Kinetic and theoretical investigation of the gas-phase ozonolysis of isoprene: Carbonyl oxides as an important source for OH radicals in the atmosphere

Article abstract of DOI:10.1021/ja970050c

Kinetic measurements as well as B3LYP/ and MP2/6-31G(d,p) calculations provide evidence that carbonyl oxides formed in the gas-phase ozonolysis of alkylated alkenes are an important source of OH radicals. In the gas-phase ozonolysis of propene, cis-2-butene, trans-2-butene, tetramethylethene, and isoprene, 18, 17, 24, 36, and 19% OH radicals (relative to reacted ozone, error margin ≤4%) are measured using CO as a scavenger for OH. The quantum chemical calculations show that OH radical production depends on syn positioned methyl (alkyl) groups and their interaction with the terminal O atom of a carbonyl oxide. For example, in the gas-phase ozonolysis of ethene only 5% OH radicals are measured while for a carbonyl oxide with syn-positioned methyl (alkyl) group, a much larger amount of OH radicals is formed. This is due to the fact that 1,4 H migration and the formation of an intermediate hydroperoxy alkene, that is prone to undergo OO bond cleavage, is energetically more favorable than isomerization to dioxirane. In the case of-syn-methyl, dimethyl, and isopropenyl carbonyl oxide calculated activation enthalpies at 298 K are 14.8, 14.4, and 15.5 kcal/mol compared to the corresponding dioxirane isomerization barriers of 23.8, 21.4, and 23.0 kcal/mol, respectively. The OO cleavage reactions of the hydroperoxy alkenes formed in these cases are just 11, 12.8, and 10.3 kcal/mol.

Full text of DOI:10.1021/ja970050c

7330
J. Am. Chem. Soc. 1997, 119, 7330-7342
Kinetic and Theoretical Investigation of the Gas-Phase
Ozonolysis of Isoprene: Carbonyl Oxides as an Important
Source for OH Radicals in the Atmosphere
Roland Gutbrod,† Elfi Kraka,*,‡ Ralph N. Schindler,† and Dieter Cremer‡
Contribution from the Institut fu¨r Physikalische Chemie, Christian-Albrechts UniVersita¨t Kiel,
Ludewig-Meyn Str. 8, 24098 Kiel, Germany, and Department of Theoretical Chemistry,
UniVersity of Go¨teborg, Kemigården 3, S-41296 Go¨teborg, Sweden
ReceiVed January 6, 1997X
Abstract: Kinetic measurements as well as B3LYP/ and MP2/6-31G(d,p) calculations provide evidence that carbonyl
oxides formed in the gas-phase ozonolysis of alkylated alkenes are an important source of OH radicals. In the
gas-phase ozonolysis of propene, cis-2-butene, trans-2-butene, tetramethylethene, and isoprene, 18, 17, 24, 36, and
19% OH radicals (relative to reacted ozone, error margin e4%) are measured using CO as a scavenger for OH. The
quantum chemical calculations show that OH radical production depends on syn positioned methyl (alkyl) groups
and their interaction with the terminal O atom of a carbonyl oxide. For example, in the gas-phase ozonolysis of
ethene only 5% OH radicals are measured while for a carbonyl oxide with syn-positioned methyl (alkyl) group, a
much larger amount of OH radicals is formed. This is due to the fact that 1,4 H migration and the formation of an
intermediate hydroperoxy alkene, that is prone to undergo OO bond cleavage, is energetically more favorable than
isomerization to dioxirane. In the case of syn-methyl, dimethyl, and isopropenyl carbonyl oxide calculated activation
enthalpies at 298 K are 14.8, 14.4, and 15.5 kcal/mol compared to the corresponding dioxirane isomerization barriers
of 23.8, 21.4, and 23.0 kcal/mol, respectively. The OO cleavage reactions of the hydroperoxy alkenes formed in
these cases are just 11, 12.8, and 10.3 kcal/mol.
1. Introduction
Identification and quantification of OH radicals generated from
these sources are important to model the chemistry of the
troposphere.
Hydroxyl radicals play an important role in the chemistry of
the atmosphere. They trigger the cyclic process that leads to
the oxidation of NOx to NO2, which is one of the criteria
pollutants of the atmosphere. The daytime atmospheric deg-
radation of volatile organic compounds (VOC) is mainly
initiated by their reaction with OH radicals, the generation of
which in the unpolluted atmosphere is attributed to the reaction
of photogenerated O(1D) atoms with water. A substantial part
of the VOCs, especially alkenes, may also be degraded by
reaction with ozone.1-3 There exists increasing evidence that
OH radicals are formed in the course of the ozonolysis.4-13
Several approaches have been adopted by the different groups
that have worked in this field to estimate OH radical production
in the course of the ozonolysis. Herron and Huie4 as well as
Martinez and Herron5,6 measured product profiles in stopped
flow studies of ozone-alkene reactions using photoionization
mass spectroscopy for the product analysis. Computer simula-
tion of their experimental results leads these authors to suggest
ester, hydroperoxide, and O atom channels, respectively, to be
operative in the reactions. Formation of OH radicals has been
implied as possible step in the ester channel, however, without
quantification. Niki and co-workers7 observed in the ozonolysis
of tetramethylethylene that the stoichiometric ratio r )
∆[alkene]/∆[O3] (∆ is the difference between starting and final
concentration, i.e., ∆ gives the consumption of the compound
in question) decreased from 1.7 to 1.0 when HCHO or CH3-
CHO were added to the system. They proposed that an OH-
initiated oxidation of the alkene was responsible for r > 1. Horie
and co-workers8,9 analyzed the product distribution in the
ozonolysis of propene and trans- and cis-2-butene by FTIR
technique and deduced OH formation yields of 14%, 24%, and
17%. Atkinson and co-workers10 employed cyclohexane as OH
radical scavenger in the ozonolysis of alkenes and terpenes.
These authors reported a value of 27% for OH radicals generated
in the ozonolysis of isoprene. Paulson and co-workers11
evaluated OH formation yields in the ozonolysis of isoprene
using methylcyclohexane as scavenger, reporting a value of 68
( 15%.
† University of Kiel.
‡ University of Go¨teborg.
X Abstract published in AdVance ACS Abstracts, July 15, 1997.
(1) (a) Wayne, R. P. Chemistry of Atmospheres; Clarendon: Oxford,
1991. (b) Leighton, P. A. Photochemistry of Air Pollution in Physical
Chemistry, A series of Monographs; Academic Press: New York, 1961;
Vol. 9.
(2) Manaham, S. E. EnVironmental Chemistry; Lewis Publishers: Chelsea,
MI, 1991.
(3) (a) Bailey, P. S. Ozonation in organic chemistry; Academic Press:
New York, 1978, 1982; Vols. 1 and 2, (b) Kuczkowski, R. L. In 1,3-Diploar
Cycloaddition Chemistry, Padwa, A., Ed.; Wiley: New York, 1984; p 197.
(4) Herron, J. T.; Huie, R. E. Int. J. Chem. Kinet. 1978, 10, 1019-1041.
(5) Martinez, R. I.; Herron, J. T. J. Phys. Chem. 1987, 91, 946-953.
(6) Martinez, R. I.; Herron, J. T. J. Phys. Chem. 1988, 92, 4644-4648.
(7) Niki, H.; Maker, P. D.; Savage, C. M.; Breitenbach, L. P.; Hurley,
M. D. J. Phys. Chem. 1987, 91, 941-946.
(8) Horie, O.; Moortgat, G. K. Atmos. EnViron. 1991, 25A, 1881-1896.
(9) Horie, O.; Neeb, P.; Moortgat, G. K. Int. J. Chem. Kinet. 1994, 26,
1075-1094.
(10) Atkinson, R.; Aschmann, S. M.; Arey, J.; Shorees, B. J. Geophys.
Res. 1992, 97, 6065-6073.
(11) Paulson, S. E.; Flagan, R. C.; Seinfeld, J. H. Int. J. Chem. Kinet.
1992, 24, 103-125.
The most extended study on OH radical production in the
gas-phase ozonolysis of alkenes under tropospheric conditions
has been presented by Atkinson and Aschmann.12 On the basis
of the known chemistry in the system cyclohexane/OH/excess
(12) Atkinson, R.; Aschmann, S. M. EnViron. Sci. Technol. 1993, 27,
1357.
(13) Gutbrod, R.; Schindler, R. N.; Kraka, E.; Cremer, D. Chem. Phys.
Lett. 1996, 252, 221-229.
S0002-7863(97)00050-4 CCC: $14.00 © 1997 American Chemical Society

InVestigation of the Gas-Phase Ozonolysis of Isoprene
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7331
Scheme 1
O2 and the experimentally determined cyclohexane/cyclohexanol
formation yields, the production of OH radicals was derived
with estimated overall uncertainties of a factor of 1.5. While
the work by Atkinson and Aschmann contains a detailed
discussion of the peroxy radical chemistry including the reaction
between HO2 and O3, possible interferences from reactions with
other intermediates, especially with carbonyl oxides and/or
dioxiranes are not addressed in ref 12. Expected sources for
OH formation are only quoted (e.g., from refs 5 and 7) in ref
12; however, no attempt is made to quantify them.
In view of the (partially contradictory) experimental observa-
tions made previously, a redetermination of the OH yields in
the ozonolysis of alkenes and especially of isoprene by new
methods appeared to be in order.
The accepted mechanism for the reaction of ozone with
alkenes is the Criegee mechanism,3 for which the first two steps
are shown in Scheme 1 for the alkenes investigated in this work.
In the first step of the ozonolysis, ozone reacts with alkene by
cycloaddition to yield in a highly exothermic reaction a 1,2,3-
trioxolane, the so-called primary ozonide. Primary ozonide in
turn decomposes into aldehyde and carbonyl oxide, the so-called
Criegee zwitterion.3 While primary ozonides have been isolated
and characterized, there is only indirect evidence for the
intermediary of carbonyl oxides in the course of the ozonolysis.3
Carbonyl oxides are formed in these reactions with large
amounts of internal energy and, therefore, it is likely that they
can lead to the generation of OH radicals during the ozonolysis
of alkenes.12
In recent theoretical work, we have investigated whether
carbonyl oxides can directly decompose to yield OH radicals
and whether such a decomposition reaction can compete with
the rearrangement of carbonyl oxide to its more stable isomer

7332 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Gutbrod et al.
Scheme 2
dioxirane.13 Using high level ab initio methods, we found that
CH2OO (1, Scheme 2) can decompose after H migration and
intermediate formation of an carbene to OH and formyl radicals.
However, the calculated activation enthalpy of this reaction (31
kcal/mol) is 13 kcal/mol higher than the activation enthalpy (18
kcal/mol) for the carbonyl oxide dioxirane isomerization (see
1-7 of paths I and II in Scheme 2). Under normal conditions,
there should be a vanishingly small probability of generating
OH radicals during the gas-phase ozonolysis of ethene.
The activation barrier for H migration becomes smaller than
that for isomerization to dioxirane in the case of alkyl carbonyl
oxides. Cremer14 was the first to point out that syn-alkylcar-
bonyl oxides are internally stabilized by interactions between a
H atom of an R-R2CH (R-RCH2) group and the terminal O atom.
be followed by a subsequent decomposition of the intermediate
hydroperoxide to OH and RCO radicals. Accordingly, OH
generation becomes faster than isomerization of carbonyl oxide
to dioxirane (see, e.g., 8-10 in Scheme 2).
In the present investigation, we provide evidence for the
generation of OH radicals during the gas-phase degradation of
alkenes by ozonolysis with carbonyl oxides being the primary
source of OH radicals. Our evidence results from both
experimental and quantum chemical investigations of the
ozonolysis of ethene (ET), propene (PR), trans-2-butene (TB),
cis-2-butene (CB), tetramethylethene (TME), and isoprene (ISP,
Schemes 1 and 2). ISP is an important biogenic trace constituent
of the atmosphere, being emitted by deciduous trees and other
vegetation leading to concentrations of ∼500 ppb within forest
regions during the summer months.1,2 In the presence of other
pollutants, especially NOx and SO2, terpenes and ISP are
involved in natural smog situations. Ozonolysis of simple
alkenes and of ISP can act as a nighttime source of OH radicals
as shall be described in the present work.
In section 2, we report on results of a kinetic investigation
of the ozonolysis of PR, TB, CB, TME, and ISP and, in this
connection, propose CO as probe to quantify OH radical
H,O interactions in the syn form should facilitate the
migration of the H atom to the terminal O atom, which should
(14) Cremer, D. J. Am. Chem. Soc. 1979,101, 7189.

InVestigation of the Gas-Phase Ozonolysis of Isoprene
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7333
Scheme 3
Table 1. Reactions Considered in the Ozonolysis of Alkenes in
2. Experimental Investigation
the Presence of Excess CO to Scavenge OHa
The quantitative determination of OH radicals generated in
the course of the ozonolysis requires a specific scavenger for
OH that does not react measurably with any stable product or
intermediate of the reaction where, as possible intermediates,
H atoms, O atoms, peroxy radicals, and the Criegee zwitterion
were considered. The scavenging should lead to a characteristic
product that can directly be related to the reaction of the
scavenger and can easily be quantified. Considering known
rates of reactions 1-19 listed in Table 1, we concluded that
carbon monoxide, CO, fulfills these requirements in the systems
under investigation in a most satisfactory way.
CO is known to react with OH radicals to yield CO2 with a
rate constant of k ) 2.3 × 10-13 cm3 molecule-1 s-1 (reaction
2 of Table 1). In air at ambient conditions, the H atoms formed
are immediately converted to HO2 according to reaction 8 of
Table 1. For this conversion at atmospheric conditions, a rate
constant of k ) 7.3 × 10-13 cm3molecule-1 s-1 was measured.15
The overall scavenging reaction is thus governed by the process
reaction no.
reaction
k
(1a)
(1b)
(1c)
(1d)
(1e)
(1f)
(2)
ET + O3 f products
1.75 × 10-18
1.3 × 10-17
1.28 × 10-16
2.2 × 10-16
1.6 × 10-15
1.43 × 10-17
2.3 × 10-13
1.14 × 10-27
1.0 × 10-21
8.2 × 10-12
3.0 × 10-11
5.6 × 10-11
6.4 × 10-11
1.1 × 10-10
1.0 × 10-10
(1.1-1.6) × 10-11
6.8 × 10-14
7.3 × 10-13 b
2.9 × 10-11
∼1.0 × 10-12
2 × 10-12 b
5 × 10-12
PR + O3 f products
CB + O3 f products
TB + O3 f products
TME + O3 f products
ISP + O3 f products
CO + OH f CO2 + H
CO + HO2 f OH + CO2
CO + O3 f products
OH + ET f products
OH + PR f products
OH + CB f products
OH + TB f products
OH + TME f products
OH + ISP f products
OH + aldehydes f products
OH + O3 f HO2 + O2
H + O2 + M f HO2 + M
H + O3 f OH + O2
(3)
(4)
(5a)
(5b)
(5c)
(5d)
(5e)
(5f)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
(18)
(19)
H + alkene f products
R + O2 + M f RO2 + M
R + O3 f RO + O2
HO2 + O3 f OH + 2 O2
HO2 + HO2 f H2O2 + O2
HO2 + RO2 f products
HO2 + OH f H2O + O2
HO2 + aldehydes f products
O + O2 + M f O3 + M
O + alkene f products
CO + OH (O2) f CO2 + HO2
(20)
2 × 10-15
2.9 × 10-11
i.e., CO2 increase in the presence of CO indicates the amount
of OH production due to carbonyl oxide decomposition. The
hydroperoxy radicals generated hereby do not react measurably
with CO to form CO2 (cf., reaction 3, Table 1). They are
removed predominantly by self-reaction and cross-reaction with
other peroxy radicals present in the oxygen-rich system and also
to a smaller extent by reactions with the aldehydes generated
in the decomposition of the primary ozonides. A small fraction
of the HO2 radicals may also react with O3 according to reaction
13 of Table 1 to regenerate OH. The significance of reaction
13 depends on the system studied. Model calculations show
that for CB, TB, and TME an OH yield of <2% and for PR
4 × 10-12
1.11 × 10-10
5 × 10-15
8.4 × 10-15
(2-4) × 10-11
a Rate constants k (in cm3 molecule-1 s-1) from ref 15. R denotes
an alkyl substituent. b At 1 atm with 20 vol % O2.
production in ozonolysis reactions. In section 3, ab initio and
DFT calculations on molecules and transition states 1-64 shown
in Schemes 2 and 3 are presented. In section 4, the results of
the quantum chemical investigations are used to rationalize the
suggested generation of OH radicals in the ozonolysis of alkenes
via an unimolecular dissociation of intermediate hydroperoxide.
Finally, a brief account of existing discrepancies in reported
experimetal OH yields is given.
(15) de More, W. B.; Sander, S. P.; Golden, D. M.; Hampson, R. F.;
Kurylo, M. J.; Howard, C. J.; Ravishankara, A. R.; Kolb, C. E.; Molina,
M. J. Chemical Kinetics and Photochemical Data for Use in Stratospheric
Modeling, EValuation Nr. 10; JPL Publication: Pasadena, 1994; p 92-20.

7334 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Gutbrod et al.
Table 2. Aldehyde and Ketone Yields Obtained in the Ozonolysis
of Ethene (ET), Propene (PR), cis-Butene (CB), trans-Butene (TB),
2,3-Dimethyl-2-butene (TME), and Isoprene (ISP) in the Presence
of 20 vol % O2 with and without Addition of External CO
yields, %a
alkene
PR
product
without CO
with CO
formaldehyde
acetaldehyde
acetaldehyde
acetaldehyde
acetone
formaldehyde
methacrolein
methyl vinyl ketone
62
46
90
94
102
54
60
43
94
95
103
55
CB
TB
TME
ISP
28
21
30
20
a With respect to reacted alkene.
Figure 1. Decrease in stoichiometric ratio r (9) and increase in CO2
yield (4) as function of CO added in the ozonolysis of isoprene. Data
are indicated with symbols; solid lines show the simulation results.
and ISP an OH yield of <4% with respect to O3 consumed can
be expected. Hence, HO2 and its reaction with O3 do not affect
the total OH yields significantly and the increase in CO2 can
be related to the primary OH, i.e., the OH radicals directly
generated in the decomposition of the intermediate hydroper-
oxide.
CO2 + H2, and CO2 + 2H.17 The corresponding reactions for
the parent dioxirane 3 are the processes 22a-22c:
Clearly, rate constants for the reactions of OH radicals with
alkenes or aldehydes (reactions 5 and 6 of Table 1) are higher
than that with CO. Hence, if OH radicals are produced in the
course of the ozonolysis in the absence of a scavenger, they
will produce hydroxy alkyl radicals, which will react with O2
and decompose or stabilize in various ways. However, these
initiation reactions of the OH radicals can largely be suppressed
by adding an excess of CO. Whereas the concentrations of
alkene and ozone in the present study were in the <25 ppm
range, the scavenger CO was added up to 40 vol %.
CO does practically not react at all with any closed-shell
species in the reaction systems investigated. For its reaction
with O3 (reaction 4, Table 1) e.g., the very low rate constant of
k ) 1.0 × 10-21 cm3 molecule-1 s-1 has been determined. Also
its reactions with H atoms and O atoms may be neglected under
present conditions. Both reactions are third order and are
suppressed through steps 8-18 of Table 1.
3 f HCOOH f H2O + CO
3 f H2 + CO2
(22a)
(22b)
(22c)
3 f 2H + CO2
Hence, CO formed in the decomposition of dioxiranes would
be available for a reaction with OH radicals formed in the
decomposition of carbonyl oxides. Thus, it might be argued
that CO2 measured in the absence of external CO might also
be due to the production of OH radicals. However, the reaction
of OH with alkenes (reactions 5a-5f, Table 1) is so much faster
than that with CO that reaction 2 will play only a role when a
large excess of CO is added to the reaction system. The CO2
generated without added scavenger will be formed in different
reaction sequences.
Figure 1 gives the increase in CO2 yields and decrease of
the stoichiometric ratio r as a function of added CO measured
for the system ISP/O3. Equivalent results were received for all
alkenes studied. In each case, the stoichiometric ratio r ) ∆
[alkene]/∆[O3] was determined by in situ spectroscopic mea-
surements. The concentration of reacted alkene was also
measured by GC/MS analysis after complete O3 consumption.
Each point in the diagram is the average of at least three kinetic
runs. The initial concentration [alkene]0/[O3]0 was 2.5 in all
sets.18 The solid lines were obtained by kinetic simulation of
the chemical system with the rate constants compiled in Table
1. It is satisfying to see that the stoichiometric ratio r decreases
from 1.2 without scavenger added to r ) 1.0, which indicates
that secondary ISP consumption by OH radicals is suppressed
by added CO. In ozonolysis experiments in the absence of
oxygen, it was established that O atoms are formed to only 4%
and thus regeneration of O3 according to reaction 18 of Table
1 can be excluded.
Most critical for the present discussion is the question whether
carbonyl oxide might act as O transfer agent and, thereby,
contribute to CO2 generation via the formal reaction
R2COO + CO f R2CO + CO2
(21)
If this reaction were to occur the observed increase in CO2 yield
upon CO addition would not be related to OH scavenging alone
and would require a correction. However, if reaction 21 would
occur to a noticeable extent, the yield in R2CO would also have
to increase. Results compiled in Table 2 clearly show that
aldehyde/ketone yields with respect to alkene consumption do
not increase upon addition of CO and, therefore, side reaction
21 cannot play a significant role in the reaction system.
Furthermore, the consistency of these yields upon addition of
excess CO must be taken as supporting evidence that these major
products are formed directly in the unimolecular decomposition
of the primary ozonides, which is considered to be not affected
by the additive. No other information is available in the
literature on products resulting from a reaction of the Criegee
biradical with CO, except a brief statement by Moortgat and
co-workers16 that no formic anhydride is formed if H2COO is
allowed to interact with excess CO.
Experiments have been carried out with [CO] ranging from
10 to 30%. It was found that in all systems the stoichiometric
ratio decreased to unity when up to 30% CO was used. In the
TB and ISP systems, 20% CO were needed to reach r ) 1.
Further addition of CO did neither decrease the stoichiometric
ratio below unity nor did it increase the CO2 yield. The increase
in the CO2 yield is considered to represent the OH radical yield
in the systems (see Table 3). Noteworthy is that the decrease
In the gas-phase ozonolysis of alkenes, a considerable amount
of intermediately formed carbonyl oxide may isomerize to the
corresponding dioxirane, which can decompose to CO + HOH,
(17) (a) Sander, W. Angew. Chem., Int. Ed. Engl. 1989, 29, 344. (b)
Bunnelle, W. H. Chem. ReV. 1991, 91, 335.
(18) Gutbrod, R; Meyer, S.; Rahman, M. M.; Schindler, R. N. Int. J.
Chem. Kinet., in press.
(16) Moortgat, G. K.; Neeb, P.; Horie, O.; Sauer, F.; Scha¨fer, C.;
Winterhalter, R. Ozonolysis of Alkenes: Recent DeVelopments; Air Pollution
Research Report 57; MPI: Mainz, Germany, 1996; pp 279-286.

InVestigation of the Gas-Phase Ozonolysis of Isoprene
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7335
detected in the gas-phase ozonolysis of ET21 and ample evidence
for the generation of dioxiranes during the ozonolysis has been
collected.17 Hence, it is the general opinion that path I
represents the preferred rearrangement route of carbonyl oxides.
Table 3. Stoichiometric Ratio r ) ∆[alkene]/∆[O3], CO2 Yields,
and OH Yields (with Respect to Reacted O3) in the Ozonolysis of
Propene (PR), cis-Butene (CB), trans-Butene (TB),
2,3-Dimethyl-2-butene (TME), and Isoprene (ISP) in the Presence
a
of 20 vol % O2
Path II is characterized by the migration of a syn positioned
H atom to the terminal O atom via TSII thus forming a
hydroperoxide, which can decompose via TSIII to OH and
formyl radicals. In the literature, there was early speculation
that path II might represent an alternative rearrangement and
decomposition route of carbonyl oxides in the gas phase.22 In
our recent ab initio investigation13 we could clarify that for the
parent carbonyl oxide 1, path II represents a high energy reaction
with an activation enthalpy of 31 kcal/mol (CCSD(T)/6-31G-
(d,p) calculations). The intermediate hydroperoxide 5, which
as a carbene would be rather unstable, does not exist on the
CCSD(T) potential energy surface. Instead, H migration leads
directly to decomposition of 1 into the formyl and OH radicals
7 (Scheme 2).
alkene
PR
p0(CO)
r
CO2 yield, %
OH yield, %
0
20
0
20
0
20
0
30
0
1.2
1.0
1.3
1.0
1.6
1.0
2.2
1.0
1.2
1.0
42
60
40
57
50
74
40
76
22
41
18 ( 4
17 ( 2
24 ( 2
36 ( 2
19 ( 4
CB
TB
TME
ISP
20
a The statistical error in the determination of r was (0.01. The error
in integrated CO2 yields was estimated to be (2%.
in stoichiometric ratio -∆r is within experimental uncertainty
and in no case smaller than the increase in CO2 production.
For PR and ISP, -∆r ≈ ∆[CO2], for CB, TB, and TME;
however, -∆r > ∆[CO2]. On the basis of the experimental
findings reported in this work, no information can be deduced
as to the additional alkene consumption (see also discussion in
section 4). It is expected that statistical calculations describing
the chemically activated systems employing the master equation
approach will provide answers to this question. Preliminary
results19 indicate that OH yields of 100% reported by Atkinson
and Aschmann12 cannot be corroborated by statistical model
calculations.
The OH yields given here are in good agreement with values
for PR (17%), CB (14%), and TB (24%) reported by Horie and
co-workers,9 where it has to be stressed that the OH yield
reported in ref 9 was based on a detailed product analysis by
FTIR spectroscopy. On the other hand, the OH yields sum-
marized in Table 3 are consistently lower than the values
reported by Atkinson and Aschmann (PR, 33; CB, 41; TB, 64;
TME, 100; ISP, 27%),12 Niki and co-workers (TME, 70; ISP,
68 ( 15%),7 and Paulson and co-workers,11 which has to be
considered when discussing the mechanism of OH formation
(see Section 4).
In case of a syn-alkyl-substituted carbonyl oxide with an R-H
atom such as in 8, an intermediate hydroperoxyalkene (see 12)
rather than an unstable hydroperoxy carbene is formed, which
in an endothermic process without energy barrier decomposes
to OH and alkyl radicals such as 14. In the case of dimethyl
carbonyl oxide 15, the activation enthalpy of path II is lower
than that of path I and, therefore, path II should represent the
preferred decomposition route.13
The question is, which of the carbonyl oxides of Scheme 1
prefer to follow path I or path II. To answer this question, we
characterize each carbonyl oxide according to the nature of the
migrating H atom and the number of atoms participating in the
cyclic transition state (TS) of the H migration. As pointed out
above, the syn-positioned H atom (encircled in Scheme 1) in
the parent carbonyl oxide 1 does not migrate easily because (a)
the intermediately formed hydroperoxy carbene is unstable and
(b) the TS involves a four-atom cycle of considerable strain
(C4-type H atom, carbonyl oxide H, 4-membered TS). In the
case of carbonyl oxide 8, there is a syn-methyl H atom that can
migrate via a five-membered cyclic TS of considerably less
strain to the closed-shell system 12 (A5-type H atom: alkyl H
atom, 5-membered TS; Scheme 2). The ozonolysis of PR, CB,
and TB will involve syn- and anti-carbonyl oxides possessing
either C4- or A5-type H atoms (Scheme 1), while the ozonolysis
of TME exclusively leads to carbonyl oxide 15 with a A5-type
H atom (Schemes 1 and 2).
While experimental results indicate a significant amount of
OH radical production during ozonolysis, it was felt that
independent confirmation for the formation of OH radicals in
the reaction systems investigated needed to be obtained. Also
a detailed computational assessment of previously implied
pathways leading to OH formation in the ozonolysis of alkenes
deemed to be in order. For this purpose, we carried out quantum
chemical calculations, which are discussed in the next section.
In the ozonolysis of ISP, the carbonyl oxides 1, 22, 29, 36,
43, 50, 54, 59, and 63 can be formed (Scheme 1). Of these, 1,
22, and 29 possess C4-type H atoms and, therefore, should not
decompose easily to yield OH radicals. However, carbonyl
oxides 36 and 43 have A5 H atoms, which should make them
prone to follow path II. Migration of a methyl H in 54 will
lead to a 6-membered TS and, accordingly, the H atom can be
denoted as A6. It has to be checked whether 54 can also lead
to an intermediate hydroperoxy alkene and, thereby, to a
preference of OH production rather than dioxirane formation.
Carbonyl oxides 50, 59, and 63 possess a vinyl H atom that
might migrate to the terminal O atom via a 5- (V5-) or
6-membered TS (V6-type H atom, Scheme 1) and it has to be
clarified whether V5 or V6 H atoms can as easy migrate as A5
H atoms.
3. Quantum Chemical Investigation
Our previous investigation13 suggests that only those carbonyl
oxides, which possess in syn position a CHR2 or CH2R group,
lead to formation of OH radicals because interactions of an H
atom of CHR2 (CH2R) with the terminal O atom of carbonyl
oxide previously described by Cremer14 in 1979 facilitate the
rearrangement to a hydroperoxy compound which decomposes
to yield OH radicals. In Scheme 1, all possible carbonyl oxides
that can be generated upon decomposition of ET, PR, TB, CB,
TME, and ISP primary ozonide are shown. For each of the
carbonyl oxides of Scheme 1, there are two possible unimo-
lecular rearrangement and decomposition paths I and II, which
are shown in Schemes 2 and 3 (molecules and TSs 1-66). Path
I corresponds to the isomerization reaction of carbonyl oxide
to dioxirane via TSI (Scheme 2).3,17,20,21 Dioxirane (3) has been
(20) Cremer, D.; Gauss, J.; Kraka, E.; Stanton, J. F.; Bartlett, R. J. Chem.
Phys. Lett. 1993, 209, 547.
(21) (a) Lovas, F. J.; Suenram, R. D. Chem. Phys. Lett. 1977, 51, 453.
(b) Suenram, R. D.; Lovas, F. J. J. Am. Chem. Soc. 1978, 100, 5117.
(22) Kafafi, S. A.; Martinez, R. I.; Herron, J. T. In Molecular structure
and energetics. UnconVentional chemical bonding; Liebman, J. F. Green-
berg, A., Eds.; VCH Publishers: New York, 1988; Vol. 6, p 283 and
references given therein.
(19) Olzmann, M.; Kraka, E.; Cremer, D.; Gutbrod, R. J. Phys. Chem.,
in press.

7336 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Gutbrod et al.
Figure 2. B3LYP/6-31G(d,p) geometries for molecules and TSs 8-14. (Distances are in Å; angles, in deg.)
Figure 3. B3LYP/6-31G(d,p) geometries for molecules and TSs 43-49. (Distances are in Å; angles, in deg.)
We have investigated molecules and TSs 1-66 with ab initio
MP2 (second-order Møller-Plesset perturbation theory)23 and
density functional theory (DFT) using the hybrid functional
B3LYP24 and the 6-31G(d,p) basis set.25 The relative stability
of ISP carbonyl oxides and some of their isomers is compared
in Table 4. Calculated reaction energies, energy barriers, zero-
point energy (ZPE) corrections, reaction and activation enthal-
pies at 298 K, and ∆∆Hf°(298) are listed in Table 5. Some
selected geometries of 1-66 are shown in Figures 2 and 3 while
a complete set of calculated geometries is given in the
Supporting Information. Relevant ∆∆Hf°(298) values calculated
for path I and path II are compared in Figures 4 and 5.
Geometries. Cremer and co-workers20 have shown that the
most critical geometrical feature of a carbonyl oxide is the ratio
of OO and CO bond length. While this is largely overestimated
by HF and semiempirical MO methods, low order electron
correlation methods tend to underestimate it.13,20 Reliable values
of the geometrical parameters of carbonyl oxides are obtained
by CCSD(T)/TZ+2P calculations (1, OO, 1.355; CO, 1.276 Å),
the results of which are surprisingly well reproduced at the DFT-
B3LYP/6-31G(d,p) level of theory (1, OO, 1.343; CO, 1.266
Å13). It has been shown13,20 that MP2/6-31G(d,p) (1, OO, 1.293;
CO, 1.298 Å) only provides reasonable starting values for
CCSD(T) or B3LYP geometry optimizations, and, therefore,
we will refer in the following exclusively to B3LYP/6-31G-
(d,p) geometries and energies.
(23) Pople, J. A.; Binkley, J. S.; Seeger, R. Int. J. Quantum Chem. Symp.
1976, 10, 1.
(24) (a) Becke, A. D.; J. Chem. Phys. 1993, 98 , 5648. (b) Stevens, P.
J.; Devlin, F. J.; Chablowski, C. F.; Frisch, M. J. J. Phys. Chem., 1994, 98
11623. (c) Bauschlicher Jr., C. W.; Partridge, H. Chem. Phys. Lett. 1995,
240, 533.
(25) Hariharan, P. C.; Pople, J. A. Chem. Phys. Lett. 1972, 66, 217.

InVestigation of the Gas-Phase Ozonolysis of Isoprene
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7337
Table 4. Relative Energies (kcal/mol) of ISP Carbonyl Oxides and
the Corresponding Hydroperoxides (B3LYP/6-31G(d,p)
Calculations, Compare with Schemes 2 and 3)
ISP carbonyl oxides
ISP hydroperoxides
no. substituent s/aa Me atb vinylc ∆Ed no. ∆Ee
∆Ef
22 2-propenyl
29 2-propenyl
36 Methyl,vinyl
43 Methyl,vinyl
50 2-propenyl
54 2-propenyl
59 Methyl,vinyl
63 Methyl,vinyl
a
a
s
s
s
s
a
a
DB
DB
COO
COO
DB
DB
COO
COO
c
t
c
t
c
t
t
9.1 26
5.9 33
14.0
14.3
34.2
31.3
0
2.3
g38
25.3
13.8
g33
1.5 40 -12.6
47 -8.8
6.4 52 g20
0
7.9 56
2.2 61
1.9 65 g20
6.3
0.5
c
a Syn (s) or anti (a) position of 2-propenyl or methyl substitutent
with regard to COO unit. b Position of methyl group at CC double bond
(DB) or COO unit. c Cis (s) or trans (t) position of vinyl group with
regard to COO unit. d Relative to most stable ISP carbonyl oxide.
e Relative to corresponding ISP carbonyl oxide. f Relative to most stable
ISP hydroperoxide.
1.361:1.272 (8, Figure 2) to 1.370:1.276 (15), indicating that
the hyperconjugative and steric effects of the methyl group(s)
support more the zwitterionic form of carbonyl oxides. In the
case of 8, the syn form is more stable than the anti form which
has been discussed by Cremer.14 Extension of the 4π system
of 1 by a vinyl group as in the ISP carbonyl oxides 22, 29, 36,
43, 50, 54, 59, and 63 (Scheme 1) leads to a planar 6π electronic
system characterized by a typical decrease of the OO/CO bond
length ratio (43 in Figure 3, 1.359:1.291) and a short C1C2
bond (1.443 Å, Figure 3). The geometry of a ISP carbonyl
oxide is also influenced by the hyperconjugative effect of the
methyl group where this effect is stronger when the methyl
group is located at the COO unit rather than at the double bond.
Electronic and steric effects of methyl or vinyl group are also
reflected by the calculated TSI and dioxirane geometries. Since
basic features of the geometry of dioxiranes and TSI have been
discussed elsewhere,14,20,26,27 we refrain from a detailed discus-
sion of calculated geometrical features and instead refer to
Figures 2 and 3 as well as the Supporting Information. As for
TSII, Figures 2 and 3 reveal that the migrating H atom takes a
position above the heavy atom plane typical of the TS of a 1,4-
sigmatropic H shift, i.e., almost in the middle between the C
atom of the methyl group and the terminal O atom. In the case
of TS 11, calculated distances H1C2 and H1O2 are 1.339 and
1.366 Å while the corresponding parameters for 46 are 1.355
and 1.323, respectively. In TS 11, the CC bond is reduced from
1.471 to 1.405 Å while the CO and OO bond lengths increase
from 1.272 and 1.361 to 1.303 and 1.409 Å. Similar changes
are calculated for 46, i.e., the geometrical features of the
hydroperoxide can already be recognized in TSII, namely the
establishment of a CC double bond and a formal C(sp2)O single
bond. After the migration atom H1 keeps its position above
the heavy-atom plane since hydroperoxides prefer the gauche
conformation because this facilitates lone pair delocalization
into the OO bond.28,29
Figure 4. B3LYP/6-31G(d,p) reaction and activation enthalpies
∆∆Hf°(298) for reaction paths I and II in the case of carbonyl oxide
(1), syn-methylcarbonyl oxide (8), and dimethylcarbonyl oxide (15).
Encircled numbers correspond to the molecules and TSs shown in
Scheme 2.
Relative Stability of ISP Carbonyl Oxides. In Table 4,
the relative energies of ISP carbonyl oxides 22, 29, 36, 43, 50,
54, 59, 63, and some of their isomers are given. They can be
understood in terms of conjugative, hyperconjugative, H bonding
and steric interactions. Considering the two resonance formulas
a and b, it becomes clear that an alkyl group at the COO unit
has a stabilizing hyperconjugative effect, which explains the
Figure 5. B3LYP/6-31G(d,p) reaction and activation enthalpies
∆∆Hf°(298) for reaction paths I and II in the case of ISP carbonyl
oxides 22, 29, 36, and 43. Encircled numbers correspond to the
molecules and TSs shown in Scheme 2.
(26) Seung-Joon, K.; Schaefer, H. F.; III; Kraka, E.; Cremer, D. Mol.
Phys. 1996, 88, 93-104.
(27) Kraka, E.; Konkoli, Z.; Cremer, D.; Fowler, J.; Schaefer, H. F.; III
J. Am. Chem. Soc. 1996, 118, 10595.
(28) Cremer, D. J. Chem. Phys. 1978, 69, 4440.
(29) Cremer, D. In The Chemistry of Functional Groups, Peroxides; Patai,
S., Eds. Wiley: New York, 1983; p 1.
Methyl substitution as in 8 (one CH3) or 15 (two CH3 groups)
increases the OO/CO bond length ratio from 1.343:1.266 (1) to

7338 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Gutbrod et al.
corrected by using projected UMP2,31 then the artificial minima
5, 26, and 33 vanish (Table 5).
In the case of the two methyl-substituted carbonyl oxides 8
and 15, the activation enthalpies for H migration are 14.8 and
14.4 kcal/mol, respectively, which is 9 and 7 kcal/mol lower
than the corresponding activation enthalpy for dioxirane forma-
tion (9, 23.8; 16, 21.4 kcal/mol, Table 5 and Figure 4). A
second methyl group has little effect on H migration while it
stabilizes TSI (21.4 compared to 23.8 kcal/mol). The interme-
diate hydroperoxy alkenes 12 and 19 are about 4 and 6 kcal/
mol higher in energy than the corresponding dioxiranes, which
means that they are also rather unstable species. The decom-
position of 12 and 19 (Figure 4) is an endothermic process
without a barrier and with an reaction enthalpy of just 11.0 and
12.8 kcal/mol, respectively.
In the case of carbonyl oxides 36 and 43, the activation
enthalpy for H migration (15.5 and 16 kcal/mol) is 7 and 8.5
kcal/mol smaller than the activation enthalpy for isomerization
to dioxirane (23 and 24 kcal/mol, Table 5, Figure 5) depending
on whether one considers the conformation with the vinyl group
in cis or trans position to the CO group. The trans form 40 of
the two possible hydroperoxybutadienes 40 and 47 is more stable
by 3.6 kcal/mol than the cis form, which reflects the confor-
mational stability of trans- and cis-butadiene and suggests that
both 36 and 43 rearrange to 40. Hydroperoxybutadiene 40 and
dioxirane 45 possess about the same relative stabilities (-12.3
and -14.2 kcal/mol, Figure 5). Decomposition of 40 to yield
42 requires 10.3 kcal/mol thus indicating that the OO bond of
the hydroperoxide is even weaker as in the case of 12 (11.0
kcal/mol) and 19 (12.8 kcal/mol, Figure 4). We note that in
cases such as 47 or 49 rapid rotation into the more stable forms
40 and 42 is most likely. However, in the present context,
determination of the most stable hydroperoxide or aldehyde form
is not of relevance.
larger stability of 36, 43, 59, and 63. Attractive interactions
between a positively charged H atom and the negatively charged
terminal O atom add to the stability of 36, 43, 50, 54, 59, and
63 where the closer approach of H and O atom explains the
higher stability of 50 compared to 54 and 63 compared to 59.
Finally, 4-electron destabilizing effects are responsible for the
lower stability of 22 compared to 29 and 36 compared to 43
(Table 4). Hence, 43 is the most stable of all carbonyl oxides
possibly generated in the ozonolysis of ISP because it has (a)
a methyl group located at the COO carbon (hyperconjugative
stabilization of resonance structure a), (b) the methyl group is
in syn position so that steric attraction between two of the CH3
H atoms and the terminal O atom becomes possible, and (c)
the vinyl group is in trans position thus avoiding steric repulsion
with the CO bond.
Clearly, attractive interactions between a methyl or vinyl H
atom and the terminal O atom facilitate the migration of H and
the formation of a hydroperoxide. However, it plays also an
important role whether the hydroperoxide to be formed repre-
sents a closed- or open-shell electron system. In the case of
ISP carbonyl oxides 22 and 29, the unstable singlet carbenes
are formed thus leading to relatively high lying TSI. For ISP
carbonyl oxides 50, 54, 59, and 63, H migration leads to the
singlet biradicals 51, 55, 60, and 64, respectively (Scheme 3),
which are difficult to calculate, but all of which should represent
high-energy forms (Table 4). We have checked this by
calculating the isomeric forms 52, 56, 57, 61, and 65, which
contain either a highly strained cyclopropyl ring or the allene
unit. Formation of 56 or 57 out of carbonyl oxide 54 is
endothermic by 6.3 and 7.1 kcal/mol, respectively, and the
corresponding TS energies are >34 kcal/mol, which excludes
H migration as an important mechanistic factor in these cases.
The situation is even worse in the case of 52 and 65, which
contain the highly strained cyclopropene ring30 [strain energies
40.9 (left) and 55.2 kcal/mol (right)]:
4. Discussion
On the basis of the results of the quantum chemical
calculations we can predict that carbonyl oxides with a A5-
type H atom prefer to rearrange to hydroperoxyalkenes followed
by OO bond cleavage to yield OH radicals (path II) rather than
to dioxiranes (path I). This is in qualitative agreement with
the experimental observation (see Table 3) that the OH yield
increases with the number of methyl substituents providing H
atoms that can interact with the terminal O atom of carbonyl
oxide via an A5-type arrangement of bonds. Although there
were early speculations in the literature that path II might play
a role in carbonyl oxide reactions,22 we provide in this work
the first evidence that path II is indeed the preferred decomposi-
tion path. It is also a novel finding that the OO bond of a
hydroperoxy alkene is so extremely labile that when possessing
an excess energy of just 10-12 kcal/mol, it immediately cleaves.
These results explain a number of experimental findings, which
so far were not well understood:
(1) In the gas-phase ozonolysis, so far only dioxirane 3 has
been observed.21 Our calculations suggest that dioxirane
formation is preferred in the case of the ET ozonolysis, however
not necessary, in the case of the ozonolysis of higher alkenes.
(2) In the ozonolysis of higher alkenes, dioxiranes were never
observed.3,17 This was always considered as a result of the fact
that aldehyde-carbonyl oxide recombination to yield final
ozonides is much faster than dioxirane formation. We note that
hydroperoxide formation with subsequent reactions at the CC
double bond or OO bond cleavage is probably more likely than
dioxirane formation in most cases.
Even in the case of the allene 61, the barrier for H migration is
too high to have a significant amount of 63 following path II
rather than path I (dioxirane formation, Scheme 3). Hence, for
the ISP carbonyl oxides 50, 54, 59, and 63 (Scheme 3) it was
sufficient to investigate just the path I.
Reaction Barriers and Energetics. The energy data col-
lected in Table 5 reveal that carbonyl oxides with a C4-type H
atom will prefer isomerization to dioxirane (path I) rather than
H migration and decomposition (path II, Scheme 2). Calculated
activation enthalpies for carbonyl oxides 1, 22, and 29 are close
to 31 kcal/mol while the activation enthalpies for path I are
just 20.4 (CCSD(T) calculations: 17.7 kcal/mol20), 24.1, and
20.3 kcal/mol, respectively (Table 5, Figures 4 and 5). In none
of these cases could calculational proof for the existence of an
intermediate hydroperoxycarbene be obtained. An artificial
local minimum is only found at the UMP2 and UDFT levels of
theory because in these cases the calculated energies for TSIII
and formyl + OH radical are too high because of spin
contamination (Table 5). If spin contamination errors are
(30) Wiberg, K. B. Angew. Chem. 1986, 98, 312.
(31) Chen, W.; Schlegel, H. B. J. Chem. Phys. 1994, 101, 5957.

InVestigation of the Gas-Phase Ozonolysis of Isoprene
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7339
Table 5. Calculated Absolute and Relative Energies for Path I and II Reactionsa
method/basis set
reaction of 1
MP2/6-31G(d,p)
energy
carbonyl oxide
TSI
dioxirane
TSII
hydroperoxide
TSIII
radical + OH
∆E
ZPE
∆∆Hf°(298)
∆E
ZPE
-189.068 65
25.0
17.8
24.3
21.5
17.9
20.4
19.2
17.7
-34.6
19.5
-33.7
-24.1
19.7
-23.5
-25.6
-24.7
33.7
15.8
31.0
34.9
15.7
31.6
34.0
30.8
10.0
17.7
9.5
10.8
17.7
9.9
8.3b
16.9
7.8
12.1
16.4
<9.9
-11.1b
13.0
-14.5
-0.2
13.0
6.6
-7.8
-11.2
18.4
0
DFT/6-31G(d,p)
-189.579 89
18.9
0
∆∆Hf°(298)
CCSD(T)/[5s3p2d/3s2p]
∆E
-189.320 44
∆∆Hf°(298)
0
reaction of 8
MP2/6-31G(d,p)
∆E
ZPE
∆∆Hf°(298)
∆E
ZPE
-228.272 97
28.0
35.1
27.2
24.9
35.1
23.8
-32.6
36.4
-31.8
-21.0
36.7
16.5
33.8
14.0
17.5
33.8
14.8
-22.0
35.5
-21.9
-16.4
36.1
2.1b
30.9
-2.0
7.3
30.8
-5.1
35.6
0
DFT/6-31G(d,p)
-228.916 04
35.9
0
∆∆Hf°(298)
-20.4
-16.1
reaction of 15
MP2/6-31G(d,p)
∆E
ZPE
∆∆Hf°(298)
∆E
ZPE
-267.471 85
24.7
51.9
23.8
22.5
52.0
21.4
-33.3
53.1
-32.9
-19.8
53.4
15.7
50.5
13.2
17.0
50.6
14.4
-20.4
52.4
-20.3
-13.6
53.1
3.3b
47.8
0.3
9.5
47.7
-0.3
52.5
0
DFT/6-31G(d,p)
-268.245 86
52.8
0
∆∆Hf°(298)
-19.4
-13.1
reaction of 22
MP2/6-31G(d,p)
∆E
-305.428 45
26.8
54.8
26.1
25.2
55.4
24.1
-32.4
56.0
-31.7
-18.1
56.7
10.4
54.7
10.0
14.0
55.2
13.3
11.9
54.5
10.9
15.3
54.2
13.5
ZPE
55.2
54.5
∆∆Hf°(298)
∆E
ZPE
0
DFT/6-31G(d,p)
-306.316 39
34.9
53.0
31.6
2.4
51.6
-1.4
56.2
0
∆∆Hf°(298)
-17.8
reaction of 29
MP2/6-31G(d,p)
∆E
-305.433 94
23.4
55.1
22.8
21.4
55.5
20.3
-32.1
56.1
-31.5
-18.1
56.8
31.0
52.9
28.3
34.3
53.3
31.1
10.7
54.9
10.4
14.3
55.4
13.6
10.0
55.9
10.6
15.9
54.3
ZPE
55.4
∆∆Hf°(298)
∆E
ZPE
0
DFT/6-31G(d,p)
-306.321 46
3.2
51.9
-0.5
56.3
0
∆∆Hf°(298)
-17.8
-11.4
reaction of 36
MP2/6-31G(d,p)
∆E
-305.439 51
26.8
54.7
25.8
25.3
55.2
24.0
-29.8
55.8
-29.5
-14.2
56.4
-20.4
55.1
-20.4
-12.6
56.3
7.3b
52.8
5.5
10.5
51.2
-2.0
ZPE
55.3
54.7
∆∆Hf°(298)
∆E
ZPE
0
DFT/6-31G(d,p)
-306.328 43
18.3
53.9
15.5
56.1
0
∆∆Hf°(298)
-14.1
-12.3
reaction of 43
MP2/6-31G(d,p)
∆E
ZPE
∆∆Hf°(298)
∆E
ZPE
-305.443 04
26.1
55.1
25.4
24.1
55.6
23.0
-28.7
55.9
-28.4
-14.3
56.5
17.0
53.5
14.4
18.9
54.0
16.0
-16.9
55.1
-17.0
-8.8
56.2
10.3b
52.7
8.3
13.3
51.1
0.3
55.5
0
DFT/6-31G(d,p)
-306.330 86
56.3
0
∆∆Hf°(298)
-14.2
-8.7
a Absolute energies in hartree, relative energies in kcal/mol. DFT/B3LYP calculations. CCSD(T) energies from ref 13. b Values are obtained
with PMP2/6-31G(d,p)//UMP2/6-31G(d,p). The relative UMP2/6-31G(d,p) energies are 31.2, -8.7, -3.6 (reaction of 1); 10.1, 6.0, 2.9 (reaction of
8); 10.0, 7.0, 6.4 (reaction of 15); 35.8, 36.4, 14.2 (reaction of 29); 15.4, 6.1 (reaction of 36); 18.8, 8.8 (reaction of 43).
(3) Sander and co-workers32 have recently reported the
formation of dimesitylcarbonyl oxide 67 in CFCl3/BrF2CCF2-
Br at -70 °C as the first carbonyl oxide in solution. Upon
raising the temperature to -50 °C the alcohol 68 is formed as
an unexpected rearrangement product (reaction 23).
calculations that in its equilibrium geometry 67 possesses one
methyl H atom close to the terminal O atom and, accordingly,
formation of an intermediate hydroperoxy derivative RR′COOH
with some biradical character via path II is likely. Decomposi-
tion of RR′COOH into ketone RR′CdO and OH radical in the
solvent cage and recapture of OH by the RR′CdO radical would
lead directly to the observed product 68.
(4) The formation of addition products of carbonyl oxides in
the presence of protic nucleophiles could involve an internally
formed hydroperoxide since nucleophilic attack at the carbonyl
oxide C atom probably causes a proton to migrate and, thereby,
facilitates formation of a hydroperoxide. Addition reactions at
the double bond could explain a number of byproducts of the
ozonolysis.3
Kraka and co-workers33 have shown on the basis of B3LYP
(32) Kirschfeld, A.; Muthusamy, S.; Sander, W. Angew. Chem., Int. Ed.
Engl. 1994, 33, 2212.
(33) Kraka, E.; Sosa, C. P.; Cremer, D. Chem. Phys. Lett. 1996, 260,
43.
As for the generation of OH radicals, three factors have to
be considered: (a) The ratio of the activation energies for path
I and path II as calculated in this work, (b) The amount of excess

7340 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Gutbrod et al.
energy produced when carbonyl oxides are formed in the course
of the gas phase ozonolysis, and (c) dissipation mechanisms of
the excess energy during the gas-phase ozonolysis.
There is also the possibility that carbonyl oxide is formed in
an electronically excited state where the lowest is just about 20
kcal/mol above the ground state and corresponds to a triplet
state.36 In this state, carbonyl oxide can easily decompose to
aldehyde and O(3P), which would represent an alternative
reaction path provided spin-orbit coupling and the chance of
intersystem crossing would be large. However, calculations
suggest that probably less than 5% carbonyl oxide follows this
reaction path.36 Apart from the unimolecular reactions of
carbonyl oxide there is also, according to our calculations, the
chance of a bimolecular reaction with O2, which adds at the C
atom of the carbonyl oxide thus forming the triplet biradical
OOCH2OO. The biradical can rearrange to dioxirane and O2
providing in this way another route to the typical dioxirane
oxidation products CO2, CO, etc. Although the addition of O2
to carbonyl oxide involves a (calculated) barrier of just 0.7 kcal/
mol, rate calculations reveal that the bimolecular reaction cannot
compete with the unimolecular decomposition routes of carbonyl
oxide.19
In the case of the ozonolysis of PR, CB, TB, or ISP, about
equal amounts of carbonyl oxides that prefer to rearrange to
dioxirane or to decompose to yield OH radicals should be
formed. If one expects about 36% of the latter to follow path
II as found in the case of the ozonolysis of TME, a OH yield
close to 20% should be expected for the ozonolysis of these
alkenes as is found in the kinetic investigation.
It should be pointed out that the predominant goal of the
quantum chemical calculations presented here is to firmly
establish that carbonyl oxides generated in the primary steps of
the alkene ozonolysis are the most important source of OH
radicals. It is suggested to label them as primary OH radicals.
Depending on the experimental techniques applied, secondary
OH radicals can also be encountered in an ozonolysis experi-
ment. At least two sources for secondary OH radicals have to
be considered:
(i) In the decomposition of intermediate hydroperoxides 12,
19, 40, and 47 (Scheme 2) vinoxy-type radicals are formed,
which are calculated to be primarily C centered. These radicals
will readily react with O2 as has been in shown in the case of
14.37 Also, formation of OH radicals38,39 as well as of glyoxal
and formaldehyde37 has been reported. No information is
available for the methyl-substituted homologues; however, an
increased efficiency for secondary OH formation can be
predicted.
(ii) Stoichiometric ratios r > 1 imply secondary alkene
consumption through OH attack. In the case that no OH
scavenger is present, the most likely reaction of OH is that with
alkene according to reaction 24 thus forming a radical which
can easily react with O2 present in the reaction system (see
reaction 25):
Apart from this one has to consider which carbonyl oxides
will preferentially be formed upon primary ozonide decomposi-
tion. Cremer34 has predicted that in the ozonolysis of CB more
anti will be produced; however, in the ozonolysis of TB more
syn methyl carbonyl oxide is formed. This is insofar confirmed
as the OH radical yield in the TB ozonolysis (Table 3) is
significantly higher than in the CB ozonolysis. Similar differ-
ences in OH yields have been reported before;9,12 however, no
explanation has been offered to account for this observation.
In the case of the ISP ozonolysis ozone can attack at the methyl-
substituted double bond or the unsubstituted double bond. From
rate comparisons,3 one knows that ozone attack at a methyl-
substituted double bond is always faster, which has to do with
the fact that electron-rich double bonds (double bonds with a
higher π-HOMO) form a stronger π-complex with ozone and,
therefore, are more prone to undergo the cycloaddition reaction
to primary ozonide. Hence, in the ISP ozonolysis carbonyl
oxides 36, 43, 59, and 63 should preferentially be formed where
36 and 43 will lead to more OH production and 59 and 63 to
more dioxirane production.
During the gas-phase ozonolysis of ET, primary ozonide is
formed with an excess enthalpy of 59 kcal/mol (B3LYP/6-31G-
(d,p)) measured relative to the enthalpy of the TS of the ozone-
ET cycloaddition reaction.34,35 Decomposition of ET primary
ozonide is endothermic by 6.6 kcal/mol, which means that both
1 and formaldehyde are formed with an excess enthalpy of about
52 kcal/mol. Since methyl substitution enhances the stability
of both carbonyl oxide and aldehyde, the amount of excess
energy increases with the number of methyl groups attached to
the alkene double bond. In the case of the TME ozonolysis,
the excess enthalpy of acetone and carbonyl oxide 15 is 82 kcal/
mol, which reflects the fact that the decomposition of TME
primary ozonide is an exothermic step (∆∆Hf° -20.6 kcal/
mol).35
One can conclude that for the ozonolysis of both PR, CB,
TB, TME, and ISP the excess enthalpy is so large that, despite
their different stability, all carbonyl oxides shown in Scheme 1
can be formed in the form of hot molecules. The more
vibrational degrees of freedom a carbonyl oxide possesses the
higher should be its chance to stabilize before a subsequent
reaction and the higher the chance of stabilization the more the
difference between path I and path II barriers will matter since
carbonyl oxide will become more selective.
From these considerations, we conclude that the largest
amount of OH radicals should be generated in the ozonolysis
of TME, which is in agreement with the experimental observa-
tion. In the TME ozonolysis, just carbonyl oxide 15 is formed,
which has a larger chance of stabilization than either 8 or 1.
Hence, 15 is more selective than 8 or 1 and will preferentially
follow path II that leads to the formation of hydroxyl radicals.
Statistical calculations describing the chemically activated
systems employing the master equations suggest that at 1013
mbar 20% stabilized carbonyl oxide can be expected in the case
of the ET ozonolysis; however, ∼40% can be expected in the
case of the TME ozonolysis.19 Furthermore, calculations predict
only a small amount of 1 to lead to OH formation; however,
about 40% of 15 do so, which may be considered as a
verification of the qualitative arguments concerning energy
dissipation in molecules with a relatively large excess energy.
On the other hand, one has to note that also vibrationally excited
15 largely prefers path II because of the lower barrier.19
CHRdCHR + OH f •CHRsCH(OH)R
•CHRsCH(OH)R + O2 f •OOCHRsCH(OH)R
(24)
(25)
(26)
•OOCHRsCH(OH)R f HOOCHRsCH(O•)R
HOOCHRsCH(O•)R f HO + •OCHRsCH(O•)R
•OCHRsCH(O•) f 2RHCdO
(27)
(28)
H migration (reaction 26) and decomposition of the hydroper-
(36) Anglada, J. M.; Bofill, J. M.; Olivella S.; Sole´, A. J. Am. Chem.
Soc. 1996, 118, 4636.
(34) Cremer, D. J. Am. Chem. Soc. 1981, 103, 3619, 3627, 3633.
(35) Andersson, S.; Cremer, D. To be published.
(37) Zhu L.; Johnston, G. J. Phys. Chem. 1995, 99, 15114.
(38) Gutman, D.; Nelson, H. H. J. Phys. Chem. 1983, 87, 3902.

InVestigation of the Gas-Phase Ozonolysis of Isoprene
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7341
oxide thus formed (reaction 27) should be similar to the carbonyl
oxide rearrangements investigated in this work. The total
process CHRdCHR + OH f 2RHCdO would be exothermic
by ∼30 kcal/mol and would be characterized by the fact that
primary OH starting the reaction sequence (reactions 24-28)
would be replaced by secondary OH from reaction 27 thus
initiating a chain reaction. The consumption of alkene would
be larger than expected from the ozone-alkene reaction and,
as a consequence, the OH yield measured relative to alkene
consumption would increase.
Hence, experimentally determined OH yields using a scav-
enger may be larger than the true OH yield in the ozonolysis
depending on the efficiency of the chain (reactions 24-27)
relative to the efficiency of scavenging primary OH or any of
the chain carriers.
Scavenging experiments have been carried out by Niki and
co-workers using HCHO and CH3CHO,7 by Atkins and co-
workers using cyclohexane,10,12 by Paulson and co-workers
employing methylcyclohexane,14 and in this work using a large
excess of CO. Some experiments with HCHO and with CO,
respectively, have also been reported by Horie and co-workers.9
Although it is beyond the scope of the present investigation to
rationalize differences in the action of these scavengers, it is
safe to say that the additional chemistry introduced into the
ozonolysis system by scavengers such as aldehyde7,9 or CO (this
work and ref 18) will be less complex than using cyclohexane
or its methyl derivative. A comparison of reported OH yields
as given by the different research groups is also hampered by
the fact that some authors relate OH yields to alkene consump-
tion10,12 while others relate it to O3 consumption.7,9,18 A direct
comparison requires that secondary processes as outlined in ii
can be neglected. For example, Horie and co-workers9 pointed
out that OH yields of 0.64 as reported by Atkinson and
Aschmann12 for the TB ozonolysis correspond to an overall OH
formation yield and may contain contributions from secondary
formation(s).
Figure 6. Schematic illustration of experimental setup used in the
ozonolysis experiments.
with White mirror systems for in situ absorption measurements
in the UV and in the IR, using optical path lengths of 8 m and
of 40 m, respectively. The setup also permitted direct sampling
for product analysis by HPLC or GC/MS as schematically
outlined in Figure 6. As standard bath gas a 20% O2/He mixture
was used. Various amounts of the OH scavenger CO were
added to the system while the O2 concentration was always kept
at 20%. Ozone was prepared by microwave discharge in pure
O2 at (20-30) mbar and collected in a cold trap over silica. It
was transported into the reactor by a flow of bath gas. Its
concentration in the reactor was determined by absorbance
measurements at 254 nm using ꢀ(O3) ) 1155 × 10-20 cm2
The OH yields reported here have been obtained in ozonolysis
experiments, in which CO was present up to 50% (see Figure
1), i.e., CO acted as both a radical scavenger and effective
collision partner suppressing secondary alkene consuming
processes. The high concentrations of the additive were needed
to secure that the stoichiometric ratio r ) ∆[alkene]/∆[O3] had
truly reached its limiting value of r ) 1.00 ( 0.02. Secondary
reactions were precluded under these conditions.
molecule-1 11
The alkene was added as the last reactant to the
.
gas mixture. It was transported into the reactor from calibrated
volumes by He. To achieve rapid mixing an effective teflon
stirrer was used during the filling operation. Concentrations
used were [O3] 8.3-9.6 ppm and [alkene] 20.8 -24.1 ppm. He
(purity >99.999 vol %), O2 (purity >99.998 vol %), and CO
(purity >99.999 vol %) from Messer-Griessheim were used.
He and CO were passed through a 7 m long molecular sieve
13× column to remove traces of impurities in particular CO2.
ET, PR, CB, TB, TME, and ISP were reagent-grade chemicals
from Aldrich. They were degassed and these vapors diluted
with He were stored in darkened flaks.
5. Conclusions
In a two-pronged approach involving both kinetic measure-
ments and quantum chemical calculations, we have shown that
substituted carbonyl oxides with a A5-type H atom in syn
position prefer to rearrange to hydroperoxy alkenes, which have
a labile OO bond and, therefore, decompose to yield OH
radicals. Using CO as scavenger for OH radicals in ozonolysis,
we could detect 18, 17, 24, 36, and 19% OH radicals in the
ozonolysis of PR, CB, TB, TME, and ISP. Since ISP is a trace
constituent of the atmosphere with a concentration of about 500
ppb, the ozonolysis of this compound can be considered as an
important nighttime source of OH radical production in the
atmosphere. Future work has to clarify under which conditions
OH radical production due to carbonyl oxide decomposition
becomes particularly critical.
Computational Methods. In previous work,13 we have used
many-body perturbation theory (MBPT), coupled-cluster (CC)
theory, and density functional theory to find a reliable but
economic method that provides a fairly accurate description of
carbonyl oxides. This method turned out to be DFT with
Becke’s three-parameter hybrid functional B3LYP and a 6-31G-
(d,p) basis set. Accordingly, we have used in the present work
second-order Møller-Plesset perturbation theory (MP2)23 with
the 6-31G(d,p) basis25 to achieve reasonable starting geometries
and reference energies, which then were utilized for energy,
geometry, and frequency calculations at the B3LYP/6-31G(d,p)
level of theory.24
6. Experimental and Computational Details
(40) Kraka, E.; Gauss, J.; Reichel, F.; Olsson, L.; Konkoli, Z.; He, Z.;
Cremer, D. COLOGNE 94, Go¨teborg, 1994.
(41) Frisch, M. J.; Head-Gordon, M.; Trucks, G. W.; Foresman, J. B.;
Schlegel, H. B.; Raghavachari, K.; Robb, M. A.; Binkley, J. S.; Gonzalez,
C.; Defrees, D. J.; Fox, D. J.; Whiteside, R. A.; Seeger, R.; Melius, C. F.;
Baker, J.; Martin, R. L.; Kahn, L. R.; Stewart, J. J. P.; Topiol, S.; Pople, J.
A.; Gaussian 92; Gaussian Inc.: Pittsburgh, PA, 1992.
Experimental Details. All reactions were carried out at
atmospheric pressure in a 70 L duran glass reactor equipped
(39) Lorenz, K.; Rha¨sa, D.; Zellner, R.; Fritz, B. Ber. Bunsen-Ges. Phys.
Chem. 1985, 89, 341.

7342 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Gutbrod et al.
Zero-point energy (ZPE) and temperature corrections for T
) 298 K were calculated by scaling harmonic MP2/6-31G(d,p)
frequencies by a factor of 0.930 and the corresponding B3LYP/
6-31G(d,p) values by 0.963. Calculations have been performed
with COLOGNE9440 and GAUSSIAN 92.41 The calculated
energies and geometries of all molecules considered in this work
are given in the Supporting Information.
(NFR). All calculations were done on the CRAY YMP/J932 of
the Rechenzentrum Kiel and on the CRAY YMP/464 of the
Nationellt Superdatorcentrum (NSC), Linko¨ping, Sweden. E.C.
and D.C. thank the NSC for a generous allotment of computer
time.
Supporting Information Available: The calculated energies
and geometries of all molecules considered in this world (12
pages). See any current masthead page for ordering and Internet
access instructions.
Acknowledgment. This work was supported at the Univer-
sity of Kiel by the CEC within the Environmental Research
Programme, contract EV5V-CT91-0038 and at the University
of Go¨teborg by the Swedish Natural Science Research Council
JA970050C