This journal is
©
the Owner Societies 2016
Phys.Chem.Chem.Phys.,
2016,
18
, 10241--10254 |
10241
Cite this:
Phys.Chem.Chem.Phys.,
2016,
18
,10241
Atmospheric fates of Criegee intermediates in the
ozonolysis of isoprene
†
Tran B. Nguyen,
‡
*
a
Geoffrey S. Tyndall,
b
John D. Crounse,
a
Alexander P. Teng,
a
Kelvin H. Bates,
c
Rebecca H. Schwantes,
a
Matthew M. Coggon,
§
c
Li Zhang,
d
Philip Feiner,
d
David O. Milller,
d
Kate M. Skog,
e
Jean C. Rivera-Rios,¶
e
Matthew Dorris,
e
Kevin F. Olson,
8
fg
Abigail Koss,
h
Robert J. Wild,
hi
Steven S. Brown,
h
Allen H. Goldstein,
fg
Joost A. de Gouw,
h
William H. Brune,
d
Frank N. Keutsch,¶
e
John H. Seinfeld
cj
and Paul O. Wennberg
aj
We use a large laboratory, modeling, and field dataset to investigate the isoprene + O
3
reaction, with the
goal of better understanding the fates of the C
1
and C
4
Criegee intermediates in the atmosphere.
Although ozonolysis can produce several distinct Criegee intermediates, the C
1
stabilized Criegee
(CH
2
OO, 61
9%) is the only one observed to react bimolecularly. We suggest that the C
4
Criegees
have a low stabilization fraction and propose pathways for their decomposition. Both prompt and non-
prompt reactions are important in the production of OH (28%
5%) and formaldehyde (81%
16%).
The yields of unimolecular products (OH, formaldehyde, methacrolein (42
6%) and methyl vinyl
ketone (18
6%)) are fairly insensitive to water,
i.e.
, changes in yields in response to water vapor
(
r
4% absolute) are within the error of the analysis. We propose a comprehensive reaction mechanism
that can be incorporated into atmospheric models, which reproduces laboratory data over a wide range
of relative humidities. The mechanism proposes that CH
2
OO + H
2
O(
k
(H
2
O)
B
1
10
15
cm
3
molec
1
s
1
)
yields 73% hydroxymethyl hydroperoxide (HMHP), 6% formaldehyde + H
2
O
2
,and21%formicacid+H
2
O;
and CH
2
OO + (H
2
O)
2
(
k
(H
2
O)
2
B
1
10
12
cm
3
molec
1
s
1
) yields 40% HMHP, 6% formaldehyde + H
2
O
2
,
and 54% formic acid + H
2
O. Competitive rate determinations (
k
SO
2
/
k
(H
2
O)
n
=1,2
B
2.2 (
0.3)
10
4
) and field
observations suggest that water vapor is a sink for greater than 98% of CH
2
OO in a Southeastern US forest,
even during pollution episodes ([SO
2
]
B
10 ppb). The importance of the CH
2
OO + (H
2
O)
n
reaction is
demonstrated by high HMHP mixing ratios observed over the forest canopy. We find that CH
2
OO does not
substantially affect the lifetime of SO
2
or HCOOH in the Southeast US,
e.g.
,CH
2
OO + SO
2
reaction is a
minor contribution (
o
6%) to sulfate formation. Extrapolating, these results imply that sulfate production by
stabilized Criegees is likely unimportant in regions dominated by the reactivity of ozone with isoprene. In
contrast, hydroperoxide, organic acid, and formaldehyde formation from isoprene ozonolysis in those areas
may be significant.
a
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California, USA. E-mail: tbn@ucdavis.edu
b
Atmospheric Chemistry Observations & Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USA
c
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California, USA
d
Department of Meteorology, The Pennsylvania State University, University Park, PA, USA
e
Department of Chemistry, University of Wisconsin at Madison, Madison, WI, USA
f
Department of Environmental Science, Policy, and Management, University of California at Berkeley, Berkeley, CA, USA
g
Department of Civil and Environmental Engineering, University of California at Berkeley, Berkeley, CA, USA
h
Earth Systems Research Laboratory, Chemical Sciences Division, National Oceanographic and Atmospheric Administration, Boulder, CO, USA
i
Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, CO, USA
j
Division of Engineering and Applied Science, California Institute of Technology, Pasadena, California, USA
†
Electronic supplementary information (ESI) available. See DOI: 10.1039/c6cp00053c
‡
Now at Department of Environmental Toxicology, University of California, Davis, Davis, CA, USA.
§
Now at Earth Systems Research Laboratory, Chemical Sciences Division, National Oceanographic and Atmospheric Association, Boulder, CO, USA.
¶ Now at Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA, USA.
8
Now at Chevron Corp, San Ramon, CA, USA.
Received 4th January 2016,
Accepted 16th March 2016
DOI: 10.1039/c6cp00053c
www.rsc.org/pccp
PCCP
PAPER
Published on 16 March 2016. Downloaded by California Institute of Technology on 27/05/2016 15:59:35.
View Article Online
View Journal
| View Issue
10242
|
Phys.Chem.Chem.Phys.,
2016,
18
, 10241--10254
This journal is
©
the Owner Societies 2016
1. Introduction
Ozonolysis is one of the main atmospheric oxidation pathways
for volatile alkenes. Reaction with ozone globally removes
B
10%
of isoprene (C
5
H
8
),themostabundantalkeneintheatmosphere.
For monoterpenes (C
10
H
16
) and sesquiterpenes (C
15
H
24
), ozonolysis
is a substantially larger sink due to their faster rate coefficients with
ozone.
1
The first steps of the alkene ozonolysis mechanism are
shown in Fig. 1.
2
Two primary ozonides (POZ) are formed from
ozone addition at either double bond of isoprene, decomposing
into methacrolein (MACR), methyl vinyl ketone (MVK), form-
aldehyde (HCHO), and potentially up to nine activated Criegee
intermediates (CI, denoted with asterisk). The C
4
Criegees
(MACR-OO* and MVK-OO*) can be formed with four conforma-
tions each that are
syn
or
anti
to methyl or vinyl groups. The CI
can experience a few unimolecular processes – most notably,
decomposition
3
into a hydroxyl radical (OH) and a beta-oxy alkyl
radical (R) and thermalization by atmospheric gases to form the
stabilized Criegee intermediate (SCI). In addition, a fraction of
SCI has been suggested to be formed through POZ decomposi-
tion.
4
Most of the OH from isoprene ozonolysis is thought to be
produced by the
syn
-methyl MVK-OO conformers (Fig. 1, g and h)
via
the formation of a vinyl hydroperoxide (VHP) intermediate.
5–9
The stabilized Criegees (SCIs) may undergo bimolecular reaction
with a number of atmospheric species, including water vapor (H
2
O),
sulfur dioxide (SO
2
), formic acid (HCOOH), carbonyls (
e.g.
,HCHO
and acetaldehyde), NO, NO
2
,O
3
,RO
2
, alkenes, among others.
10–16
Even if a substantial fraction of CIs are stabilized, they may still
experience unimolecular losses. The structure, and even confor-
mation, of the SCI dictate their unimolecular and bimolecular
reactivities,
9
with
syn
SCI more susceptible to decomposition. The
simplest SCI (CH
2
OO) has special importance in atmospheric
chemistry as it is produced by all exocyclic alkenes, including
isoprene. Unlike other SCIs, however, CH
2
OO is non-
syn
(
i.e.
, not facing any hydrocarbon groups), which greatly
reduces its unimolecular reactivity.
17
Fig. 2 shows a simplified reaction scheme between CH
2
OO
and water (or water clusters),
18–21
sulfur dioxide (SO
2
), and
formic acid (HCOOH). The CH
2
OO + (H
2
O)
n
reaction (where
n
=1,2,
...
) produces hydroxymethyl hydroperoxide (HMHP),
hydrogen peroxide (H
2
O
2
) + formaldehyde (HCHO), and formic
acid (HCOOH) + H
2
Oasmainproducts.
22–28
The CH
2
OO + HCOOH
reaction produces hydroperoxy methylformate (HPMF).
23,27,29
Finally, the CH
2
OO + SO
2
reaction produces SO
3
, which then
reacts with water to form H
2
SO
4
.
30
Certain populations of SCIs may produce OH,
31
perhaps
analogously to the hot Criegee VHP channel, among other
products. Decomposition of SCIs is rarely discussed within
the scope of the atmospheric fates; however, it is an important
consideration in understanding their total reactivity. Previously
published unimolecular decomposition rates for larger Criegees
have high uncertainty, so the following is only a qualitative
discussion. SCI decomposition rates has been shown to increase
with size.
11,32
For example, even though the thermalized acetone
oxide ((CH
3
)
2
COO) has been recently reported to undergo a
diffusion-limited reaction with SO
2
,
33
its short unimolecular
lifetime due to its all-
syn
conformation
, i.e.
, both sides facing
methyl groups, severely limits the atmospheric relevance of its
bimolecular reactions (
t
uni
= 0.001–0.004 s).
32–34
It should be
noted that experimental determinations of unimolecular life-
times (
e.g.
,0.002s)
33
may have some contribution from Criegee
self-reaction; thus,
t
uni
may be closer to the higher end of the
reported range. The ratio
k
decomp
/
k
SO
2
for
syn
-CH
3
CHOO and
(CH
3
)
2
COO have been measured to be 1 and 2 orders of magni-
tude higher than that for CH
2
OO, respectively.
32
Thus, even in a
polluted atmosphere (
B
10 ppb SO
2
and 50% RH at room
temperature), decomposition using the Olzmann
et al.
(1997)
lower-limit rate coefficient accounts for the majority (
B
76%) of
the acetone oxide fate, while the reaction with SO
2
is minor
(
B
8%). The Newland
et al.
(2015) relative rate coefficient (using
k
SO
2
of Huang
et al.
(2015))predictsanevenhigherdecomposition
fraction.
Fig. 1
The first steps of the Criegee mechanism of ozonolysis, shown for
isoprene. Criegee intermediates are drawn as zwitterions; however,
depending on the chemical structure, they may also have diradical
character.
Fig. 2
The reaction of CH
2
OO with HCOOH, SO
2
,andH
2
O (and possibly
water clusters). The production of HCOOH + H
2
O and HCHO + H
2
O
2
from the water reaction has been suggested to (at least partially) result
from surface-mediated decomposition of HMHP;
27
however, it is not clear
if there are direct routes to these products from CH
2
OO + H
2
O.
Paper
PCCP
Published on 16 March 2016. Downloaded by California Institute of Technology on 27/05/2016 15:59:35.
View Article Online
This journal is
©
the Owner Societies 2016
Phys.Chem.Chem.Phys.,
2016,
18
, 10241--10254 |
10243
The present work focuses on understanding the bimolecular
reactive channels of CH
2
OO and, more generally, the mechanism
of isoprene ozonolysis in the atmosphere. We neglect the
unimolecular reactions for CH
2
OO, as it has a long lifetime
with respect to decomposition at 298 K and 1 atm (
B
3 s).
34
Furthermore, we provide suggestions for a unifying reaction
scheme that may be incorporated into atmospheric models.
2. Experimental
2.1. Chamber methods
Experiments were conducted in the Caltech dual 24 m
3
Teflon
environmental chambers at
B
295 K and
B
1 atm. A subset of
the work was performed as part of the FIXCIT campaign and
the overview manuscript
35
provides an in-depth description
of the chamber and relevant experiments. Product yield studies
were investigated with isoprene and ozone mixing ratios of
B
100 and 600 ppb, respectively, and relative humidity (RH) in
the approximate range of
o
4–76%. The production of OH was
investigated in the absence of a chemical scavenger, but all
other studies were performed with excess cyclohexane (50 ppm)
to scavenge OH. Although excess cyclohexane is used, a minor
fraction of the products will result from OH chemistry. Relative
rate experiments were used to investigate the competition
between SO
2
and H
2
O at lower isoprene and ozone mixing
ratios (
B
25 ppb and 100 ppb, respectively) and 10 ppm
cyclohexane.
RH inside the reaction chambers was adjusted to the desired
level at the beginning of each experiment with a Nafion
membrane humidifier (Permapure, LLC) and recirculating
ultra-purified water (Millipore Milli-Q, 18 M
O
,
o
3 ppb TOC).
The RH was stable throughout each experiment, as verified by
a Vaisala HMT221 probe that was calibrated in the range of
11–95% with saturated salt solutions. Water vapor in the range
of RH
o
10% was measured by chemical ionization mass
spectrometry (CIMS, see Section 2.2). However, the accuracy
of RH measurements degrades in the lower range due to the
difficulty in determining small mixing ratios of H
2
O; thus we
quote ‘‘dry’’ RH as ‘‘
o
4%’’. Although RH in dry conditions is
quoted as an upper bound, we estimate the actual RH in the
chamber is closer to 1%. Ozone was introduced into the
chamberbyflowingairthroughac
ommercial UV ozone generator.
Reagents,
e.g.
, isoprene (Aldrich,
Z
99%) cyclohexane (CHX,
Aldrich
4
99%), were introduced into the chamber by volumetric
injection of liquid material using Hamilton gas-tight syringes.
In general, the order of introduction was water vapor, ozone,
cyclohexane, and then isoprene. For relative rate studies,
gaseous SO
2
(standard mixture 10 ppm in N
2
, Scott Specialty
Gases) was introduced into the chamber using a calibrated mass
flow controller. After injection of isoprene, several short high-
pressure pulses of air were introduced into the chamber to
homogenize the contents of the chamber so that the reaction
can start immediately and uniformly. We verified that injected
gases were well-mixed in
o
5 minutes using this method. The
duration of a typical experiment was 5–7 hours.
2.2. Analytical quantification
Isoprene, methacrolein (MACR), methyl vinyl ketone (MVK),
and cyclohexane (CHX) were quantified by gas chromatography
with a flame ionization detector (GC-FID). The GC-FID was
calibrated with commercial standards (Aldrich) in the range of
20–200 ppb by use of volumetric gas-tight syringes and a
calibrated mass flow of N
2
into a 100L Teflon calibration bag.
Additionally, the absolute quantities of ISOP, MACR, and MVK
was cross calibrated using Fourier
transform infrared spectroscopy
(FT-IR) in the range of 1–20 ppm
via
a similar method. The ppm-
level calibration bags were quantified with FT-IR using tabulated
absorption cross sections
36
before sampling with GC-FID. The
mixing ratio of ozone was quantified by a calibrated ozone
absorption monitor (Horiba APOA-360). The mixing ratios of
NO and NO
2
were quantified with a commercial NO
x
monitor
(Teledyne T200). NO was observed at baseline level (limit of
detection 0.5 ppb) and NO
2
remained below 5 ppb during
ozonolysis experiments. Sulfuric acid aerosols were measured
using a time-of-flight aerosol mass spectrometer (AMS, Aerodyne)
37
and data processing was performed using the Pika 1.14D analysis
module in Igor Pro. The instrumental ionization efficiency was
calibrated with 350 nm ammonium nitrate particles.
Formaldehyde and HO
x
(OH and HO
2
) were measured
in situ
by two laser-induced fluorescence (LIF) instruments during
the FIXCIT campaign. The University of Wisconsin (UW) LIF
instrument
38,39
quantified formaldehyde from the difference
between its online (353 nm) and offline signal. The Pennsylvania
State University (PSU) Ground-based Tropospheric Hydrogen
Oxides Sensor (GTHOS)
40
measured OH and HO
2
by the fluores-
cent assay by gas expansion (FAGE) technique. OH was quanti-
fied spectroscopically (near 308 nm) and the zero background is
determined using hexafluoropropene (C
3
F
6
)asanOHscrubber
in the instrument inlet. HO
2
was measured after its chemical
conversion in the instrument inlet to OH using pure NO (HO
2
+
NO
-
OH + NO
2
). The known interference of HO
2
measurement
by RO
2
radicals
41
was corrected in the following manner: the NO
addition to GTHOS was modified to reduce the reaction time
and the amount of NO added. Although the conversion of HO
2
to
OH was decreased from
B
90% to less than 10%, the conversion
of RO
2
to OH was reduced to less than 1%, so that more than
95% of the signal was due to converted HO
2
and only a few
percent was due to RO
2
.
41
These conversion rates were measured
with GTHOS in the Brune laboratory at PSU and are similar to
those determined by Fuchs
et al.
(2011).
Gas-phase hydroperoxides (H
2
O
2
, HMHP, MHP,
etc.
), acidic
compounds (SO
2
, HCOOH,
etc.
), and other volatiles with more
than one polar functional group (
e.g.
, hydroxy carbonyls) are
quantified with a triple-quadrupole chemical ionization mass
spectrometer (CIMS) using CF
3
O
as an ionization reagent.
42,43
The sample flow from the chamber was diluted by a factor
of 12 with dry N
2
before mass spectrometry analysis. The dry
(RH
o
4%) sensitivity of triple-quadrupole CF
3
O
CIMS to
different analytes was cross-calibrated with a CF
3
O
time-of-flight
(ToF) CIMS during the FIXCIT campaign. The ToF CIMS was
calibrated for a variety of compounds (H
2
O
2
, HMHP, HCOOH, SO
2
,
peracetic acid (PAA), acetic acid (AA), hydroxyacetone (HAC),
etc.
)
PCCP
Paper
Published on 16 March 2016. Downloaded by California Institute of Technology on 27/05/2016 15:59:35.
View Article Online
10244
|
Phys.Chem.Chem.Phys.,
2016,
18
, 10241--10254
This journal is
©
the Owner Societies 2016
with commercial or synthesized standards based on gravimetric
or spectrometric techniques (see Section S3 of ref. 44 for more
information). Table S1 (ESI
†
) provides more information about
CIMS detection of the major compounds discussed in this work.
The CIMS measurement uncertaint
ies are approximately 20–30% for
calibrated compounds (
e.g.
,HCOOH)and
B
50% for uncalibrated
compounds (
e.g.
,HPMF).
In addition to the dry sensitivity, the dependence of the ion
chemistry on water vapor is unique to each CIMS instrument
and is critical for the accurate interpretation of RH-dependent
yields. We obtained the water-dependent calibration curves in
the experimental RH range by introducing a sample stream
(containing a stable gaseous source of each compound) and a
dilution stream that has tunable water vapor content to the
CIMS flow tube region. The water vapor mixing ratio of the
dilution stream was achieved by mixing flow-controlled quan-
tities of a humid air stream ([H
2
O]
B
3%, quantified by FT-IR)
and a dry N
2
stream ([H
2
O]
o
100 ppm). A stable source of
HMHP, for which a commercial standard is unavailable, was
synthesized in the Teflon chamber using the HCHO + HO
2
reaction,
45
which produces a low yield of HMHP. The photolysis
of HCHO (
B
2 ppm) generates the HO
2
that is needed to react
with HCHO. The UV lights were turned off after approximately
1 hour, and the
B
6 ppb HMHP formed during the photolytic
period was stable in the dark indefinitely. A typical water-
dependent calibration alternates a dry data point with several
humid points and zeros (where sample flow is shut off), after
each period is allowed to stabilize (Fig. S1, ESI
†
). Water curves
were obtained for HCOOH, H
2
O
2
, and SO
2
using commercial
standards as the sample source, in an identical manner. The
sensitivity of the CIMS toward HPMF was not measured, but
was assumed to be similar to HMHP based on the molecular
characteristics of these two compounds.
46
2.3. Wall loss corrections
Alpha-hydroxy hydroperoxides like HMHP have a propensity to
participate in heterogeneous reactions on humid surfaces.
27
Thus,
we measured wall loss rates for HMHP, HCOOH, and H
2
O
2
as a
function of RH to correct for this effect. HMHP was synthesized
via
an alternative method to the one described in Section 2.2:
a gaseous mixture o
f formaldehyde/N
2
(produced by flowing dry
N
2
past heated paraformaldehyde solid) was bubbled into an
aqueous H
2
O
2
solution (50% v/v). The outflow of the bubbler
(containing HCHO, HMHP, HCOOH, and H
2
O
2
) was introduced
into the chamber until the signal of HMHP in CIMS was adequate,
after which the flow w
as stopped and the wal
llosswasmonitored
for 8–10 hours. The production of HCOOH from HMHP conver-
sion may obscure the HCOOH wall loss to a degree. However, by
virtue of the synthesis method (high water content in the H
2
O
2
bubbler), the HCOOH mixing ratio in the chamber was more
abundant than HMHP by a factor of 100, so that even if all of the
HMHP were converted to HCOOH, the production yield signal
would impact
k
wall
of HCOOH by only 1%. We did not observe
noticeable wall loss of HMHP, HCOOH, or H
2
O
2
under dry
conditions (Fig. S2, ESI
†
); however, the wall loss rates become
non-negligible at the highest RH investigated (72%), where HMHP
was removed at a rate of approximately 0.1% per minute. The
humidity-dependent wall loss rates (
k
wall_HMHP
=
1.4
10
5
RH min
1
,
k
wall_H
2
O
2
=
9.6
10
6
RH min
1
,and
k
wall_HCOOH
=
2.2
10
6
RH min
1
) were used to correct the CIMS data.
3. Results and discussion
3.1. Humidity-dependent product yields
The molar yields of products from the isoprene ozonolysis in
the RH range of
o
4–76% are reported in Table 1. Fig. 3 shows
Table 1
Molar yields of major products for different RH and SO
2
conditions. Initial conditions of ISO and O
3
in B are similar to A, except CHX was not
added as a reagent. CH
2
OO yields for A are obtained by summing the yields of CH
2
OO + (H
2
O)
n
and CH
2
OO + HCOOH products measured using CIMS.
In C, CH
2
OO yield determinations include the measurement of H
2
SO
4
aerosol using AMS. See Table S2 (ESI) for details on quantification with CIMS.
Missing entries indicate unavailability of instruments or irrelevancy of data. Measurement uncertainties are: SO
2
(
20%), CH
2
OO (
15 AMS,
30% CIMS),
MACR (
15%), MVK (
33%), OH (
25%), HCHO (
20%), HCOOH (
10%), H
2
O
2
(
20%), HMHP (
20%), HPMF (
50%). BDL = below detection limit for
GC-FID
Experiment type
(
T
= 295 K)
SO
2
(ppb)
RH
(%)
Yield
(CH
2
OO)
Yield
(MACR)
Yield
(MVK)
Yield
(OH)
Yield
(HCHO)
Yield
(HCO
2
H)
Yield
(H
2
O
2
)
Yield
(HMHP)
Yield
(HPMF)
(A) Product yields
(scavenger)
[ISO]
B
100 ppb
[O
3
]
B
600 ppb
[CHX]
B
50 ppm
0
o
4
0.30
0.39
0.15
—
0.79
0.05
0.012
0.17
0.064
0
6
0.40
0.39
0.14
—
—
0.07
0.014
0.26
0.060
0
13
0.51
0.41
0.17
—
—
0.09
0.019
0.38
0.030
0
27
0.61
0.41
0.16
—
—
0.13
0.038
0.44
0.013
0
37
0.61
0.41
0.18
—
0.83
0.13
0.044
0.44
0.001
0
44
0.62
0.42
0.16
—
—
0.15
0.046
0.41
0.005
0
51
0.61
0.40
0.21
—
—
0.20
0.054
0.34
0.004
0
73
0.62
0.44
0.19
—
—
0.28
0.066
0.24
0.002
0
76
0.60
0.41
0.19
—
—
0.28
0.065
0.22
0.003
(B) Product yields
(no scavenger)
0
4
—
—
—
0.28
—
—
—
—
—
0
52
—
—
—
0.28
—
—
—
—
—
(C) Relative rate
[ISO]
B
25 ppb
[O
3
]
B
120 ppb
[CHX]
B
10 ppm
15
20
0.60
BDL
BDL
—
—
—
—
—
—
15
3
0.65
BDL
BDL
—
—
—
—
—
—
75
3
0.66
BDL
BDL
—
—
—
—
—
—
Paper
PCCP
Published on 16 March 2016. Downloaded by California Institute of Technology on 27/05/2016 15:59:35.
View Article Online
This journal is
©
the Owner Societies 2016
Phys.Chem.Chem.Phys.,
2016,
18
, 10241--10254 |
10245
the trends in yields of select gas-phase organic products measured
by GC-FID and CIMS. As expected, the ‘‘prompt’’ products,
i.e.
,
those formed primarily from the decomposition of primary
ozonides (POZ) such as HCHO, MACR, and MVK, do not exhibit
a strong dependence on water vapor (Fig. 3A and Table 1). This
is also true for yields of OH radicals, which are produced from
decomposition channels. The observation that OH yields from
isoprene ozonolysis are independent of humidity has been
reported in other works.
5,47
Further insights on the sources of
OH and HCHO yields are obtained by model simulations (see
Section 3.2). The yields of carbonyls and OH from this work are
not significantly different from those reported elsewhere.
8,9,48–51
The trends in yields for carbonyls are slightly positive with
humidity, possibly supporting a minor production from SCI +
(H
2
O)
n
reaction (H
2
O
2
ascoproduct).However,themeasurement
uncertainties are significant (10–30%) and, thus, this channel was
treated as minor in the development of our mechanism.
Stabilized CH
2
OO yields obtained by a chemical scavenging
method were similar whether H
2
OorSO
2
was used as the
Criegee scavenger (Table 1,
Y
SCI
B
0.60 using CIMS, and
B
0.64
using AMS). As the detection of scavenged products utilized two
independently-calibrated instruments, their agreement lends
further confidence to the yield results. Our CH
2
OO determination
is consistent, within uncertainties, with those reported recently
(0.56–0.60).
32,52
However, it is in poor agreement with the 0.27
yield determined by Hasson
et al.
(2001).
49
We believe the
discrepancy is due to the fact that HCOOH and H
2
O
2
were
not counted as CH
2
OO + (H
2
O)
n
products in the Hasson
et al.
(2001) work, and the offline HMHP determination may have
experienced aqueous losses. The CH
2
OO yield reported here is
supported by independent observations of its co-products, MVK
and MACR (Fig. 1). Fig. 3C shows that the CH
2
OO yield
determined here is in good agreement with the C
4
carbonyl
sum at high water vapor mixing ratios where CH
2
OO is fully
scavenged. The inferred CH
2
OO yield in our work does not
include formaldehyde as a product due to limited data. Form-
aldehyde formation becomes important at low RH because of
competing reactions such as CH
2
OO + O
3
; thus, a significant
deviation in the inferred CH
2
OO yield compared to the C
4
carbonyl sum occurs in the low RH range.
The products derived from CH
2
OO bimolecular reactions
have a strong relationship with water vapor mixing ratio due
to competition from the CH
2
OO + (H
2
O)
n
reaction (Fig. 3B).
Hydroperoxy methylformate (HPM
F),seeminglythesoleproduct
of the CH
2
OO + HCOOH reaction (Fig. S3, ESI
†
), is observed only
under very dry conditions in accordance with previous reports.
27,53
This is because formic acid in ozonolysis experiments is rarely
present at the levels needed to compete with water vapor. In
addition to compounds reported i
n Table 1, RH-dependent yields
of minor species like acetic acid were also observed (
o
0.06).
Representative CIMS mass spectra showing all products are given
in Fig. S4 (ESI
†
). Acetic acid has not been identified in past
isoprene ozonolysis studies, but serves as an important clue in
deducing the fragmentation patterns of C
4
Criegees.
HMHP is the most abundant CH
2
OO + (H
2
O)
n
product,
followed by formic acid, then H
2
O
2
(+ HCHO). The maximum
HMHP yield is determined to be
B
44% from isoprene (
B
73%
from CH
2
OO), somewhat higher than other values reported in
the literature. Insightful comparisons with literature values
prove challenging, however, due to the poor agreement in
CH
2
OO + (H
2
O)
n
product yields. For example, single-point
‘‘humid’’ HMHP and H
2
O
2
yields are reported to be 0.09–0.30
and 0.01–0.12, respectively.
27,49,50,54,55
Some of the inconsistencies
in past experiments have been attributed to the challenge of
quantifying hydroperoxides with offline aqueous methods
(
e.g.
, high-performance liquid chromatography (HPLC)).
Interestingly, we find HMHP yields decrease above RH
B
40%
(Fig. 3B). The reduction in yield for HMHP at high humidity is
almost fully compensated by an increase in yield for HCOOH
(Fig. 3D). Although wall-mediated reaction is a convenient
explanation, our RH-dependent corrections for wall loss using
authentic compounds should account for this chemistry (Fig. S2,
ESI
†
). Instead, model simulation results in Section 3.2 support
the idea that the RH-dependent yields of HMHP and HCOOH are
controlled by reactions of both the water monomer and dimer.
The dimer becomes exceedingly more abundant at higher RH.
As the model simulations fit concentration data that have been
wall-loss corrected, the heterogeneous reaction is not included in
the mechanism. The atmospherically-relevant reaction of water
dimer with CH
2
OO was first suggested by Ryzhkov and Ariya
18,21,28
and later confirmed by experimental works.
19,20
Ryzhkov and Ariya
suggested the decomposition products to be H
2
O
2
and HCHO;
however, our data are more consistent with HCOOH as the major
product from this reaction.
Past studies explored a large range in water vapor mixing
ratio (9000–20 000 ppm, RH
B
28–63% at 298 K) while reporting
only a single ‘humid’ yield for products. Thus, it is possible that
poor literature agreement may be due to snap-shot observations
along different humidity points in the HMHP yield curve.
These disagreements are likely exacerbated by the absence of
wall loss corrections, which depend on the reaction vessel.
Fig. 3
Molar yields of the isoprene + O
3
reaction products (A–C) at
several RH conditions. The CH
2
OO yield in panel C is inferred from the
sum of the scavenged products of CH
2
OO with water vapor and formic
acid. Panel D shows the fraction of water-scavenged CH
2
OO that is
observed as HMHP (
f
HMHP
) and HCOOH (
f
HCOOH
). Solid lines indicate
least-squares fits, when applicable, and dashed lines only serve to guide
the eyes. HMHP and HCOOH in panel B can each be fit by two exponential
curves delineated at RH
B
40%, but no singular relationship.
PCCP
Paper
Published on 16 March 2016. Downloaded by California Institute of Technology on 27/05/2016 15:59:35.
View Article Online
10246
|
Phys.Chem.Chem.Phys.,
2016,
18
, 10241--10254
This journal is
©
the Owner Societies 2016
To our knowledge, only two other HMHP yield studies have
been performed at multiple RH conditions.
24,49
Hasson
et al.
(2001) and Huang
et al.
(2013)didnotreportyieldtrends
similar to this work,
i.e.
, their data reported a rise-to-maximum
relationship of HMHP with RH (maximum yields of 16% and
25%, respectively). Yet, despite the fitting function used by
Hasson
et al.
(2001), their data show that the average HMHP
yield at RH
B
80% (
B
0.12
0.03) is lower than its yield at
RH
B
40% (
B
0.16
0.04) for isoprene – congruent with our
observed trends.
We are unsure of the reasons for discrepancies with the
Huang
et al.
(2013) work. In addition to the plateauing HMHP
yield, Huang
et al.
(2013) reported a declining yield of HCOOH
with humidity (
e.g.
, 40% yield of HCOOH at RH 5% that
decreases to 30% yield at RH 90%). It is difficult to understand
how HCOOH can be produced in higher yields under dry
conditions when HCOOH formation from Criegee isomerization
is minor compared to the major channel of CH
2
OO + (H
2
O)
n
.
56,57
Again, given the lower HMHP yields reported by studies using
offline analysis techniques, it is possible that aqueous losses
may have occurred and direct comparisons are not possible.
Furthermore, the bis-hydroxymethyl peroxide (BHMP) reported
in Huang
et al.
(2013) (and absent in this work) may hint
at side reactions that are symptomatic of the high reagent
concentrations (ppm level) used in that work or condensed-
phase chemistry of H
2
O
2
and HCHO.
3.2. Toward a unifying mechanism
Major atmospheric models either do not represent ozonolysis
chemistry or provide a significantly abridged version that
generally neglects the formation of major compounds such as
HMHP.
58,59
Here, we describe a detailed chemical mechanism
based on the new data presented in this work and those
available in the literature. The
in situ
observations of oxyge-
nated volatile organic compounds and HO
x
enable us to place
new constraints on many aspects of the mechanism. Mecha-
nism simulations of HCHO assumes that are no observational
interferences from ROOH or other compounds, which is
currently unverified for the LIF instrument used here but has
been identified for proton-transfer-reaction (PTR-) and GC-based
instruments.
60
The proposed mechanism provides enough
chemical specificity to capture the RH-dependent yields of
OH, carbonyls (HCHO, MACR, MVK), and major products of
CH
2
OO + (H
2
O)
n
chemistry. Although uncertainties persist
along several channels in the ozonolysis chemistry, especially
in the fate of the C
4
Criegees, the proposed scheme is a good
starting point for further development and use in atmospheric
models.
Fig. 4
Overall scheme of isoprene ozonolysis and reactions of Criegee intermediates, with proposed isomerization and decomposition pathways of C
4
Criegees. Observed product species are shown in red. The reaction of CH
2
OO with O
3
and isoprene, although present in the model mechanism, are not
shown in the figure. Literature values: [a] ref. 8, [b] this work, [c] ref. 61, [d] ref. 62, [e] ref. 5, [f] ref. 63–65, [g] ref. 66, [h] ref. 67.
Paper
PCCP
Published on 16 March 2016. Downloaded by California Institute of Technology on 27/05/2016 15:59:35.
View Article Online
This journal is
©
the Owner Societies 2016
Phys.Chem.Chem.Phys.,
2016,
18
, 10241--10254 |
10247
3.2.1. POZ and C
4
Criegee reactions.
Fig. 4 shows the
proposed reaction scheme for isoprene ozonolysis. Compounds
observed in this work are shown in red. The chemical structures
of some of the minor oxygenated species may not be unique,
as this CIMS technique cannot distinguish isobaric species. We
used the branching ratios for POZ formation that was suggested
by Aschmann and Atkinson,
8
which implies that the lower steric
hindrance from the 3,4-addition of ozone is more important
than the effect of the electron-donating CH
3
group in the
1,2-addition.
61
It is assumed that there is negligible conforma-
tional interconversion between Criegees due to their zwitterionic
character,
68,69
i.e.
, the barrier to interconversion is expected to be
large.
70
We note that data available for Criegees with allylic
groups, which would presumably facilitate interconversion,
is still scarce. Thus, the assumptions and reaction channels
discussed here may need to be re-evaluated in future work.
Evidence of bimolecular reactions of the C
4
SCI is not
significant. For example, the signals of C
4
a
-hydroxyalkyl hydro-
peroxides that are analogous to HMHP,
e.g.
, from the reaction
of
anti
MACROO + (H
2
O)
n
, were not observed here. Further-
more, MACR + MVK yields did not significantly increase
following SO
2
addition,
e.g.
, as would be expected from the
MACROO + SO
2
-
MACR + SO
3
reaction. The insignificant
production of MACR from the MACROO + SO
2
reaction and
the fast
anti
-SCI + H
2
O rate coefficient determined recently
(2.4
10
14
cm
3
molec
1
s
1
)
71
favor the hypothesis that the
stabilization fraction of the C
4
Criegees is small, as opposed a
larger population of SCI where bimolecular reactions are non-
competitive. Thus, we assumed a Criegee stabilization fraction
of 0.03 as suggested by Kuwata and Valin.
62
However, accessible
unimolecular pathways of CIs and SCIs are often identical, so it
is not possible for this study to fully distinguish the two fates.
An SCI unimolecular rate constant of
B
250 s
1
would also be
consistent with observations. The 0.03 ‘‘stabilization fraction’’
can be viewed as an effective fraction of Criegees that react
bimolecularly under H
2
O-dominated conditions. Extensive C
4
Criegee decomposition (hot or thermalized) is consistent with
the high yields (
4
80%) of HCHO that are observed in this work
and elsewhere.
48
The production of HCHO from the prompt
POZ decomposition is constrained by MVK + MACR yields to be
approximately
B
40% by mole with respect to isoprene loss.
CH
2
OO is fully scavenged by water in most of our experiments,
so little additional HCHO can originate from side reactions of
CH
2
OO at atmospherically-relevant RH.
After the POZ decomposition, the distribution of the
syn
/
anti
conformers of the C
4
Criegees is thought to be asymmetric. We
used the branching ratios suggested by Kuwata and coworkers,
5,62
with the caveat that the MVK-OO* conformer distribution is
loosely based on the hot acetaldehyde oxide, for lack of direct
information. Unimolecular reactions of the C
4
Criegees have
been suggested to occur
via
5-member dioxole or 3-member
dioxirane intermediates.
5,62,72
The model mechanism allows C
4
Criegees that are
syn
and
anti
to vinyl groups to form dioxole
and dioxirane intermediates, respectively, using the theoretically-
predicted dioxole/dioxirane branching ratios.
62
Dioxoles have
been suggested to isomerize into products containing carbonyl
and epoxide functional groups, which may further decompose;
5
however, the CIMS technique used in this work is likely not
sensitive to these specific compounds. The dioxole products were
not traced in the model because they represent an insignificant
fraction of the carbon in the mechanism (
B
3%) and are not
thought to impact OH or HCHO yields. A minor fraction of the
anti
MACROO* is allocated toward hot acid formation, yielding
methacrylic acid (which may also occur
via
a dioxirane inter-
mediate).
73
The dioxirane channels represent a larger fraction of
the carbon in the ozonolysis. We followed the recommendations
of Peeters, Vereecken, and coworkers,
63,64
in conjunction with
observations derived from acetaldehyde oxide,
65
to assign the
majority of the dioxirane fate to the decarboxylation pathway
(products CO
2
+HO
2
+ alkyl radical for dioxiranes in the primary
position). A decarboxylation branching ratio of
B
0.7 gave good
agreement with observations. As these dioxiranes have allylic
functionality, we assign the balance of the carbon to a proposed
isomerization pathway (Scheme S2, ESI
†
) that may form a stable
product. The alkyl radical that is produced in the decarboxylation
step in Route A is the methylvinyl radical, which is known to
generate HCHO + peroxyacetyl radical (
B
0.35) or HCHO +
methylperoxyl radical + CO (
B
0.65) in the presence of oxygen.
66
It is probable that the observed acetic acid (0.02–0.06 from dry to
humid) is produced from the reaction of peroxyacetyl radicals
with HO
2
or RO
2
.
74–77
We speculate that the higher acetic acid
yield under more humid conditions may be due to unidentified
wall reactions. The methylperoxyl radical is a precursor to methyl
hydroperoxide under HO
2
-dominant conditions. Methyl hydro-
peroxide has been identified in previous works,
50,78
but without
complete mechanistic knowledge of its chemical source.
The
syn
MVKOO* will decompose to OH and a beta-oxy alkyl
radical
via
a vinylhydroperoxide intermediate (B route). The
further reactions of the beta-oxy alkylperoxyl radical (RO
2
) are
much more uncertain. In the mechanism suggested here, this
chemistry is proposed to proceed similarly to the RO
2
radicals
found in MVK + OH chemistry that have analogous function-
ality.
67
We followed the recommendations in Praske
et al.
(2015) for the branching ratios of the three product channels
with HO
2
: beta-oxy hydroperoxide, 1,2-dicarbonyl + OH + HO
2
,
and alkoxyl radical (RO) + OH + O
2
. The RO radical fragments to
formaldehyde and an acyl radical and promptly reacts with O
2
to produce an acylperoxyl (RC(O)OO) radical. The acylperoxyl
radical may undergo three fates upon reaction with HO
2
(Fig. 4B),
modeled after reactions of peroxyacetyl.
79,80
These data suggest
that decarboxylation is an important fate for this particular
acylperoxyl radical, which affects both OH and formaldehyde in
the process (
via
the chemistry of the vinyl radical.)
81
3.2.2. C
1
Criegee + water reaction.
The mechanism illu-
strated in Fig. 4 was integrated into a kinetic model. Most of the
rates and branching ratios available in the literature were
imported for use in the model mechanism and assumed to
be accurate. The product yields and rate coefficients of Criegee
reactions labeled [b] in Fig. 4 were empirically tuned to provide
satisfactory agreement with observational data within the full
RH range, as shown in Fig. 5. The reaction inputs into the
kinetic model are shown in Scheme S1 (ESI
†
). We find that
PCCP
Paper
Published on 16 March 2016. Downloaded by California Institute of Technology on 27/05/2016 15:59:35.
View Article Online