0.7) ꢁ 10ꢂ17 cm3 moleculeꢂ1
s
ꢂ1; they are at
average 12 times slower than the reaction of ozone with the
first-generation products and nearly 400 times slower than
with the first DB of a-humulene. At typical tropospheric ozone
mixing ratios of 30 ppb the lifetime of a-humulene is about
2 min,48 whereas the first- and second-generation products
have average lifetimes of 1 h and 12.5 h, respectively. The
ozonolysis of the first-generation products is still faster than
the one of the sesquiterpenes a-copaene (2.5 h lifetime) and the
monoterpenes a-pinene and sabinene (4.3 h), 3-carene (10 h)
The observed stability of the secondary ozonides provides
important fundamental insight in their unimolecular forma-
tion. SOZs are formed by the intramolecular reaction of the
Criegee moiety with the carbonyl group. This suggests that
during the ozonolysis of a-humulene at atmospheric pressures
the initially formed POZ will decompose rapidly giving excited
Criegee Intermediates. A large fraction of the exited CI will be
c
10998 Phys. Chem. Chem. Phys., 2011, 13, 10970–11001
This journal is the Owner Societies 2011

View Online
Table 6 Identification of third-generation products
Compound class
C4
C5
C6
Identification
Dicarbonyl
Evidence from PTR-MS
Carbonyl carboxylic
acids part 1
Hints by PTR-MS
Carbonyl carboxylic
acids part 2
Tentatively identified by HPLC-MS
Dicarboxylic acids
Identified by HPLC-MS with reference
of a-humulene is a minor source of OH-radicals in the
troposphere.
SOA yields have been also determined. In the case of 4-fold
excess of ozone nearly 45% SOA yield was obtained. a-Humulene
in excess lowered the yield to about 14 to 17%. In the first case the
SOA is properly dominated by second- and third-generation
products, while the excess of a-humulene caused the formation
of most first-generation products.
Products of all three product generations could tentatively
be identified: those of the first-generation contained 13 to 15
carbon atoms, those of the second-generation 8 to 11 and
those of the third-generation 4 to 6. Because of the unavailability
of reference substances only the C6-dicarboxylic acid XIII
(2,2-dimethyl-succinic acid) could be positively identified by
comparison with an authentic standard.
The PTR-MS measurements allowed investigating on-line
the progress of the ozonolysis of a-humulene in the reactor.
Besides the measurement of cyclohexane only some third-
generation products could be tentatively identified. All C4- and
C5-compounds could only be observed in the gas-phase, also
the C6-dicarbonyl compound XV. The three other tentatively
identified third-generation products could only be observed in
the aerosol phase, which was investigated by HPLC-ESI(ꢂ)-MS.
The focus in this part of the study was laid on the carboxylic
acids and the experimental setup was optimised for this
class of compounds. Therefore only carboxylic acids could
be identified with this technique. As mentioned above, the
C6-dicarboxylic acid XIII could definitely be identified, the
corresponding oxocarboxylic acids XIVa and XIVb only
tentatively. In total eight products of the third-generation
could be tentatively identified.
Fig. 22 Quasi-molecular ion signals of XIX (m/z 87), XVII (m/z 101)
and XV (m/z 115): displayed are examples of background signal, and
conditions of excess of a-humulene (2.5 : 1), 1.7-fold excess of ozone,
and high excess of ozone (taken from three different experiments).
and b-phellandrene (7.9 h). Even the average lifetime of
second-generation products is with 12.5 h shorter than that
of the sesquiterpenes a-cedrene (14 h) and the monoterpene
b-pinene (24.7 h).23,48
The OH-radical yield was determined via the yield of cyclo-
hexanone measured by PTR-MS. OH yields of (10.5 ꢀ 0.7)%
for the first double bond and (12.9 ꢀ 0.7)% for the second
double bond could be calculated from the measurement
data. Thus it can be concluded that, first the hydroperoxide
channel is not the major process, and second the ozonolysis
c
This journal is the Owner Societies 2011
Phys. Chem. Chem. Phys., 2011, 13, 10970–11001 10999