Direct studies on the decomposition of the tert-butoxy radical and its reaction with NO

Article abstract of DOI:10.1039/a806524a

The first laser induced fluorescence (LIF) spectrum for the tert-butoxy radical is reported following radical generation by excimer laser photolysis of tert-butyl nitrite. The laser flash photolysis-LIF technique is used to measure the temperature dependence of the rate coefficient for the reaction with NO (T = 200-390 K): (CH₃)₃CO + NO → (CH₃)₃CONO (3) which can be represented in either Arrhenius or AT(-n) format: k₃ = (7.8 ± 1.8) x 10^{- 12} exp(2850 ± 290 J mol^-1/RT) cm³ molecule^-1 s^-1 k₃ = (4.17 ± 0.12) x 10^-11(T/200)(-1.27±0.07) cm³ molecule^-1 s^-1 and the tert- butoxy decomposition. (CH₃)₃CO→CH₃COCH₃ + CH₃ (2). Whilst the former reaction shows no pressure dependence (p = 70-500 Torr of helium), the latter reaction is in the fall-off regime over the entire range of experimental conditions (T = 303-393 K, p = 13-600 Tort helium). The fall-off data were fitted using an inverse Laplace transformation/master equation model to give the following Arrhenius parameters: k₂/(∞)(T) = (1.4±0.6) x 10¹³ exp[- (57±2) kJ mol^-1/RT]. These Arrhenius parameters are significantly lower than previous indirect measurements or calculations and the atmospheric and mechanistic implications are discussed. Finally, reaction (3) was also studied by monitoring the temporal dependence of the NO LIF signal following the photolysis of tert-butylnitrite with no additional NO. The results are in good agreement with the tert-butoxy monitoring and allow for an estimation of the rate parameters for the tert-butoxy self reaction.

Full text of DOI:10.1039/a806524a

View Article Online / Journal Homepage / Table of Contents for this issue
Direct studies on the decomposition of the tert-butoxy radical and its
reaction with NO
M. Blitz,a M. J. Pilling,a S. H. Robertsonb and P. W. Seakins*a
a School of Chemistry, University of L eeds, L eeds, UK L S2 9JT
b Molecular Simulations Inc, T he Quorum, Barnwell Road, Cambridge, UK
Recei¿ed 18th August 1998, Accepted 12th October 1998
The Ðrst laser induced Ñuorescence (LIF) spectrum for the tert-butoxy radical is reported following radical
generation by excimer laser photolysis of tert-butyl nitrite. The laser Ñash photolysis-LIF technique is used to
measure the temperature dependence of the rate coefficient for the reaction with NO (T \ 200È390 K):
(CH ) CO ] NO ] (CH ) CONO
(3)
3 3
3 3
which can be represented in either Arrhenius or AT~n format:
k \ (7.8 ^ 1.8) ] 10~12 exp(2850 ^ 290 J mol~1/RT ) cm3 molecule~1 s~1
3
k \ (4.17 ^ 0.12) ] 10~11(T /200)~1.27B0.07 cm3 molecule~1 s~1
3
and the tert-butoxy decomposition.
(CH ) CO ] CH COCH ] CH
3 3 3
(2)
3
3
Whilst the former reaction shows no pressure dependence (p \ 70È500 Torr of helium), the latter reaction is in
the fall-o† regime over the entire range of experimental conditions (T \ 303È393 K, p \ 13È600 Torr helium).
The fall-o† data were Ðtted using an inverse Laplace transformation/master equation model to give the
following Arrhenius parameters:
k=(T ) \ (1.4 ^ 0.6) ] 1013 exp[[(57 ^ 2) kJ mol~1/RT ]
2
These Arrhenius parameters are signiÐcantly lower than previous indirect measurements or calculations and
the atmospheric and mechanistic implications are discussed.
Finally, reaction (3) was also studied by monitoring the temporal dependence of the NO LIF signal
following the photolysis of tert-butylnitrite with no additional NO. The results are in good agreement with the
tert-butoxy monitoring and allow for an estimation of the rate parameters for the tert-butoxy self reaction.
tolysis for alkoxy generation coupled with laser induced Ñuo-
rescence (LIF) detection, primarily of the reaction products.
The tert-butoxy radical is an obvious starting point as the
lack of a-hydrogens and the geometry of the radical means
I
Introduction
Alkoxy radicals are mechanistically important in the low tem-
perature oxidation of hydrocarbons.1,2 In general three reac-
tions are possible: decomposition (1a), isomerization (1b) and
reaction with oxygen (1c)
that reactions with
O
and internal abstraction
2
(isomerization) are extremely unlikely, and hence the only
reaction of atmospheric signiÐcance is the decomposition to
acetone and methyl radical [reaction (2)].
RO ] carbonyl ] alkyl radical
RO ] ~R@OH
(1a)
(1b)
(1c)
(CH )CO ] CH COCH ] CH
(2)
3
3
3
3
RO ] O ] HO ] carbonyl
2
2
The study of tert-butoxy reactions is further simpliÐed as it
These di†erent reaction pathways can have signiÐcant e†ects
determining, for example, the number of NO to NO conver-
sions that can take place during the oxidation of the parent
hydrocarbon in the atmosphere and hence a†ecting tropo-
spheric ozone production. Despite the importance of alkoxy
reactions, only reactions of the methoxy radical have been
reasonably well characterised; there have been very few direct
studies of larger alkoxy radicals. This dearth of information
stems both from difficulties in cleanly generating, and sensiti-
vely monitoring, alkoxy radicals.
was possible to obtain a LIF spectrum of C H O allowing
4
9
direct detection of the radical reagent.
2
The majority of the previous work on tert-butoxy decompo-
sition has been performed by Batt and co-workers.4h8 Either
thermal decomposition or photolysis was used to generate
alkoxy radicals and rate coefficients were extracted following
end product analysis by gas chromatography. The Arrhenius
parameters generated in this current work are di†erent from
BattÏs earlier work. The mechanistic implications of the low A
factors and activation energies are discussed in Section V,
along with a consideration of the atmospheric implications.
In the atmosphere, reactions (1aÈc) dominate the kinetics of
alkoxy removal. However, in the laboratory, alkoxy reactions
As part of a larger consortium3 investigating the reactions
of biogenic hydrocarbons, we have started a systematic study
of the reactions of C alkoxy radicals using laser Ñash pho-
4
Phys. Chem. Chem. Phys., 1999, 1, 73È80
73

View Article Online
often occur in the presence of NO, either from the photolysis
or decomposition of a nitrite alkoxy precursor or as an added
reagent, and hence reactions of alkoxy radicals with NO are
of interest. Although there are no direct studies of the NO
reaction with tert-butoxy, the reactions with methoxy, ethoxy
and isopropoxy radicals have been studied and hence provide
a bench mark against which we can judge the validity of our
technique. The rate coefficients are relatively invariant with
the size of the alkoxy species and Atkinson1 has suggested a
common Arrhenius expression for the reaction of all alkoxy
radicals with NO. The reaction was studied in two ways,
Ðrstly by direct observation of the reaction of tert-butoxy with
an excess of NO; and secondly by monitoring the concentra-
tion of NO following the photolysis of the tert-butoxy nitrite
precursor. In conjunction with the decomposition measure-
ments this second method allows for the simultaneous estima-
tion of the tert-butoxy self reaction.
The paper is structured as follows. In Section II we describe
the experimental apparatus and techniques used to monitor
the decomposition and reaction of tert-butoxy with NO.
Section III describes the observed laser-induced Ñuorescence
spectrum and Section IV details the results of the studies of
the reaction of tert-butoxy with excess NO. In Section V the
results of the study of the tert-butoxy decomposition reaction
between 303 and 393 K are presented and the implications of
the resulting Arrhenius parameters are discussed. The results
of NO monitoring experiments are presented in Section VI
and Ðnally the work is summarised in Section VII.
The pressures in the cell were monitored using capacitance
manometers, and varied by throttling the exit valve on the
cell. High total Ñow rates (D750 sccm) were used in the NO
experiments, the bu†er gases for the experiments being either
helium or heliumÈmethane mixtures. For the tert-butoxy
experiments lower Ñow rates could be used and pressures were
varied between 15È600 Torr using helium as the bu†er gas.
Sub ambient temperatures were obtained by Ñowing suitable
coolants around the outside of the reactor, high temperatures
were obtained using cartridge heaters located in the metal
block situated around the centre of the reaction cell. Tem-
peratures were measured using two thermocouples (type K)
located just above and below the reaction zone. Temperatures
could be controlled to better than ^1 K.
tert-Butylnitrite (Aldrich, 98%) was thoroughly degassed,
diluted in helium and stored in blackened 5 l glass bulbs. NO
(Linde, 99%) was distilled at 196 K to remove NO and then
2
diluted in helium before storage in darkened bulbs. Helium
(CP grade, 99.999%, BOC) and methane (UHP, 99.995%
Linde) were used directly from the cylinders.
III tert-Butoxy spectrum
Photolysis of tert-butylnitrite, (60 mTorr in 70 Torr of helium)
produced a species that Ñuoresces with considerable red shift
when probed between ca. 330È360 nm. The lifetime of the
observed signal was ca. 500 ns. The use of a Perspex Ðlter
which transmits light at wavelengths greater than 390 nm
removes any scattered light. SigniÐcantly red shifted Ñuores-
cence is characteristic of lower alkoxy radicals.10h12 A spec-
trum of the observed species was recorded: the signal was
obtained with a 50 ls delay between the photolysis and probe
lasers and the cell was cooled to 203 K in order to sharpen the
spectrum which is displayed in Fig. 1. This spectrum is
broadly similar to other alkoxy spectra but quantitatively dif-
ferent, i.e. it does not arise from lower alkoxy radicals formed
in alternative photochemical reactions. Whilst the spectrum is
broad, it does appear to have vibrational structure, with two
progressions, and a separation within each progression of
approximately (500 ^ 20) cm~1. This type of vibrational pro-
gression is observed in other alkoxy radicals and has been
assigned to the CÈO stretch in the excited state. There appears
to be a gradual decrease in the frequency of stretch with the
size of the alkoxy radical (ethoxy 590 cm~1,11 isopropoxy
560 ^ 10 cm~1 10) which may be due to a combination of
weakening bond strength and change in mass of the alkyl
fragment. The apparent Ðne structure is an artefact of the
doubling crystal autotracker.
II Experimental
All experiments were undertaken in a conventional slow Ñow,
laser Ñash photolysis-laser induced Ñuorescence apparatus,
similar to that used in previous experiments.9 tert-Butoxy and
NO were generated via the pulsed excimer laser (Lambda
Physik LPX 100) photolysis of tert-butyl nitrite at 351 nm
(XeF). In experiments where tert-butoxy was monitored the
laser repetition rate was 3 Hz, but in experiments monitoring
the time dependence of the NO concentration a slower repeti-
tion rate of 0.3 Hz was used to ensure that the complete con-
tents of the cell were Ñushed between laser shots. The
photolysis energy was approximately 100 mJ pulse~1 con-
tained within a beam ca. 2 ] 1 cm in size.
The probe light was generated from a Nd:YAG (Spectron
SL803, doubled output at 532 nm) pumped dye laser, using
appropriate doubling/mixing (KDP) crystals. For tert-butoxy
observation, the light output from a Pyridine 1 dye (670È720
nm) was doubled to probe a range of tert-butoxy features (See
below). For NO observation, the output from the dye laser
(Rhodamine 6G) was doubled and then mixed with the funda-
mental (1064 nm) from the Nd:YAG laser to generate light ca.
226 nm. NO was probed in the P2 branch head of the
A 2& ^ X 2% band system near 226.87 nm. The probe light
was O1 mJ pulse~1 and intersected the photolysis laser beam
at right angles.
The time dependence of the intensities of the 5 strongest
peaks of the spectrum in the presence of added NO all gave
decay constants that were the same within experimental error
Laser induced Ñuorescence was collected by a photomulti-
plier (EMI 9813 QKB) perpendicular to both the photolysis
and probe beams and the resulting signal was integrated using
a boxcar averager (SRS SR250) minimising scattered light.
For NO monitoring the Ñuorescence was Ðltered by a mono-
chromator centred on the probe beam wavelength and for the
tert-butoxy experiments, a Perspex Ðlter collected light at
wavelengths greater than 390 nm. The integrated signal was
subsequently digitised and stored on a PC. The radical time-
proÐles were accumulated by scanning the delay between the
photolysis and probe lasers with each kinetic trace containing
100 time points, and each point being the average of 4È10
laser samples.
Fig. 1 tert-Butoxy excitation spectrum monitored by recording the
total Ñuorescence passing through a Perspex Ðlter. Conditions: 203 K,
D100 mTorr of tert-butoxynitrite, 70 Torr total pressure of helium.
Gas Ñows were measured using calibrated Ñow controllers
and mixed in a manifold prior to entering the reaction cell.
74
Phys. Chem. Chem. Phys., 1999, 1, 73È80

View Article Online
indicating that all of the spectrum shown in Fig. 1 arises from
a single species.
Photolysis of the C ÈC alkyl nitrites is an established
1
3
method of generating the corresponding alkoxy radical and
there is no reason to expect that tert-butylnitrite should be
any di†erent. This is supported by earlier end product analysis
studies following the photolysis of tert-butylnitrite8 where the
results are entirely consistent with reactions of the tert-butoxy
radical.
In the absence of a complete spectral assignment it is
impossible to positively identify the observed spectrum with
the tert-butoxy radical, however, spectroscopic and chemical
evidence indicate that this is a reasonable assumption and it is
difficult to postulate any other species which could give the
observed Ñuorescence spectrum with the time dependence pre-
sented in subsequent sections.
IV Results for the reaction of tert-butoxy + NO
Initial experiments on reaction (3) were performed in a con-
ventional manner, with the
(CH ) CO ] NO ] (CH ) CONO
3 3 3 3
(3)
decay of the tert-butoxy LIF signal being monitored in the
prescence of a large excess of NO. Further information on the
reaction obtained by monitoring the time dependence of NO
produced in the photolysis pulse is given in Section VI. A
typical decay trace and a range of bimolecular plots are
shown in Fig. 2. In all cases the decays were exponential and
Fig. 2b shows that the bimolecular plots are linear as [NO] is
varied, typically by a factor of 10. No pressure dependence
was observed [p(He) \ 70È600 Torr].
Fig. 2 a, Typical tert-butoxy decay trace following the photolysis of
60 mTorr of tert-butyl nitrite in the presence of 9.1 mTorr of NO at a
total pressure of 71 Torr of helium at 333 K; b, typical bimolecular
plot showing the variation of pseudo-Ðrst-order rate coefficient with
[NO] at 333 K.
Experiments were repeated over the temperature range
200È390 K and the temperature dependence of the reaction is
depicted in Fig. 3. The data can conveniently be represented
in either Arrhenius or AT~n format:
earlier work on the smaller alkoxy reactions showed that the
reaction with NO was at the high pressure limit at 15 Torr,11
and the lack of any pressure dependence in this study is con-
sistent with these observations.
Given that recombination is the only possible outcome of
the tert-butoxy ] NO reaction, the measured activation
energy could, in theory, be used in conjunction with the
reverse decomposition data to estimate the overall reaction
enthalpy and hence the heat of formation of the tert-butoxy
radical. However, in practice there are difficulties with this ap-
k \ (7.8 ^ 1.8) ] 10~12
3
] exp(2850 ^ 290 J mol~1/RT ) cm3 molecule~1 s~1
k \ (4.17 ^ 0.12) ] 10~11(T /200)~1.27
3
^ 0.07 cm3 molecule~1 s~1
Unlike primary and secondary alkoxy radicals, abstraction of
an a hydrogen by NO is impossible in the tert-butoxy system
(and b abstraction will probably be extremely slow), however,
instead of recombination to reform the nitrite, it could be pos-
sible to form tert-C H NO but this is considered to be
4
9
2
unlikely. Zellner14 has studied the recovery of CH ONO and
3
saw no evidence for the formation of CH NO , which would
be expected to require a signiÐcant energy barrier for forma-
tion.
3
2
Table 1 shows that the room temperature value determined
in this study is consistent with other alkoxy reactions. The
temperature dependence of the reaction is slightly greater than
that predicted by AtkinsonÏs universal parameters1 and the
results from Table 1 do seem to suggest a very slight increase
in the temperature dependence of the reactions with increasing
size of the alkoxy radical, a trend which would be predicted
from current theories of recombination reactions.15 The
Fig. 3 Temperature dependence of the reaction of tert-butoxy with
NO.
Table 1 Arrhenius parameters and k
values from direct studies of the reaction of alkoxy radicals with NO
298
1011 A/cm3
1011 k /cm3
298
Radical
C H O
molecule~1 s~1
E/kJ mol~1
molecule~1 s~1
Reference
4.4 ^ 0.4
3.8
3.5
2.5
3.8
11
13
10
2
5
C H O
1.9
1.22
0.8
[1.7
2
5
iso-C H O
[2.6
3
7
tert-C H O
All
[2.85
[1.25
This work
1
4
9
2.3
Phys. Chem. Chem. Phys., 1999, 1, 73È80
75

View Article Online
proach. Firstly, there is considerable variation (^5 kJ mol~1)
in the activation energy for the decomposition reaction of the
nitrite16h18 and secondly, the enthalpy of formation of the
tert-butylnitrite molecule has not been measured. A value for
the enthalpy of formation of the nitrite could be estimated
from group additivity,19 however, a comparison of the mea-
sured and estimated values for the heat of formation of methyl
and ethyl nitrite indicates an approximate uncertainty of ^3
kJ mol~1 in the value of the OÈNO/C group. The precision of
any resulting calculation will therefore not be an improvement
on the current recommended heat of formation for the tert-
butoxy ([90.7 kJ mol~1) proposed by McMillan and
Golden.20
program.24 The universal force-Ðeld was used together with
the charge equilibrium25 module to construct a potential
energy surface. The molecular geometry was obtained by
minimising the potential energy (charges were recalculated
periodically). Once the minimum structure was obtained, the
frequencies were obtained in the usual way.26 The lower three
frequencies were discarded as these are associated with the
large amplitude motion of internal rotation. The barrier
heights of the hindered internal rotation were calculated by
altering the torsion angle and calculating the energy, and
determining the di†erence between highest and lowest values.
Variation of calculated tert-butoxy vibrational frequencies by
^10% produces only a slight variation (\5%) in the best Ðt
parameters which is within other experimental and calcu-
lational errors.
The vibrational frequencies, internal and external moments
of inertia were used to calculate a density of states for the
tert-butoxy reagent.27 Microcanonical rate coefficients [k(E)]
were obtained by inverse Laplace transformation of an Arrhe-
nius expression for the decomposition reaction and incorpor-
ated into a master equation to calculate the canonical rate
coefficient of the reaction under experimental conditions. Col-
lision frequencies were calculated using tabulated data28 for
helium and for tert-butoxy, by analogy with similar molecules.
The calculated Ðts were insensitive to changes in these esti-
mated parameters. The other parameter required in the
V
tert-Butoxy decomposition
The decomposition reaction has been studied from 303È393
K. Within this range the reaction
(CH ) CO ] CH COCH ] CH
3 3 3
(2)
3
3
was found to be pressure dependent and for each temperature
the reaction was studied between pressures of 13 and 600 Torr
of helium. For a given pressure and temperature the decays
were always exponential in nature. At low temperatures a
slight correction had to be introduced due to the reaction of
the alkoxy radical with residual NO, however, this correction
was always small. The highest temperature was limited by the
speed of the reaction and the need to ensure that vibrational
relaxation of the tert-butoxy radical was complete before
decomposition occurred. Examples of the pressure dependence
are shown as the solid points in Fig. 4.
master equation is S*E T, the average energy transferred in a
d
downward collision. This parameter was allowed to Ñoat in
the Ðt.
The Ðt was performed by comparing the calculated (k
)
calc
and experimental (k ) data over the complete T , p range
expt
with the use of a s2 function
From our study it is not possible to unequivocally state that
we are observing the decomposition reaction, there is a small
possibility that some reagent is being removed by an alterna-
tive Ðrst order process such as a 1,4 isomerization. However
decomposition is by far the most likely route as it will have a
signiÐcantly lower activation energy. Previous end product
analyses of the decomposition reaction8 have shown no evi-
dence of an alternative Ðrst order removal process.
1
s2 \ ;
T, p
(k [ k )2
calc expt
p2
where p2 is the variance associated with each experimental
measurement. The nature of the s2 surface and complexity of
the calculation (numerical derivatives required) renders con-
ventional Marquardt techniques inefficient. We have therefore
conducted a grid search varying A=, E= and S*E T over a
In order to investigate the pressure dependence of this reac-
tion and to extract Arrhenius parameters for the limiting high
2
2
d
range of conditions compatible with previous experimental
pressure rate coefficient, k=, we have modelled the experimen-
determinations. Fig. 5 shows the s2 surface as a function of
2
tal data using two methods: the inverse Laplace transform-
A=, E= for S*E T \ 150 cm~1, and the solid lines in Fig. 4
d
master equation technique (ILT-ME)15 and the Troe
are the Ðt to the experimental data with the best Ðt param-
factorisation formulation.21h23
eters of: A= \ 1.4 ] 1013 s~1, E= \ 57 kJ mol~1 and
For the ILT-ME method the geometry, frequencies and
rotational barriers were calculated using the Cerius2
S*E T \ 150 cm~1. The wells reÑect the relatively coarse grid
d
used; the insert in Fig. 5 shows the surface in more detail
Fig. 4 Pressure dependence of the Ðrst order rate coefficient at 303, 323, 343, 363 and 383 K with ILTÈME (solid lines) and Troe (dashed lines)
Ðts to the data.
76
Phys. Chem. Chem. Phys., 1999, 1, 73È80

View Article Online
studies, for example Wagner and Wardlaw,33 have used more
complex parameterizations of F , however, these have gen-
cent
erally been for association reactions covering a much wider
range of temperature. For the limited temperature covered in
this dissociation study only a simple parameterization of F
is justiÐed.
cent
The resulting Ðts are shown as the dotted lines in Fig. 4 and
the Ðtting parameters are summarised in Table 2. Fits of com-
parable quality can also be achieved by Ðxing the high pres-
sure limiting values of A= and E= at the values determined in
2
2
the ILT-ME method and Ñoating only F
and k0 .
cent
2
One of the surprising features of this reaction is that it is
indeed pressure dependent. Gutman and co-workers29 have
studied the decomposition of the neopentyl radical;
(CH ) CCH ] (CH ) C : CH ] CH
3 3 3 2 3
(4)
2
2
which might be expected to show similarities with the tert-
butoxy system, and found it to be at 65% of the high pressure
limit at 18 Torr of helium at 560 K. At 400 K (close to the
highest temperature in this study) 98% of high pressure limit
is predicted to be reached at a mere 1 Torr. In contrast the
tert-butoxy system is well into the fall-o† regime at lower tem-
peratures and higher pressures.
The di†erences in fall-o† characteristics for the potentially
similar neopentyl and tert-butoxy systems can be rationalised
by examination of the RRKM expression for the micro-
canonical rate coefficient k(E):15
Fig. 5 s2 surface for a Ðxed value of *E \ 150 cm~1. A has been
d
2
varied from 1 ] 1012È2.5 ] 1013 in steps of 1 ] 1012 s~1. E has been
2
varied from 53È60 kJ mol~1 in steps of 0.5 kJ mol~1.
around the minimum. The quality of the Ðt is excellent, the s2
value corresponding to an average deviation for each datum
point of less than 6%, comparable with the experimental error
associated with that measurement (D10%). The s2 surface is
relatively Ñat close to the minimum and the nature of the
surface shows that the Ðtting parameters are correlated. Based
on the shape of the s2 surface and uncertainties in the experi-
ments and calculations, we estimate the following range of
best Ðt parameters from the ILT-ME method as:
W (E`)
k(E) \
ho(E)
where W (E`) is the number of rovibrational levels in the tran-
sition state and o(E) is the density of states of the reagent
radical. For a given energy above the threshold, W (E`) will be
similar for both systems, however, because of the much
greater threshold energy for neopentyl (129 kJ mol~1)29 the
density of states in the neopentyl radical will be orders of
magnitude greater than for the tert-butoxy and the micro-
canonical rate coefficient correspondingly lower. With a much
lower dissociation rate, a Boltzmann distribution of energised
states can more readily be achieved and hence the high pres-
sure limit is reached at signiÐcantly lower pressures in the
neopentyl system.
A= \ (1.4 ^ 0.6) ] 1013 s~1;
2
E= \ (57 ^ 2) kJ mol~1;
2
S*E T \ (150 ^ 20) cm~1
d
Table 2 summarises the current measurements and the liter-
ature values. The Arrhenius parameters are lower than pre-
vious determinations and recommendations and this is
discussed following the Troe analysis. The resulting value of
S*E T is lower than the values of D200 cm~1 observed in the
d
higher temperature alkyl decompositions.29h31 However,
careful studies32 have indicated that S*E T shows a positive
d
temperature dependence and the values of S*E T from the
d
Semi-quantitative conÐrmation of this argument can be
achieved with the current ILTÈME model. As a Ðrst-order
approximation we use the same frequencies for the two
systems, but input the Arrhenius parameters to match the tert-
tert-butoxy system are probably compatible with the higher
temperature work suggesting that S*E T is independent of the
d
nature of the radical.
butoxy (A \ 1.4 ] 1013 s~1, E \ 57 kJ mol~1) and neopentyl
An alternative method of Ðtting is via the methodology pro-
posed by Troe.21h23 The data are treated globally (i.e. all fall-
o† curves Ðtted simultaneously) with the Troe parameters k0,
a
(A \ 8 ] 1013 s~1, E \ 129 kJ mol~1)29 systems. Micro-
a
canonical rate coefficients are calculated using the ILT formu-
lation of k(E):15
k= modelled by an Arrhenius function and F , the param-
cent
eter describing the broadening of the Lindemann Hinshel-
Ao(E [ E0)
k(E) \
wood fall o†, being Ðtted as
o(E)
F
\ C(T /300)n
cent
where E0 is the threshold energy and Fig. 6(a) shows the
reduction in magnitude of k(E) for the neopentyl system.
Each fall o† curve could have been Ðtted individually, but this
introduces too much Ñexibility into the Ðtting. Previous
Inputting the k(E)s into the ME with a constant S*E T of 150
d
Table 2 Arrhenius parameters for tert-butoxy decomposition
Data
A/1013 s~1
E /kJ mol~1
k
/s~1
298
a
k= from ILT-Me Ðtting
1.4 ^ 0.6
0.35 ^ 0.04a
2.1c ^ 0.5a
39.8
11
13
57 ^ 2
53.1 ^ 0.3a
32.5 ^ 0.7a
67.0
1430
1722
2
k= from Troe Ðttingb
2
k0 from Troe Ðttingb
2
Batt and Robinson 19877
717
1218
762
814
Batt et al. 19898
62.5 ^ 0.6
Choo and Benson 198134
Atkinson 19971
64.0
60
67.7
a The reported errors here are statistical only (^1p). b F \ (0.395 ^ 0.025)(T /300)(1.13B0.22). c Units 10~9 cm3 molecule~1 s~1.
cent
Phys. Chem. Chem. Phys., 1999, 1, 73È80
77

View Article Online
Fig. 7 Arrhenius plot for tert-butoxy decomposition. Lines are
extrapolations to lower temperatures (solid line, this work; dashed
line, Batt et al.8; dotted line, Atkinson1)
removal process for the tert-butoxy radical, it is unlikely that
any re-evaluation of the Arrhenius parameters will have any
e†ect on the predicted atmospheric decomposition mechanism
for the tert-butoxy radical. The signiÐcance of the current Arr-
henius parameters is for other alkoxy radicals where there is a
signiÐcant competition between decomposition and either
isomerization or reaction with oxygen. Further experiments
Fig. 6 (a) Microcanonical rate coefficients for the tert-butoxy system
returned by the ILT model as a function of input Arrhenius param-
eters, showing considerable reduction in k(E) for values corresponding
to neopentyl decomposition. (b) Reduced fall o† curves at 303 and 383
K for Arrhenius parameters corresponding to tert-butoxy and neo-
pentyl decomposition.
on other C alkoxy radicals are currently in progress.
4
The relatively low value of the pre-exponential factor has
mechanistic implications, implying as it does, that the tran-
sition state is not as loose as normally expected for a disso-
ciation reaction. The dissociation is more complex than that
occurring on a Type II potential such as for ethane decompo-
sition. As well as the cleavage of a CÈC bond there is a com-
pensating tightening of the CÈO bond and therefore very high
A factors would not be predicted, but nevertheless the experi-
mentally determined A factors are considerably lower than
thermochemical estimates.34
Possible insights into the mechanism come from an exami-
nation of the reverse reaction, the addition of methyl radicals
to acetone. One might expect the electron deÐcient methyl
radical to preferentially attack the oxygen atom, the tert-
butoxy radical only being formed following an internal
rearrangement. By the principle of microscopic reversibility,
this rearrangement must be the Ðrst step in the overall decom-
position reaction and would be expected to have a signiÐ-
cm~1 gives the reduced fall o† curves shown in Fig. 6(b) con-
Ðrming the pressure independence of the neopentyl system
under these conditions.
The Arrhenius parameters returned from the ILTÈME and
Troe Ðts are signiÐcantly lower than previous measurements,8
thermochemical calculations34 or the review values of Atkin-
son.1 Support for the signiÐcantly lower Arrhenius parameters
measured in this work comes from a recent direct study on
isopropoxy decomposition using
a
similar laser Ñash
photolysis-LIF technique.35
(CH ) HCO ] CH CHO ] CH
(5)
3 3
3
3
An Arrhenius analysis of data taken over the pressure range
10È50 000 Torr gives the following parameters:
cantly lower
A factor than a simple decomposition.
Thermochemistry would appear to support this mechanism as
the enthalpy of formation of the hydroxyalkyl radicals is
lower than that of the corresponding alkoxy species. There
have been very few studies of addition reactions to carbonyl
species, however, one very relevant paper by Sosa and
Schlegel37 investigates the routes of H atom addition to the
carbonyl bond in formaldehyde. Here, despite the enhanced
stability of the hydroxymethyl radical, addition is predicted to
occur via attack on the carbon atom of the double bond, evi-
dence against the mechanism postulated above. Alkoxy for-
mation via addition reactions would appear to be good
subjects for theoretical investigation as the relatively facile
abstraction process generally dominates over addition making
experimental determinations difficult.
The only reported experimental measurement of methyl
addition to acetone is an isotopic study by Knoll et al.38 who
looked for the acetone products of methyl scrambling follow-
ing the addition of methyl radicals to deuterated acetone.
However, given that the heat of formation of the tert-butoxy
radical ([90.7 kJ mol~1 giving a reaction endothermicity of
23 kJ mol~1, ref. 20) is correct, the activation energy deter-
mined during this study is incompatible with the activation
energy of 48 kJ mol~1 determined by Knoll et al. for the addi-
tion of methyl radicals to acetone.
k= \ 9.3 ] 1013 exp([63.7 kJ mol~1/RT ) s~1
5
A possible partial explanation for the higher parameters
reported in the previous relative rate work8 can be found by
examining the temperature dependence of the reference reac-
tion with NO. Batt et al. have used a temperature independent
rate coefficient for this reaction of 4.2 ] 10~11 cm3
molecule~1 s~1 calculated using thermochemical methods,
whereas the direct measurements reported above show that
the reaction has a negative temperature dependence and
values lower than that used by Batt et al. over their experi-
mental temperature range of 303È393 K.
Park et al.36 measured a single value of k at 413 K and a
2
pressure of 200 Torr of argon. The reported value of 1.2 ] 106
s~1 is 50% greater than the high pressure limiting value calcu-
lated from the current Arrhenius parameters and it is therefore
difficult to reconcile the two measurements.
Over a tropospheric temperature range of 220È305 K the
earlier Arrhenius equation of Batt8 predicts lower decomposi-
tion rate coefficients than the present work, varying from
D38% of the value predicted from this work at 220 K (0.16
s~1 vs. 0.4 s~1) to 90% at 305 K (2174 s~1 vs. 2420 s~1). A
graphical depiction of the Arrhenius parameters is given in
Fig. 7. As decomposition is the only signiÐcant atmospheric
78
Phys. Chem. Chem. Phys., 1999, 1, 73È80

View Article Online
Table 3 Mechanism used to Ðt NO decay traces
Reaction
k
k
/cm3 molecule~1 s~1
282
Reference
This work
t-C H O ] NO ] t-C H ONO (3)
Ñoated
4
9
4
9
3
t-C H O ] CH ] CH COCH (2)
k \ 565 s~1
4
9
3
3
3
2
t-C H O ] t-C H O ] products (6)
k
Ñoated
4
9
4
9
6
CH ] CH ] C H (7)
k \ 6 ] 10~11
41
42
3
3
2
6
7
CH ] NO ] CH NO (8)
k \ 3 ] 10~12 or 9.7 ] 10~12
3
3
8
k
NO ]
diff
mm at the reaction zone. This enabled concentrations of
between 0.2 and 1.25 mTorr of radicals, ROÈNO, to be gener-
ated. It was observed that there was a signiÐcant residual NO
signal present (O0.35 mTorr) in the absence of the photolysis
laser. This was attributed to degradation of the nitrite precur-
sor either in the storage bulb or as it passed through the metal
manifoldÈreaction cell. This residual concentration of NO was
accounted for within the kinetic analysis. After each experi-
ment the nitrite Ñow was turned o† and a gas mixture of
known NO concentration was Ñown in its place. This enabled
the NO signal obtained for each experiment to be converted
to an absolute NO concentration. In total, eight experiments
were performed at 282 K.
VI tert-Butyl nitrite photolysis: NO monitoring
In an attempt to further validate the experimental method we
have performed a series of experiments at 282 K monitoring
the NO time dependence following the photolysis of tert-butyl
nitrite. The decay traces showed a residual NO concentration
in comparison to the pretrigger baseline indicating an alterna-
tive removal reaction for tert-butoxy other than recombi-
nation with NO. A portion of this loss will be due to
tert-butoxy decomposition which can be accounted for using
the Arrhenius parameters generated in section IV. However,
the methyl radicals produced in the decomposition reaction
may also react with NO and this needs to be considered in the
analysis. Decomposition does not fully account for the
residual level of NO which in these experiments has been
attributed to the tert-butoxy ] tert-butoxy reaction.
Whilst tert-butyl nitrite has been used in several studies8 as
a photolytic precursor for tert-butoxy radicals, Kades et al.39
have also investigated the vibrational distribution of the NO
generated by photolysis around 350 nm. Under these condi-
tions NO is produced vibrationally ““hotÏÏ with little (D5%)
direct population of ground state v \ 0. This observation was
conÐrmed during our preliminary experiments. When tert-
butyl nitrite was photolysed in 100 Torr He little NO (l \ 0)
signal was observed initially, an observation attributed to the
inefficiency of helium in vibrationally relaxing NO.40 There-
fore a bu†er gas is required that efficiently relaxes the ““hotÏÏ
The data were Ðtted globally (solid line in Fig. 8) to the
kinetic scheme shown in Table 3. From the Arrhenius expres-
sion determined in Section V, k \ 565 s~1 CH3 ] CH is
2
3
well determined and can be Ðxed, 6 ] 10~11 molecule~1 cm3
s~1.40 There is a difficulty with the CH ] NO rate constant
in that under the experimental conditions the reaction is in
3
the fall-o† region. The value for CH ] NO in 100 Torr is
3
3 ] 10~12 molecule~1 cm3 s~1, but in our experiments we
also added ca. 2 Torr of CH , and this might signiÐcantly
4
push the rate constant towards its high pressure limit of
9.7 ] 10~12 molecule~1 cm3 s~1.42 Di†usion was approx-
imated to a Ðrst-order (exponential) loss process; while this is
an approximation it does not signiÐcantly a†ect the overall
result because di†usional losses are small (D10 s~1) compared
to chemical losses.
The kinetic data were Ðtted using a program that numeri-
cally integrated the above scheme and adjusted the rate
parameters, using a Marquardt algorithm, to best Ðt the data.
The program also allowed all eight kinetic traces to be Ðtted
simultaneously, in a global analysis procedure. The datum
points were assigned equal weighting. Two kinetic schemes
were used. Firstly (Scheme 1), the lower value of reaction (7),
corresponding to reaction within the fall-o† region was used,
secondly (Scheme 2), the addition of 2 Torr of methane was
assumed to promote reaction to the high pressure limit. The
results of the analysis are presented in Table 4.
NO but does not chemically a†ect the system. CH is known
4
to efficiently relax vibrationally ““hotÏÏ NO40 and will not
chemically interfere with the chemistry of the system. There-
fore ca. 2 Torr CH was additionally added to the system.
4
Under these conditions the system is thermalized faster than
the chemical change.
Examples of a number of [NO] vs. time traces are shown in
Fig. 8. These traces show the decay of NO to a non-zero base-
line. In order for reaction on the timescale of the experiment a
sizeable radical concentration was required. This was achieved
by having a high nitrite concentration (ca. 25È75 mT), and
also by focusing the photolysis light, using a 76 cm focal
length lens, such that the photolysis beam measured ca. 6 ] 3
The returned values for k are in excellent agreement with
3
the earlier direct determinations (k \ 2.5 ] 10~11 cm3
298
molecule~1 s~1) and do not appear to depend signiÐcantly on
the values used for the CH ] NO reaction. The Ðtted values
3
for k on the other hand are strongly dependent on the
6
residual level of NO and hence on the competition between
reactions (7) and (8). The rate coefficient for reaction (6) is
therefore poorly determined in these experiments. Better esti-
mates may be obtained by working at a lower temperature
where decomposition is insigniÐcant and any residual NO
Table 4 Returned parameters from numerical Ðts of NO decay
traces
Scheme
1011k a
1011k a
k
/s~1
3
6
diff
1
2
2.44 ^ 0.19
2.54 ^ 0.20
0.60 ^ 0.16
1.33 ^ 0.19
16 ^ 1
15 ^ 1
Fig. 8 Typical NO decays following the photolysis of tert-butyl
nitrite at 282 K. The solid line shows a Ðt to the data using the
parameters generated by a global analysis of all experiments based on
Scheme 1.
a Units are in cm3 molecule~1 s~1.
Phys. Chem. Chem. Phys., 1999, 1, 73È80
79

View Article Online
5
6
7
8
L. Batt, Int. J. Chem. Kinet., 1979, 11, 977.
should be more directly attributable to the tert-butoxy recom-
bination. The time dependence of [NO] following nitrite pho-
L. Batt and G. N. Robinson, Int. J. Chem. Kinet., 1982, 14, 1053.
L. Batt and G. N. Robinson, Int. J. Chem. Kinet., 1987, 19, 391.
L. Batt, M. W. M. Hisham and M. Mackay, Int. J. Chem. Kinet.,
1989, 21, 535.
tolysis at low temperatures is currently under study for all C
alkoxy isomers and will be reported in a subsequent paper.
4
9
M. A. Blitz, D. G. Johnson, M. Pesa, M. J. Pilling, S. H.
Robertson and P. W. Seakins, J. Chem. Soc., Faraday T rans.,
1997, 93, 1473.
VII Conclusions
10 R. J. Balla, H. H. Nelson and J. R. McDonald, Chem. Phys., 1985,
99, 323.
11 M. J. Frost and I. W. M. Smith, J. Chem. Soc., Faraday T rans.,
1990, 86, 1751 and M. J. Frost and I. W. M. Smith, J. Chem. Soc.,
Faraday T rans., 1990, 86, 1757.
12 X. Q. Tan, J. M. Williamson, S. C. Foster and T. A. Miller, J.
Phys. Chem., 1993, 97, 9311.
13 P. Devolder SARBVOC 1st Annual Report, European Union,
Brussels, 1997.
14 R. Zellner, J. Chim. Phys., 1987, 84, 403.
15 K. Holbrook, M. J. Pilling and S. H. Robertson, Unimolecular
Reactions, J. Wiley, Chichester, UK, 1997.
16 L. Batt and R. T. Milne, Int. J. Chem. Kinet., 1974, 6, 945.
17 G. D. Mendenhall, D. M. Golden and S. W. Benson, Int. J.
Chem. Kinet., 1975, 7, 725.
18 L. Batt and R. T. Milne, Int. J. Chem. Kinet., 1976, 8, 59.
19 S. W. Benson, T hermochemical Kinetics, WileyÈInterscience, New
York, 1976.
A laser-induced Ñuorescence spectrum has been recorded fol-
lowing the photolysis of tert-butylnitrite. Without spectro-
scopic assignment it is not possible to conclusively attribute
the spectrum to the tert-butoxy radical, however, the photoly-
sis of alkyl nitrites is reasonably well characterised and the
tert-butoxy radical is the major expected radical product.
Support for the identiÐcation of the spectrum comes from the
excellent agreement for the rate coefficient for the recombi-
nation of the tert-butoxy radical with NO as determined by
direct monitoring of the LIF signal or from the time depen-
dence of the NO co-photoproduct. Studies on the tert-
butoxy ] NO reaction conÐrm
a negative temperature
dependence and yield a room temperature rate coefficient of
2.5 ] 10~11 cm3 molecule~1 s~1. These results add to the
relatively small body of knowledge on alkoxy ] NO reactions
and suggest that the recommendations of Atkinson1 may
over-estimate ambient rate coefficients for the reactions of
larger alkoxy radicals with NO.
20 D. F. McMillan and D. M. Golden, Ann. Rev. Phys. Chem., 1982,
33, 493.
21 J. Troe, J. Phys. Chem., 1979, 83, 114.
22 J. Troe, Ber. Bunsen-Ges. Phys. Chem., 1983, 87, 161.
23 R. G. Geiber, K. Luther, and J. Troe, Ber. Bunsen-Ges. Phys.
Chem., 1983, 87, 169.
The decomposition of the tert-butoxy radical has been
studied directly as a function of pressure and temperature.
The reaction is in the fall-o† regime under the experimental
conditions using both a monoatomic bath gas and with an
additional polyatomic bath gas (methane). Limiting high pres-
sure rate coefficients were estimated via an ILT-ME technique
and the Troe factorisation technique. Due to the increased
sophistication of the ILT-ME method, we consider the param-
eters returned from this methodology as our best estimates for
24 Cerius2, version 3.7, Molecular Simulations Inc., 9685 Scranton
Road, San Diego, CA 92121È3752.
25 A. K. Rappe and W. A. Goddard, J. Phys. Chem., 1991, 95, 3358.
26 E. B. Wilson, J. C. Decius and P. C. Cross, Molecular V ibrations,
Dover, New York, 1980.
27 J. Gang, M. J. Pilling and S. H. Robertson, J. Chem. Soc.,
Faraday T rans., 1996, 92, 3509.
28 R. B. Bird, W. E. Stewart and E. N. Lightfoot, T ransport Pheno-
mena; John Wiley and Sons, New York, 1960, 511.
29 I. R. Slagle, L. Batt, G. W. Gmurczyk, D. Gutman and W. Tsang,
J. Phys. Chem., 1991, 95, 7732.
30 M. A. Hanning-Lee, N. J. B. Green, S. H. Robertson and M. J.
Pilling, J. Phys. Chem., 1993, 97, 860.
31 P. W. Seakins, S. H. Robertson, M. J. Pilling, I. R. Slagle, G. W.
Gmurczyk, A. Bencsura, D. Gutman and W. Tsang, J. Phys.
Chem., 1993, 97, 4450.
32 Y. Feng, J. T. Niiranen, A. Bencsura, V. D. Knyazev, D. Gutman
and W. Tsang, J. Phys. Chem., 1993, 97, 871.
33 A. F. Wagner and D. M. Wardlaw, J. Phys. Chem., 1988, 92,
2462.
calculating k=. The Arrhenius parameters from this study are
2
signiÐcantly lower than those from previous determinations or
calculations. High-level theoretical calculations are required
for alkoxy radical decompositions as the results of this study
seem to indicate that the mechanism could be more complex
than expected. Such calculations would be of immense help
for both the atmospheric and combustion communities as
higher alkoxy and polyfunctional alkoxy radicals (e.g. the b-
hydroxy alkoxy radicals formed in the atmospheric oxidation
of alkenes) are difficult to study experimentally.
34 K. Y. Choo and S. W. Benson, Int. J. Chem. Kinet., 1981, 13, 833.
35 P. Devolder, Ch. Fittschen, A. Frenzel, H. Hippler, G. Poskreby-
shev, O. F. Stiebel and B. Viskolcz, Phys. Chem. Chem. Phys.,
1999, submitted.
Acknowledgements
The authors are grateful for funding from NERC (GST/02/
1049) and the European Union (Contract No PL950731).
36 J. M. Park, N. W. Song and K. Y. Choo, Bull Korea, Chem. Soc.,
1990, 11, 343.
37 C. Sosa and H. B. Schlegel, Int. J. Quant. Chem., 1986, 29, 1001.
38 H. Knoll, G. Richter and R. Schliebs, Int. J. Chem. Kinet., 1980
12, 623.
References
1
2
R. Atkinson, Int. J. Chem. Kinet., 1997, 29, 99.
R. W. Walker and C. Morley, ““Basic Chemistry of CombustionÏÏ,
in L ow T emperature Combustion and Autoignition, ed. C. Morley
and M. J. Pilling, Comprehensive Chemical Kinetics, Elsevier, BV,
Amsterdam, 1997, vol. 35.
39 E. Kades, M. Rosslein and J. R. Huber, J. Phys. Chem., 1994, 98,
13556.
40 I. J. Wysong, J. Chem. Phys., 1994, 101, 2804.
41 I. R. Slagle, D. Gutman, J. W. Davies and M. J. Pilling, J. Phys.
Chem., 1988, 92, 2455.
42 J. W. Davies, N. J. B. Green and M. J. Pilling, J. Chem. Soc.,
Faraday T rans., 1991, 87, 2317.
3
4
Structure Activity Relationships for the Degradation of Biogenic
Volatile Organic Compounds (SARBVOC) EU Project
PL950731.
L. Batt, R. D. McCulloch and R. T. Milne, Int. J. Chem. Kinet.,
1975, 7, 441.
Paper 8/06524A
80
Phys. Chem. Chem. Phys., 1999, 1, 73È80