Picosecond IR study of UV-induced peroxide decomposition: Formation and vibrational relaxation of CO2 in CH2Cl2 solution

Article abstract of DOI:10.1524/zpch.2001.215.6.717

Picosecond UV pump (248.5 nm)/IR probe spectroscopy has been applied to the study of the decomposition of several aromatic diacyl peroxides, peroxycarbonates, and of a peroxyester dissolved in CH₂Cl₂. Measuring the IR transient absorbance in the 2100 to 2450cm^-1 range allows to monitor the formation and vibrational relaxation of the photoproduct CO₂ via the asymmetric stretching mode (ν₃) with a time resolution of 1.8 ps. With each of the six peroxides a CO₂ molecule is released at delay times below 10ps. The energy relaxation of the initially formed vibrationally hot CO₂ is followed over the time range up to 500 ps. Analysis of the transient IR spectra, via an anharmonic oscillator model proposed by Hamm et al. [1], shows a monoexponential decay of internal energy. Irrespective of the type of peroxide a single relaxation time of 67±5 ps is found to adequately represent the cooling behavior of CO₂ in liquid CH₂Cl₂. The initial temperatures of vibrationally hot CO₂ at a delay time of 20 ps after applying the pump pulse differ considerably, between 1400 and 2700 K for the decompositions of tert-butyl benzoyl peroxide and tert-butyl benzoyl carbonate, respectively. These remarkably high temperatures are assumed to originate from energy release associated with structural relaxation of the bent OCO moiety to form the linear CO₂ molecule. by Oldenbourg Wissenschaftsverlag, Muenchen.

Full text of DOI:10.1524/zpch.2001.215.6.717

Zeitschrift für Physikalische Chemie, 215, 6, 717–735 (2001)
by Oldenbourg Wissenschaftsverlag, München
Picosecond IR Study of UV-Induced Peroxide
Decomposition: Formation and Vibrational
Relaxation of CO2 in CH2Cl2 Solution
∗
By Michael Buback , Matthias Kling, Marco Thomas Seidel,
Frank Dieter Schott, Jörg Schroeder * and Ulrich Steegmüller
Institut für Physikalische Chemie der Universität Göttingen, Tammannstr. 6,
D-37077 Göttingen, Germany
Dedicated to Prof. Dr. Dr. h.c. mult. J. Troe
on the occasion of his 60th birthday
(Received August 3, 2000; accepted September 20, 2000)
Picosecond UV pump (248.5 nm)/IR probe spectroscopy has been applied to the study
of the decomposition of several aromatic diacyl peroxides, peroxycarbonates, and of
a peroxyester dissolved in CH2Cl2. Measuring the IR transient absorbance in the 2100
to 2450 cm−1 range allows to monitor the formation and vibrational relaxation of the
photoproduct CO2 via the asymmetric stretching mode (ν3) with a time resolution of
1.8 ps. With each of the six peroxides a CO2 molecule is released at delay times below
10 ps. The energy relaxation of the initially formed vibrationally hot CO2 is followed
over the time range up to 500 ps. Analysis of the transient IR spectra, via an anharmonic
oscillator model proposed by Hamm et al. [1], shows a monoexponential decay of internal
energy. Irrespective of the type of peroxide a single relaxation time of 67 5 ps is found
to adequately represent the cooling behavior of CO2 in liquid CH2Cl2. The “initial”
temperatures of vibrationally hot CO2 at a delay time of 20 ps after applying the pump
pulse differ considerably, between 1400 and 2700 K for the decompositions of tert-butyl
benzoyl peroxide and tert-butyl benzoyl carbonate, respectively. These remarkably high
temperatures are assumed to originate from energy release associated with structural
relaxation of the bent OCO moiety to form the linear CO2 molecule.
Introduction
Peroxides are widely used as initiators in free-radical polymerizations [2]. Re-
liable rate coefficients for thermally and photochemically induced peroxide
decomposition are required for modeling polymerization processes. A large
body of literature on the decomposition kinetics of organic peroxides has ac-
cumulated [3]. Nevertheless, the mechanistic details of their decomposition dy-
*
Corresponding authors. E-mail: mbuback@gwdg.de, jschroe2@gwdg.de

718
M. Buback et al.
namics are not yet adequately understood. Particular interest centers around the
mode of bond scission in peroxides which upon decomposition release a stable
small molecule, such as carbon dioxide. The primary dissociation event in per-
oxyesters, diacyl peroxides, and peroxycarbonates may be either the scission of
the O−O bond (with subsequent decarboxylation of the OCO containing inter-
mediate) or concerted two-bond scission associated with the instantaneous re-
lease of CO2. Picosecond UV pump/IR probe experiments are perfectly suited
for the analysis of such processes, as the formation of CO2 can be monitored
directly via the transient absorbance of the asymmetric stretching mode (ν3).
Among the four peroxides studied so far by this technique with excita-
tion at 308 nm, tert-butyl-peroxy-2-naphthylcarbonate (TBNC) was found to
decompose in a stepwise manner with a CO2 formation time of 31 ps [4],
whereas for di(1-naphthoyl) peroxide [4], di(2-naphthoyl) peroxide [5], and
for tert-butyl-9-methylfluorene-9-percarboxylate [6] CO2 appeared within the
experimental time resolution of 1.8 ps, suggesting concerted bond scission.
As the absorptivity of most peroxides is fairly low at 308 nm, it ap-
peared mandatory to modify the experimental system such that UV exci-
tation may be carried out at shorter wavelengths. The present paper illus-
trates the modification of the assembly for UV excitation at 248.5 nm, which
allowed us to carry out pump/probe experiments on dibenzoyl peroxide
(DBPO), di(4-methoxy benzoyl) peroxide (DMBP), tert-butyl benzoyl perox-
ide (TBPB), and tert-butyl benzoyl carbonate (TBBC). In order to study the
effect of excitation energy, tert-butyl-peroxy-2-naphthylcarbonate (TBNC) [4]
and di(2-naphthoyl) peroxide (DNPO) [5] which have already been investi-
gated at an excitation wavelength of 308 nm, were reinvestigated at 248.5 nm.
The structures of the individual peroxides are given in Scheme 1. As can be
seen in this scheme, DBPO, DMBP and DNPO belong to the diacyl perox-
ide family whereas both TBBC and TBNC are peroxycarbonates. It should
be noted that the diacyl peroxides contain two –C(O)O– moieties which may
release CO2. Within the present study only the formation of the first CO2,
which occurs in the ps or sub-ps range, is of interest. With diacyl peroxides
a second CO2 molecule may be released by thermal decarboxylation of the
intermediate aroyloxy free radical. This process occurs on the ns to s time
scale and has already been investigated by UV pump/VIS probe and EPR
experiments [7].
The modes of bond scission are illustrated in Scheme 2. k1 refers to
the rate coefficient for single-bond scission, whereas kcon denotes the rate
coefficient for concerted two-bond scission. krec and k2 characterize the re-
combination and decarboxylation rates of the intermediate free radical, re-
spectively. Beyond the fundamental interest to understand the detailed mech-
anism of the unimolecular fragmentation process, the mode of dissociation,
sequential or concerted, may strongly influence initiator efficiency in the
technical process of free-radical polymerization. In addition to studying the
mode of bond scission, the dynamics of energy relaxation for the reaction

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
719
Scheme 1. Peroxides under investigation: dibenzoyl peroxide (DBPO), di(4-methoxy ben-
zoyl) peroxide (DMBP), di(2-naphthoyl) peroxide (DNPO), tert-butyl benzoyl carbonate
(TBBC), tert-butyl-peroxy-2-naphthoylcarbonate (TBNC) and tert-butyl benzoyl peroxide
(TBPB).
products is also fundamental to the understanding of unimolecular reactions.
By monitoring the time evolution of the IR-band shape of the asymmet-
ric stretching mode (ν3), such relaxation measurements may be carried out
on CO2. The preceding UV pump/IR probe studies provided clear evidence
for the production of vibrationally hot CO2 (vide infra). The CO2 spectra
measured in CCl4 are in close agreement with data determined on vibra-
tionally hot CO2 in shock waves [8], in heated gas cells [9], and in su-
personic burners [10]. To deduce a vibrational temperature from the tran-
sient spectra, an anharmonic oscillator model applied by Hamm et al. appears
to be perfectly suited [1]. Vibrational relaxation in condensed fluid phase
has been extensively studied during recent years [11–13], but information
about the energy relaxation of highly excited small polyatomic molecules
like CO2 in solution is scarce [12, 14]. To the best of our knowledge we
report for the first time on energy tranfer of vibrationally excited CO2 in
a polar liquid.

720
M. Buback et al.
Scheme 2. Reaction scheme for the photoinduced peroxide decarboxylation: diacyl perox-
ide (R = aryl, Rꢀ = aroyl); peroxyester (R = aryl, Rꢀ = tert-butyl), and peroxycarbonates
(R = aryloxy, Rꢀ = tert-butyl); for explanation of the rate coefficients, k, see text.
Experimental section
The experimental setup for the time-resolved study of peroxide decompos-
ition induced by UV pump pulses at 308 nm and monitored via tunable IR
probe pulses in the 2000 to 5000 cm−1 region was described previously [15].
This device was modified for excitation at 248.5 nm. The main components
and characteristics of the new UV pump/IR probe spectrometer are illustrated
in Fig. 1.
The oscillator tube (filled with 70 mbar Xe, 90 mbar HCl (5% in He) and
2.34 bar Ne) of a twin tube excimer laser (Lambda Physik EMG 150E) is op-
Fig. 1. Schematic view of the laser setup with SHG: second harmonic generation, DFDL:
distributed feedback dye laser and DFG: difference frequency generation.

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
721
erated on the XeCl* line at 308 nm with a repetition rate of 10 Hz. The laser
radiation pumps a cascade dye laser system containing a distributed feedback
dye laser ( -DFDL) [16]. The -DFDL (0.1 M solution in dimethyl formamide
of the dye C307) provides laser pulses of about 1 ps width at 497 nm. These
pulses were amplified to 150 J by passing twice through both an off-axis am-
plification cell [17] (7 mM solution in propylene carbonate of the dye C102)
and a Bethune type [18] cell (3 mM solution in ethanol of the dye C307) with
a 2 mm bore.
The fundamental pulses at 497 nm were used for second harmonic gener-
ation (SHG) in a 0.3 mm thick BBO crystal (cut at an angle of 53◦ relative
to the optical axis). Double pass amplification in the second tube of the ex-
cimer laser operated on the KrF* line (gas mixture: 150 mbar Kr, 120 mbar F2
(5% in He) and 1.83 bar He) provides UV pump pulses at 248.5 nm with single
pulse energies up to 5 mJ. Pulse-to-pulse fluctuations are less than 5% and the
amplified spontaneous emission background is below 10%.
The remainder of the fundamental pulse at 497 nm (from SHG) serves for
pumping the prismatic DFDL [19] and for difference frequency generation
(DFG). The tunable prismatic DFDL, which was originally developed for the
emission range between 700 and 850 nm by Jasny [15], has been modified,
again in cooperation with Jasny, for the tuning range from 550 to 595 nm.
This modification required the construction of a laser cuvette with a front
prism made from fused silica at an angle of 60◦. Also the diffraction grat-
ing had to be changed, to 1400 lines/mm (blazed at 500 nm). The tunable
wavelength range from 552 to 566 nm is obtained with only one dye solu-
tion (5.8 mM solution in 1,1,1,3,3,3-hexafluoro-2-propanol of the dye R6G).
This signal pulse is amplified in an off-axis cell (13 mM solution in propylene
carbonate of the dye C153) to an energy of around 0.5 J at pulse durations
of 3 to 4 ps.
IR probe wavelengths ranging from 3 to 5.1 m are produced by DFG with
type-I phase matching of signal and fundamental pulses in a LiIO3 crystal of
5 mm thickness (cut at an angle of 22◦ with respect to the optical axis). Phase
matching upon changing signal wavelength was achieved by a computer con-
trolled rotation unit. Visible pulses behind the DFG are blocked with Si plates
(Fig. 2) adjusted at the Brewster angle (73.7◦). The IR probe pulses are focused
(lens L2, f = 300 mm) on a Ge beamsplitter, which divides the IR beam into
a sample and a reference beam of equal energy. Sample (S) and reference (R)
beam are transmitted through identical stainless steel flow cells with 100 m
optical path length.
The UV pump and IR probe beams overlap under an angle of 10◦ with the
focus of the UV beam (lens L1, f = 1000 mm) being slightly behind the sam-
ple cell. The polarization between pump und probe pulses was adjusted to 54.7◦
(magic angle) with a λ/2 plate. Delay times between pump and probe pulses
were adjusted by a computer controlled translation stage. By means of a dielec-
tric mirror (DM) pump energy was reduced to 2 mJ in order to avoid transient

722
M. Buback et al.
Fig. 2. Schematic view of the pump/probe assembly with DM: dielectric mirror, BS:
beamsplitter, PD: photodetector, S: sample cell and R: reference cell.; Inset: transient IR
absorption of Germanium at 2300 cm−1 after excitation at 248.5 nm (see text).
signals and damage at the front window (MgF2, 2 mm thick) of the sample cell.
The backside window of the sample cell and both windows of the reference cell
consisted of 5 mm thick CaF2.
Probe and reference IR beams were collimated with CaF2 lenses (L3,
f = 100 mm) and focused onto liquid-nitrogen-cooled InSb photo-detectors
PD (EG&G J10D M-204). Electronic signals from the photodiodes were am-
plified and digitized with an A/D converter. Typically 50 individual laser shots
were averaged to obtain a single absorbance data point for either constant delay
time or constant wavenumber. The detection limit for changes in optical density
was about OD = 0.001.
The cross-correlation width and the zero point delay between pump and
probe pulses were deduced by independently measuring the transient absorp-
tion of Ge plates at 2300 cm−1 after excitation at 248.5 nm. Experimental data
points together with a simulated curve are shown in the inset in Fig. 2. The
simulation consists of a convolution of the system response with the Gaussian
cross-correlation function. From the rise time of the absorption measurement,
a time resolution for the pump/probe experiment of 1.8 ps has been deduced.
The spectral width of the IR pulses is 10 to 12 cm−1, the wavenumber accur-
acy is 5 cm−1. The precision within one experimental spectrum is better
than 1 cm−1.
The peroxide samples (Scheme 1) were synthesized and provided by
AKZO Nobel (Research Center Deventer, The Netherlands) in high purity.

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
723
With CH2Cl2 (Merck, Uvasol) being used as the solvent, the transient sig-
nals at high pump intensities were very small. The entire set of experiments
was carried out at ambient pressure and temperature. The concentration of
the sample solutions was adjusted to an optical density at the pump wave-
length of OD(248.5 nm) = 10 for di(2-naphthoyl) peroxide (DNPO) and of
OD (248.5 nm) = 2 for the other peroxides. Under these conditions, the num-
ber of molecules in the pumped volume exceeds the number of photons in the
pump pulse by at least a factor of 10, which assures that bleaching effects will
not occur.
Results
Transient IR absorption measured after ps photo-excitation of tert-butyl ben-
zoyl carbonate (TBBC) in CH2Cl2 at delay times between −40 and 500 ps
are shown in Fig. 3. Decay curves were measured at fixed wavenumber every
20 cm−1 across the ν3-band and combined to yield the 3-dimensional plot
shown in Fig. 3. Transient background absorption by the solvent and the win-
dow material were subtracted according to the procedure described below.
At negative delay times the signal is dominated by fluctuations within the
detection limit of OD = 0.001. Slightly enhanced transient signals occur
between 2320 and 2350 cm−1. They are due to effects of atmospheric CO2
on the pulse profile [20]. At zero delay time, an instantaneous rise in optical
density is seen in the spectral range from 2170 to 2330 cm−1. With increasing
delay time the absorption band simultaneously shifts to higher wavenumbers
and narrows. After 500 ps the band maximum is located at 2335 cm−1 (with
a spectral width of 19 cm−1) with this position being in good agreement with
the one measured for a stationary CO2 spectrum in CH2Cl2 at ambient tem-
perature. Studying the same solution in an FTIR spectrometer (Bruker IFS 88)
yields a band maximum position of 2338 cm−1 and a band width of 8 cm−1,
which is in full agreement with literature values [21]. The difference in the ab-
solute band position is within the limits of accuracy of the ps spectrometer. The
spectral width of the ps band is reproduced by convolution of the FTIR band
profile with the spectral broadening function of the ps IR pulses (11 cm−1). The
measured absorption band at longer delay times thus can unambiguously be
assigned to the dissociation product CO2.
Fig. 4 shows spectra measured at delay times of 20, 100, and 500 ps after
photolysis of TBBC in CH2Cl2. The lines fitted to the data will be explained in
the discussion section. Shifting and broadening of CO2 absorption as in Figs. 3
and 4 was observed for all peroxides under investigation. Absorption bands
recorded at delay times greater than 300 ps remain unchanged for all peroxides.
At wavenumbers between 2410 and 2450 cm−1, time dependent signals
were observed which could not be assigned to the absorption of CO2. Fig. 5
shows signals for this range as a function of delay time after photolysis of

724
M. Buback et al.
Fig. 3. Temporal (−40 to 500 ps) and spectral (2170 to 2410 cm−1) evolution of the ν3
absorbance (optical density) of CO2 after UV-induced decomposition of TBBC in CH2Cl2.
Fig. 4. Transient spectra of the ν3 absorbance of CO2 at three delay times (20, 100, and
500 ps) after UV-induced decomposition of TBBC in CH2Cl2.

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
725
Fig. 5. Background signal to the blue of the CO2 ν3-band: time-dependent transient ab-
sorbance in the 2410 to 2450 cm−1 range after UV-induced decomposition of DBPO in
CH2Cl2.
DBPO in CH2Cl2. Almost the same signals are observed in ps experiments car-
ried out on pure CH2Cl2. They are assigned to transient absorption of the front
window and of the solvent. The pure solvent signals cannot be used directly
for correction of the transient spectra recorded during peroxide decomposition
as the absorbed energy is different in both experiments. The time-dependent
transients in Fig. 5 were averaged and fitted biexponentially to yield the wave-
length independent transient background signal given as the intense solid line
in Fig. 5. The shorter one of the two half-lives associated with the biexponential
fit is 8 2 ps. The second half life is in the nanosecond region. A simu-
lated background signal is subtracted from each of the time-dependent traces
recorded during the peroxide decomposition experiments.
Fig. 6 presents time-dependent traces measured for the decarboxylation of
DBPO at 2335 cm−1 (which is the peak position at delay times > 300 ps) and at
2250 cm−1. The simulated background signal from Fig. 5 is included in Fig. 6.
It is clearly seen, that the fast component of the background signal is significant
only up to 10 ps but remains below 40% of the overall signal intensity at 2250
and 2335 cm−1. After 20 ps, the background absorption gives rise to a nearly
constant offset of OD = 0.001. This offset agrees well with the signal intensity
at 2250 cm−1 and long delay times. The signal at 2335 cm−1 and negative delay
times is primarily due to the action of atmospheric CO2.
The time dependent signals at both 2250 and 2335 cm−1 reflect the spec-
tral and temporal evolution of the CO2 band: Toward increasing delay time,
the band shifts to higher wavenumbers, resulting in an absorption decay at

726
M. Buback et al.
Fig. 6. Time-dependent transient absorbance at 2250 cm−1 and 2335 cm−1 after decompos-
ition of DBPO; the solid line denotes the simulated signal from Fig. 5.
2250 cm−1 (after the instantaneous rise at time zero) and in an absorption
increase at 2335 cm−1.
Discussion
The absorption observed in the wavenumber region from 2100 to 2400 cm−1
(Figs. 3 and 4) is exclusively assigned to CO2 from photolysis of the perox-
ides. The temporal and spectral evolution after photolysis of the asymmetric
CO2 stretching vibration (ν3) is determined by both the mode of bond scission
and, more importantly, by the energy relaxation of the vibrationally excited
molecule. In order to clarify the decomposition mechanism, the integrated ab-
sorbance of the entire spectral region from 2100 to 2400 cm−1 will first be
studied. Subsequently the band profile of the transient spectra will be investi-
gated in an attempt to derive the internal vibrational energy of CO2 and the time
constant of vibrational relaxation in the liquid CH2Cl2 solvent.
A. Decomposition mechanism
In contrast to molar absorption coefficients, ε(ν), at a fixed wavenumber ν,
˜
˜
ꢀ
which are temperature dependent, the vibrational band intensity B =
ε(ν)dν
˜
˜
ν
3
is constant or only weakly temperature dependent [22]. Thus integrated ab-
A(ν)dν , may serve as a measure for CO concentration.
Integrated absorbances, B*, are determined as the area under the optical density
ꢀ
sorbance, B* =
˜
˜
2
ν
3
curves measured at fixed delay time.

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
727
In Fig. 7, the resulting numbers are plotted versus delay time for the pho-
tolysis of TBBC and DBPO. Subtraction of the background signal may not
completely eliminate the transient background absorption within the first 10 ps
of the detection period.
In Fig. 7, integrated absorbances are seen also at negative delay times. They
result from effects on the pulse profile of the atmospheric CO2 that is contained
in the IR light path [20]. The transient background signal reduces the time reso-
lution in the initial period after applying the pump pulse. The maximum value
of integrated absorbance is reached within 10 ps for each of the two peroxides.
A weak decay is seen toward larger delay times.
The other peroxides studied also reach maximum integrated absorbance
within the first 10 ps. This type of behaviour, with respect to the available time
resolution of the experimental setup, is considered as instantaneous decompos-
ition. Absorption of CO2 that is formed during (successive) thermal decarboxy-
lation of intermediates on a ns – s time scale [7] will not contribute in the time
Fig. 7. Integrated absorbance of the ν3 band of CO2 as a function of delay time for DBPO
and TBBC decomposition in CH2Cl2.

728
M. Buback et al.
range up to 500 ps [4, 6]. The slight decrease of B* in the 20 to 500 ps range
(Fig. 7) most likely results from a weak temperature dependence of B.
The B* vs. delay time data in Fig. 7 indicate that with TBBC the formation
of CO2 and thus the breaking of two bonds takes place in either a concerted or
in a very fast sequential manner. With DBPO decomposition, one of the two
CO2 molecules that may be produced, is released in a similar concerted or fast
sequential way whereas the second CO2 molecule will be formed thermally on
a much longer time scale. Fast CO2 release within the first 10 ps is also seen
in the UV-induced decomposition of DMBP, DNPO, TBNC and TBPB. A de-
layed formation of CO2, as has been observed in the previous UV pump/IR
probe study into TBNC at a pump wavelength of 308 nm, does not occur at
an excitation wavelength of 248.5 nm. This difference is most likely due to the
larger excess energy in the 248.5 nm experiment.
It needs to be kept in mind that the notation “concerted or very fast sequen-
tial” decomposition used within the present paper refers to the time resolution
of a few picoseconds. It is obviously a matter of priority to further enhance the
time resolution of the UV pump/IR probe experiment into the sub-ps range.
Such efforts are currently underway.
B. Modeling of transient CO2 spectra
The CO2 spectra recorded at delay times up to 300 ps are due to vibrationally
hot CO2. It appears interesting to estimate the vibrational temperature from the
measured ν3-band profiles. This may be achieved via an anharmonic oscillator
model presented by Hamm et al. [1], in which band broadening is interpreted as
being due to anharmonic coupling among vibrationally excited modes. These
authors pointed out that the experimental spectra may serve as a molecular ther-
mometer for a dissolved species. Rotational excitation is not included in the
model which is certainly justified under conditions where rotational relaxation
is extremely fast.
The spectrum A (ω) of a particular normal mode ω may be calculated
˜
k
k
in situations where the anharmonicity constants, xik, and the relative popula-
tions of the n levels of mode k, a(nk), are known. For the ν3 mode of CO2, the
expression reads:
ꢂ
S
S
S
ꢁ
ꢁ ꢁ
A (ω) =
[a(n3)−a(n3 +1)](n3 +1)·
a(n1) a(n2)
˜
3
n
=0
n
=0 n =0
3
1
2
ꢃ
· g [ω−ν(n → n +1)] ,
(1)
˜
˜
3
3
3
where S is the highest vibrational level that is considered in the estimate. The
calculations included vibrational levels up to S = 10. The line shape function,
g (ν), accounts for the band contour in an environment of liquid CH Cl . FTIR
˜
3
2
2

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
729
studies on the ν3 mode of CO2 in CH2Cl2 at ambient temperature revealed that
a Lorentzian profile with a half-width of 8 cm−1 is a suitable line shape func-
tion. g (ν) is assumed to be independent of temperature. The populations a(n )
˜
3
k
depend on the internal temperature of the molecule according to the Boltzmann
distribution. Above 1000 K, excited levels of the (high-frequency) ν3 mode
are less populated, in contrast to the excited levels of the two low-frequency
modes, ν1 and ν2. The contour of the overall band varies with temperature as
a consequence of changes in the populations a(n ). The wavenumber ν for the
˜
k
nk → nk +1 transition of mode k is coupled to the other modes i
ꢁ
= k according
to Eq. (2):
ꢁ
ꢁ
xik
ν(n → n +1) = ω +2x +
+2xkknk +
xikni .
(2)
˜
˜
k
k
k
kk
2
i
ꢁ
=k
i
ꢁ
=k
Via Eq. (1), absorbance spectra A (ω) were calculated as a function of
˜
3
temperature. The harmonic frequencies are [23]: ω = 1354.07 cm−1, ω =
˜
˜
1
2
672.95 cm−1 and ω = 2396.30 cm−1; the anharmonicity constants are [24]:
˜
3
x11 = −3.10 cm−1, x22 = 1.59 cm−1, x33 = −12.50 cm−1, x12 = −5.37 cm−1,
x13 = −19.27 cm−1 and x23 = −12.51 cm−1. A frequency shift of 11 cm−1 has
been applied to the gas phase phase frequencies given above to adjust for the
effect of solvent environment [21].
Shown in Fig. 8 is the calculated absorbance band contour (in arbitrary
units) for the ν3 band of CO2 at 1000 K. The non-diagonal anharmonic terms
strongly contribute to the overall bandshape. In addition to the maximum peak
at 2336 cm−1 distinct side bands occur toward lower wavenumbers. As a con-
sequence of the negative sign of the anharmonicity constants, the peak position
is red-shifted relative to the spectrum taken at 300 K. The spectra calculated
via Eqs. (1) and (2) were convoluted with a Lorentzian line shape function of
half-width 11 cm−1 to account for the bandwidth of the IR probe pulse. The
procedure results in a significantly broadened spectrum given by the solid line
in Fig. 8.
Fig. 9 shows a series of spectra calculated for temperatures between 300
and 3300 K at intervals of 200 K. Each spectrum is convoluted with the IR
pulse spectral profile. With decreasing temperature, a narrowing of the ab-
sorption band and a shift to larger wavenumbers are observed. The arrows
indicate the direction of spectral change upon lowering temperature. The se-
ries of spectra resembles measured sequences of spectra as the ones shown
in Fig. 3.
Comparison of simulated and measured spectra allows to assign a vibra-
tional temperature to indiviual spectra determined at different delay times.
In doing so, the temperatures for the experimental TBBC spectra in Fig. 4
taken at delay times of 20, 100, and 500 ps are obtained to be 2700 200 K,
1000 100 K, and 300 50 K, respectively. The uncertainties in estimated
(model-dependent) temperature, of about 10%, reflect the quality by which

730
M. Buback et al.
Fig. 8. Simulated CO2 spectrum for 1000 K before and after convolution with the IR pulse
profile.
Fig. 9. Simulated CO2 spectra (convoluted with IR pulse profile) between 300 and 3300 K,
the arrows indicate the change with decreasing temperature.
the experimental spectra are simulated via the anharmonic oscillator model.
The “true” temperature certainly may be off by more than 10%. The lines
in Fig. 4 are the simulated curves for 2700, 1000 and 300 K. The agreement
between measured and simulated spectra is very satisfactory within the entire
range of the ν3 band. It is particularly noteworthy that the anharmonic oscillator
model fully accounts for the influence of temperature on both bandshape and

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
731
Fig. 10. Time dependence of the vibrational temperature of CO2, released in the UV-
photoinduced decomposition of TBBC and DBPO in CH2Cl2; the solid lines are monoex-
ponential fits (see text).
peak position. It may be concluded that the band analysis of the CO2 ν3 mode
indeed provides a reasonable estimate of internal temperature.
According to the procedure described above, a temperature is assigned to
each of the transient spectra. The results for DBPO and TBBC are shown in
Fig. 10. For both peroxides the internal temperature vs. delay time data are ad-
equately represented by a monoexponential fit (solid lines). The relaxation half
lives are 67 4 ps (DBPO) and 64 2 ps (TBBC). The vibrational tempera-
tures at a delay time of 20 ps are determined to be 1800 200 K (DBPO) and
2700 200 K (TBBC).
In Fig. 11, the time dependence of CO2 internal temperature after peroxide
decomposition in CH2Cl2 is plotted for the entire set of experiments. The re-
laxation data clearly reflects peroxide structure. The peroxycarbonates TBBC
and TBNC produce CO2 with highest temperatures, around 2700 and 2600 K at
t = 20 ps. For the three diacyl peroxides, the CO2 temperature at t = 20 ps is in
the 2000 K (DNPO and DMBP) to 1800 K (DBPO) range. For TBPB the low-
est “initial” CO2 temperature is derived, which is close to 1400 K at t = 20 ps.
Again a monoexponential decay of temperature vs. delay time is observed.
A single such decay curve (solid line in Fig. 11) is sufficient to represent the
relaxation behavior of CO2 formed in the fragmentation of peroxides which are
similar in the relevant part of their molecular structure.
The monoexponential relaxation times τ and the “initial” temperatures for
t = 20 ps as obtained from the analysis of the individual transient absorbance
spectra are summarized in Table 1.
Irrespective of “initial” temperature, which varies between 1400 K (TBPB)
and 2700 K (TBBC), within experimental error the cooling of vibrationally hot

732
M. Buback et al.
Fig. 11. Time dependence of the vibrational temperatures of CO2, released in the UV-
photoinduced decomposition in CH2Cl2 of the entire set of peroxides studied within the
present work; the solid lines represent monoexponential fits for peroxides which belong to
the same family.
CO2 in CH2Cl2 occurs with a time constant of 67 5 ps. This result is consis-
tent with the observation of exclusively monoexponential cooling curves. The
monoexponential cooling behavior is not surprising, since the initial tempera-
ture assigned to CO2 corresponds to a relatively small degree of excitation of
the CO2 normal modes, i.e. more than half of the population are typically found
at n3 = 0, n2 = 1, and n1 = 0. In this range of excess energies vibrational re-
laxation is expected to follow a monoexponential decay [25, 26]. That such
identical cooling behavior is observed, provides strong support for the internal
temperature determination via the anharmonic coupling model (Eq. 1).
Consequently the “initial” temperatures at delay times of 20 ps can also be
regarded with some confidence. These temperatures, which differ significantly
for the three groups of peroxides, must be related to the maximum available
Table 1. Estimated temperatures T (at a delay time of 20 ps) and cooling relaxation times
τ of CO2 from peroxide decomposition.
CO2 source
T (at t = 20 ps)/K
τ/ps
TBBC
TBNC
DNPO
DMBP
DBPO
TBPB
2700 200
2600 200
2000 200
2000 200
1800 200
1400 100
64
2
4
4
6
4
9
70
64
67
67
69

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
733
excess energy and the extent of vibrational energy redistribution within the
peroxide molecule prior to bond scission.
Details of the photodissociation dynamics on a sub-ps scale are not ac-
cessible with the current experimental setup. Moreover, information about the
electronic surface, on which the dissociation takes place, is not available. Never-
theless, a rough estimate of the maximum CO2 temperature attainable in the
reaction can be made. If one assumes that intramolecular vibrational energy
redistribution (IVR) in the peroxide molecule prior to the primary dissocia-
tion step is complete, the vibrational excess energy in the ground state may be
obtained from a harmonic oscillator model considering the pump photon en-
ergy [4]. With vibrational frequencies taken from ab initio calculations [27],
the peroxide internal energies after excitation correspond to vibrational tem-
peratures of the ground state, e.g., of about 1400 K for DBPO and 1300 K
for TBBC. Subtraction of approximate energies associated with the disso-
ciation of the peroxy-bond (100 kJ mol−1) and the C (α)–C (carbonyl)-bond
(30 kJ mol−1) [3, 5], yields, as an upper limit for the internal temperature of CO2
produced from fully thermalized excited peroxide molecules, values around
1100 K for both peroxides. This temperature is significantly lower than the vi-
brational temperatures estimated from the band profile at a delay time of 20 ps.
A reason for this discrepancy might be that dissociation occurs from a not
fully thermalized state. It is, however, fairly unlikely that IVR within the perox-
ide molecule will take longer than a few picoseconds [11]. In addition, previous
UV pump (308 nm)/IR probe experiments on TBNC showed that even in case
of stepwise decomposition (where CO2 is formed with a time constant of about
31 ps, when IVR will definitely be completed) the vibrational temperature is
1600 K above the value estimated for dissociation from a thermalized state [4].
Another explanation for the high “initial” temperature assumes that a sig-
nificant fraction of the energy release, associated with linear CO2 formation
from a bent precursor species, remains in the CO2 molecule. Rough estimates
of the energy release associated with angle and bond length relaxation during
CO2 formation indicate that temperatures as high as the ones deduced from the
transient spectra may easily be reached. Similar conclusions have been derived
from the photodissociation at 266 nm of s-triazine where a particularly high
excitation of the bending mode of the HCN product is seen [28]. Zhang and
Hossenlopp investigated the photoinduced (at 222 nm) decarboxylation of per-
fluorinated anhydrides which leads to the formation of CO2 [29]. Analysis of
the population of the CO2 bending levels, from 0110 up to 0440, yields a Boltz-
mann temperature of 3100 1000 K whereas the vibrational temperature de-
duced from the asymmetric stretching mode (0001) was noticeably lower.
The argument that the high vibrational temperatures of CO2 essentially re-
sult from the energy associated with structural relaxation of the bent O−C−O
moiety can also easily explain the very high temperatures observed after
photo-decomposition (at 308 nm) of TBNC. With the O−C−O moiety and the
aromatic ring being separated by two single bonds, the dissipation of excess

734
M. Buback et al.
energy from CO2 formation is less favored than in situations with conjugated
double bonds as in the diacyl peroxides and the aromatic peroxyester. The
structural relaxation argument also explains why the peroxides clearly exhibit
a family-type behavior with respect to the “initial” temperature of the vibra-
tionally hot CO2.
Conclusion
Analysis of the transient IR spectra of CO2 released during ps photodecom-
position of several aromatic peroxides, via an excited anharmonic oscillator
model, provides the time dependence of vibrational temperature. Irrespective
of the type of peroxide and thus of “initial” temperature, a single monoexpo-
nential relaxation adequately describes the cooling of CO2 in the environment
of liquid CH2Cl2. Further peroxide decomposition experiments should be car-
ried out with other solvents to quantify the influence of the molecular surround-
ing on intermolecular energy relaxation. Preliminary studies in liquid CCl4
indicate that the relaxation rate is much slower and occurs with a time con-
stant of about 180 20 ps. Lower cooling rates in CCl4 as compared to other
chlorinated hydrocarbons have also been reported for energy relaxation of sev-
eral small [12] and large [11] molecules in these solvents. The remarkably high
“initial” temperatures can be best understood by assuming that the large excess
energy mainly results from structural relaxation of the bent O−C−O moiety.
It appears to be a matter of priority to further improve the time resolution of
the UV pump/IR probe experiment. Experiments with sub-ps time resolution
are required in order to unambiguously distinguish between single bond scis-
sion and concerted two-bond scission.
Acknowledgement
This work has been generously supported by the Deutsche Forschungsgemein-
schaft within the Sonderforschungsbereich 357 (“Molekulare Mechanismen
unimolekularer Prozesse”). Helpful comments by Dr. D. Schwarzer are grate-
fully acknowledged. The authors are also grateful to AKZO NOBEL for the
synthesis and provision of the peroxide samples.
References
1. P. Hamm, S. M. Ohline and W. Zinth, J. Chem. Phys. 106 (1997) 519.
2. A. Rudin, The Elements of Polymer Science and Engineering, Academic Press,
Orlando (1982).
3. R. Hiatt, in: Organic Peroxides, Vol. II, ed. D. Swern, Wiley Interscience, New York
(1971) 799; S. Patai, in: The Chemistry of Peroxides, Wiley, New York (1983);

Picosecond IR Study of UV-Induced Peroxide Decomposition ...
735
K. Fujimori, in: Organic Peroxides, ed. W. Ando, Wiley, New York (1992) 319;
Y. Sawaki, in: Organic Peroxides, ed. W. Ando, Wiley, New York (1992) 425;
M. Buback and C. Hinton, Z. Phys. Chem. 193 (1996) 61; M. Buback and J. Sand-
mann, Z. Phys. Chem. 214 (2000) 583; and references cited in these articles.
4. J. Aschenbrücker, M. Buback, N. P. Ernsting, J. Schroeder and U. Steegmüller, Ber.
Bunsenges. Phys. Chem. 102 (1998) 965.
5. U. Steegmüller, Dissertation, Göttingen (1997).
6. J. Aschenbrücker, M. Buback, N. P. Ernsting, J. Schroeder and U. Steegmüller,
J. Phys. Chem. B 102 (1998) 5552.
7. S. Yamauchi, N. Hirota, S. Takahara, H. Sakuragi and K. Tokumaru, J. Am. Chem.
Soc. 107 (1985) 5021; L. Grossi, J. Lusztyk and K. U. Ingold, J. Org. Chem.
50 (1985) 5882; G. C. Flowers and J. E. Leffler, J. Org. Chem. 54 (1989) 3995;
D. E. Falvey and G. B. Schuster, J. Am. Chem. Soc. 108 (1986) 7419; J. Wang,
T. Tateno, H. Sakuragi and K. Tokumaru, J. Photochem. Photobiol. A: Chem. 92
(1995) 53; and references cited in these articles.
8. M. Steinberg and W. O. Davies, J. Chem. Phys. 34 (1961) 1373.
9. R. H. Tourin, J. Opt. Soc. Am. 51 (1961) 175.
10. C. C. Ferriso, J. Chem. Phys. 37 (1962) 1955.
11. T. Elsaesser and W. Kaiser, Annu. Rev. Phys. Chem. 42 (1991) 83.
12. J. C. Owrutsky, D. Raftery and R. M. Hochstrasser, Annu. Rev. Phys. Chem. 45
(1994) 519.
13. A. Tokmakoff, B. Sauter, A. S. Kwok and M. D. Fayer, Chem. Phys. Lett. 221
(1994) 412; T. P. Dougherty and E. J. Heilweil, Opt. Lett. 19 (1994) 129; A. Tok-
makoff and M. D. Fayer, J. Chem. Phys. 103 (1995) 2810; D. Schwarzer, J. Troe,
M. Votsmeier and M. Zerezke, J. Chem. Phys. 105 (1996) 3121; P. Hamm, M. Lim
and R. M. Hochstrasser, J. Chem. Phys. 107 (1997) 10523; L. K. Iwaki, J. C. Deàk,
S. T. Rhea and D. D. Dlott, Chem. Phys. Lett. 303 (1999) 176; G. Seifert, R. Zürl,
T. Patzlaff and H. Graener, J. Chem. Phys. 112 (2000) 6349; and references cited in
these articles.
14. D. Budac and P. Wan, J. Photochem. Photobiol. A: Chem. 67 (1992) 135, and refer-
ences therein.
15. J. Aschenbrücker, U. Steegmüller, N. P. Ernsting, M. Buback, J. Schroeder and
J. Jasny, Appl. Phys. B 65 (1997) 441.
16. S. Szatma´ri and F. P. Schäfer, Opt. Commun. 68 (1988) 196.
17. T. Bultmann, D. Bingemann, N. P. Ernsting, D. Schwarzer and L. Nikowa, Rev. Sci.
Instrum. 66 (1995) 4393.
18. D. S. Bethune, Appl. Opt. 20 (1981) 1897.
19. J. Jasny, Opt. Commun. 53 (1985) 238; J. Jasny, Rev. Sci. Instrum. 57 (1986) 1303.
20. M. Woerner, A. Seilmeier and W. Kaiser, Opt. Lett. 14 (1989) 636.
21. D. B. Cunliffe-Jones, Spectrochim. Acta A 25 (1969) 779.
22. M. Buback, J. Schweer and H. Tups, Z. Naturforsch. 41a (1986) 505; M. Buback,
J. Schweer and H. Tups, Z. Naturforsch. 41a (1986) 512.
23. I. Suzuki, J. Mol. Spectrosc. 25 (1968) 479.
24. H. R. Gordon and T. K. McCubbin, J. Mol. Spectrosc. 19 (1966) 137.
25. C. Heidelbach, V. S. Vikhrenko, D. Schwarzer and J. Schroeder, J. Chem. Phys. 110
(1999) 5286.
26. I. I. Fedchenia, G. Käb and J. Schroeder, to be published.
27. Method and basis set: HF/6-31G(d) (frequencies scaled with 0.8929), Gaussian 98,
Revision A.5, Gaussian, Inc., Pittsburgh PA (1998).
28. J. Park, Chem. Phys. Lett. 293 (1998) 383.
29. Y. Zhang and J. M. Hossenlopp, Chem. Phys. Lett. 275 (1997) 319.