Você está na página 1de 9

JOURNAL OF CATALYSIS 116, 531-539 (1989)

Mechanism of the Oxidation of o-Xylene to Phthalic Anhydride

Department of Chemistry, Brunei University, Oxbridge UB8 3PH, United Kingdom

Received April 12, 1988; revised October 21, 1988

A mechanism is proposed for the selective oxidation of o-xylene to phthalic anhydride catalysed
by V205/Ti02. The chief catalytic species is identified as an oxohydroxy-vanadium group forming a
monolayer on the TiO, (anatase) surface, and possible unit reactions involving this and related
groups with o-xylene and with organic species formed from it are presented: these account for the
formation of intermediate products (o-tolualdehyde, phthalide, etc.) as well as of phthalic anhy-
dride itself. The five-membered ring products (phthalide, phthalic anhydride) may arise by dehy-
dration of noncyclic precursors. The unique catalytic properties of the V205/Ti02 system arise from
the conjunction in the same active centre of a reducible V=O bond and an acidic hydroxyl
group. 0 1989 Academic Press, Inc.

INTRODUCTION resulting from the over oxidation of PA.

The structures of these molecules are given
The selective oxidation of o-xylene to below. Other minor by-products have also
phthalic anhydride catalysed by VzOs/Ti02 been reported, and carbon oxides are
is an important industrial process (I), the formed at all stages of the reaction.
free-world production in 1986 being about It is a complicated reaction. To convert
2.3 million tonnes. The reaction proceeds at o-xylene to phthalic anhydride, 12 bonds
near atmospheric pressure and at about have to be broken, and 12 formed; a num-
380-400C to give almost complete conver- ber of other bond-breaking and bond-form-
sion of o-xylene and selectivities to phthalic ing acts will doubtless also occur within the
anhydride of 70-75% (2). Catalysts contain catalytically active species at the surface.
about 6% V20s supported on TiO;? (ana- In view of the considerable importance of
tase), the two in turn being carried on ce- the reaction it is strange that no greater de-
ramic pellets or rings; small amounts of gree of curiosity has been shown concern-
other oxides, acting as promoters, may also ing its mechanism. Mechanistic discussions
be present. were originally confined to defining path-
Work on this reaction performed before ways by which the observable gas-phase
1979 has been reviewed by Wainwright and components might be transformed (5). In
Foster (2), although many of the earlier an earlier paper (6), we outlined a rake
studies employed catalysts that were much mechanism to describe the course of the
less selective than those now used. There reaction, and suggested structures for pos-
have been a number of more recent ac- sible adsorbed intermediates. More re-
counts of the reaction, some employing in- cently, FTIR studies of the interaction of
dustrial catalysts (3) and others laboratory toluene (7, 8) and of o-xylene (8), and of
preparations (4-7). Its principal features intermediate products, with V205/Ti02 cat-
are therefore well known. The major initial alysts have led to more refined mechanistic
product is o-tolualdehyde (TAL), which is speculations, but in both cases these appear
subsequently converted via o-phthalide to be based on unrealistic models for the
(PL) to phthalic anhydride (PA): o-toluic catalytically active surface: a critique of
acid is a minor intermediate, while maleic these ideas will be developed below. An
anhydride (MA) is a significant by-product even less sophisticated discussion of the
Copyright 8 1989 by Academic Press, Inc.
All rights of reproduction in any form reserved.

mechanism of these and related reactions this is not to say, however, that other
has also recently appeared (9). phases approximating more closely to V205
Equally strange is the lack of effort to that may exist at high V205 concentrations
understand how the nonselective reaction (23, Z4), or even unsupported V205 (Z5),
proceeds. There seems to be a blithe as- are necessarily inactive, but the evidence
sumption that o-xylene and any product de- (5, 12) clearly suggests that such phases
rived from it is capable of being deeply ox- play only a minor role in the reaction. Sec-
idized (5), but how and why this happens is ond, the active surface is now well estab-
rarely considered. It is not even known lished to have both Lewis and Bronsted
whether deep oxidation occurs at some spe- acidity, to extents depending on the amount
cific type of site on the catalyst surface, or of water and of V205 present (Z6), for
whether it necessitates some particular which the Vejux-Courtine model makes no
type of adsorbed intermediate. It has, how- provision. Third, direct examination by
ever, been noted that at low temperatures electron microscopy (17, 18) and the use of
and low conversion part of the o-xylene dis- other methods of physical characterization
appears as a strongly adsorbed carbo- (7, 14, 19) provide no support for their con-
naceous deposit (5, 6), which acts as a cept of the existence of a multilayer having
catalyst poison and which under some the structure of V205 in register with the
circumstances vacates the surface to con- TiOz support.
dense as an oily liquid downstream of the We may therefore assume for the pur-
reactor (6). This phenomenon has been im- poses of this paper that the catalytically ac-
plicated in the formation of carbon oxides tive species reside in a monolayer of
(6). V205 upon the TiOz surface and that
The purpose of this paper is to present a they are capable of exhibiting both Bron-
series of unit steps that can account for the sted and Lewis acidity, as well as electron-
formation of the selectively oxidized prod- accepting character (7): they are easily re-
ucts. It is now possible to do this, since ducible by H2 to a V3+ species in a one-step
there is reasonable certainty as to the na- process (Z4), and do not show the charac-
ture of the catalytically active species. The teristic strong Raman band for the
first step is therefore to review briefly what V=O.--V vibration (14, 16). Bearing in
is known about this species, and to suggest mind EXAFS studies (20) and much other
how it may act to effect the observed reac- work (2Z), the principal surface species is
tions . formulated as a tetrahedral oxohydroxy va-
nadium complex (Structure A of Scheme
l), where both the lower oxygen atoms
THE NATURE OF THE CATALYTICALLY form part of the anatase surface. The sur-
face of a monolayer catalyst is therefore
The reason for the peculiar efficiency of quite different from that of V205 itself. Al-
the V205/Ti02 combination for the selective though Structure A is proposed as the chief
oxidation of o-xylene has long been a mat- species present under reaction conditions,
ter for debate. The notion (10) that there when water from the partial combustion of
exists a precise crystallographic fit between the o-xylene is present, under anhydrous
the surface of Ti02 (anatase) and the layer conditions there may occur an intermolecu-
structure of V205 has commanded much at- lar dehydration, leading to linear chains of
tention, but is open to criticism on a num- V-O-V bridged species. Evidence pointing
ber of grounds. First, catalytic activity is in this direction (16, 22) does not therefore
now definitely attributable to a single VO, conflict with what is proposed here, since it
monolayer in which the three-dimensional has been obtained under anhydrous condi-
structure of V205 cannot exist (5, 6, II, 12): tions. The hydration of Lewis acid centres

more bonds to atoms connected to the sub-

stituent groups on the aromatic ring, as
shown by the broken lines. Hereafter the 0
atoms linking these species to the TiOz sur-
face will be omitted for simplicity, as the
breaking of V-O-T1 bonds is not envis-
aged. Only species A and D are assumed to
be present at the outset; C is formed from A
by reduction involving the organic reac-
tants and is reoxidizable to A by molecular
OZ. Ample amounts of O2 and HZ0 are
present to allow the vertical transitions to
SCHEME 1. Proposed structures of catalytic species take place freely: the mode of chemisorp-
having single V atoms. The numbers are the oxidation tion of O2 is not to be discussed in detail,
states of the V atoms in the species shown, and the but may involve species C. All the struc-
broken lines denote possible points of attachment of
atoms connected to the substituent groups on the aro- tures shown in Scheme 1 bear either an 0
matic ring. atom or an OH group, or both, and are
identified as those species which initiate
unit reactions. We shall encounter addi-
tional species that are the products of such
to Bronsted centres has been observed ex- reactions and these may only have a bond
perimentally (14). linking the V atom directly or indirectly to
the substituent C atom. It will usually be
quite clear what has to be done to restore
Whereas both TiO;! and V205 predomi- the initial structure of the active site. To
nantly catalyse dehydration of isopropanol describe the later stages of the reaction se-
to propene, the surface complex gives quence, it will be necessary to invoke pairs
about 25% dehydration and 75% dehydro- of active sites having structures such as
genation to acetone (23). While the former those shown in Scheme 1. No use is made
reaction is easily rationalized in terms of a of V-H bonds, and molecular H2 is not sug-
proton-assisted elimination of water, the gested as a product.
latter requires the addition of the isopro- Species A and B in Scheme 1 will show
pan01 molecule at the OH bond across the Bronsted acidity and Species D Lewis acid-
V=O, with the formation of a V-O-C ity. Both types of acid centre have been
species. Internal electron movement then detected in calcined catalysts (16, 21, 24,
gives acetone and HZ. These reactions will 251, but under reaction conditions, where
be shown to be archetypal of those occur- as noted above water is amply present,
ring during o-xylene oxidation. The con- such V4+ species as are present (24) are ex-
junction of the acidic and oxidizing (i.e., H- pected to be predominantly in the hydrated
accepting) functions in the V-OH and form B. The presence of O2 will ensure that
V=O bonds, respectively, accounts well the V4+ and V3+ concentrations remain low,
for the unique catalytic properties of the so the net oxidation state of the V atoms
V205/Ti02 system. will depend chiefly on the concentration of
In formulating a reaction mechanism, adsorbed radicals formed by reductive pro-
only four variants of the catalytic species cesses. Of the species depicted in Scheme
are needed (Structures A-D in Scheme 1); 1, species A is considered to be the most
these have vanadium oxidation states of 5, abundant, and for these reasons, chemi-
4 and 3 as shown in the scheme, and three sorptive reactions of the o-xylene and of
of the four are able to accommodate one or the intermediate products are represented

o-xytene o-tolualdetyde tioluic acid phthalide anhydride

pq pq pg l-@q
vll v/l vll v
SCHEME 2. Simplified form of the rake mechanism based on five surface intermediates.

as involving the Brprnsted acid centre A then be classified into one of three types,
rather than either B or C. viz., (i) H atom abstraction; (ii) 0 atom in-
sertion (i.e., O=V-Cz to V-O-C-);
A GENERAL OUTLINE OF THE and (iii) 0 atom rearrangement (i.e.,
PHTHALIC ANHYDRIDE comes apparent that each unit step can be
The sensible way to represent the reac- written in more than one way; those shown
tion mechanism is as a rake comprising a below have been selected as being the most
series of adsorbed intermediates, each con- likely, but may not be unique. The bases on
verted to the next by partial oxidation, and which the selection has been made include
some if not all of which can desorb, if nec- (a) the use where possible of species A
essary by further chemical reaction, to give rather than D on the grounds that it is likely
the observed gas-phase products. Scheme 2 to be more abundant; (b) the use where pos-
illustrates the principle and resembles that sible of a single V species, rather than an
given before (6), save that o-toluic acid is adjacent pair; (c) inherent chemical proba-
now included as an intermediate product. bility; and (d) simplicity.
Routes to polymer, carbon oxides, ma-
leic anhydride, and other products are not H ATOM ABSTRACTION
included. It should be easily possible to
evaluate the relative magnitudes of the unit H atom abstraction occurs in one of three
rate constants that define a given pattern of ways: (i) from a -CH3 group in a molecule
product yields as a function of conversion, not already attached to the surface, as in
but this has apparently not been tried and the initial chemisorption of o-xylene; (ii) by
will not be undertaken in this paper. breaking a C-H bond where the carbon
The chief purpose of the ensuing discus- atom is already attached to the surface; and
sion is to try to identify probable structures (iii) from a -CH3 group where the molecule
for the adsorbed species involved and pos- is attached to the surface via the other sub-
sible mechanisms for the various compo- stituent group.
nent reactions. The C-H bonds in the -CHJ groups of u-
xylene are weakened by the presence of the
aromatic ring. The chemisorption of o-xy-
THE GROUND RULES lene is therefore most economically shown
Every unit reaction is formulated as be- as in Scheme 3. In this and subsequent
ing radical rather than ionic in nature: schemes, each structure is denoted by a let-
although some of the bonds involved are ter identifying its form as shown in Scheme
undoubtedly highly polar, the radical repre- 1, and a number defining the nature of the
sentation is probably nearer the truth. All bond(s) to the substituent C atom. Struc-
unit reactions of the organic molecules can ture E has only the bond(s) to this C atom.

4, ,OH I HO$OH a ON ,lH2


The same reaction will describe the chemi- 0 ATOM INSERTION AND REARRANGEMENT
sorption of the second -CHx group, where
the other substituent group is attached to a The next and very important step is to
neighbouring V atom (see below). A second construct a C-O bond. This may occur by
method of H atom abstraction will also be 0 atom insertion into the V-C bond of
considered later. Note that when the re- Structure Dl (Scheme 4). Rearranging this
mainder of the organic molecule is not un- structure to make a>C=O group necessi-
dergoing reaction, it is not shown in the re- tates the elimination of two H atoms; this
action schemes. might proceed by the sequence shown in
There seems to be general agreement that Scheme 5. These reactions are regarded as
the reaction is initiated by activation of the essentially irreversible. The carbonyl spe-
-CHs group, although both Miyata et al. (7) cies E2 then leads preferentially to o-tolu-
and van Hengstum et al. (8) prefer to depict aldehyde in the gas phase. Since it can
this as a simultaneous reaction of two C-H apparently also readsorb readily and un-
bonds with V=O groups and the formation dergo further reaction, its desorption and
of a V-O-C bond with the aid of a third readsorption must be formulated with due
V=O group, involving the reduction of regard to microscopic reversibility (Scheme
three V5+ species to V4+. The likely separa- 6). o-Tolualdehyde therefore chemisorbs
tion between V=O groups on the surface is on what is taken to be the most abundant
such as to render this mechanism improba- surface species. There is incidentally no ev-
ble. Failure to recognise Structures Bl or idence for the formation of the o-toluyl al-
Dl in FTIR may not be surprising: they cohol, from which we conclude that the V-
may quickly undergo the next reaction 0-CHZ- species El shown in Scheme 5 is
step. There is the additional suggestion for not hydrolysed to V-OH + HOCH*-: this
the adsorption of toluene (7) that the aro- is as expected for an ether-like bond.
matic ring may be n-bonded to a The work of van Hengstum et al. (8)
neighbouring Lewis acid centre (e.g., shows clearly that the adsorbed form of o-
Structure D in Scheme 1). Even if this were tolualdehyde does not possess a normal C-
to apply with o-xylene, it would have no H bond, and so our representation of it as
effect on reactions at the substituents. Structures B2 or D3 in Scheme 6 is consis-



- 0
<- $


tent with their observation. It has to be said to the -COO. stage (with the dot * repre-
that the significance of the partial bonds in- senting bonding to the surface) before oxi-
dicated by these authors by dashed or dot- dation of the second -CH3 starts. The
ted lines is by no means clear. It seems dialdehyde is not observed as an intermedi-
preferable to employ normal, full covalent ate product, and indeed it has been sug-
bond symbolism to represent any structures gested (6) that premature oxidation of the
which may have some degree of stability. second -CH3 group (even just to the extent
The next step is to form a carboxy spe- of the breaking of a C-H bond with conse-
cies and hence to provide a route to o-toluic quent formation of a V-C bond) is likely to
acid in the gas phase. The carboxy species lead to deep oxidation. It appears that,
may be formed from the carbonyl species once one -CH3 group has achieved the
by a further 0 insertion step (Scheme 7). -COO- stage, oxidation of the second may
This then with low probability may form o- safely commence. It is doubtful whether
toluic acid in a manner analogous to that by progressive oxidation of the second -CHj
which o-tolualdehyde is formed (Scheme group can occur at the same V atom as that
8). Structure E3 in these schemes is ex- to which the first fully oxidized substituent
actly as Structure D in the formulation of is already bonded: it is more likely that a
van Hengstum er al. (8), and it is accepted second neighbouring atom is involved, and
that it could equally .be shown as a mono- that the species participating at the start of
dentate or bidentate ligand, as suggested by the second phase is best represented as
them, if this were more consonant with the Structure A, E3 in Scheme 10. Reaction
infrared evidence. There is therefore no di- may then proceed on the second -CH3
rect conflict between the proposals in this group as outlined above for the first.
paper and the relevant experimental evi-
dence available to date. THE FORMATION OF FIVE-MEMBERED
One of the strangest features of this reac-
This then completes the first phase of the tion, and one that has been little com-
reaction. There are very clear indications mented on, is the ease with which the five-
that one -CH, group must be fully oxidized member ring products, viz., phthalide (PL)
and PA, are formed. Ring closure by inser-
tion of an 0 atom into a dicarbonyl species
I seems inherently improbable, and some al-
c=o Lo
Cl ] - 1 ternative route to ring closure therefore
needs to be found.
A@ A E3 o-Phthalic acid readily undergoes ring
closure by dehydration even at its melting
point (21OC), and it is therefore possible
SCHEMES that PA is formed after oxidation of both



substituents to -COOH. Under reaction electron movements postulated closely re-

conditions, the dehydration would probably semble those that would describe ring clo-
proceed so rapidly that the nonappearance sure in free o-phthalic acid, and exactly the
of o-phthalic acid need not cause surprise. same sequence applied to the structure El,
We will consider further below whether E3 above would account for the formation
its actual formation is necessary or not. of PL. Two further remarks are in order. (i)
Arguing by analogy, PL could result from An additional advantage of having a mono-
dehydration of another unobserved in- adsorbed species in the cyclisation step lies
termediate, 2-hydroxymethylbenzoic acid in its greater flexibility vis-a-vis its diad-
(Scheme 9). Ring closure by dehydration sorbed precursor; this may assist formation
seems distinctly more plausible than invok- of the transition state. (ii) PL is an observed
ing 0 atom insertion. intermediate that can be converted to PA;
Returning to the reaction sequence, the reversal of the cyclisation, as depicted
reaction will continue as shown in Scheme above for PA, to return it to the reaction
10: here the codes beneath each structure sequence involves surface species (A and
define the bonding about the V atoms, and C) that are thought to be reasonably popu-
the numerals by the arrows relate to the lous.
previous schemes, where details concern-
ing the proposed reaction mechanisms may
be sought. Now it is possible to envisage Maleic anhydride (MA) is a significant
the structures El, E3 and E3, E3 forming, by-product at high conversion of o-xylene.
respectively, PL and PA. This would, how- It is usually thought to be formed from PA,
ever, leave one of the V atoms in the 2-t but more probably it is formed from one of
oxidation state, a possibility ruled out by an the last adsorbed intermediates in the reac-
admittedly arbitrary assumption. It is there- tion sequence. The critical step is likely to
fore worth exploring whether these struc- be the breaking of a C-H bond attached to
tures might afford the cyclic products with- the benzene ring, followed by ring opening
out the postulated noncyclic intermediates of the latter. Once begun, the process goes
being completely desorbed from the sur- to completion, that is, the four carbon at-
face. One possibility is illustrated in oms not involved in the anhydride ring or
Scheme 11. The -OH groups attached to its precursor form carbon oxides: traces of
each V atoms are introduced to facilitate citraconic anhydride (i.e., methyl-substi-
the elimination of HZ0 and to maintain the tuted MA) have, however, been reported.
oxidation state of the V atoms at a mini-
mum of 3. These -OH groups could arise as
shown in Scheme 12, i denoting any appo- COOH
priate numeral. Cyclisation then proceeds
through an adsorbed species tied to the
CH,OH - a
o CH:2

surface at only one point (Scheme 11). The SCHEME 9


fiov5 fJo
i$ /\/\ w co
9 7 /\
A (EZ,E3) (EZ,E3)

-0 52
O\Y cy/fpH HO&g0

/\ /\ /\ /\

(Q.C2) (n.o) (

I Ia O\T 0 a HO< i,OH

/\ A I-H.O /\ /\


SCHEME 13. Reactions leading to observed C8 products.


Comment was made earlier on the impor- Galantowicz, M., Gasior, M., Grzybowska, B.,
tance of understanding the mechanism of and Sloczynski, J., Przem. Chem. 62, 87 (1983).
Saleh, R. H., and Wachs, I. E., Appl. Catal. 31,87
carbon oxide formation if positive steps are (1987).
to be taken to suppress it. The reactant and Bond, G. C., and Konig, P., J. Catal. 77, 309
each observable intermediate have at some (1982).
time been implicated (2, 5), and as just Miyata, H., Mukai, T., Ono, T., Ohno, T., and
noted the overoxidation of PA to MA must Hatayama, F., J. Chrm. Sot. Faraday Trans. I
84, 2465 (1988).
afford further carbon oxides. Mention was
8. van Hengstum, A. J., Pranger, J., van Hengstum-
made above of the suggestion (6) that Nijhuis, S. M., van Ommen, J. G., and Gellings,
chemisorption of o-xylene simultaneously P. J., J. Catal. 101, 323 (1986).
through both -CH3 groups might lead to a 9. Andersson, S. L. T., J. Catal. 98, 138 (1986).
strongly adsorbed species that might ulti- 10. Vejux, A., and Courtine, P., J. Solid State Chem.
23, 93 (1978).
mately (e.g., after polymerisation) suffer
11. Bond, G. C., and Briickman, K., Faraday Disc.
deep oxidation. A connection between the Chem. Sot. 72, 235 (1981).
carbon imbalance and the formation of car- 12. Saleh, R. Y., Wachs, I. E., Chan, S. S., and
bon oxides gave (6) some credence to this Cherisch, C. C., J. Catal. 98, 102 (1986).
belief, but a quantitative analysis of the 13. Cavani, F., Centi, G., Foresti, E., Trifirb, F., and
product concentrations as a function of Busca, G., J. Chem. Sot. Faraday Trans. 1 84,
237 (1988).
conversion is needed to conform this idea. 14. Bond, G. C., Perez Zurita, J., Flamerz, S., Gel-
There are also strong indications that the lings, P. J., Bosch, H., van Ommen, J. G., and
nonselective process is associated, at least Kip, B. J., Appl. Catal. 22, 361 (1986).
in part, with bare patches of TiOz surface 15. Gasior, M., Grzybowska, B., Haber, J., Machej,
(4-7, 12). T., and Ziolkowski, J., J. Cata/. 58, 15 (1979).
16. Miyata, H., Fujii, K., and Ono, T., J. Chem. Sot.
In conclusion, the main outline of the Faraday Trans. 184, 3121 (1988).
proposed reaction scheme is shown in I7 Backhaus, K. O., Haase, R., Illgen, U., Richter-
Scheme 13. However speculative the de- Mendau, J., Scheve, J., Schulz, I. W., Thiessen,
tails of the various transformations may be, K., and Vetter J., Mater. Sci. Monogr. 28B, 1041
it seems inevitable that the reaction should
18. Vejux, A., and Courtine, P., J. Solid State Chem.
follow a path similar to that shown. It is to 63, 179 (1986).
be hoped that this theoretical analysis may 19. Bond, G. C., Perez Zurita, J., and Flamerz, S.,
stimulate further experimental work de- Appl. Catal. 27, 353 (1986).
signed to unravel the course of this fasci- 20. KozXowski, R., Pettifer, R. F., and Thomas, J. M.,
J. Phys. Chem. 87, 5176 (1983).
nating reaction.
21. Gellings, P. J., in Specialist Periodical Reports-
Catalysis (G. C. Bond and G. Webb, Eds.), Vol.
REFERENCES 7, p. 105. Royal Society of Chemistry, London,
1. Patterson, W. R. in Catalysis and Chemical Pro- 1985; Busca, G., Centi, G., Marchetti, L., and
cesses (R. Pearce and W. R. Patterson, Eds.), p. Trifiro, F., Langmuir 2, 568 (1986).
251. Blackie, Glasgow, 1981. 22. Kijenski, J., Baiker, A., Glinski, M., Dollenmeier,
2. Wainwright, M. S., and Foster, N. R., Catal. Rev. P., and Wokaun, A., J. Catal. 101, 1 (1986).
Sci. Eng. 19, 211 (1979). 23. Bond, G. C., and Flamerz, S., Appl. Catal. 33,219
3. Nikolov, V. A., Klissurski, D. G., and Had- (1987).
jiivanov, K. I., in Catalyst Deactivation, 1987 24. Busca, G., Langmuir 2, 577 (1986).
(B. Delmon and G. F. Froment, Eds.), p. 173. 25. Ivanovskaya, F. A., and Sembaev, D. K., Zh. Fiz.
Elsevier, Amsterdam, 1987. Khim. 61, 494 (1987).