Temperature dependent kinetics of the OH/HO2/O3 chain reaction by time-resolved IR laser absorption spectroscopy

Article abstract of DOI:10.1021/jp9934984

The catalytic odd hydrogen (HO_x) O₃ cycle, OH + O₃ → HO₂ + O₂ and HO₂ + O₃ → 2 O₂ is one of the most important atmospheric processes leading to the natural destruction of atmospheric O₃. This cycle is active at lower stratospheric altitudes (20-30 km, 190-230 K) in the mid-latitudes. The HO_x chain reaction is estimated to be responsible for ~ 50% of the global O₃ loss in the stratosphere. This makes it important to have accurate knowledge of the temperature dependent rate constants k₁ and k₂ for reliable modeling of atmospheric O₃ phenomena. An extensive temperature dependent kinetic study of the catalytic HO_x O₃ cycle based on time-resolved, Doppler limited direct absorption spectroscopy of OH with a single mode (Δν = 0.0001/cm) high-resolution IR laser was presented. The sum of the two rate constants, k₁ + k₂, was measured at 190-315 K and can be accurately described by an Arrhenius-type expression (k₁ + k₂(cc/sec) = 2.26(40) x 10^-12 exp[-976(50)/T]. These results agreed excellently with studies by Ravishanka et al. and Smith et al but were higher than the values currently accepted for atmospheric modeling. These studies reflected the first of such rate measurements to access the 190-230 K range relevant to kinetic modeling of O₃ chain loss in the lower stratosphere.

Full text of DOI:10.1021/jp9934984

3964
J. Phys. Chem. A 2000, 104, 3964-3973
Temperature Dependent Kinetics of the OH/HO2/O3 Chain Reaction by Time-Resolved IR
Laser Absorption Spectroscopy
Sergey A. Nizkorodov, Warren W. Harper, Bradley W. Blackmon, and David J. Nesbitt*
JILA, National Institute of Standards and Technology and UniVersity of Colorado, and Department of
Chemistry and Biochemistry, UniVersity of Colorado, Boulder, Colorado 80309-0440
ReceiVed: September 30, 1999; In Final Form: January 11, 2000
This paper presents an extensive temperature dependent kinetic study of the catalytic HO
OH + O f HO + O and (2) HO + O f OH + 2 O , based on time-resolved, Doppler limited direct
absorption spectroscopy of OH with a single mode (∆ν ) 0.0001 cm ) high-resolution infrared laser. The
sum of the two chain rate constants, k + k , is measured over the 190-315 K temperature range and can be
accurately described by an Arrhenius-type expression: k
x
ozone cycle, (1)
3
2
2
2
3
2
-
1
1
2
3
-12
1
+ k
2
(cm /s) ) 2.26(40) × 10 exp[-976(50)/T].
These results are in excellent agreement with studies by Ravishankara et al. [J. Chem. Phys. 1979, 70, 984]
and Smith et al. [Int. J. Chem. Kinet. 1984, 16, 41] but are significantly higher than the values currently
accepted for atmospheric modeling. In addition, these studies also reflect the first such rate measurements to
access the 190-230 K temperature range relevant to kinetic modeling of ozone chain loss in the lower
stratosphere.
I. Introduction
The so-called catalytic odd hydrogen (HOx) ozone cycle,1,2
and 5. Subsequent measurements of k1 have also been obtained
via laser induced fluorescence by Smith and co-workers (T )
7
2
40-295 K) and laser magnetic resonance methods by Zahniser
8
and Howard (298 K), with results in substantially good
OH + O f HO + O
(1)
(2)
3
2
2
agreement with the higher values of Ravishankara et al.6
HO + O f OH + 2 O
2
It is important to note at the outset that although these
previous studies nominally obtain values of k1 by analysis of
OH removal as a single exponential process, this is not strictly
correct for a two center chemical chain reaction. Indeed, as
2
3
is one of the most important atmospheric processes leading to
the natural destruction of atmospheric ozone. This cycle is
particularly active at lower stratospheric altitudes (20-30 km,
9
described in a previous kinetic analysis, the loss of OH is
rigorously given by (i) a single exponential induction decay
characterized by the sum of the two chain rates (k1 + k2)
followed by (ii) a steady-state propagation regime where OH
is maintained at a constant level ≈ k2/(k1 + k2)[OH]0 by chain
reactions 1 and 2. However, since this induction rate is typically
dominated by the first step (i.e., k1 >> k2), direct observation
of these finite, steady-state OH levels in the propagation regime
requires especially high signal to noise methods. Such methods
have been recently demonstrated in our laboratory, based on
shot noise limited direct absorption of a single mode laser to
1
90-230 K) in the mid-latitudes. Indeed, the HOx chain reaction
is estimated to be responsible for nearly half of the global ozone
loss in the stratosphere. This makes it especially important to
3
have accurate knowledge of the temperature dependent rate
constants k1 and k2 for reliable modeling of atmospheric ozone
phenomena.
Reactions 1 and 2 have been studied by a number of research
groups. The first direct experimental measurement of k1 was
4
performed in 1973 by Anderson and Kaufman, who reported
this rate constant between 220 and 450 K. These studies reacted
H atoms with NO2 to prepare OH radicals in a conventional
flow-tube apparatus, with the HOx chain reaction monitored as
a function of time (i.e., the distance for a well-characterized
9
detect the OH radicals. Exponential analysis of the fast chain
induction kinetics yields a room temperature rate sum of k1 +
-
14
3
k2 ) 8.4(8) × 10
(cm /molecule)/s, which is in excellent
6
2
agreement with the 300 K data of Ravishankara et al.
flow velocity) via resonance OH fluorescence on the A Σ r
2
5
X Π band. Shortly thereafter, Kurylo performed room tem-
perature studies of k1 using pulsed UV laser photolysis of O3
Measurements of the HO2 + O3 reaction rate constant have
proven more challenging since k2 is at least an order of
magnitude smaller than k1 under usual stratospheric and
laboratory temperatures (200 K e T e 400 K), and thus any
OH formed in reaction 2 quickly regenerates HO2 via reaction
1. One approach to decouple the effects of reaction 1 is to use
an efficient scavenger which reacts quickly with OH but is inert
to HO2. This method has been successfully demonstrated by
1
followed by the fast reaction of O( D) photofragments with H2
to generate OH, which again was detected via resonant
6
fluorescence. In a later study by Ravishankara et al., k1 was
measured via similar flash photolysis/resonance fluorescence
methods over a temperature range of 238-357 K. Of particular
relevance to the present work, however, these authors were first
to point out the profound complications caused by vibrationally
excited OH on the observed kinetics (see below). The results
reported by Ravishankara et al. indicate k1 to be systematically
higher (by up to 40%) compared with the earlier data of refs 4
8
10
11
Zahniser and Howard, Manzanares et al., and Wang et al.,
who used either C3H8 or C2F3Cl for chemical removal of OH.
1
2
One very clever method by Sinha et al. exploits isotopically
labeled H O2 in excess O3 reagent with laser magnetic
1
8
16
1
0.1021/jp9934984 CCC: $19.00 © 2000 American Chemical Society
Published on Web 04/04/2000

Kinetics of the OH/HO2/O3 Chain Reaction
J. Phys. Chem. A, Vol. 104, No. 17, 2000 3965
be measured over a substantially extended range of O3 densities
compared to previous studies of the reaction. Under these
conditions of increased O3 density, the chain nature of the HOx
cycle, which had proven difficult to detect unambiguously in
previous kinetic studies of the reaction, becomes experimentally
9
quite important. (iv) IR detection allows the probing of a much
broader range of rovibrational levels of OH, without interference
2
from rapid predissociation effects in the upper A Σ (V g 2 and/
or high N) states of OH. (v) Most importantly, this broad tuning
capability permits direct monitoring of vibrationally excited OH-
(
V,N) populations, and thus quantitative characterization of these
important rovibrational relaxation processes. In our previous
publication (hereafter noted as paper I), direct IR absorption
9
methods have been successfully applied to the measurement of
k1 + k2 at 298 K. The main thrust of this paper is to extend
these measurements to the lower temperature domain relevant
to atmospheric modeling of the stratosphere. A recipe for
separating this result into individual components (k1 and k2) is
provided at the end of the paper, and the next efforts in this
laboratory will concentrate on this challenging task.
Figure 1. Schematic cut through the temperature-controlled flow cell
(top view). The gas inlet and outlet manifolds are designed such that
the gas flows across the cell through multiple entry and exit points,
which serves to reduce temperature and concentration gradients along
the length of the cell. Surrounding the reaction chamber is a cooling
jacket through which a temperature-controlled fluid is pumped. Circles
indicate the placements of a baratron and two thermocouple gauges
for in situ measurements of the gas pressure and temperature. The
bottom drawing shows how the thermocouples are positioned (side
view).
The method and apparatus are essentially the same as those
used previously, with suitable modifications to the flow cell
for achieving variable, low-temperature flow conditions. The
reaction sequence is initiated by a short pulse of UV radiation
from an excimer laser. This pulse photodissociates a small
18
resonance detection of the H O2. Via this detection scheme,
neither the 18OH nor OH product of reaction 1 is able to
1
fraction of O3 in the reaction flow chamber into O( D), which
16
rapidly abstracts hydrogen atoms from H2O (or H2 for the low-
temperature studies) to form OH. The nascent rovibrational-
state distribution of OH generated during this reaction sequence
18
16
regenerate H O2 via reaction with O3, thus isotopically
breaking the chain cycle. The combination of these measure-
ments has yielded results over a 233-450 K range, but which
show upward curvature in the Arrhenius plot at the lowest
temperatures. Results obtained by different groups are in fairly
is generally quite “hot”, particularly when H2 is used as the
20-23
reactant.
However, under favorable conditions, equilibration
by collisions with the buffer gas is rapid compared to the
characteristic time scales for reactions 1 and 2. Thus, formation
of OH can be regarded for all practical purposes as essentially
instantaneous, leading to substantial simplifications in the kinetic
analysis. The time dependent densities of OH (and HO2) are
monitored in real time via high-resolution direct absorption
spectroscopy, which is capable of fully resolving each vibra-
tional, rotational, spin-orbit, and lambda-doublet state. The rate
constants are then obtained from fitting the observed time
dependence of OH densities to an appropriate kinetic model
representation.
1
1
good agreement particularly at room temperature and below.
In spite of this wealth of experimental studies,4-12 the values
of k1 and k2 in the temperature range relevant to the lower
stratosphere (190-230 K) are still partly based on extrapola-
tions. For example, the value of k1 used currently in stratospheric
modeling13 represents an average of results obtained in refs 4-8.
6
Ravishankara et al. determined k1 in the 238-357 K temper-
ature window, but their attempts to reach lower temperatures
were precluded by strong kinetic interference from vibrationally
excited OH radicals. Though the lowest temperatures (223 K)
4
studied by Anderson and Kaufman sample the top of this
II. Experimental Section
stratospheric window, their results at higher temperatures are
unfortunately not in good agreement with the data of Ravis-
The experimental strategy is described in detail in paper I,
and, except for a description of the modifications required for
temperature dependent measurements, only a brief summary is
6
hankara et al. Similarly, k2 has been determined down to 233
K, but extrapolation to lower temperatures is even more
1
1,12
9
problematic due to strong curvature in the Arrhenius plots
given here. The heart of the experimental apparatus is a newly
mentioned above. In summary, there is a need for extension of
the available experimental data on OH/HO2/O3 chemical chain
reactions down to temperatures characteristic of the lower
stratosphere.
The lack of reliable low-temperature data has prompted us
to use alternative methods based on high-resolution infrared (IR)
designed temperature controlled flow cell shown in Figure 1.
The reaction chamber is a tube, 106 cm long, 2.5 cm in diameter,
with a volume of 0.52 L. The windows into the chamber are
made of IR grade quartz optics, with a transmittance limited
-1
by surface reflection to >90% in both the near-IR (≈3700 cm )
and UV (≈308 nm). Each entrance to the reaction chamber has
two such windows separated by an evacuated space, which
serves to protect the inner optics from atmospheric water vapor
condensation at reduced temperatures. The gas mixture enters
and leaves the reaction chamber through the gas inlet/outlet
manifold system that runs along the entire length of the tube
(Figure 1). The net direction of the flow is across the reaction
chamber as opposed to the previous design where the gas flows
along the tube. The main advantage of such an arrangement is
that fresh gas flows into each area of the chamber on a pulse
by pulse basis, with the temperature and density gradients along
the tube reduced to a minimum.
direct absorption spectroscopy to study the kinetics of reactions
9
1
and 2. There are several important advantages in direct
absorption versus fluorescence detection of OH. (i) The method
is quite general and equally applicable to detection of both HO2
and OH radicals, which are the key players in the HOx ozone
cycle. (ii) IR absorption line strengths are accurately known
for both species;1
4-19
thus, the measurements can easily be
converted into absolute number densities of each radical species.
iii) IR radiation in this range is not absorbed by O3 as opposed
(
to the near-UV light at around 308 nm used to probe OH in the
fluorescence-based methods. This permits the rate constant to

3
966 J. Phys. Chem. A, Vol. 104, No. 17, 2000
Nizkorodov et al.
TABLE 1: Rate Constants Determined in This Studya
cross-section
The gas inlet/outlet systems and the reaction chamber are
surrounded by a cooling jacket, which contains a flowing fluid
with a controlled temperature. For the present experiments,
ethanol, n-pentane, and water are used as the cooling fluids,
depending on the desired temperature range. The fluid is
circulated by a 32 L/min pump between the cooling jacket of
the flow cell and a copper heat-exchanger coil, immersed in a
constant temperature bath. The temperature of the gas in the
reaction chamber is measured by two K-type thermocouples in
the center and at the edge of the cell (Figure 1). The
thermocouples and the thermometer are periodically tested
against temperature standards and found to be accurate to <1
K over the 180-350 K temperature range. For all of the
experiments reported here, the two thermocouple readings in
the cell are consistent both with each other and with the
temperature of the circulating fluid to within ∆T ≈ 1 K (1.5 K
for T < 210 K).
O
3
k
1
-14
+ k
2
-
19
2
3
mixture
T (K)
(×10 cm )
(×10 cm /s)
a
a
a
b
a
b
a
b
a
b
b
b
b
b
b
b
b
b
b
b
b
314.0
295.0
287.0
277.1
273.0
266.9
266.0
256.4
253.0
250.0
248.3
240.0
239.9
234.8
223.4
217.7
212.6
207.1
201.8
195.8
190.6
1.487
1.350
1.310
1.271
1.257
1.240
1.237
1.215
1.208
1.203
1.200
1.188
1.188
1.181
1.170
1.165
1.161
1.158
1.155
1.152
1.150
10.9
8.94
8.28
5.96
6.52
5.20
6.20
4.44
5.16
4.60
4.07
3.99
3.77
3.67
3.18
2.10
2.25
1.86
1.90
1.51
1.62
Ozone is produced from oxygen prior to the experiments with
a commercial corona discharge ozone generator and trapped on
silica gel beads at about 170 K. Ozone is delivered into the cell
by passing a regulated flow of Ar through the trap warmed to
190-220 K. Apart from the O3/Ar mixture, two other gases
a
The first and the second columns specify the flow cell mixture (a
) Ar/H O/O ; b ) CO /H /O ) and temperature for a given experi-
flow into the reaction chamber, one containing the OH precursor
molecule (H2 or H2O) and the other a buffer gas. The partial
flows and thus the partial pressures of the combined gases
entering the reaction chamber are measured and controlled with
a set of calibrated flow meters. The choice of the buffer gas
and OH precursor molecule depends on the temperature range
of interest. For T > 245 K, the OH precursor is formed by
passing a known flow of Ar through a porous water bubbler,
with Ar also as the buffer gas diluent. Under typical experi-
mental conditions the cell can contain up to 3 Torr of O3 and
up to 5 Torr (1 Torr ) 133.32 Pa) of H2O at a total pressure of
2
3
2
2
3
mental run. The fourth column provides the resulting rate constant, k
1
+
k
2
, and the third column contains O
3
absorption cross-sections at
27
3
08 nm used to deduce the O
3
density from absorption measurements.
Estimated 95% confidence intervals amount to (1 K for temperature
(1.5 K at T < 210 K), 15% for the rate constants, and (4% for the
UV cross-section ((10% for T ) 314 K).
(
O
3
at eleVated temperatures (T > 298 K) are based on extrapolations
of data from ref 27 and are thus somewhat less accurate (10%)
than the cross-sections at room temperature and below (4%).
As a second independent test, the density of O3 in the flow is
2
0-30 Torr, most of which is the Ar buffer. At lower
additionally verified by absorption of a mercury UV lamp at
temperatures, where the H2O vapor pressure becomes too low,
Ar/H2O is replaced by H2, while CO2 is used as the buffer gas.
CO2 is chosen by virtue of its relatively efficient quenching of
OH(V,N) states (the quenching rate constants increase nearly
30
2
53.7 nm before the gas is injected into the flow cell. The O3
cell densities measured at these two wavelengths agree quan-
titatively after correcting for the temperature and pressure
difference in the flow cell and Hg lamp cell (see paper I for
more details).
-
13
exponentially with vibrational quantum number from 2 × 10
3
-11
3
24,25
cm /s (for V ) 1) to 5 × 10
cm /s (for V ) 9)
and its
The temporal profile of the OH radical density is probed via
chemical inertness to both OH and HO2 chain radicals. For these
lower temperature experiments, the flow cell normally contains
31
absorption of a continuous single mode F-center laser (FCL)
tunable between 2.5 and 3.3 µm, with a line width of 2 MHz.
The FCL radiation is split into several separate beams, one of
which is directed through the flow cell to a signal InSb detector,
another is focused onto a reference detector, and the rest is used
for laser beam diagnostics. Both detectors are carefully imped-
ance-matched and have ≈400 ns rise times. The centers of
individual rovibrational transitions of OH are located to better
8
-10 Torr H2 and 9-14 Torr CO2 in addition to the O3/Ar
mixture, again at a total pressure of 20-30 Torr. The total flow
into the 0.52 L reaction chamber is normally of the order of
2
-3 (Torr‚L)/s, which means that the chamber is completely
refreshed every 5 s (25-50 excimer laser pulses at the 5-10
Hz repetition rate of the experiment). With <2 (mJ/cm )/pulse
2
of UV photolysis laser the chain reaction loss of ozone due to
prior pulses in the flow cell is insignificant.
-1
than 0.002 cm accuracy using a home-built traveling Mich-
elson interferometer (i.e., λ-meter) constructed after the design
3
2
Because of the important role the absolute concentration of
ozone plays in the determination of the rate constant, the O3
density is always measured via in situ absorption of the XeCl
excimer UV radiation used to dissociate O3. This method has
the advantage of excellent dynamic range and readily permits
by Hall and Lee. With the F-center laser centered on top of
an OH absorption feature, the time dependent absorbance is
measured as a transient imbalance between the signal and
reference detector outputs. This difference is digitized by an
oscilloscope and averaged for 500-2000 UV laser pulses. About
10-20% of 2500 sampled time points occur before the UV laser
pulse to determine the zero level. The noise density in the
1
5
17
-3
the determination of [O3] in the range of 10 -10 cm
through the 106 cm long flow cell. The UV absorbance can be
converted into absolute density using O absorption cross-
3
-9
-1/2
resulting trace is (6-9) × 10 Hz
at about 20 µW of IR
sections2
6-28
for T ) 195-298 K. To account for small
laser power; this is equivalent to an absorption sensitivity of
-5
variations in the O3 absorbance over the spectral output of the
excimer laser, the ozone UV cross-section is taken as a weighted
average of the two major XeCl emission lines at 307.9 and 308.2
<10 in the 1 MHz detection bandwidth. At room temperature
for a Doppler broadened peak, this results in a detection
sensitivity of e10 cm for the total OH density (or e109
1
0
-3
2
9
-3
nm. Table 1 provides the resulting cross-sections for each
relevant temperature point. Note that the cross-section values
cm
per quantum state). Depending on the O3 density,
temperature, and UV laser power (nominally kept below 2 mJ/

Kinetics of the OH/HO2/O3 Chain Reaction
J. Phys. Chem. A, Vol. 104, No. 17, 2000 3967
2
cm /pulse to avoid secondary radical-radical processes), peak
quencher for vibrationally excited OH (hereafter denoted simply
-
4
-3
34,35
OH absorbances of 10 to 5 × 10 are routinely observed,
which translates into a peak signal-to-noise ratio of ≈10-500.
as OH*).
This rapid vibrational relaxation is essential for
reliable kinetic analysis and interpretation since OH* tends to
3
6,37
react with O3 much faster than the ground-state OH.
-
10
3
11
III. Kinetic Analysis
Reaction 5 is so fast (k (298K) ) 2.2 × 10
cm /s) that
5
vibrational relaxation of the resulting OH represents by far the
rate-limiting step for the experimental appearance of thermalized
OH absorption and must occur in the shortest possible time.
A detailed kinetic treatment of the HOx chain reaction
appropriate for the present experimental conditions has been
described in paper I; here we only provide a brief summary of
9
(iii) H2O is completely inert toward OH, with the exception of
this analysis along with necessary modifications for the low-
temperature studies. The analysis is based on the following
kinetic conditions being satisfied. (i) The processes leading to
initial generation of OH are very rapid compared to the rates
of reactions 1 and 2 and can be approximated by an instanta-
neous rise of OH density to [OH]0 at t ) 0. Furthermore,
rovibrational relaxation with the buffer gas is sufficiently fast
to maintain OH(V,N) in a thermal distribution of quantum states.
the rather high-barrier, irrelevant symmetric exchange reaction
H2O + OH T OH + H2O. Due to the combination of these
favorable factors, OH can be produced and fully equilibrated
under typical experimental conditions in less than 3-6 µs, which
is 1-3 orders of magnitude smaller than the time scale of the
chain induction period (100 µs to 20 ms). Thus, most kinetic
measurements at T > 245K (including the room temperature
study of paper I) have been performed with H2O/O3/Ar flow
using reaction 5 to prepare OH.
(
ii) Total radical density is negligibly small compared to that
of O3 and other stable species, i.e., [OH]0 , [O3], which allows
treatment of reactions 1 and 2 as pseudo-first-order processes
and neglect of radical-radical processes, at least on the fast
time scales of the chain induction. (iii) On a much longer time
scale, radicals OH and HO2 undergo slower irreversible decay
due to secondary reactions, diffusion from the laser probe region,
and deactivation on the walls, which can be simply ap-
proximated by a first-order decay. Under these conditions, one
can solve analytically for [OH(t)] as a simple double exponential
decay9
At sufficiently low cell temperatures, however, the H2O vapor
pressure becomes vanishingly small, thus requiring other
methods of OH generation. In this study we make use of reaction
6
1
, which has a nearly gas kinetic rate constant of k6 ) l.0 ×
-
10
3
11
0
cm /s at room temperature. The disadvantage with this
approach is that it results in an extremely “hot” nonequilibrium
rovibrational distribution of OH. Specifically, reaction 6 pro-
2
0-23
duces OH with significant vibrational population
)
up to V
4 and rotational excitation up to N > 25, which can take
considerable time to relax in the absence of an efficient OH*
3
8
3
2
quencher such as H2O. To complicate matters further, the H
atom product of reaction 6 undergoes fast secondary reaction
[
OH](t)
k
1
k
2
kind[O3]t
e-ktermt
)
e-
+
(3)
[
OH]0
k1 + k
k1 + k
2
2
-11
3
13
with O3 [k7(298K) ) 2.9 × 10
cm /s],
where the chain induction rate constant is given by kind ) k1 +
k2 and the chain termination rate constant,
H + O f OH(V e 9) + O
(7)
3
2
which is known from previous studies to generate a nascent
OH
irr
HO2
irr
k
k2
k
k1
39-41
OH product state distribution
some indications of a minor channel
peaking at V ≈ 7-9, with
kterm.
)
+
(4)
42,43
k1 + k2 k1 + k2
for forming OH(V)0-
3) with extremely high rotational energies (N g 33). This highly
depends on the irreversible first order decay rates for OH and
excited OH rovibrational distribution necessarily requires
extensive energy transfer with the flow cell gas mixture in order
to reach thermal equilibrium. Thus, any time dependent kinetic
probe of HOx chain kinetics monitored via OH(V)lr0) absorp-
tion can be contaminated by time dependent cascade relaxation
processes, population in the upper state OH(V)1), and premature
formation of HO2 by vibrationally accelerated reactions of OH-
(V) with O3.
As an explicit example of such behavior, Figure 2a demon-
strates the temporal evolution of OH absorbance signals in an
Ar/H2/O3 mixture monitored on specific V ) 3 r 2, V ) 2 r
1, and V ) 1 r 0 rovibrational transitions, which demonstrates
several interesting effects. Most dramatically, stimulated emis-
sion is initially observed on the V ) 1 r 0 transition, which
indicates the presence of a vibrational population inVersion
between OH(V)0) and OH(V)1) states at t < 50 µs. This
emission strength initially grows with time, reflecting an increase
in OH(V)1) vs OH(V)0) populations due to cascade relaxation
from higher lying vibrational states. Eventually, however, this
situation is reversed, and the absorbance signal becomes positive,
eventually rising to a maximum level some 220 µs after the
laser pulse and then slowly decaying due to competition between
cascade relaxation and the HOx chain cycle. For the V ) 2 r
1 transition, no stimulated emission is observed and the signal
reaches a maximum much sooner than for the V ) 1 r 0
transition, with subsequent decay due to reactive removal by
OH(V)1) + O3 and cascade relaxation. It is interesting to note
-1
HO2. The chain induction period is defined by τind ) (kind[O3]) ,
which is equal to the 1/e time of the induction process at a
given O3 density. Once again, since k1 . k2, the second term
in eq 3 is normally much smaller than the dominant induction
term; thus the double exponential decay behavior has been
commonly ignored in previous kinetic analyses of the OH +
O3 reaction rates.
The validity and relative importance of these kinetic assump-
tions for the present study will be examined in more detail
throughout the next sections. Here only the conditions necessary
for a swift generation of thermally equilibrated OH are
discussed. The primary chemical initiation step following the
O3 photolysis and resulting in generation of OH involves rapid
1
reaction between O( D) and a hydrogen containing species; e.g.,
1
O( D) + H O f OH + OH
(5)
(6)
2
1
O( D) + H f OH + H
2
As explained in detail in paper I and in the study of Ravishan-
6
kara et al., based on a similar chain initiation sequence, water
is the optimal choice of OH precursor for several important
3
3
reasons. (i) The degree of OH vibrational excitation from
reaction 5 is relatively modest (V e 2), which limits the extent
of subsequent relaxation necessary to achieve thermalized kinetic
conditions. (ii) In addition, H2O is an extremely efficient

3968 J. Phys. Chem. A, Vol. 104, No. 17, 2000
Nizkorodov et al.
Figure 3. Sample OH transient absorbance recorded on the V ) 1r0
-
2
P
1
1
(2.5) transition (UV power ) 0.9 (mJ/cm )/pulse, [O
5
3
] ) 8.0 ×
1
0 , T ) 196 K). The solid line inside the experimental data points
represents a fit to the double exponential decay model described in the
text. The inset displays the line shape of this transition, best described
by a Voigt profile with roughly comparable Doppler and pressure
broadened components. The peak absorbances (0.17%) at early times
correspond to a total density (summed over all quantum states) of
+(2.5)
1
Figure 2. Time dependence of OH absorbances monitored on R
12
-3
+
-
[OH]total ≈ 4 × 10 cm
.
V ) 3r2, R
transitions at [O
pulse, T ) 243 K. (a) Ar/H
1
(2.5) V ) 2r1, and P (2.5) V ) 1r0 rovibrational
1
1
6
-3
2
3
] ) 3.1 × 10 cm , UV photolysis ) 2.3 mJ/cm /
17 -3
2 3 2
/O
mixture ([Ar] ) 6.0 × 10 cm , [H
]
times for the various V ) 3r2, 2r1, and 1r0 signal
components. Most importantly, this makes collisional thermal-
ization of the OH faster than the subsequent chain induction
rate by at least an order of magnitude under all experimental
conditions sampled. This clean separation in time scales
effectively breaks the parameter correlation between the OH-
(V)1r0) rise and fall components, and thereby permits reliable
extraction of the τind decay times from a least squares analysis.
This has been independently tested by comparison measurements
of kind with H2O and CO2 buffers at T > 245 K, which indeed
produce identical results. Conversely, the use of Ar buffer in
this same temperature regime resulted in values of kind system-
atically low by as much as 2-fold, which is consistent with
similar qualitative observations of Ravishankara et al.6
1
7
-3
)
3.6 × 10 cm ). For these mixtures, relaxation of vibrationally
excited OH* is quite slow and occurs on a time scale comparable to
induction of the chain reaction. Negative absorbance (i.e., stimulated
emission) signals observed on the V ) 1r0 transition after the UV
laser pulse indicate vibrational population inversion. (b) CO
2
/H
2
/O
3
1
7
-3
17
-3
mixture ([CO
1
2
] ) 4.8 × 10 cm , [H
2
] ) 3.6 × 10 cm , [Ar] )
1
7
-3
.2 × 10 cm ). Due to the much greater vibrational quenching
efficiency of CO vs Ar buffer gas, the OH* signals are now strongly
2
suppressed, which allows for a clean separation between the OH
equilibration and chain reaction time scales.
that the OH(V)2r1) absorbance signal does not decay com-
pletely to zero, suggesting that either the HOx cycle or one of
the secondary reactions can regenerate vibrationally excited OH-
(
V)l), thereby further complicating the kinetics. The time
evolution of the V ) 3r2 absorbance is similar to the V ) 2r1
transition, except that everything occurs on an even faster time
scale since the probe intercepts the OH(V) cascade relaxation
process at a higher vibrational level.
IV. Results
Figure 3 displays a typical time dependence of the OH
absorbance signal on the V ) 1r0 P1 (2.5) transition recorded
-
Figure 2a demonstrates that in the absence of an efficient
relaxing agent such as H2O or CO2, the vibrational relaxation
of OH prepared via reactions 6 and 7 can take as long as 200-
at T ) 196 K. Under the conditions of Figure 3 ([O3] ) 8.0 ×
1
5
-3
17
-3
17
-3
10 cm ; [CO2] ) 5.l × 10 cm ; [H2] ) 4.4 × 10 cm ;
1
6
-3
2
[Ar] ) 4.5 × 10 cm ; UV power ) 0.9 mJ/cm /pulse) the
OH thermalizes to its maximum signal in about τrise ) 60 µs,
after which it undergoes double exponential decay with an
induction period of the order of τind ) 2 ms. The stimulated
emission signals discussed above are also present in the very
earliest times (t < 10 µs) but are not visible on the scale of
Figure 3. The inset of Figure 3 shows the line shape of the V )
3
00 µs. This may be only a factor of 2-3 different from the
time scale for the chain induction process at high O3 densities,
which can therefore introduce substantial systematic uncertain-
ties in extracting accurate values of τind. In order to eliminate
this problem, the buffer gas for experiments at T < 245 K is
either CO2 or O2. Both of these species (i) have more than an
order of magnitude larger cross-sections for rovibrational
-
-1
1r0 P1 (2.5) transition (centered at 3484.599 cm ) recorded
under such conditions. The line shape has a characteristic Voigt
profile with roughly equal contributions from Doppler and
pressure broadening. All kinetic OH traces reported in this work
are recorded with the probe laser at line center, though tests for
the probe laser shifted to either side of the Voigt profile result
in negligible changes in the time domain signals and/or derived
44,45
relaxation of OH* compared to Ar and H2,
(ii) are
chemically inert to both OH and HO2 chain radicals, and (iii)
have vapor pressure greater than 20 Torr down to e190 K. For
the present low-temperature studies, CO2 is the more preferable
quenching buffer gas simply because it is 2-fold more efficient
than O2 at quenching OH*,2
with H2.
4,25
as well as being nonreactive
rate constants.
From (i) known line strengths for OH transitions,1
experimentally determined line shape parameters, and (iii) a
rotational thermalized Boltzmann distribution (see below), the
time dependent OH absorbance can be directly converted into
4-16
By way of example, Figure 2b demonstrates the result of
replacing Ar by an equivalent pressure of CO2 in the flow cell,
with all other conditions held constant. The OH* relaxation rate
increases significantly, as evident in the greatly diminished rise
(ii)

Kinetics of the OH/HO2/O3 Chain Reaction
J. Phys. Chem. A, Vol. 104, No. 17, 2000 3969
Figure 4. Dependence of the induction rate, τind-1, on the O
density
3
at several selected temperatures. From eq 3, the slopes of such curves
are equal to the sum of the chain rates, k
1
2
+ k .
absolute OH densities. Indeed, this represents one major
advantage of direct absorption IR laser methods for such radical
studies and permits direct verification of pseudo-first-order
kinetic rate conditions. For example, the 0.17% peak OH
absorbance of Figure 3 translates into a total [OH] density
Figure 5. Transient OH absorbances monitored on different V ) 1r0
3
rovibrational transitions originating from JOH ) 1.5-5.5, Ω ) /
2
at
16
-3
2
[O
3
] ) 3.3 × 10 cm , UV power ) 1.5 (mJ/cm )/pulse, and T )
1
2
-3
(
summed over all rotational levels at 196 K) of 4 × 10 cm ,
243 K. Within experimental uncertainty, transitions from the associated
+
-
which is 3-4 orders of magnitude smaller than the O3 density.
The [OH]0/[O3] ratio is determined primarily by the UV laser
power density and is maintained at the 10 -10 level for all
kinetic measurements.
lambda-doublet components [e.g., P
guishable; thus, only one is shown. The R
1
(3.5) and P
1
(3.5)] are indistin-
(4.5) signals
1
(4.5) and Q
1
-
3
-4
have been scaled by 0.67 and 1.70 for clarity. All the traces are parallel
on the logarithmic scale, indicating that rotations are fully equilibrated
on the time scale of the OH/HO /O chain reaction.
2 3
On the basis of the previous kinetic analysis in section III,
the time dependent [OH] can be well-represented by a double
exponential decay; i.e.,
[
OH]/[OH] ) R exp(-t/τ ) + â exp(-t/τterm.) (8)
0 ind
where R ) k1/(k1 + k2) and â ) k2/(k1 + k2). Here, the faster
and much larger decay component is due to induction of the
-
1
chain reaction, τind ) kind[O3] ) (k1 + k2)[O3], whereas the
longer and much smaller decay component represents slow
termination of the chain by radical-radical and diffusive
processes. This functional form does not take into account OH*
rovibrational relaxation at the earliest times; thus, the least
squares fitting process is only started after t g 3τrise, by which
time rovibrational relaxation is >90% complete. The solid curve
on Figure 3 shows the result of such a fit. According to the
-
1
model, τind should be a linear function of [O3], with a slope
Figure 6. Dependence of induction rate (τind-1) on the internal energy
of the OH probe state at T ) 243 K (all other conditions are the same
as in Figure 2b). The middle diagram displays the relevant OH energy
levels. This graph demonstrates the absence of nonequilibrium rotational
or spin-orbit effects on the chain kinetics.
-1
of k1 + k2. By way of example, linear least squares fits of τind
vs [O3] are presented in Figure 4 for several selected temper-
atures. The observed intercepts are indistinguishable from zero
within the experimental uncertainties, and fixing the intercepts
at zero in the fits produces a less than (3% change in the least
squares slopes. This is consistent with a negligibly small rate
of OH (and HO2) deactivation due to the second-order chemical
processes compared to the rate with which a dynamic steady
state is achieved in our OH/HO2/O3 system. Table 1 reports a
summary of k1 + k2 values derived from these least squares
fits over the temperature range of 191-314 K.
such diagnostic tests at T ) 243 K, which sample a series of
2
lower rotational levels JOH ) 1.5-5.5 in the Π3/2 spin-orbit
state of OH(V)0) via different P-, Q-, and R-branch transitions.
The absorbances shown in Figure 5 vary by more than an order
of magnitude over this J range due to a combination of changes
in the line strengths and the lower state populations; the
logarithmic slopes are robustly independent of the state probed.
This is reflected in excellent parallelism of the individual traces
observed, even during the very early stages of the chain reaction.
Thus, one can safely conclude that internal rotational energy
relaxation within the OH manifold is much faster than the OH/
HO2/O3 chain reaction rates under investigation, and monitoring
the kinetics on a single OH rovibrational transition is fully
justified.
The above data rely on spectroscopic detection of OH(V)0)
via direct IR absorption on a single rovibrational transition,
-
-1
typically P1 (2.5) at 3484.599 cm . It is important to demon-
strate an insensitivity to this choice of the probed state, or, stated
differently, that internal energy relaxation within the OH(V)0)
manifold is much faster than reaction. The independence of the
reaction kinetics on the rotational and spin-orbit state probed
has been explicitly verified by monitoring the OH decay times
for a series of 1r0 rovibrational transitions under identical flow
and temperature conditions. Figure 5 illustrates an example of
The last points are further illustrated by Figure 6, which shows
-
1
the dependence of the fitted τind on the internal energy of the

3970 J. Phys. Chem. A, Vol. 104, No. 17, 2000
Nizkorodov et al.
TABLE 2: Simplified Sequence of Chemical Processes
Taking Place in the Flow Cell Used in the Numerical
Simulationa
3
reaction
k(298K) (cm /s)
1
(1∆)
O
O
3
+ hV f O( D) + O
2
80%
20%
3
3
+ hV f O( P) + O
2
1
3
3
-10
O( D) + O
3
3
2
2
f O
f O
2
+ O( P) + O( P)
+ O
1.2 × 10
1
-10
O( D) + O
2
2
1.2 × 10
1
3
-11
O( D) + O
f O( P) + O
O f OH + OH
2
4.0 × 10
1
-10
O( D) + H
2.2 × 10
1
3
-10
O( D) + CO
2
f O( P) + CO
f OH + H
f OH + O
OH + O f HO + O
HO f OH + O
f OH + OH
f H O + H
f OH + O
+ M f HO + M
f O( P) + 2O
f OH + HO
2
1.1 × 10
1
-10
O( D) + H
2
1.0 × 10
-
-
-
-
-
11
14
15
11
15
H + O
3
2
2.9 × 10
8.2 × 10
2.0 × 10
8.1 × 10
6.7 × 10
5.9 × 10
3
2
2
2
+ O
3
2
H + HO
2
OH + H
2
2
3
-11
-32
O( P) + HO
2
2
2
Figure 7. Sample Boltzmann plot for OH( Π3/2,V)0,J) populations
in the temperature controlled flow cell. The slope corresponds to a
rotational temperature of Trot. ) 256 ( 15 K, in agreement with the in
situ thermocouple measurement of Tgas ) 243 ( 1 K.
H + O
2
2
5.7 × 10 [M]
l
3
-15
O
2
( ∆) + O
3
2
3.8 × 10
3
-15
O( P) + H
2
O
2
2
1.7 × 10
3
-12
OH + OH f H O + O( P)
OH + O( P) f H + O
2
1.9 × 10
3
-11
2
3.3 × 10
3
-15
O( P) + O
OH + H
HO + HO
O( P) + O
OH + HO
3
f O
f H
f H
+ M f O
f H O + O
OH + OH + M f H
2
+ O
O + HO
+ O
+ M
2
8.0 × 10
-
-
12
12
2
O
2
2
2
1.7 × 10
1.7 × 10
2
2
2
O
2
2
3
-34
2
3
6.0 × 10 [M]
-10
2
2
2
1.l × 10
-31
2
O
2
+ M
6.9 × 10 [M]
a
Rate constants are taken from refs 9 and 13.
8
00 µs after the photolysis laser pulse and scaled by the
appropriate line strengths so that the vertical scale is proportional
to state-resolved densities. The slope corresponds to a rotational
temperature of Trot ) 256 ( 15 K, which is in agreement with
in situ thermocouple temperature measurements of T ) 243 (
1
K. Since the individual logarithmic time profiles for each
transition in Figure 6 are parallel, this value of Trot. clearly
remains constant throughout the chain reaction. We also note
2
that Trot measured for the other spin-orbit component ( Π1/2)
-
1
Figure 8. Dependence of the induction rate (τind ) on the UV
2
2
and the relative populations between the Π3/2 and Π1/2
manifolds are also consistent with these thermocouple temper-
ature measurements.
16
photolysis laser power measured at T ) 218 K ([O
3
] ) 5.9 × 10
-
3
17
-3
17
-3
cm , [CO
×
2
-3
] ) 5.3 ×10 cm , [H ] ) 4.4 × 10 cm , [Ar] ) 1.7
2
1017 cm ). Higher UV laser powers correspond to higher relative
Finally, especially for such slow rate processes, it is important
to verify the assumption that radical-radical reactions do not
play a significant role. Such radical-radical reactions often
occur with no activation barrier at the gas kinetic rate; thus their
rates can become competitive with the much slower radical-
closed shell chain reactions 1 and 2 even at extremely low
radical to O3 density ratios. To explicitly address this issue, we
have performed both (i) empirical UV power dependent studies
abundance of free radicals in the mixture and hence to higher radical-
-
1
radical reaction rates. The graph shows that τind is independent of
2
the laser power up to 2.5 (mJ/cm )/pulse; thus, nonlinear radical-radical
processes can be ignored under normal experimental conditions of 1-2
2
(
mJ/cm )/pulse.
probed OH level. For this comparison, the flow and temperature
conditions are again held constant while the probe transition is
varied. Figure 6 includes data from both spin-orbit components
(Figure 8) and (ii) numerical kinetic simulations (Table 2 and
2
2
of OH ( Π1/2 and Π3/2) and covers the internal energy range
Figure 9). Based on tabulations of atmospherically relevant
-
1
from EOH ) 0-770 cm . The diagram of the energy levels
probed is shown in the inset. The lambda-doubling splittings
for each J and Ω level are too small to be shown on this scale;
experimentally, however, we also observe no differences in the
magnitude and time dependence kinetics for pairs of lambda-
13
rates, the most important radical-radical processes affecting
the chain kinetic model are found to be
3
O( P) + HO f OH + O
(9)
2
2
-
+
doublet transitions [e.g., P1 (2.5) and P1 (2.5)]. Within experi-
OH + HO f H O + O
2
(10)
2
2
-
1
mental uncertainty, the least squares fitted values of τind are
robustly independent of lower state energy, confirming that that
not only rotational but also spin-orbit and lambda-doublet
relaxation occurs on significantly faster time scales than the
desired chain reaction kinetics.
In effect, reaction 9 interconverts the less reactive HO2 chain
radical back into the more reactive OH. This reaction involves
3
O( P) atoms, which are formed in the initial photolysis of O3
(
20% yield at 308 nm), as well as from subsequent nonreactive
1
The above measurements can also be used to independently
verify rotational and spin-orbit temperatures of the OH
manifold. By way of example, a sample Boltzmann plot for
quenching of O( D) and other processes (Table 2). Reaction
10 represents a HOx chain termination process that becomes
important at sufficiently long times and high radical densities.
However, the combined influence of these radical-radical
2
the OH( Π3/2, V ) 0, J) manifold is shown in Figure 7, measured

Kinetics of the OH/HO2/O3 Chain Reaction
J. Phys. Chem. A, Vol. 104, No. 17, 2000 3971
Figure 10. Arrhenius plot for the experimentally determined k
1
+ k
2
rate data (filled circles). Fits to an Arrhenius expression (not shown)
-12
3
result in a preexponential factor of 2.26(40) × 10 cm /s and activation
energy of 976(50) K, where numbers in parentheses represent 95%
6
confidence intervals. Data of Ravishankara et al. (open circles) and
7
Smith et al. (open triangles) are in good agreement with present results.
4
The fit to the data of Anderson and Kaufman (dashed line) is in
reasonable agreement, but lies consistently 20-25% below the present
1
1
1
results. Values of k recommended for use in stratospheric modeling
(thick solid line) are also 15-20% lower than the present results.
rate constants and (ii) at least one of these rate constants (k2)
3
7,43
exhibits non-Arrhenius behavior.
However, since k1 is at
Figure 9. Influence of UV photolysis laser power on OH/HO
2
/O
3
least an order of magnitude greater than k2 over the temperature
range of the present study (315 to 190 K), the temperature
dependence of k1 + k2 will be primarily determined by k1.
kinetics. (a) Simulated [OH]/[OH]
dash), 1 (long dash), and 5 (dash-dot) mJ/cm /pulse of UV radiation
0
traces at 0.01 (solid), 0.1 (short
2
1
6
-3
17
-3
flux at 298 K ([O
4
3
] ) 5 × 10 cm , [CO
2
] ) 4 × 10 cm , [H
× 10 cm ). The decay shifts from double exponential to
multiexponential due to radical-radical reactions as the UV laser power
2
] )
6
Ravishankara et al. observed excellent linearity in the Arrhenius
1
7
-3
plot for k1 + k2 down to 238 K, which suggests that reaction 1
has normal Arrhenius behavior. The present data also strongly
support this conclusion. Figure 10 displays kind ) k1 + k2
obtained from the present work as a function of inverse
temperature on a semi-logarithmic scale. Within the uncertainty
of the experiment, the data points fall on a line corresponding
and therefore radical density are increased. (b) Induction rate (τind
derived by fitting the simulated traces to a double exponential decay.
)
2
Under 1-2 mJ/cm /pulse conditions, τind is obtained to within 5%
uncertainty. (c) Fitted â/R ratios for slow and fast decay components.
2
Even for 1-2 mJ/cm /pulse conditions, the â/R ratio from the numerical
2 1
simulation deviates from k /k by as much as a factor of 2-3.
-
12
3
to a preexponential factor of A ) 2.26(40) × 10
cm /s and
activation energy of Eact/R ) 976(50) K, where numbers in
channels on the fast induction part of the OH decay is quite
small. This independence on radical density can be experimen-
tally verified; this is clearly demonstrated in Figure 8 for
parentheses represent 95% confidence intervals.
The present results can be compared with several previous
temperature dependent chain reaction studies. Excellent agree-
-
1
15
-3
measurements of τind at a fixed [O3] ) 5.9 × l0 cm , with
6
ment is observed with the data of Ravishankara et al. and data
2
UV photolysis energy varied from 0.3 to 2.5 mJ/cm /pulse. By
7
of Smith et al., which are almost indistinguishable from the
way of additional confirmation, numerical kinetic simulations
including radical-radical processes demonstrate a completely
negligible influence (<5%) on the fast OH decay rates for UV
photolysis pulse energies in this regime (Figure 9b). All rate
constants reported in this work have been performed in the 1-2
present data within the experimental uncertainties (Figure 10).
When one considers that the present experiment and the LIF
studies of refs 6 and 7 are based on fundamentally different
detection principles, the agreement is outstanding. On the other
4,46
hand, data of Anderson and Kaufman appear to lie systemati-
2
mJ/cm /pulse range.
cally below the present results by ≈20-25%. Table 3 sum-
marizes the activation energies, preexponential factors, and 298
K values for k1 + k2 obtained by different groups. Inspection
of Table 3 shows that the preexponential factor obtained by
V. Discussion
Temperature dependence of rate constants is often character-
ized in terms of Arrhenius plots of ln{k} vs 1/T. A priori, there
is no reason to expect that the measured rate constant sum kind
6
Ravishankara et al. is somewhat lower and outside of the 95%
-
12
confidence intervals obtained in this work (1.82(30) × 10
3
-12
3
)
k1 + k2 will follow any simple Arrhenius form, i.e., k(T) )
cm /s vs 2.26(40) × 10
cm /s), but this is compensated for
by a slightly lower value of Eact/R (930(50) K vs 976(50) K).
A exp(-Eact./RT), because (i) it is composed of two individual

3972 J. Phys. Chem. A, Vol. 104, No. 17, 2000
Nizkorodov et al.
TABLE 3: Comparison between k
1
(or k
1
+ k
2
) Measured in
[HO2]t)0 * 0 as opposed to the [HO2]t)0 ) 0 assumed to derive
eq 3. Under such conditions, the decay of the OH population is
still double exponential, but the preexponential factor ratio
becomes â/R ) k2/k1{1 + [HO2]t)0/[OH]t)0}. The fact that
larger values of â/R are routinely observed in H2/O3/Ar mixtures,
which exaggerate the kinetic influence of OH*, supports this
explanation.
This and Previous Studiesa
quantity
measured
A
value at 298 K
-
12
3
-14
3
(×10 cm /s) Eact/R (K) (×10 cm /s)
ref.
b
k
k
k
k
k
k
1
1
1
1
1
1
+ k
+ k
+ k
2
2
2
1.6
960
6.5(15)
6.5(10)
8.2
6.5(10)
7.46(16)
8.4(8)
6.8
4
5
6
8
7
9
13
1.82(30)
1.52(10)
1.6
930(50)
890(60)
940(300)
976(50)
Second, though the fast induction decay component is
insensitive to UV power and radical densities (Figure 8), this
is not true for the slow propagation/termination part of the
experimental OH decay. Indeed, the results of numerical
simulations shown in Figure 9c indicate that, even at the very
lowest UV laser power levels employed in this study, the radical
densities are still sufficient to influence the amplitude of this
secondary decay component and thus the experimentally
observed â/R ratios. By way of example, Figure 9 shows the
+ k
+ k
2
2
recommended
value for k
1
k
1
+ k
2
2.26(40)
8.5(13)
this work
a
The rate constant sum has been fit to an Arrhenius form, k(T) ) A
b
14
3
exp(-Eact/RT). Normalized to k
suggested by Chang and Kaufman.
1
46
(298K) ) 6.5 × 10 cm /s as
The same is also true for comparison with data of Smith et al.
1
6
-3
-
12
7
results for [O3] ) 5 × 10 cm as a function of photolysis
pulse intensity; the fast OH decay rates obtained from least
squares fits to the numerical simulations start to noticeably
(
A ) 1.52(10) × 10 , Eact/R ) 890(60) K), although the
discrepancies are somewhat greater. The activation energy of
4
,46
Eact/R ) 960 K obtained by Andersen and Kaufman
is in
deviate from the predicted double exponential behavior above
quite good agreement with present results, with the 20-25%
2
1
mJ/cm /pulse. While such least squares fits to a double
deviations in Figure 10 primarily due to a lower preexponential
-
12
3
-12
3
exponential form reproduce the correct values of τind to better
factor (1.6 × 10
cm /s vs 2.26(40) × 10
cm /s). The
than 5% accuracy (Figure 9b) for the experimental conditions
studies by Zahniser and Howard and Kurylo report only 298 K
2
data,5,8 which are about 25% lower than the present k1 + k2
of this study (1-2 mJ/cm /pulse), the â/R ratios are predicted
to deviate from k2/k1 ) 0.025 by as much as a factor of 2-3
value at room temperature.
(Figure 9c).
Also plotted in Figure 10 are the current temperature
dependent values of k1 recommended for use in stratospheric
modeling,13 which is significantly below the current results for
k1 + k2 for all temperatures studied. This 15-20% discrepancy
cannot be simply ascribed to neglect of k2, which is at most
The most important contributor to these effects is reaction 9
3
3
of O( P) with HO2 to form OH + O2. The O( P) atoms arise
mainly from the minor channel in the O3 photolysis, physical
quenching of O( D) and slow reaction between O3 and O2( ∆),
1
1
5% of k1 over this temperature range. In fact, the recommended
which is also formed in the O photolysis as a companion to
3
3
1
-4
values are obtained as an average of data obtained in refs 4-8,
all but one of which reflect measurement of the rate constant
O( D). The concentration of O( P) is actually quite low (<10
of [O ]); however, the radical-radical reaction 9 is 4 orders of
3
-
12
3
3
9
sum k1 + k2. The recommended A factor of 1.6 × 10
cm /s
magnitude faster than (2), and thus k [O( P)] can nevertheless
is nearly 40% smaller than the results of this study, but with a
be comparable to the slow second step of the chain reaction,
sufficiently large uncertainty in the activation energy (Eact/R )
i.e., k [O ]. In essence, this radical-radical reaction “side-steps”
2
3
9
40(300) K) to include the current value of Eact/R ) 976(50)
reaction 2, i.e., converts HO2 radicals back into OH radicals.
The basic effect of this is to artificially increase the amplitude
of the long-term OH decay component and thereby increase
the â/R ratio obtained from double exponential fits to the
experimental data.
K. The results of this study suggest that the recommended values
for these chain reaction rates should be revised upward.
Finally, we would like to comment on the possibility of
determining the rate constant k2 from the present experimental
results. According to the kinetic model (eqs 3 and 8), the ratio
â/R of preexponential factors determined from the double
exponential fit of the OH decay traces should equal the ratio of
the rate constants; i.e.,
Figure 9c demonstrates that a reliable determination of k2/k1
via measurement of â/R ratios requires extremely low UV
2
photolysis powers (i.e., , 0.1 mJ/cm /pulse at 308 nm), resulting
1
0
-3
in long-time OH concentrations of , 10 cm , which is not
experimentally feasible even with our shot noise limited IR
absorption capabilities. An alternative approach would involve
high-resolution absorption methods to observe both HO2 and
OH populations at very long times (t . τind), for which the
â/R ) k /k1
(11)
2
If both the sum and the ratio of the two rate constants can be
determined, individual k1 and k2 values should also be obtainable
from the data. Extensive experimental tests based on this kinetic
analysis have been performed; the results yield a room tem-
perature ratio of k2/k1 = 0.05, which oVerestimates the currently
accepted value by a factor of 2. Furthermore, this empirical ratio
is obtained from the behavior of the [OH] time traces at very
long times, which is now found to be experimentally sensitive
to the photolysis energy, i.e., concentration of radicals in the
kinetic mixtures.
This breakdown in the simple chain kinetic model arises for
the following several reasons. First of all, if reaction of
vibrationally excited OH* with O3 occurs much faster than
relaxed OH with O3, then production of HO2 from OH occurs
on a time scale competitive with OH* vibrational relaxation. If
so, then the initial conditions effectively become [OH]t)0 * 0,
3
ratio [OH]/[HO2] asymptotically goes to k2/k1 as O( P) is
irreversibly consumed. It is worth noting that the necessary
rovibrational line strengths for OH and HO2 required to convert
absolute absorbances into absolute population densities have
1
4-19
already been experimentally determined.
Initial measure-
ments of quantum-state-resolved HO2 populations from the OH/
HO2/O3 chemical chain reaction are currently underway, with
which one can hope to provide additional kinetic information
on the slow step of this chain in the 200-230 K temperature
regime relevant to the lower stratosphere.
Acknowledgment. This work has been supported by grants
from the National Aeronautics and Space Administration and
the Air Force Office of Scientific Research.

Kinetics of the OH/HO2/O3 Chain Reaction
J. Phys. Chem. A, Vol. 104, No. 17, 2000 3973
References and Notes
(23) Huang, Y.; Gu, Y.; Liu, C.; Yang, X.; Tao, Y. Chem. Phys. Lett.
986, 127, 432.
1
1
1
1
(
(
(
1) Lary, D. J. J. Geophys. Res. 1997, 102, 21515.
2) Hunt, B. G. J. Geophys. Res. 1966, 71, 1385.
3) Wennberg, P. O.; Cohen, R. C.; Stimpfle, R. M.; Koplow, J. P.;
(24) Dyer, M. J.; Knutsen, K.; Copeland, R. A. J. Chem. Phys. 1997,
07, 7809.
(25) Dodd, J. A.; Lipson, S. J.; Blumberg, W. A. M. J. Chem. Phys.
Anderson, J. G.; Salawitch, R. J.; Fahey, D. W.; Woodbridge, E. L.; Keim,
E. R.; Gao, R. S.; Webster, C. R.; May, R. D.; Toohey, D. W.; Avallone,
L. M.; Proffitt, M. H.; Loewenstein, M.; Podolske, J. R.; Chan, K. R.; Wofsy,
S. C. Science 1994, 266, 398.
991, 95, 5752.
(
26) Daumont, D.; Brion, J.; Charbonnier, J.; Malicet, J. J. Atmos. Chem.
992, 15, 145.
27) Malicet, J.; Daumont, D.; Charbonnier, J.; Parisse, C.; Chakir, A.;
Brion, J. J. Atmos. Chem. 1995, 21, 263.
(
(
(
(
4) Anderson, J. G.; Kaufman, F. Chem. Phys. Lett. 1973, 19, 483.
5) Kurylo, M. J. Chem. Phys. Lett. 1973, 23, 467.
6) Ravishankara, A. R.; Wine, P. H.; Langford, A. O. J. Chem. Phys.
(
(
(
28) Molina, L. T.; Molina, M. J. J. Geophys. Res. 1986, 91, 14501.
29) Mount, G. H. J. Geophys. Res. 1992, 97, 2427.
30) Mauersberger, K.; Barnes, J.; Hanson, D.; Morton, J. Geophys. Res.
1
979, 70, 984.
7) Smith, C. A.; Molina, L. T.; Lamb, J. J.; Molina, M. J. Int. J. Chem.
Kinet. 1984, 16, 41.
(
Lett. 1986, 13, 671.
31) Nelson, D. D.; Schiffman, A.; Lykke, K. R.; Nesbitt, D. J. Chem.
(
(
8) Zahniser, M. S.; Howard, C. J. J. Chem. Phys. 1980, 73, 1620.
Phys. Lett. 1988, 153, 105.
(9) Kulcke, A.; Blackmon, B.; Chapman, W. B.; Kim, I. K.; Nesbitt,
D. J. J. Phys. Chem. A 1998, 102, 1965.
10) Manzanares, E. R.; Suto, M.; Lee, L. C.; Coffey, D. J. J. Chem.
Phys. 1986, 85, 5027.
(32) Hall, J. L.; Lee, S. A. Appl. Phys. Lett. 1976, 29, 367.
(33) Butler, J. E.; Talley, L. D.; Smith, G. K.; Lin, M. C. J. Chem.
Phys. 1981, 74, 4501.
(
(
(
11) Wang, X.; Suto, M.; Lee, L. C. J. Chem. Phys. 1988, 88, 896.
12) Sinha, A.; Lovejoy, E. R.; Howard, C. J. J. Chem. Phys. 1987, 87,
(34) Kliner, D. A. V.; Farrow, R. L. J. Chem. Phys. 1999, 110, 412.
(35) Raiche, G. A.; Jeffries, J. B.; Rensberger, K. J.; Crosley, D. R. J.
Chem. Phys. 1990, 92, 7258.
2
122.
13) DeMore, W. B.; Sander, S. P.; Golden, D. M.; Hampson, R. F.;
(
(36) Streit, G. E.; Johnston, H. S. J. Chem. Phys. 1976, 64, 95.
(37) Teitelbaum, H.; Aker, P.; Sloan, J. J. Chem. Phys. 1987, 119, 79.
(38) Holtzclaw, K. W.; Upschulte, B. L.; Caledonia, G. E.; Cronin, J.
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 Number 12; Jet Propulsion Laboratory: Pasedena,
CA, 1997.
F.; Green, B. D.; Lipson, S. J.; Blumberg, W. A. M.; Dodd, J. A. J. Geophys.
Res. 1997, 102, 4521.
(14) Goldman, A.; Schoenfeld, W. G.; Goorvitch, D.; Chackerian, C.,
(
39) Ohoyama, H.; Kasai, T.; Yoshimura, Y.; Kimura, H.; Kuwata, K.
Chem. Phys. Lett. 1985, 118, 263.
40) Charters, P. E.; Macdonald, R. G.; Polanyi, J. C. Appl. Opt. 1971,
0, 1747.
Jr.; Dothe, H.; Melen, F.; Abrams, M. C.; Selby, J. E. A. J. Quant. Spectrosc.
Radiat. Transfer 1998, 59, 453.
(
(
15) Holtzclaw, K. W.; Person, J. C.; Green, B. D. J. Quant. Spectrosc.
Radiat. Transfer 1993, 49, 223.
16) Nelson, D. D.; Schiffman, A.; Nesbitt, D. J.; Orlando, J. J.;
Burkholder, J. B. J. Chem. Phys. 1990, 93, 7003.
17) Buchanan, J. W.; Thrush, B. A.; Tyndall, G. S. Chem. Phys. Lett.
983, 103, 167.
1
(
41) Klenerman, D.; Smith, I. W. M. J. Chem. Soc., Faraday Trans. 2
(
1
987, 83, 229.
(
42) Smith, D. R.; Blumberg, W. A. M.; Nadile, R. M.; Lipson, S. J.;
Huppi, E. R.; Wheelr, N. B.; Dodd, J. A. Geophys. Res. Lett. 1992, 19,
93.
43) Dodd, J. A.; Lipson, S. J.; Lowell, J. R.; Armstrong, P. S.;
(
1
1
5
(
(
18) Zahniser, M. S.; Stanton, A. C. J. Chem. Phys. 1984, 80, 4951.
19) Zahniser, M. S.; McCurdy, K. E.; Stanton, A. C. J. Phys. Chem.
(
Blumberg, W. A. M.; Nadile, R. M.; Adler-Golden, S. M.; Marinelli, W.
J.; Holtzclaw, K. W.; Green, B. D. J. Geophys. Res. 1994, 99, 3559.
(44) Rensberger, K. J.; Jeffries, J. B.; Crosley, D. R. J. Chem. Phys.
1989, 90, 2174.
(45) Chalamala, B. R.; Copeland, R. A. J. Chem. Phys. 1993, 99, 5807.
(46) Chang, J. S.; Kaufman, F. J. Phys. Chem. 1978, 82, 1683.
984, 93, 1065.
(
20) Aker, P. M.; Sloan, J. J. J. Chem. Phys. 1986, 85, 1412.
(21) Butler, J. E.; Jursich, G. M.; Watson, I. A.; Wiesenfeld, J. R. J.
Chem. Phys. 1986, 84, 5365.
22) Cleveland, C. B.; Jursich, G. M.; Trolier, M.; Wiesenfeld, J. R. J.
Chem. Phys. 1987, 86, 3253.
(