In situ Raman spectroscopy of the decomposition of tertiary butyl peroxy pivalate under high pressure and high temperature

Article abstract of DOI:10.1366/0003702011951669

The decomposition of tertiary butyl peroxy pivalate (TBPPI), in n-hexane, under a pressure of 1900 bar and at a temperature of 100 °C, was studied in situ with the use of a transportable high-through-put fiber-optic Raman spectrometer. The Raman probe was conveniently coupled to an optical high-pressure cell containing the TBPPI. Time-resolved experiments with a 10 wt% solution of TBPPI in n-hexane were carried out, and information on the mechanism of decomposition and the (first-order) kinetics was obtained. It is shown that the main decomposition products of TBPPI are CO₂ and t-butanol. From first-order kinetic fits, of the Raman profiles of intensity vs. time of reactants and products, the rate constant for the decomposition of TBPPI and the formation of CO₂ is found to be 1.65-10^-3s^-1 (half-life time for TBPPI of approximately 420 s). t-Butanol formation occurs at a slower rate of approximately 1.22.10^-3s^-1 (characteristic time of 566 s).

Full text of DOI:10.1366/0003702011951669

In Situ Raman Spectroscopy of the Decomposition of
Tertiary Butyl Peroxy Pivalate under High Pressure and
High Temperature
¨
´
DETLEF FREITAG, CHRISTIAN GOTZ, GERHARD LUFT, ANDRE VAN DER POL,*
and KLAAS WIERDA
Department of Chemical Technology, Darmstadt University of Technology, Petersenstraße 20, D-64287 Darmstadt, Germany (D.F.,
C.G., G.L.); and DSM Research, P.O. Box 18, 6160 MD Geleen, The Netherlands (A.V., K.W.)
The decomposition of tertiary butyl peroxy pivalate (TBPPI), in n-
hexane, under a pressure of 1900 bar and at a temperature of 100
8C, was studied in situ with the use of a transportable high-through-
put Ž ber-optic Raman spectrometer. The Raman probe was con-
veniently coupled to an optical high-pressure cell containing the
wavenumber were obtained. In the present study com-
TBPPI. Time-resolved experiments with a 10 wt % solution of
TBPPI in n-hexane were carried out, and information on the mech-
beneŽ t prompted our attempt to characterize the kinetics
and the mechanism of TBPPI decomposition.
The decomposition of TBPPI has been studied previ-
ously;2 however, in that study intensity proŽ les at a Ž xed
plete Raman spectra are measured at each time interval.
Careful analysis of such spectral data sets offers insight
into the reactions and the reaction mechanisms.
anism of decomposition and the (Ž rst-order) kinetics was obtained.
It is shown that the main decomposition products of TBPPI are
CO2 and t-butanol. From Ž rst-order kinetic Ž ts, of the Raman pro-
Ž les of intensity vs. time of reactants and products, the rate constant
for the decomposition of TBPPI and the formation of CO2 is found
EXPERIMENTAL
2
2
1
to be 1.65·10 3s (half-life time for TBPPI of approximately 420 s).
8
Previous experiments were carried out in a 90 scat-
tering conŽ guration for which two optical windows are
necessary.2 The new Raman spectrometer, which is de-
scribed below, operates in a back-scattering conŽ gura-
tion, in which only one window is needed. Our revised
optical cell, consisting of corrosion-proof steel, is illus-
trated in Fig. 1. The cell, of 2 mL volume, can bear
pressures of up to 2000 bar and temperatures of up to
t-Butanol formation occurs at
a slower rate of approximately
2
2
1
1.22·10 3s (characteristic time of 566 s).
Index Headings: Raman; Peroxide; Decomposition; High pressure.
INTRODUCTION
In situ
monitoring of reactions in an optical cell by
8
200 C (2) and contains only one pressure-resistant sap-
means of vibrational spectroscopy is an elegant and rel-
atively fast way of investigating kinetics. Accurate mea-
surements of high reaction rates are possible. For pro-
cesses at high pressures (kbar), infrared and near-infrared
have been applied by Buback and Hinton,1 while Raman
spectroscopy has been applied by Kessler et al.2 Raman
spectroscopy has the advantage that sapphire cell win-
dows can be used. Sapphire possesses the advantage of
stability up to pressures of 12 000 bar, but has the dis-
advantage of being transparent for electromagnetic radi-
phire window (thickness 10 mm) mounted in a special
alloy (1). The temperatures of the steel mantle and within
the reaction chamber are measured by means of Ni–Cr/
Ni thermocouples, which are Ž xed in a drill hole of the
mantle and in the reaction chamber through one of the
bores (4). Electrical heating is facilitated through four
heating rods (3) placed in the jacket. The bores (4) are
also used for Ž lling and draining the cell.
Through the boring of the high-pressure screw joint
(5) the probe of the spectrometer is attached to the cell,
with the focusing lens positioned in front of the sapphire
window. The spectrometer is a Hololab system, manu-
factured by Kaiser Optical Systems, Inc. Wavelength cal-
ibrations are done automatically. An optical Ž ber of 2 m
length was connected to a Raman probe (HoloProbe,
Model IMO-SS-0.1, Kaiser Optical Systems, Inc.). The
laser light leaves the probe as a parallel beam. A home-
made insertion probe was constructed of brass. The probe
consisted of a pipe of outer diameter of 4.5 mm and inner
diameter of 2.7 mm. At the end of the pipe a diode laser
plano-convex glass lens (Melles Griot, LXP001) was
Ž xed with cyanoacrylate glue. The insertion probe was
designed to Ž t onto the Holoprobe, via a 0.800-36 (in.)
thread, and into the opening of the high-pressure cell win-
dow, close to the sapphire window (see Fig. 1, item 5).
The lens focuses the laser light at 14 mm, through the
sapphire window into the reaction volume. The Raman
light is also collected through the insertion probe (back-
scattering geometry), and a second optical Ž ber delivers
2
1
ation down to 2000 cm only, masking a large part of
the infrared and near-infrared region. In this study we
combine a modern Ž ber-optic Raman spectrometer with
an optical high-pressure cell into an easy setup for the
study of high-pressure reactions.
The potentials of this technique will be demonstrated
by a study of the decomposition of tertiary butyl peroxy
pivalate (TBPPI) at a pressure of 1900 bar and a tem-
8
perature of 100 C. Organic peroxides are used as sources
of free radicals to initiate polymerization of oleŽ ns in
high-pressure polymerization.3 The process typically
takes place at pressures of 1000 to 3000 bar and temper-
8
atures of 130 to 300 C. The peroxides decompose ex-
tremely rapidly at such high temperatures. In order to
understand, or even predict, the polymerization kinetics
and the reactor behavior it is desirable to have access to
reliable kinetic parameters of all reactions involved. This
Received 10 April 2000; accepted 13 October 2000.
* Author to whom correspondence should be sent.
0003-7028 / 01 / 5502-0136$2.00 / 0
136
Volume 55, Number 2, 2001
APPLIED SPECTROSCOPY
q
2001 Society for Applied Spectroscopy

FIG. 1. Optical cell for measurement under high pressure (max. pressure: 2000 bar; max. temperature: 200 8C). (1) Sapphire window with mounting;
(2) steel block; (3) heating rod; (4) inlet; (5) screw joint with boring for brass probe.
the Raman light to the spectrometer entrance of the Hol-
olab. The excitation wavelength is 784.68 nm. The laser
power leaving the insertion probe is 110 mW. The spec-
vessel. The nitrogen pressure of 5 bar also serves as the
prepressure for the syringe pump by means of which the
peroxide solution is compressed to 1900 bar. With the
opening of the inlet valve, the peroxide solution  ows
into the measuring cell until full pressure equilibration is
achieved. The temperature and the pressure are adjusted
automatically within 30 to 45 s. The acquisition of the
measured values as well as the pressure and temperature
in the cell is controlled by a PC.
2
1
tral range is 100 to 2000 cm with a spectral resolution
2
1
of 5 cm
.
The entire setup with the optical cell placed in the
middle is shown in Fig. 2. The device for Ž lling the cell
(look at the top of Fig. 2) consists of a storage vessel and
a syringe pump. The gas in the cell can be released
through an outlet valve or a rupture disk. The peroxide
solution is prepared by diluting 10 g TBPPI (supplied by
Peroxid-Chemie as 75 wt % solution in isododecane;
After the reaction, the gases can be collected and sub-
sequently analyzed with gas chromatography mass spec-
trometry (GC-MS). A Fisons GC8000 gas chromatograph
n
m
commercial name: TBPPI-75-AL) and 65 g -hexane
with a 50 m FFAP column (i.d. 0.32 mm, layer 0.23 m)
(puriŽ ed by distillation over sodium), resulting in a 10
wt % solution of TBPPI.
and a Fisons MD 800 mass spectrometer was used. The
8
temperature program is 3 min at 60 C, followed by heat-
8
8
Before the start of each experiment, the entire setup
ing up at 20 C/min to 160 C. The scan time is 0.45 s
per scan.
n
was purged with -hexane and nitrogen and then evacu-
ated. The evacuated cell is heated. The solution of per-
For time-resolved experiments, a set of spectra is re-
8
corded after constant temperature (100 C) and pressure
n
oxide in -hexane is stored under nitrogen in the storage
FIG. 2. Flow sheet of the experimental setup.
APPLIED SPECTROSCOPY
137

to reactants, are Ž tted nonlinearly by a function of the
l 5 a 1 b* 2c*t
shape
exp(
where is the intensity, is the time, and
positive Ž tted constants. Similarly, proŽ les that increase,
) (pseudo-Ž rst-order kinetics),
l
t
a, b,
c
and are
a 2
due to products, are Ž tted by a function of the form
b*
2c*t
c,
). The rate constants are given by and the
characteristic half-life time for the reactant TBPPI is giv-
exp(
c.
en by ln(2)/ Note that, because of baseline effects which
are not fully compensated for, as well as small errors in
a
the Ž tting procedure, an off-set (constant ) might be pre-
sent in the product and reactant proŽ les, which must be
taken into account to obtain a reliable Ž t result.
RESULTS AND DISCUSSION
FIG. 3. Raman spectra of 75 wt % TBPPI in iso-dodecane (top) and
n-hexane (bottom) recorded at T 5 25 8C and ambient pressure.
Spectra of TBPPI and n-Hexane. Figure 3 shows the
(1900 bar) conditions are reached. The time needed for
scanning and recording one spectrum is 32.2 s. Typically
a set of 40 spectra is recorded, covering a time window
of about 1200 s.
iso
Raman spectra of a 75 wt % solution of TBPPI in
-
8
n
dodecane at 25 C and ambient pressure and of pure
-
hexane. Both spectra have been baseline corrected to
compensate for broad baseline features. A comparison
The spectra recorded at different time intervals were
exported to and processed by the Labspec 2.08 program
(Dilor, S.A., France). All spectra are stored in a single
multidimensional data set with a wavenumber dimension,
a time dimension, and a spectral intensity dimension. The
data set is baseline corrected in the wavenumber dimen-
sion Ž rst, with the application of linear local baselines
intersecting local minima. Fluorescence can lead to  uc-
tuations in baseline height in the spectral dimension. The
 uorescence might originate from degradated particles,
which move into the focal volume of the Raman probe
occasionally under the action of the pulsating high-pres-
sure pump. Although the baseline can be Ž tted and sub-
tracted in the wavenumber dimension, it is very difŽ cult
to correct entirely because the precise shape of the base-
8
with Raman spectra recorded at 25 C and high pressure
up to 1900 bar revealed only a slight shift of most Raman
2
signals, smaller than 1 cm 1, which we will neglect here.
n
The spectra of the peroxide and the -hexane are com-
pletely different, enabling the detection of TBPPI dis-
solved in -hexane without much effort. Raman spectra
from samples of low concentration of TBPPI in -hexane
showed that, for concentrations below 2 wt %, no repro-
ducible quantiŽ cations were obtained. Therefore, time-
resolved experiments were carried out with a 10 wt %
solution of TBPPI in -hexane at 1900 bar and 100 C.
From Fig. 3, strong bands can be identiŽ ed due to TBPPI,
n
n
n
8
n
which do not overlap with bands due to -hexane. These
bands are located at wavenumbers 871, 801, 591, and 558
2
1
cm
.
a priori.
line is not known
The remaining uncompensated
Assignment of Peaks in the Time-Dependent Ra-
man Data Set. In Fig. 4, an entire time-dependent data
set of spectra is shown in one plot. The spectra are pre-
sented as difference spectra, created by subtracting the
Ž nal spectrum from all other spectra. In this way the
 uctuations will appear in a noise-like pattern, containing
spikes and discontinuities, superimposed on proŽ les of
the Raman peak intensities vs. time. Therefore, after the
baseline correction, the data set is Ž ltered in the time
dimension, with removal of the spikes. The next step of
the data treatment is Ž ltering in the wavenumber dimen-
sion with the application of a 9-point moving average
Ž lter.
n
spectrum due to -hexane (the solvent) is almost com-
pletely cancelled, Raman peaks due to reactants appear
as positive peaks, and peaks due to products appear as
negative peaks. Both the negative and positive peaks be-
come weaker in this representation as time proceeds.
Note that difference traces like this cannot be analyzed
kinetically, because the intensity of all peaks is artiŽ cially
brought to zero as the result of the subtraction procedure.
The positive peaks, due to reactants, are due to TBPPI,
as can be very easily veriŽ ed by comparing Figs. 3 and
4. The positions of the most prominent peaks are 249,
478, 558, 592, 802, 872, 1031, 1100, 1264, and 1455
The major components of the decomposing system of
n
TBPPI dissolved in -hexane are the peroxide and the
hexane themselves and the products 2-methyl-2-propanol
t
( -butanol) and CO2 (see the Results and Discussion sec-
n
t
tion). Spectra of the peroxide, the -hexane, and the -
butanol components in the pure form are measured. Each
spectrum from the time-dependent data set is next Ž tted
as a linear combination of these three spectra of the com-
ponents. The spectrum of CO2 contains only two very
narrow lines and is not considered in the Ž t. From the
Ž tting coefŽ cients, proŽ les for the components are cre-
ated, and the Raman signal intensity is then essentially
represented by a Ž tting coefŽ cient. The proŽ le for the
CO2 contribution is created by integrating the intensity of
one of the two CO2 peaks for each time interval.
2
2
1
cm 1. The very strong peak at 872 cm is due to the
O–O stretching vibration of the peroxide group.2
In order to establish an assignment of peaks due to
products (negative peaks), possible steps in the decom-
position reaction TBPPI should be considered. A very
simple pathway for the decomposition of TBPPI is given
in Scheme I.
ProŽ les of Raman signal intensities vs. time were ex-
ported to Excel 97 (Microsoft). ProŽ les that decrease, due
138
Volume 55, Number 2, 2001

FIG. 4. Difference Raman spectra of the data set overlaid in one plot. The spectra are generated by subtracting from each spectrum the Ž nal
spectrum. In the upper right corner a zoom-in of a reactant (left) and a product peak (right) are shown.
2
1
5
(C O
acteristic peak due to acetone at 1706 cm
group) is not observed. The concentration of aceton
must be very low or zero. Negative peaks of low inten-
sity located at 291, 391, 437, 521, 688, 982, 1077, and
2
1
1405 cm
remain unassigned. However, the simulta-
neous presence of three peaks located at 437, 982, and
2
1
1077 cm tempts one to speculate on a compound con-
taining an aliphatic ether group. The three observed
2
peaks are then due to COC deformation (437 cm 1),
2
COC symmetric stretching (982 cm 1), and COC asym-
2
metric stretching (1077 cm 1).4
On the basis of this qualitative discussion of the oc-
currence of peaks only, much information is obtained.
t
Undoubtly, -butanol and CO2 could be detected, but no
peaks due to acetone are observed. This observation im-
t
plies that for the -butoxy radical, hydrogen abstraction
b
dominates over -scission.
The GC-MS analysis of the gaseous products revealed
the formation of hexene and acetone (concentrations not
quantiŽ ed). These compounds are not detected in the Ra-
man spectra. Apparently the concentrations are too low.
Analysis of the Time-Dependent Raman Intensities
by First-Order Kinetics. Kinetic information was ob-
tained from the Raman intensity vs. time proŽ les. Each
spectrum of the Raman data set was Ž tted as a sum of
SCHEME I. Pathway for the decomposition of TBPPI.
The most prominent product peak is located at 754
2
1
t
(see Fig. 4). We assign this peak to -butanol,
2
cm
which shows a strong Raman peak at 751 cm 1. A closer
look also reveals negative peaks at 347, 914, 1208, and
2
1
t
t
1239 cm , all due to -butanol. -Butanol peaks, located
2
at 474, 1025, and 1453 cm 1, overlap with peaks due to
n
the spectra of the pure components -hexane. TBPPI, and
n
-hexane and TBPPI. These peaks can be observed in
t
-butanol. A representative Ž t result is shown in Fig. 5.
In Fig. 6 the proŽ les for the relative contributions, ob-
Fig. 4, but the intensity is low. Apparently due to the
overlap with peaks due to TBPPI, the intensity is par-
tially cancelled. A doublet of peaks located at 1280 and
tained from the spectral Ž t, to the spectra due to TBPPI,
2
1
t
-butanol, and CO2 are shown. The Ž gure also shows sim-
1380 cm is also clearly visible in Fig. 4. These peaks
are assigned to CO2.3 Further weak peaks are observed
ulated proŽ les (Ž rst-order kinetics), calculated as de-
scribed in the Experimental section. The decomposition
2
at 703, 1150, and 1655 cm 1. These peak positions could
3
t
of TBPPI and the formation of -butanol and CO2 can be
cis trans
match those of
-
-hexadiene; however, the inten-
very well described by using a (pseudo)-Ž rst-order ki-
netic equation; note the high correlation coefŽ cients for
cis
sities are weak, and supposedly the other peaks of
-
trans
n
-hexadiene overlap with peaks due to -hexane and
.
all three Ž ts ( 0.99). Identical characteristic times of 420
TBPPI. The formation of 2,4-hexadiene is therefore not
evidenced by our observations, and the weak peaks re-
main unassigned. It is noteworthy that the very char-
2
2
6
k 5
20 s (
position of TBPPI and the formation of CO2. For -bu-
1.65·10 3s 1) are obtained for the decom-
t
APPLIED SPECTROSCOPY
139

FIG. 5. A representative Ž t result for the Ž tting of the time-dependent
Raman spectra, showing (from top to bottom) an experimental spec-
trum, the total Ž t result, the n-hexane component, the TBPPI compo-
nent, and the t-butanol component. Two arrows mark peaks in the ex-
perimental spectrum, due to CO2. These peaks are not taken into ac-
count in the Ž tting procedure.
6
k 5
20 s is found (
tanol, a characteristic time of 566
2
2
1.22·10 3s 1).
t
At Ž rst sight it is surprising that the formation of -
butanol occurs at a lower rate than the rate of TBPPI
decomposition. However, a simple kinetic rate law cla-
k1
k2
k1
k3
(re-
riŽ es this result. Assuming
and
t
sulting in a quasi-stationary concentration of -BuO·),
we can derive from Scheme I the pseudo-Ž rst-order rate
law:
t 2
d[
BuOH]
5 k9k
1[TBPPI]
t
d
with
FIG. 6. ProŽ les of the Raman intensity of TBPPI (top), t-butanol (mid-
dle), and CO2 (bottom). Dots represent experimental data points; the
solid line represents the Ž tted Ž rst-order proŽ le.
1
k9 5
k3
1
1
k
2 [RH]
n
The solvent -hexane (RH) is abundantly present;
therefore its concentration can be assumed to be con-
n
of products in -hexane, because the products are ex-
pected to be stable. Once the measurement of concentra-
tions is possible, then useful information such as the yield
of CO2 or -butanol per unit amount of TBPPI is within
k 9
k 9
is smaller
stant, leaving
also constant. Evidently
t
than 1, which explains the lower rate of -BuOH for-
mation. With an -hexane concentration of 7.78 mol/L
and the two experimentally observed rate constants,
is found to be 0.742, resulting in a ratio of rate con-
stants for hydrogen addition and -scission (
t
n
reach.
k 9
CONCLUSION
b
k k
/ 2) of
3
in situ
Raman spectroscopy
2.7 mol/L.
It was demonstrated that
It is not the purpose of this study to actually measure
concentrations of various compounds. However, we
would like to point out the potential usefulness of such
an approach. The Raman scattering efŽ ciency for TBPPI
is a suitable technique for the study of peroxide decom-
position reactions at high temperature and high pressure.
In a single experiment a wealth of data can be obtained,
and information on reactant and product identiŽ cation
and quantiŽ cation, as well as on the mechanism, is ac-
quired.
t 5
can be obtained if
of the product proŽ les,
can then be used to calculate the intensity due to TBPPI
t 5 t 5
0 is known. By back-extrapolation
t 5
0 can be obtained. This result
It was shown that the main decomposition products of
TBPPI are CO2 and -butanol. From Ž rst-order kinetic Ž ts
t
at
0. Since at
0, the concentration of TBPPI is
10 wt %; the scattering efŽ ciency can be estimated. The
Raman scattering efŽ ciency for the products can be de-
termined by measuring samples of known concentrations
of the Raman intensity vs. time proŽ les of reactants and
products, the characteristic times for TBPPI decomposi-
tion and CO2 formation were found to be approximately
140
Volume 55, Number 2, 2001

istry and Physics: A Practical Approach. W. B. Holzapfel, Ed. (Ox-
ford University Press, Oxford, 1997), pp. 151–186.
2. W. Kessler, G. Luft, and W. Zeiß, Ber. Bunsenges. Phys. Chem.
101, 698 (1997).
3. B. Schrader and W. Meier, Raman/IR Atlas of Organic Compounds
(Verlag Chemie, GmbH, Weinheim, 1975).
4. F. R. Dollish, W. G. Fateley, and F. F. Bentley, Characteristic Ra-
man Frequencies of Organic Compounds (John Wiley and Sons,
New York, 1974).
t
420 s. -Butanol formation occurs at a slower rate of ap-
proximately 566 s.
ACKNOWLEDGMENT
The authors thank Dr. Thomas Pollock (Peroxid Chemie) for valuable
discussions and for providing samples of TBPPI.
1. M. Buback and C. Hinton, in High-Pressure Techniques in Chem-
APPLIED SPECTROSCOPY
141