N-Hydroxyphthalimides and Metal Cocatalysts for the Autoxidation of p-Xylene to Terephthalic Acid

Article abstract of DOI:10.1021/jp035870s

N-Hydroxyphthalimide (NHPI) and its derivatives, such as 3-F-NHPI, 4-Me-NHPI, N-acetoxyphthalimide, and N,N-dihydroxypyromelitimide, were used as promoters with Co(OAc)₂ catalyst for the autoxidation of p-xylene (pX) and other methyl arenes. All the promoters gave acceptable rates and yields of terephthalic acid. The initial reaction rates, measured by the rate of oxygen uptake, were analyzed by a rate equation in terms of [pX], [Co(II)], and [NHPI]. The metal cocatalysts Mn(II) and Ce(III) accelerated the reaction significantly at millimolar concentrations. The reaction occurs by a chain mechanism that involves formation of the phthalimide N-oxyl radical, PINO ^. (that is, R₂NO^.), which abstracts a hydrogen atom from the methyl group of p-xylene to form the carbon-centered radical ArCH₂^.. In a stepwise fashion, the sequence progresses through alcohol, aldehyde, and carboxylic acid; at each stage, C-H abstraction by PINO^. is involved. A significant kinetic isotope effect on the overall oxidation of p-xylene was found, v_i(H)/v_i(D) = 3.4. The activity of the substituted NHPI promoters follows the order NHPI > 3-F-NHPI > 4-Me-NHPI, which can be interpreted in terms of kinetic stability of the corresponding PINO radical.

Full text of DOI:10.1021/jp035870s

J. Phys. Chem. A 2004, 108, 425-431
425
N-Hydroxyphthalimides and Metal Cocatalysts for the Autoxidation of p-Xylene to
Terephthalic Acid
Basudeb Saha, Nobuyoshi Koshino, and James H. Espenson*
Ames Laboratory and Department of Chemistry, Iowa State UniVersity of Science and Technology,
Ames, Iowa 50011
ReceiVed: June 30, 2003; In Final Form: NoVember 11, 2003
N-Hydroxyphthalimide (NHPI) and its derivatives, such as 3-F-NHPI, 4-Me-NHPI, N-acetoxyphthalimide,
and N,N-dihydroxypyromelitimide, were used as promoters with Co(OAc)2 catalyst for the autoxidation of
p-xylene (pX) and other methyl arenes. All the promoters gave acceptable rates and yields of terephthalic
acid. The initial reaction rates, measured by the rate of oxygen uptake, were analyzed by a rate equation in
terms of [pX], [Co(II)], and [NHPI]. The metal cocatalysts Mn(II) and Ce(III) accelerated the reaction
significantly at millimolar concentrations. The reaction occurs by a chain mechanism that involves formation
of the phthalimide N-oxyl radical, PINO• (that is, R2NO•), which abstracts a hydrogen atom from the methyl
•
group of p-xylene to form the carbon-centered radical ArCH2 . In a stepwise fashion, the sequence progresses
through alcohol, aldehyde, and carboxylic acid; at each stage, C-H abstraction by PINO• is involved. A
significant kinetic isotope effect on the overall oxidation of p-xylene was found, Vi(H)/Vi(D) ) 3.4. The
activity of the substituted NHPI promoters follows the order NHPI > 3-F-NHPI > 4-Me-NHPI, which can
be interpreted in terms of kinetic stability of the corresponding PINO radical.
Introduction
CHART 1: Structural Formulas of Substituted
N-Hydroxyphthalimides
The catalytic oxidation of hydrocarbons by molecular oxygen
is an important reaction for the production of commodity
chemicals,1,2 such as terephthalic acid, which is used to
manufacture poly(ethylene terephthalate), a polymer commonly
made into fibers, resins, films, etc. The most widely used catalyst
package for the aerobic oxidation of p-xylene (pX) to tere-
phthalic acid combines Co(OAc)2, Mn(OAc)2, and HBr.3-7
Bromide, which corrodes expensive titanium reactors and forms
CH3Br(g),8 which can deplete the ozone layer, is, however, an
undesirable component.9,10
Recently, Ishii et al.11 have developed an efficient bromide-
free catalyst package, which combines Co(OAc)2 and N-
hydroxyphthalimide (NHPI). This catalyst has been reported to
be highly efficient for the aerobic oxidation of organic com-
pounds containing sufficiently reactive C-H bonds, such as
alkanes,12 alkylbenzenes,2 aromatic compounds containing ben-
zylic groups,13,14 and so on. The Co(OAc)2/NHPI catalyst
resembles the classical Co(OAc)2/HBr combination that cata-
lyzes oxidation of hydrocarbons in the sense, that under
oxidative conditions, NHPI is converted to PINO•. Like the
dibromide radical15 formed from the bromide-containing system,
the PINO radical abstracts a hydrogen atom from a C-H bond
of the substrate and propagates radical chain branching.
The NHPI-based catalyst also offers the possibility that one
may be able to tune the catalyst performance by introducing
substituents on the aryl ring of the NHPI. In this study, we
describe the kinetics of autoxidation of pX catalyzed by the
combination of NHPI and Co(OAc)2. The promoters investigated
are shown in Chart 1. Kinetic studies have also been carried
out in the presence of metal cocatalysts Mn(II) and Ce(III) to
define their mechanistic roles.
Experimental Section
Reagents. The following materials were used as obtained
commercially without purification: cobalt(II) acetate tetrahy-
drate, glacial acetic acid, NHPI, manganese(II) acetate, cerium-
(III) acetate, pX, pX-d10 (99+% D), m-xylene, pseudocumene,
durene, p-toluic acid, 4-methylphthalic anhydride, 3-fluo-
rophthalic anhydride, 1,2,4,5-benzenetetracarboxylic anhydride,
hydroxylamine hydrochloride, and acetic anhydride.
The derivatives 4-Me-NHPI and 3-F-NHPI‚H2O were syn-
thesized from the phthalic anhydrides and hydroxylamine
hydrochloride according to a literature procedure.16 The purity
1
of each product was checked by H NMR, and the observed
1
values of the H NMR parameters in DMSO-d6 agreed with
those reported.16 Anal. Found (Calcd) for 4-Me-NHPI,
C9H7O3N: C, 60.19 (61.02); H, 3.89 (3.98); N, 8.28 (7.91).
Anal. Found (Calcd) for 3-F-NHPI‚H2O, C8H6FO4N: C, 48.21
(48.25); H, 2.86 (3.04); N, 6.89 (7.03).
N-Acetoxyphthalimide, NAPI, was prepared by dissolving
NHPI (0.7 g, 4.2 mmol) in 10 mL of acetic anhydride and
continuously stirring the mixture for 4 h. The solution was
filtered to remove any insoluble NHPI, and the clear filtrate
was allowed to cool in the refrigerator, whereupon NAPI
* Corresponding author.
10.1021/jp035870s CCC: $27.50 © 2004 American Chemical Society
Published on Web 12/23/2003

426 J. Phys. Chem. A, Vol. 108, No. 3, 2004
Saha et al.
TABLE 1: Maximum Absorption Wavelengths and Molar
1
precipitated as white crystals. H NMR (δ, HOAc-d4): 2.36
Absorptivities for PINO• in HOAc
(CH3, s); 7.84 (2H, d); 7.9 (2H, d). Anal. Found (Calcd) for
C10H7O4N: C, 58.43 (58.54); 3.09 (3.44); N, 7.40 (6.83).
N,N-Dihydroxypyromelitimide, NDHPI, was prepared from
1,2,4,5-benzenetetracarboxylic anhydride and hydroxylamine
hydrochloride. The latter (1.34 g, 19.2 mmol) and Et3N (2.6
mL, 19 mmol) were dissolved in 60 mL of ethanol. After the
solution had been stirred for 10 min, 1,2,4,5-benzenetetracar-
boxylic anhydride (2.18 g, 9.8 mmol) was added. The mixture
was refluxed for 8 h. As the reaction progressed, the clear
solution gradually changed to yellow and then dark red. The
resulting red solution was poured into ca. 100 mL of H2O. The
product precipitated as a yellow powder, which was filtered and
dried under vacuum. Yield: 52%. 1H NMR (δ, CD3CN): 8.25
(2H, br s, NOH); 8.15 (2H, s). Anal. Calcd for C10H4N2O6‚
2H2O: C, 42.2; H, 2.84; N, 9.89. Found: C, 42.2; H, 2.88; N,
9.56.
A solution of Co(OAc)3 in glacial acetic acid was prepared
by passing ozone through a freshly prepared solution of Co-
(OAc)2‚4H2O.17,18 Excess ozone was purged from the solution
with a vigorous stream of argon. Cobalt(III) acetate exists in
acetic acid in a number of forms,19 but in these circumstances
the species is the hydroxo-bridged dimetallic Co(III) complex
known as Co(III)s,20 as confirmed by its characteristic UV-vis
spectrum.21
General Procedure. The progress of the autoxidation reac-
tions was monitored by the oxygen uptake method using a
manometric apparatus similar to the one described in the
literature.1,22 The reactor, which contains an impeller to maintain
oxygen saturation of the solution, was thermostated at 70 °C
by means of a circulating water bath. Oxygen consumption was
measured by monitoring the decrease in volume, at a constant
1 atm pressure of pure oxygen, in a buret connected to the
reactor. The apparatus is shown in the Supporting Information.
The initial reaction rates were calculated from the slope of the
linear plots of the volume of oxygen consumption against time.
The oxidation products of some reactions were monitored
by HPLC, having calibrated the method with known compounds,
qualitatively and quantitatively. For these analyses, a 20 µL
aliquot was removed from the reactor at different times during
the reaction and diluted to 1 mL with 1:4 DMSO/CH3CN (v/
v). The diluted solution was then run through the HPLC column.
A Waters model 501 solvent delivery system, Waters 996
photodiode array detector, and Novapak C18 3.9 × 150 mm
column were used for this method. A binary solvent of 50%
H2O/0.5% CH3COOH and 50% CH3CN with a flow rate of 0.7
mL/min was used in the isocratic mode. Identification of
oxidation products was performed by comparing the retention
time of the HPLC chromatogram peaks with those of authentic
samples of p-tolualdehyde, p-toluic acid, 4-carboxybenzalde-
hyde, and terephthalic acid. Each peak in the HPLC chromato-
gram was properly integrated, and the actual concentration of
each component was obtained from the precalibrated plot of
peak area against concentration, as presented in Figure S1 in
the Supporting Information.
radical
λmax
ꢀa
3-F-PINO•
PINO•
367
380
397
1.35
1.36
1.31
4-Me-PINO•
a Units 103 L mol-1 cm-1
.
full constitution of many metallic species has not been estab-
lished.) The formation of PINO• is accompanied by a large
increase in absorbance in the vicinity of 380 nm. Values of the
maximum wavelength and molar absorptivity are given in Table
1.
The absorbance changes for formation and decomposition of
these radicals were measured at 380 nm by UV-vis spectro-
photometry, using Shimadzu UV-2101 and 2501 spectropho-
tometers. The absorbance-time data for the rapid formation
reaction were to fitted to pseudo-first-order kinetics, eq 2. The
subsequent self-decomposition step then set in, following
second-order kinetics in accord with eq 3, where Y represents
absorbance, kψ the pseudo-first-order rate constant for PINO•
formation, kd the second-order rate constant for self-decomposi-
tion, and ꢀ the molar absorptivity of PINO•.
Yt ) Y∞ + (Y0 - Y∞)e-kψt
(2)
(3)
Y0 - Y∞
Yt ) Y∞ +
1 + (Y0 - Y∞)kdt/ꢀ
Results
The overall oxidation of pX to terephthalic acid is given by
the stoichiometry
C6H4(CH3)2 + 3O2 f C6H4(COOH)2 + 2H2O
(4)
Various partially oxidized forms of pX are produced during the
course of the complex free radical chain reaction. One can use
the oxygen uptake rate to monitor the reaction progress, but a
cautionary note is in order: -d[O2]/dt ) -3 d[pX]/dt only in
the limit in which the organic intermediates do not accumulate,
which has not proved to be the case. Despite the numerical
uncertainty so introduced, we have chosen to take the oxygen
uptake rate as a convenient measure of the reaction rate. These
intermediates may be 4-methylbenzyl alcohol, p-tolualdehyde,
p-toluic acid, 4-carboxybenzyl alcohol, and 4-carboxyben-
zaldeyde.
Oxidation of pX with NHPI as Promoter. A preliminary
study of the oxidation was carried out with 820 mM pX, 0.3
mM NHPI, and 40 mM Co(OAc)2 in HOAc. The initial change
in the volume of O2 consumption was a linear function of time.
The slope of the linear plot gives the initial reaction rate; when
converted to concentration units, Vi ) 61.3 × 10-6 mol L-1
s-1 as compared to 7.1 × 10-6 mol L-1 s-1 22 with 10 mM
NaBr in place of NHPI. Thus, the NHPI promoter is superior
to bromide. However, the NHPI-promoted reaction almost
stopped at ca. 1500 s, after consuming ca. 28 mL of O2, which
amounts to the oxidation of only 7% of pX conversion to
p-tolualdehyde; by way of comparison, the bromide-promoted
reaction resulted in ca. 1% pX oxidation in the same time. It
was noticed that the NHPI reaction stopped as the solution took
on a green color with a UV-vis spectrum that matched that of
Co(III)s. Further addition of 0.4 mM NHPI immediately turned
the solution to a slightly darker pink color than that of the
starting Co(II) solution. As shown in Figure 1, the reaction
Substituted PINO radicals were generated in glacial acetic
acid by the oxidation of substituted NHPI with Co(III)s:
R2NOH + Co(OAc)3 f R2NO• + Co(OAc)2 + HOAc (kf)
(1)
(In equations such as the one shown here, and others written
subsequently, it is convenient to show stoichiometric formulas;
ions are inappropriate in this low dielectric medium, and the

Autoxidation of p-Xylene to Terephthalic Acid
J. Phys. Chem. A, Vol. 108, No. 3, 2004 427
Figure 3. Dependence of Vi on the concentration of pX or pX-d10 under
the conditions 820 mM pX, 40 mM Co(OAc)2, and 0.4 mM [NHPI] at
70 °C in HOAc.
Figure 1. Volume of O2 consumed as a function of time in an
experiment with the following concentrations: 820 mM pX, 40 mM
Co(OAc)2, and (a) 0.3 mM or (b) an additional 0.4 mM NHPI at 70
°C in HOAc.
[NHPI], [Co(II)], and [pX]. From the series in which two of
the three concentrations were held constant, the reaction is
second-order with respect to both [Co(II)] and [pX], Figures
S2 and S3, and first-order with respect to [NHPI], Figure S4.
Combining those observations, we start to write V ) k[Co(II)]2-
[pX]2[NHPI], but that did not fit the data.
A more elaborate form was used, which is actually in good
accord with the kinetic equations for the scheme put forth by
Zakharov.23 This rate law, eq 5, where k2, k4, and kH are the
d[O2]
2{k2[pX] + k4[Co(II)] + kH[NHPI]}2
-
) V )
(5)
dt
k6
second-order rate constants for the reaction of the peroxyl radical
ArCH2OO• with pX, Co(II), and NHPI, respectively, is based
on the free radical chain mechanism described in the Discussion.
The second-order rate constant for the self-reaction of peroxyl
radical, which is the termination step, is designated k6.
Least-squares fitting of the 19 experimental data (Table S1)
Figure 2. Volume of O2 consumed as a function of time in an
experiment utilizing a syringe pump to introduce a 20 mM NHPI
solution at a flow rate of 0.82 mL/h. Other concentrations: 820 mM
pX and 40 mM Co(OAc)2 at 70 °C in HOAc.
resumed with Vi ) 83 × 10-6 mol L-1 s-1, but then stopped
again after consuming another 30 mL of O2 as Co(III)s again
accumulated. Starting with a higher concentration of NHPI at
the outset led to a reaction that stopped sooner.
to eq 5, with k6 fixed at 1.5 × 108 L mol-1 s-1 24
, gave k2 ) 9.4
( 1.9 L mol-1 s-1, k4 ) (2.5 ( 0.1) × 103 L mol-1 s-1, and kH
) (7.1 ( 0.3) × 103 L mol-1 s-1 at 70 °C. The experimental
and fitted initial rates agree within (18%. A comparison of
experimental and fitted rates is given in Figure S5, which shows
that the variations are random with respect to the concentration
variables.
From these results we concluded that small, continuous
additions of NHPI might give better results. Thus, a syringe
pump was used to deliver the NHPI solution into the reactor to
test this idea. Two experiments were performed by injecting a
stock NHPI solution (20 mM) into the reaction mixture
containing 820 mM pX and 40 mM Co(OAc)2 at 70 °C, at flow
rates of 0.39 and 0.82 mL/h. As long as the NHPI solution was
continuously added, in both cases the volume of O2 was
consumed linearly as a function of time, with Vi values of 26 ×
10-6 and 57 × 10-6 M s-1, respectively. However, the reaction
stopped when the syringe pump was turned off, but started again
when it was turned on. As shown in Figure 2, the reaction
continued even after consuming 160 mL of O2, a total of 1.24
mM NHPI having then been added. Without use of the syringe
pump, the reaction stopped after consuming only 45 mL of O2
when the 1.24 mM NHPI had been used.
Kinetics. Interestingly, at the low concentrations of Co(II),
the reaction continued for a longer time, even without use of a
syringe pump. An experiment under the conditions of 5 mM
Co(OAc)2, 5 mM NHPI, and 820 mM pX at 70 °C consumed
275 mL of O2 in 2.5 h, which amounts to ca. 70% oxidation of
pX to p-tolualdehyde. A series of such experiments was carried
out, varying the concentrations of NHPI, Co(II), and pX in the
ranges 1.0-7.9 mM, 5-25 mM, and 0.2-4.0 M, respectively.
Table S1 shows the dependence of the initial reaction rates on
Kinetic Isotope Effects. The kinetic isotope effect (KIE) for
the autoxidation of pX was measured by performing an
experiment with 820 mM pX-d10, 40 mM Co(II), and 0.4 mM
NHPI. The result, as shown in Figure 3, yielded a significant
KIE value: Vi(H)/Vi(D) ) 3.4 ( 0.2, which accounts for a
hydrogen atom abstraction mechanism.
Different Substrates. Other methyl arenes, m-xylene, pseudoc-
umene, and durene, were also studied with the Co(OAc)2/NHPI
combination, with 820 mM substrate, 40 mM Co(OAc)2, and
0.4 mM NHPI. The initial reaction rates increased with the
number of methyl groups. The results are summarized in Table
2. The rate differences are rather small, but perhaps the increase
follows the order of the bond dissociation energies of the C-H
bonds, which decreases from m-xylene to durene.
Oxidation of p-Toluic Acid. The intermediate product of
the oxidation of pX with the Co(II)/NHPI catalyst is p-toluic
acid. Once one methyl group of pX has been oxidized, the other
methyl group is deactivated by the electron-withdrawing effect
of the -CO2H group. Thus, the application of a catalyst in the
manufacture of terephthalic acid depends on its effectiveness
to oxidize p-toluic acid. To explore this further, p-toluic acid

428 J. Phys. Chem. A, Vol. 108, No. 3, 2004
Saha et al.
TABLE 2: Initial Rates of Oxygen Uptake for the
Autoxidation of 820 mM Methylarene, Catalyzed by 40 mM
Co(OAc)2 and 0.4 mM NHPI at 70 °C in HOAc
substrate
Via
V1/nb
substrate
Via
V1/nb
m-xylene
p-xylene
67
83
11
14
pseudocumene
durene
93
98
16
8.2
a Units 10-6 M s-1; b ) number of aliphatic hydrogen atoms (taken
as 6 for pseudocumene).
TABLE 3: Comparison of the Initial Rates of O2
Consumption Obtained from the NHPI- and
Bromide-Promoted Autoxidation of pX under the Conditions
820 mM pX and 40 mM Co(OAc)2 at 70 °C in HOAc
Figure 5. Yields of terephthalic acid formed from the autoxidation of
100 mM pX with the Co(OAc)2 catalyst, NHPI promoters, and metal
cocatalysts, 0.3 mM Mn(OAc)2 or 0.15 mM Ce(OAc)3 at 70 °C in
HOAc.
promoter (mM)
Via
20
1.9
NHPI (0.4)
Br- (10)
a Units 10-6 M s-1
.
80% p-toluic acid in less than 1 h. The subsequent oxidation of
p-toluic acid led to the formation of 4-carboxybenzaldehyde and
then terephthalic acid. The formation of intermediate oxidation
products and terephthalic acid as a function of time is shown
in Figure S6. Figure 5 compares the yield of the final product,
terephthalic acid, for different catalysts.
Figure 5 illustrates a number of interesting features. First,
the Co(II)/Mn(II)/NHPI catalyst is more efficient than the Co-
(II)/Ce(III)/NHPI catalyst in terms of yield of terephthalic acid.
Second, the rate of formation of terephthalic acid with the
unsubstituted NHPI is more effective than with the substituted
one, the activity order being NHPI > 3-F-NHPI > 4-Me-NHPI.
Under comparable conditions, an experiment with p-toluic acid
as a starting substrate afforded a yield of 92% terephthalic acid
in 3.5 h. The p-toluic acid oxidation, however, proceeded
smoothly with the addition of 5 mM NHPI in one portion at
the beginning of the reaction.
Figure 4. Dependence of the initial rates (Vi) of O2 uptake for the
Co(II)/NHPI-catalyzed autoxidation of pX on metal cocatalysts at 70
°C in HOAc.
was directly oxidized with the Co(OAc)2/NHPI catalyst. A
reaction with 820 mM substrate, 40 mM Co(OAc)2, and 0.4
mM NHPI resulted in Vi ) 20 × 10-6 mol L-1 s-1. The same
experiment with 10 mM NaBr in place of NHPI had Vi ) 1.9
As mentioned above, pX oxidation stopped after 42-50%
of total reaction with the combination of 5 mM substituted
NHPI, 5 mM Co(II), and 0.3 mM Mn(II); it was therefore
necessary to add more NHPI to complete the oxidation. Starting
the reaction with the total amount of NHPI (10 mM) in one
portion did not improve the yield either, except that the initial
rate of O2 consumption was higher. Cessation of the reaction
after 42-50% may be due to the decomposition of NHPI.
Decomposition of NHPI occurs rapidly during the oxidation of
p-tolualdehyde to p-toluic acid, concurrently with the buildup
of the green color of Co(III)s. If the total quantity of NHPI used
can be reduced by a simple modification, the present oxidation
would be more desirable. In such efforts, the oxidation of pX
was carried out with NAPI instead of NHPI. Under comparable
conditions and at 70 °C, the reaction with NAPI was extremely
slow, particularly in the first 20 h. However, complete oxidation
to terephthalic acid was found using the NAPI in one portion
at the beginning, without buildup of Co(III)s. Moreover, the
mother liquor of the first run was equally active for the second
run, and was able to catalyze the oxidation of another 100 mM
pX. The first and second runs produced 98% and 95%
terephthalic acid in 70 and 74 h, respectively (Figure 6), the
balance being p-toluic acid and 4-carboxybenzaldehyde.
NDHPI Promoter. This promoter contains two hydroxyimide
moieties on the benzene ring. Therefore, we anticipated that it
might be more active. Under comparable conditions 100 mM
pX, 5 mM Co(II), 5 mM NDHPI, and 0.3 mM Mn(II) at 70 °C
in HOAc, the beneficial effect of NDHPI was immediately
realized from the initial reaction rate measurement, Vi ) 105 ×
10-6 mol L-1 s-1, 33% higher than that of the NHPI-promoted
oxidation. With a total of 6.5 mM NDHPI, the yield of
× 10-6 mol L-1 s-1 22
.
Thus, the efficiency of the NHPI-
promoted oxidation of p-toluic acid is better than that of the
bromide-promoted oxidation (Table 3).
Kinetic Effects of Metal Cocatalysts. The effectiveness of
the Co(OAc)2/NHPI catalyst was studied in the presence of a
cocatalyst. Mn(OAc)2, Ce(OAc)3, and Zr(IV) (as a solution of
zirconyl chloride in acetic acid) are known to accelerate
reactions in which bromide25,26 and bromoanthracene22 are used
to promote oxidation in a manner that shows strongly synergistic
rate enhancements. The synergistic effects of Mn(OAc)2 and
Ce(OAc)3 with the NHPI promoter are also quite significant.
Experiments were performed at varying concentrations of metal
cocatalysts, in the range of 0.1-2.0 mM for Mn(OAc)2 and
0.1-3.0 mM for Ce(OAc)3, with fixed concentrations of Co-
(OAc)2 (5 mM), NHPI (1.0 mM), and pX (820 mM). The results
are shown in Figure 4 and Table S2. In each case the rate attains
a maximum value at a particular concentration of cocatalyst
and then declines slowly. One experiment with 10 mM Zr(IV)
showed no effect on the reaction rate.
Product Identification. HPLC determinations were per-
formed on reactions starting with 100 mM pX, 5 mM Co(OAc)2,
5 mM promoter, and the optimum concentration of Mn(OAc)2
(0.3 mM) or Ce(OAc)3 (0.15 mM). However, all reactions came
to a halt after 42-50% of total oxidation due to decomposition
of NHPI, which therefore was added at 5-7 mM concentration
in one or two further portions. HPLC monitoring indicated that
p-tolualdehyde was formed in the early stage, followed by 70-

Autoxidation of p-Xylene to Terephthalic Acid
J. Phys. Chem. A, Vol. 108, No. 3, 2004 429
hydrogen abstraction, likely by the contributions from reactions
11 and 13, as given later.
The peroxyl radical formed from reaction 10 reacts with
ArCH3 and Co(II) to propagate and branch the radical chain
reactions (eqs 11 and 12),24,28 and a single peroxyl radical
becomes three, by a sequence of pathways:
ArCH2OO• + ArCH3 f ArCH2OOH + ArCH2 (k2) (11)
•
ArCH2OO• + Co(OAc)2 + HOAc f
ArCH2OOH + Co(OAc)3 (k4) (12)
The peroxyl radical can also abstract a hydrogen atom from
the NHPI and participate in the chain-branching reactions:
Figure 6. Yields of terephthalic acid (solid bar) and p-toluic acid
(hatched bar) formed from the autoxidation of pX with 5 mM Co-
(OAc)2, 0.3 mM Mn(OAc)2, and 10 mM NAPI in both experiments at
70 °C in HOAc.
ArCH2OO• + R2NOH f ArCH2OOH + R2NO• (kH) (13)
The chain termination step is the self-reaction of the peroxyl
radical by the Russell mechanism:30
terephthalic acid was 85% in 4 h. As shown in Figure S6, the
same reaction with the NHPI promoter required 10 mM NHPI
to produce 85% terephthalic acid in 4 h. The beneficial effect
of NDHPI has also been reported for the oxidation of nitro-
toluenes.27
2ArCH2OO• f ArCHO + ArCH2OH + O2 (k6) (14)
When the rates of chain branching and termination become
equal, the concentration of the peroxyl radical and the steady-
state oxidation rate can be expressed by eqs 15 and 16,
respectively.
Discussion
Ishii et al. have reported a mechanism for the Co(II)/NHPI-
catalyzed oxidation of methyl arenes.11 It has been amply
demonstrated that the metal-catalyzed autoxidations of methyl
arenes are free radical chain reactions.22,24,28 The initiator is the
small concentration of an aralkyl hydroperoxide that is inevitably
present in the hydrocarbon. The initiation step generates two
reactive species, cobalt(III) and aralkoxyl radicals:
k2[ArCH3] + k4[Co(II)] + kH[NHPI]
[ArCH2OO•]ss )
(15)
k6
d[O2]
2{k2[ArCH3] + k4[Co(II)] + kH[NHPI]}2
-
) V )
dt
k6
(16)
Recently, Amorati et al.31 reported the value of kH as 7.2 ×
103 L mol-1 s-1 at 30 °C in benzene for the autoxidation of
cumene catalyzed by NHPI and 2,2′-azobis(2,4-dimethylvale-
ronitrile). The O-H BDE of NHPI in polar solvents is higher
than that in non-hydrogen-bonding solvents such as benzene.32,33
Therefore, the O-H BDE of NHPI can be expected to be larger
in HOAc than in benzene; thus, kH at 70 °C in HOAc seems
reasonable. Zakharov et al. reported values of k2 ) 4.5 L mol-1
s-1 and k4 ) 7.2 × 102 L mol-1 s-1 for the Co(II)/bromide-
catalyzed autoxidation of pX at 70 °C.24 Thus, our values of k2
and k4 are not in exact agreement, but each is within a factor of
2-3. The source of disagreement may be related to the
accumulation of experimental error, with 3 parameters being
determined from 14 data points.
As noted above, NHPI decomposition occurred during pX
oxidation, particularly when O2 was being consumed rapidly,
indicative of a fast chain reaction. Consequently, PINO• is also
being formed quite rapidly, and it may suffer partial self-
decomposition (Figure 7), in parallel with its involvement in
the oxidation of substrates. Owing to the gradual decomposition
of PINO•, the catalytic cycles become slower. Hence, the
monomeric Co(III) generated in the chain reaction cannot reenter
the catalytic cycle as efficiently given the net loss of NHPI.
Thus, monomeric Co(III) gradually accumulates in the solution
and then forms Co(III)s.19 Once fresh NHPI has been added,
PINO• is again generated in the reaction between Co(III)s and
NHPI, and the catalytic cycle resumes.
ArCH2COOH + Co(OAc)2 + HOAc f
ArCH2O• + CoIII(OAc)3 + H2O (6)
The highly reactive ArCH2O• then reacts with pX (here, ArCH3)
to form an aralkyl radical:
•
ArCH2O• + ArCH3 f ArCH2OH + ArCH2
(7)
As mentioned above, the aralkyl radical can also be formed in
the following pair of reactions:
Co(OAc)3 + R2NOH f Co(OAc)2 + R2NO• + HOAc (8)
•
R2NO• + ArCH3 f R2NOH + ArCH2
(9)
The subsequent oxygenation of ArCH2• produces the important
alkylperoxyl radical:
ArCH2 + O2 f ArCH2OO•
(10)
•
To justify the occurrence of reaction 8, independent studies
of the reduction of Co(III)s with NHPI were carried out in HOAc
at variable temperatures. PINO• was formed quite rapidly, with
t1/2 ) 33 s for a reaction between 0.3 mM Co(III)s and 2.0 mM
NHPI at 15 °C (Figure S7). The reaction between pX and PINO•,
eq 9, exhibits a large KIE value, kH/kD ) 25.0 at 25 °C.29 Thus,
the hydrogen abstraction in reaction 9 can be justified. Under
the present catalytic condition, the KIE is Vi(H)/Vi(D) ) 3.4 (
0.2, which indicates that the catalytic cycle also involves
Metal Cocatalysts. The roles of the redox-active metals Mn-
(II) and Ce(III) can be explained by their participation in the
chain-branching mechanism, Scheme 1, similar to the one

430 J. Phys. Chem. A, Vol. 108, No. 3, 2004
Saha et al.
Co(III) oxidation and its subsequent decomposition were
determined. The results are as follows: (i) the formation of
unsubstituted PINO• is faster than that of 4-Me-PINO•; (ii) the
decomposition of unsubstituted PINO• is slower than its
substituted analogues, increasing in the order PINO• < 3-F-
PINO• < 4-Me-PINO•. Thus, it can be concluded that the kinetic
stability, and hence the available concentration of each type of
PINO• for the reaction with pX, follows the order PINO• >
3-F-PINO• > 4-Me-PINO•. This combination of reactivities
explains our observed activity order.
The NAPI-promoted oxidation of pX is slower than that with
NHPI, but NAPI remained active even after recycling. This can
be explained by gradual hydrolysis of NAPI to the active form
of promoter, NHPI, utilizing the water resulting from the
oxidation. Therefore, NAPI is resistant to the rapid decomposi-
tion at an early stage of the reaction when the chain reaction
takes place rapidly.
Figure 7. Decomposition of PINO radical generated from Co(III)s and
NHPI at 60 °C in HOAc. Inset: ln(kd/T) versus 1000/T, giving ∆Hq )
40.7 ( 0.6 kJ mol-1 and ∆Sq ) -109 ( 2 J K-1 mol-1. kd ) second-
order rate constant for self-decomposition of PINO•.
SCHEME 1: Chain Branching of Peroxyl Radicals in
the Autoxidation of pX with Co(II)/NHPI
Acknowledgment. We are grateful for support from the
National Science Foundation (Grant CHE-0240409) and BP
Chemicals. This research was in part carried out at the facilities
of the Ames Laboratory of the Office of Basic Energy Sciences,
U.S. Department of Energy, under Contract W-7405-Eng-82
with the Iowa State University of Science and Technology.
Supporting Information Available: Tables showing the
effects of reactant and catalyst concentrations and of cocatalysts
on the initial rates of O2 consumption, a diagram of the
nanometric device for oxygen uptake measurements, a HPLC
chromatogram of pX oxidation products, plots of initial rates
of O2 consumption versus concentration of reactant and catalysts,
comparison of experimental and calculated initial rates of O2
consumption, yields of oxidation products as a function of time,
and formation of phthalimide N-oxyl radical. This material is
available free of charge via the Internet at http://pubs.acs.org.
SCHEME 2: Inhibition of Chain Branching by Excess
Mn(II)
References and Notes
(1) Emanual, N. M.; Denisov, E. T.; Mazius, Z. K. Liquid-Phase
Oxidation of Hydrocarbons; Plenum Press: New York, 1967; p 350.
(2) Yoshino, Y.; Hayashi, Y.; Iwahama, T.; Sakaguchi, S.; Ishii, Y. J.
Org. Chem. 1997, 62, 6810-6813.
(3) Brill, W. F. Ind. Eng. Chem. 1960, 52, 837-840.
(4) Raghavendrachar, P.; Ramachandran, S. Ind. Eng. Chem. Res. 1992,
31, 453-462.
reported for the Co(II)/Mn(II)/bromide-catalyzed oxidation of
pX.18 The second-order rate constant for the step ArCH2OO•
+ Mn(II) is 103 times faster than the corresponding reaction
with Co(II).24 This can be explained by the lower reduction
potentials of metal cocatalysts, 1.54 V for Mn(III/II) and 1.61
V for Ce(IV/III), compared with Co(III/II) (2.0 V).34 Thus, in
the presence of Mn(II) or Ce(III) step A of Scheme 1 becomes
very fast, leading to rapid chain branching. Hence, ArCH2OO•
is formed quickly. Step A with Mn(II) involves formation of
an intermediate ion pair, ArCH2OO-‚MnIII.24 As shown in
Scheme 2, the ion pair reacts with excess Mn(II), resulting in
one ArCH2OO• instead of three. Therefore, the rate passes
through a maximum at an optimal Mn(II) concentration.
Substituted NHPI Promoters. From Figure 5, the activities
of substituted NHPI follow the order NHPI > 3-F-NHPI >
4-Me-NHPI. Wentzel et al.16 reported the Co(II)/NHPI-catalyzed
autoxidation of ethylbenzene. They suggested that 3-F-NHPI
is most active, particularly at the initial stage of the reaction.
The authors proposed a polar transition state between PINO•
and substrate, where a partially negatively charged hydroxyl
oxygen atom was stabilized by the electron-withdrawing F of
3-F-PINO. While such a polar transition state model does not
support our experimental order of activity of substituted NHPI,
we interpret our result in terms of the kinetic stability of the
substituted PINO•.
(5) Jacob, C. R.; Varkey, S. R.; Ratnasamy, P. Appl. Catal., A 1999,
182, 91-96.
(6) Cincotti, A.; Orru, R.; Cao, G. Catal. Today 1999, 52, 331-
347.
(7) Partenheimer, W. Catal. Today 1995, 23, 69-157.
(8) Broeker, J. L.; Partenheimer, W.; Rosen, B. I. U.S. Patent No.
5453538, 1995.
(9) Environmental Protection Agency. Fed. Regist. 2003, 68.
(10) Li, C. M.; Hou, J.; Yu, Y.; Hou, H. J. EnViron. Sci. 2000, 12, 266-
269.
(11) Ishii, Y.; Sakaguchi, S.; Iwahama, T. AdV. Synth. Catal. 2001, 343,
393-427.
(12) Ishii, Y.; Iwahama, T.; Sakaguchi, S.; Nakayama, K.; Nishiyama,
Y. J. Org. Chem. 1996, 61, 4520-4526.
(13) Ishii, Y.; Nakayama, K.; Takeno, M.; Sakaguchi, S.; Iwahama, T.;
Nishiyama, Y. J. Org. Chem. 1995, 60, 3934-3935.
(14) Ishii, Y. J. Mol. Catal. A 1997, 117, 123-137.
(15) Metelski, P. D.; Espenson, J. H. J. Phys. Chem. A 2001, 105, 5881-
5884.
(16) Wentzel, B. B.; Donners, M. P. J.; Alsters, P. L.; Feiters, M. C.;
Nolte, R. J. M. Tetrahedron 2000, 56, 7797-7803.
(17) Jiao, X.; Espenson, J. H. Inorg. Chem. 2000, 39, 1549-1554.
(18) Metelski, P. D.; Adamian, V. A.; Espenson, J. H. Inorg. Chem.
2000, 39, 2434-2439 and references therein.
(19) Jones, G. H. J. Chem. Res., Synop. 1981, 228-229.
To determine the kinetic stability of the PINO radicals, the
rate constants for the formation of each type of radical35 by

Autoxidation of p-Xylene to Terephthalic Acid
J. Phys. Chem. A, Vol. 108, No. 3, 2004 431
(20) Chipperfield, J. R.; Lau, S.; Webster, D. E. J. Mol. Catal. 1992,
75, 123-128.
(21) Partenheimer, W.; Gipe, R. K. In Nature of the cobalt-manganese-
bromine catalyst in the methylaromatic compounds process; Oyama, S. T.,
Hightower, J. W., Eds.; ACS Symposium Series; American Chemical
Society: Washington, DC, 1993.
(22) Saha, B.; Espenson, J. H. J. Mol. Catal. A 2004, 207, 121-127.
(23) Zakharov, I. V.; Geletti, Y. V. Petrol. Chem. 1979, 18, 145-
153.
(24) Zakharov, I. V. Kinet. Catal. 1998, 39, 523-532.
(25) Scientist: A Nonlinear Least-Squares Program, Version 2.01,
MicroMath Scientific Software, Inc.
(28) Zakharov, I. V.; Geletii, Y. V.; Adamian, V. A. Kinet. Catal. 1991,
32, 31-37.
(29) Koshino, N.; Saha, B.; Espenson, J. H. J. Org. Chem., accepted
for publication.
(30) Russell, G. A. J. Am. Chem. Soc. 1957, 79, 3871-3877.
(31) Amorati, R.; Lucarini, M.; Mugnaini, V.; Pedulli, G. F.; Minisci,
F.; Recupero, F.; Fontana, F.; Astolfi, P.; Greci, L. J. Org. Chem. 2003,
68, 1747-1754.
(32) Amorati et al. have estimated the O-H BDE as 369 and 364 kJ/
mol in t-BuOH and in non-hydrogen-bonding solvents. The estimated BDE
in DMSO is even larger, 375 kJ/mol (ref 31).
(33) Koshino, N.; Cai, Y.; Espenson, J. H. J. Phys. Chem. A 2003, 107,
4262-4267.
(34) Adamian, V. A. Private communication, as cited in ref 18.
(35) Cai, Y.; Koshino, N.; Saha, B.; Espenson, J. H. Manuscript in
preparation.
(26) Chester, A. W.; Scott, E. J. Y.; Landis, P. S. J. Catal. 1977, 46,
308-319.
(27) Sawatari, N.; Sakaguchi, S.; Ishii, Y. Tetrahedron Lett. 2003, 44,
2053-2056.