Alkane Dehydrocyclization Mechanism

Download as pdf or txt
Download as pdf or txt
You are on page 1of 74

Catalysis Today 53 (1999) 443–516

Alkane dehydrocyclization mechanism


Burtron H. Davis ∗
Center for Applied Energy Research, University of Kentucky, 2540 Research Park Drive, Lexington, KY 40511, USA

Abstract
A review of results impacting the dehydrocyclization mechanisms for monofunctional catalysts proposed to date has been
made. The data indicate that, while alkane conversions at low temperatures (about 300◦ C or less) may involve reversible
adsorption, irreversible adsorption of the alkane is the rate determining step at higher temperatures representative of naphtha
reforming. The extent of irreversible adsorption depends upon hydrogen partial pressure, with alkane adsorption being
irreversible near atmospheric pressure and gradually changing to reversible adsorption at naphtha reforming pressure. At
all pressures, primary aromatic products are formed by a mechanism involving direct six-carbon ring formation. At low
pressures, the cyclopentane cyclization product yields carbon on catalyst and cracked products. At higher pressures, the
cyclopentane structures undergo nonselective hydrogenolysis to produce isomers of the alkane feed by a monofunctional
metal catalyzed pathway. A common cyclization pathway is proposed for the formation of C5 - and C6 -ring structures. With
bifunctional catalysts, the acid catalyzed cyclization is more rapid than the monofunctional metal cyclization. ©1999 Elsevier
Science B.V. All rights reserved.
Keywords: Mechanism; Alkane dehydrocyclization; Alkane; Hydrogenolysis; Naphtha reforming; Isotopic tracer studies

1. Introduction dehydrocyclization catalysts as indicated by the recent


introduction of the Pt-KL zeolite catalyst [9,10]. Look-
Dehydrocyclization of alkanes to produce aromatics ing to the future, coal and natural gas will probably
was discovered independently by a number of groups provide an increasing fraction of the liquid hydrocar-
[1–7]. It was recognized as an important discovery bon fuels. The Fischer–Tropsch synthesis, where coal
since the aromatics that were produced had a high oc- is first reacted with water to form the CO + H2 synthe-
tane number whereas the alkanes they were produced sis gas, is one of several routes which offer promise
from had a low octane number. Thus, the announce- for converting coal to liquids. Since the liquids from
ment by UOP workers was delayed because of the im- the Fischer–Tropsch Synthesis contain a large fraction
portance attached to it by the US government due to of alkanes, dehydrocyclization could assume an even
fears arising from the political crisis that eventually more important role in the reforming process.
led to World War II [4]. The discovery of naphtha re- Examples of the reaction classes occurring in the
forming with the bifunctional chlorided Pt-Al2 O3 cat- reforming process which impact the octane number
alyst increased the interest in dehydrocyclization [8]. and yield of the gasoline are:
Even today, the search continues for more selective 1. Dehydrogenation

∗ Tel.: +1-606-257-0251; fax: +1-606-257-0302.

E-mail address: [email protected] (B.H. Davis)

0920-5861/99/$ – see front matter ©1999 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 0 - 5 8 6 1 ( 9 9 ) 0 0 1 3 6 - 4
444 B.H. Davis / Catalysis Today 53 (1999) 443–516

2. Isomerization

3. Aromatization

4. Cracking (hydrocracking)

(‘A’ refers to an acid site and ‘M’ to a metal site). predominantly Somorjai and coworkers [18], have ad-
The first three reactions lead to higher octane num- vanced the view that much of the surface of platinum
ber products but reaction (4) may be undesirable since single crystal catalyst is covered by carbonaceous de-
it decreases the yield of gasoline. It should also be re- posits that play a major role in protecting some small
alized that these reaction schemes may not be unique fraction of active centers of the catalyst.
in a reaction network where a large number of re- Carbon atoms have been generated by the ‘carbon
actants may lead to over 100 reaction products. For arc’ method [19] and these were deposited in, for ex-
example, there may be an acid catalyzed cyclization ample, benzene held at −196◦ C [20]. Upon warming
which competes with reaction (3) [11]. In fact, under to room temperature the carbon atoms react with ben-
reforming conditions, the bifunctional acid catalyzed zene to form a variety of compounds. With saturated
cyclization pathway be dominant [12–14]; however, hydrocarbons, the atomic carbon reacts by insertion
the above reactions are probably the dominant ones in a C–H bond to produce a product containing one
for the process monofunctional pathway [15–17]. more carbon than the reactant. When benzene is the
Invariably, during hydrocarbon conversion, car- reactant, toluene and other aromatic products can be
bonaceous deposits are formed on the metallic func- formed. Thus, a free carbon atom would be unstable
tion and the acidic support. The nature of these carbon on a platinum surface even at room temperature, let
deposits has not been defined in much detail. Several, alone at 480◦ C where dehydrocyclization reactions are
B.H. Davis / Catalysis Today 53 (1999) 443–516 445

Scheme 1.

effected. While dehydrocyclization may produce mi- Table 1


nor amounts of aromatics with one, or even two, more Aromatic distribution reported by early workers using chromia
catalysts
carbon atoms than the original feed, they must arise
using other pathways or reactive intermediate surface Reactant C8 aromatics (mol%)
species than atomic carbon (e.g., [21]). We should con- EB OX MX PX
clude that the carbonaceous deposits on the platinum
n-octane 33 33 27 4
function must be multiply bonded to platinum and/or 3-methylheptane 15 25 – 60
to hydrogen. 2-methylheptane – – 100 –
3-ethylhexane 100 – – –
2,3-dimethylhexane 25 75 – –
2. Aromatic distribution and early mechanisms
Approach (b) has been used extensively with a
Two general approaches are available for investi- variety of reactants in studies directed toward eluci-
gating heterogeneous catalysis: (a) study one reactant dating the cyclization steps. The aromatic distribution
over a wide variety of catalysts and (b) study many re- obtained by early workers from the conversion of
actants over a single catalyst. An elegant example of some representative alkanes over chromia are pre-
the single reactant approach was conducted by Kokes sented in Table 1. The accuracy of these distributions
and coworkers (e.g., [22]) who showed that a metal ox- have been confirmed many times using more accurate
ide catalyst, ZnO, resembled in many respects Group GC analytical techniques. In spite of an abundance of
VIII metal catalysts in the hydrogenation of alkenes. aromatic isomers that could not result from a direct
For the single reactant approach to mechanistic stud- C6 -ring formation, the early thinking materialized as
ies, it is generally desirable to have a chemical se- the Twigg mechanism [23] which incorporated the
lectivity in addition to the kinetic data. As outlined ‘half-hydrogenated’ state that was being applied in
in Scheme 1, n-octane and 3-methylheptane provide the 1930s to explain alkene hydrogenation and iso-
a chemical selectivity (in Scheme 1 we anticipate our tope exchange reactions (e.g., [24]).

conclusions by showing the selectivity resulting from 2.1. Twigg mechanism


a direct six-carbon ring formation).
The chemical selectivity of these, and other reac- Numerous schemes, including bicyclic interme-
tants, are used to compare several catalyst systems and diates, were advanced to account for the aromat-
to infer reaction pathways. ics not allowed by direct C6 ring closure. These
446 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 2.

early studies are covered in excellent review articles first approach, a specific position of the reactant is la-
[15–17,25]. beled with either a stable or radioactive isotope. The
The Twigg mechanism implied that cyclization positions of the product that are labeled with the iso-
leads to an adsorbed cyclohexane; that is, dehydro- tope are determined and a mechanism consistent with
genation to form the aromatic occurred after ring the label distribution is deduced. This approach has
formation. been called the ‘labeled molecule’ approach [27]. In
The aromatics from the conversion of several another approach, a possible intermediate compound
C8 -alkanes with one catalyst (Pt-nonacidic alumina) is labeled with an isotope and is then converted in
provide an example of approach (b), and are illus- competition with another reactant, the kinetic isotope
trated in Scheme 2. approach. Another technique is to make an abrupt
Many of the early mechanistic studies were based switch from a labeled reactant to an unlabeled reac-
on kinetic measurements. Alkenes were deduced to be tant, or vice versa, and then to follow the isotopic
an intermediate since their rate of formation did not composition of the products for increasing time fol-
extrapolate to zero at short reaction times whereas the lowing the switch, the transient method [28]. To date,
rate of aromatics formation did approach zero [26]. the transient approach has not been applied to dehy-
However, for heterogeneous catalysis, kinetic stud- drocyclization but both the labeled molecule and the
ies can, at best, be suggestive. Thus, one must resort kinetic isotope method have been widely utilized. This
to other techniques to unravel details of the reaction paper will critically review the results obtained using
mechanism. One powerful technique involves the use these approaches and place them within a mechanistic
of isotopic tracers. framework.
There are three general approaches for the utiliza- Experimental expediency has caused most mecha-
tion of isotopes in dehydrocyclization studies. In the nistic studies to be carried out at atmospheric pressure.
B.H. Davis / Catalysis Today 53 (1999) 443–516 447

However, the commercial reforming process uses a to- the development of the Pt-Al2 O3 catalyst, the support
tal pressure of 60–500 psi (0.41–3.4 MPa) to retard a became as important as the metal since the isomeriza-
rapid decline in catalyst activity. At another extreme, tion activity was partially, or completely, provided by
many spectroscopic techniques that are now used to the support. The early work of Pines and coworkers
characterize catalysts must operate at high vacuum [32] have called attention to the effect of the support
conditions (ca. 10−8 to 10–10 atm). For example, So- even for the chromia-alumina catalyst.
morjai and coworkers (e.g., [29]) have carried out The influence of the support on the aromatic selec-
the dehydrocyclization of n-heptane in parallel with tivity is clearly demonstrated by the results presented
LEED studies at these low pressures. Consequently, an in Fig. 1. With the Pt-acidic alumina catalyst the aro-
assessment of the influence of pressure on the mech- matic distribution is at, or closely approaches, the equi-
anism is needed. librium value; however, when the alumina support is
made nonacidic by incorporation of alkali metal ions,
the two isomers allowed by direct C6 -ring closure ac-
3. Dehydrocyclization catalysts counted for 95% or more of the aromatic products.
It is widely accepted that Pt on acidic alumina pro-
Both metal and metal oxides have been widely used vides a bifunctional reaction pathway in which the
as catalysts to convert alkanes into aromatics. metal serves as a site for cyclization and for alkene
formation; the acidity serves to convert the n-alkene
to iso-alkenes and cyclic naphthenes by a typical acid
3.1. Metal oxide catalysts
catalyzed carbenium ion reaction. This means that the
reforming process must be concerned with a delicate
Chromia, the catalyst that provided the initial impe- balance of metal and acid catalysis in order to maxi-
tus for dehydrocyclization studies, suffered from two mize the yield of high octane product [33]. The acid-
shortcomings: (1) at atmospheric pressure rapid ‘coke’ ity function of bifunctional catalysts, as shown by the
formation prevented sufficient catalyst life for com- data in Fig. 1, prevents the use of chemical selectiv-
mercial use and (2) the catalyst was not sufficiently ity to study reactions over the metal function. In the
active at the higher hydrogen pressures where the rate present report, we will emphasize studies where the
of ‘coking’ should be retarded. These catalysts have acid function has been drastically reduced by ‘base
been extensively reviewed [15–17,25] and will only neutralization’.
be covered in this report in the section on radiotracers. To prepare nonacidic catalysts that have repro-
Supported molybdena catalysts also received attention ducible properties is not a simple process. Except
but when their commercial use resulted in an explo- for using a KOH containing solution for the final
sion/fire at Amoco’s plant [30] the interest in them wash to prepare Pt-1, the same recipe was used to
waned. prepare two supported catalysts, Pt-1 and Pt-2 [34].
1,1-Dimethylcyclohexane can undergo conversion by
3.2. Pt catalysts isomerization to form xylenes or by demethylation to
form toluene. For a similar total conversion level, the
In the 1950s, the use of the Pt-Al2 O3 catalyst for difference between Pt-1 and Pt-2 is obvious (Fig. 2);
naphtha reforming became a commercial success for clarity, only toluene and o- and m-xylene are
[8]. More recently, the Pt-Al2 O3 catalyst has been shown. On the other hand, Pt-1 and Pt-2 gave a sim-
displaced in the commercial process by supported ilar aromatic distributions for the dehydrocyclization
bimetallic or even multimetallic catalysts; the first of n-octane. This indicates that the selectivity is not
of these was the Pt/Re combination developed by absolutely determined by the catalyst but may also
Chevron workers [31]. depend on the reactant.
Nearly all of the naphtha reforming catalysts have Results for the dehydrocyclization of n-octane
the catalytically active material dispersed on a support. show that as the metal loading on nonacidic alumina
Initially, little attention was paid to the support other increased from 0.05 to 0.6 wt.% Pt, the ethylben-
than as a means of providing a high surface area. With zene: o-xylene (OX : EB) ratio changed from a 1.4 : 1
448 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 1. Comparison of the aromatics from the conversion of n-octane at 482◦ C and atmospheric pressure over Pt on acidic (left) and
nonacidic (right) alumina support . (Equilibrium C8 -aromatic composition indicated in center of figure.) (EB, ethylbenzene; OX, o-xylene;
MX, m-xylene; PX, p-xylene).

Table 2
The selectivity (OX : EB) for the conversion of n-octane over
Pt-Al2 O3 -K catalysts containing various Pt loadinga
Pt loading (wt.%) o-xylene/ethylbenzene

1b 2b 3b

0.05 1.4 – –
0.14 1.3 1.3 1.2
0.22 1.4 1.2 1.3
0.35 1.6 1.0 1.0
0.6 1.0 1.0 1.0
a 482◦ C, LHSV = 0.3, no added hydrogen, 1 atm.
b Sample numbers.
Fig. 2. Conversion of 1,1-dimethylcyclohexane over two different
Pt-Al2 O3 -K catalysts (see text for catalyst description). in these selectivity studies due to the high activity
of the catalyst for hydrogenolysis of aromatic prod-
selectivity to favor OX at low loadings to an equal ucts. These results show that the selectivity changed
selectivity for each isomer at the two highest load- with increasing metal loading from favoring one
ings (Table 2) [35]. Likewise, if we neglect the small isomer to a nearly equal selectivity for all isomers
amount of m-xylene in calculating the aromatic dis- allowed by direct C6 -ring closure. In addition, the
tribution from 3-methylheptane dehydrocyclization hydrogenolysis-demethylation activity increased from
with Pt-C catalysts, it is apparent that the selectiv- a negligible level at the lowest loading to such a high
ity changed from favoring p-xylene at the low Pt level as to preclude selectivity studies at about 2 wt.%
loading to the formation of nearly equal amounts of Pt or higher.
each isomer at 1.2% Pt loading (Fig. 3). Loadings Many of the Pt catalysts are prepared by impreg-
higher than 1.2% Pt on carbon could not be used nation with chloroplatinic acid so there is a concern
B.H. Davis / Catalysis Today 53 (1999) 443–516 449

Table 3
C8 -aromatic distribution for the dehydrocyclization of 3-methylheptane at 482◦ C (from [39])
Catalyst Sample Time on stream (min) C8 -aromatic products (mol%)

Exhylbenzene Xylene isomers


p- m- o-

Pt-Al2 O3 -K, 0.6% 1 33 22 47 4.3 27


3 102 19 52 <1.0 29
Pt-Cl-Al2 O3 -K, 0.6% 1 39 15 42 19 24
3 117 16 45 16 23
Pt-Cl-Al2 O3 -K, 1.7% 1 40 17 38 22 22
3 130 18 35 25 22
Pt-SiO2 , 0.6% 1 36 19 48 4.2 28
2 83 16 51 5.0 28

Table 4
C8 -aromatic isomer distribution calculated by disregarding
m-xylenea
Catalyst Ethylbenzene o-xylene p-xylene
Pt-Al2 O3 -K 23 28 49
Pt-Cl-Al2 O3 -K, 1.7% 22 29 49
Pt-Cl-Al2 O3 -K, 0.6% 18 30 52
Pt-SiO2 20 31 50
Cr2 O3 -Al2 O3 -A (482◦ C) 17 36 47
Cr2 O3 -Al2 O3 -A (500◦ C) 18 35 47
Cr2 O3 -Al2 O3 -B (482◦ C) 19 31 50
Cr2 O3 -Al2 O3 -B (500◦ C) 18 31 51
a For first sample for each run.

rine (Table 3) [39]. The amount of isomerization of


Fig. 3. C8 -aromatic distribution over Pt-C catalysts with increas- m-xylene, as well as for each of the other three aro-
ing Pt content (neglecting the m-xylene; EB, ethylbenzene; OX, matic products, when passed over the catalyst, with or
o-xylene; PX, p-xylene). without added hydrogen, is negligible in comparison
to the amount of alkane dehydrocyclization under the
whether the retained chloride influences the cycliza- same reaction conditions. Thus, secondary isomeriza-
tion pathway. McHenry and coworkers [36] found tion of the initially formed aromatics is not the source
that the dehydrocyclization activity of Pt-Cl-Al2 O3 of m-xylene.
catalysts was directly related to the amount of plat- m-Xylene is not allowed by direct C6 -cyclization
inum that was soluble in hydrofluoric acid. They felt of 3-methylheptane but is the only C8 -product from
that the soluble platinum was a chloroplatinum com- the cyclization of both 2- and 4-methylheptane. The
plex; however, later work showed that Pt was solu- aromatic distribution shown in Table 4 was calculated
ble only after exposure of the catalyst to air [37,38]. by assuming that m-xylene is a result of isomeriza-
In view of the later work, it does not seem likely tion of the reactant and should be neglected in the
that a dehydrocyclization–soluble Pt relationship was aromatic distribution; neglecting the m-xylene led
demonstrated in [36]. to identical aromatic compositions for a variety of
The conversion of 3-methylheptane at atmospheric metal and metal oxide catalysts. This suggests that
pressure demonstrates that the chlorine containing cyclization follows a similar pathway over both types
catalyst, even on nonacidic alumina, yields an iso- of catalysts; the chlorine merely provides an iso-
mer distribution that differs from one without chlo- merization pathway. Since 0.8 wt.% Cl, added to the
450 B.H. Davis / Catalysis Today 53 (1999) 443–516

Pt-Al2 O3 -K catalyst as NH4 Cl, did not significantly


change the isomer distribution, the chlorine added as
chloroplatinic acid is probably a specific type. Also
chromia, when the assumption is made concerning
the formation of m-xylene from 3-methylheptane,
yielded a C8 -aromatic distribution that is essentially
the same as all of the Pt catalysts (Table 4). Thus, it
is concluded that the metal oxide and metal catalysts
produce a similar aromatic product distribution.

3.3. Promoted Pt catalysts

Sulfur is a severe poison for Pt-Al2 O3 catalysts.


H2 S poisons our Pt-Al2 O3 catalyst even when it was Fig. 4. Conversion of n-octane over Pt- and Pt-Sn-Al2 O3 with the
same reaction conditions.
present in the ppm level. However, when 15 wt.% thio-
phene was added to the n-octane charge, the atmo-
spheric pressure dehydrocyclization activity actually Workers at Chevron found that a Pt-Al2 O3 catalyst
increased as long as there was incomplete conversion modified by rhenium had a higher activity and retained
of the thiophene. Thus, we have the surprising situ- the activity better than Pt-Al2 O3 [41]. Initially, it was
ation where H2 S, present from the partial conversion believed that the Pt-Re formed an alloy which was
of thiophene, does not act to poison the catalyst in more resistant to sintering and this accounted, at least
the presence of thiophene [40]. Furthermore, as shown partially, for the superiority of the Pt-Re catalyst. Prior
in Table 5, thiophene changed the Pt-Al2 O3 selectiv- to the discovery announcement by Chevron, we were
ity from equal amounts of o-xylene and ethylbenzene trying to modify Pt-Al2 O3 catalysts with metallic and
to one where the o-xylene : ethylbenzene ratio devi- non-metallic promoters. We were guided, or perhaps
ates significantly from 1 : 1 and may even approach misguided, by results from homogeneous catalysis to
the 2 : 1 ratio expected for ring closure based on the try the Pt-Sn combination. As can be deduced from
number of pathways leading to each isomer. It also Fig. 4, the Pt-Sn catalyst was more active as well as
appeared that a large quantity of alkene or alkyne much more resistant to aging than Pt alone. Other
(when the pure alkene without added hydrogen is the studies showed that there was a maximum conversion
reactant) effected a similar alteration of the aromatic for a Pt : Sn ratio of about 1 : 4 for a 0.6% Pt loading.
distribution. As can be seen in Table 5, the Pt-Sn catalyst had a
o-xylene/ethylbenzene selectivity of nearly 2 : 1.
Table 5 It was not effective to use the sodium or potas-
Product distribution from dehydrocyclization of n-octane over sium containing nonacidic alumina for the Pt-Re cata-
Pt-Al2 O3 (nonacidic) and promoted Pt-Al2 O3 after approximately lyst since the alkali metal was a severe poison for the
3 h on stream metal function. However, adding sec-butyl amine to
Catalyst C8 -Aromatic (mol%) the feed to the Pt-Re on acidic alumina catalyst poi-
Ethylbenzene p-xylene m-xylene o-xylene soned the support acidity and nearly eliminated the
conversion to aromatics that were not allowed by a
0.6% Pt 48.3 1.0 1.9 48.8
direct C6 -ring closure. As seen in Table 5, the 1.4 : 1
Pt-Sn 32.2 2.9 6.3 59.0
Pt-thiophene 43.0 Trace Trace 57.0 o-xylene : ethylbenzene ratio with Pt-Re is similar to
0.2% Pt 43.0 57.0 other promoted Pt catalysts.
Olefin 43.0 0.4 0.6 56.0 Overall, both gaseous and metallic promoters may
Alkyne 42.0 1.0 1.5 54.0 be used to increase the catalytic activity and to al-
Pt-Rea 40.2 1.2 1.5 57.2
ter the aromatic selectivity. The selectivity data for
a sec-butyl amine added to the charge. the conversion of n-octane, at later time on stream,
B.H. Davis / Catalysis Today 53 (1999) 443–516 451

Scheme 3.

Table 6
Conversion of n-nonane in the presence and absence of thiophenea (from [40])
Catalyst Time on Conversion to Thiophene in C9 -aromatics (mol%)
stream (min) aromatics (mol%) product (mol%)

PrBz Toluene Indan Indene


o-ethyl m-ethyl p-ethyl

Pt-Al2 O3 -K (no thiophene) 37 40 25.6 47.7 6.1 0.6 15.2 5.5


74 31 25.2 49.8 6.5 1.2 12.5 4.8
111 24 24.8 51.0 6.7 1.9 11.4 4.3
149 21 23.5 50.2 7.9 0.2 11.7 4.6
186 18 23.6 52.8 7.3 1.6 10.6 4.1
Pt-Al2 O3 -K (with thiophene) 39 45 7.2b 30.3 65.5 Trace Trace 4.2 –
79 31 8.2b 30.8 69.2 Trace Trace Trace Trace
123 28 8.6b – – – – – –
161 22 8.6b 30.1 68.5 Trace Trace Trace –
a 482◦ C, 3.5 cm3 of H2 /min; LHSV = 0.3.
b 15.1% thiophene in the n-nonane charge.

Table 7
Selectivity for bicyclic aromatic formation over Pt and moderated Pt catalysts for the first sample collected at about 50 mina (from [40])
Reactant Pt metal moderator Total aromatic (glc area %) Bicyclic aromatic (glc area %) Bicyclic/total cyclic
n-nonane None 42 21.0 0.5
Thiophene 45 2.0 0.04
Tin 66 1.2 0.02
n-decane None 23 15.0 0.4
Thiophene 45 4.3 0.1
Tin 72 4.8 0.07
a 482◦ C, atmospheric pressure; LHSV = 0.3.

are summarized in Table 5 and clearly demonstrate first sample collected; the data in Table 6 suggest that
that a variety of promoters alter the aromatic selectiv- the difference would have been greater at later time
ity of o-xylene : ethylbenzene = 1 : 1 for Pt-Al2 O3 to on stream. If Re has the same influence on the sec-
approach o-xylene : ethylbenzene = 2 : 1 for all of the ond cyclization activity as Sn and thiophene, the im-
promoted catalysts. proved catalyst life of the commercial Pt-Re catalyst
For n-nonane and higher carbon number alkanes, a may be due to a decrease in the rate of bicyclic (and
second cyclization to form a bicyclic indan or naph- polycyclic coke) ring compounds.
thalene structure is possible; e.g. Scheme 3. A most striking alteration of catalytic property by
As can be seen by the data in Tables 6 and 7, both promoters was shown in the work by Sinfelt [42].
tin and thiophene behave similarly and drastically de- Using a Ni-Cu alloy catalyst, it was shown that the
crease the extent of the second cyclization to produce specific activity for dehydrogenation of cyclohexane
a bicyclic aromatic. The data in Table 7 represent the remained nearly constant as the Ni : Cu alloy became
452 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 5. Data on the reforming 99–171◦ C boiling range naphtha showing the temperature required to produce 100 octane number product
as a function of time on stream for alumina-supported platinum, platinum-rhenium, and platinum-iridium catalysts at 14.6 atm pressure
(from [43]).

richer in Cu. However, the activity for the hydrogeno- a hydrogen flow needed to maintain activity during
lysis of ethane to methane decreased very rapidly naphtha conversion, almost all of the Te was lost in
as the Cu content of the alloy increased. Sinfelt a matter of hours or days of operation. This type of
has shown that similar results can be obtained with catalyst only has academic interest.
bimetallic mixtures whose solubilities do not allow The initial report [48] of the selectivity of Pt-KL ze-
alloy formation. These have been termed ‘cluster’ olite catalyst for converting n-hexane to benzene was
catalysts by Sinfelt and are presumably the basis of a extended and eventually developed to a commercial
commercial reforming catalyst [43]. catalyst by Chevron workers [9,10]. An early view of
Sinfelt and coworkers [43] developed a Pt-Ir alu- the reason for the selectivity of this catalyst was that
mina catalyst that was considered to exist as small highly dispersed Pt, atom or small cluster, selectively
bimetallic clusters. The authors presented extensive adsorbed n-hexane by a terminal carbon and that the
characterization data to indicate that the clusters are shape selectivity of the pore structure associated with
reasonably uniform in composition and have proper- the Pt allowed only direct 6-carbon ring formation
ties expected for uniform alloy catalysts. The incor- [49]. However, this model could not account for a sim-
poration of Ir modifies the catalytic properties of Pt, ilar selectivity for a catalyst made by the addition of
especially in decreasing the hydrogenolysis activity. Pt to a support of Al stabilized MgO [50].
Furthermore, these catalysts are claimed to have even
greater stability than the Pt-Re catalyst (Fig. 5).
A catalyst that exhibited early promise as a selec- 4. Structure of the catalyst
tive dehydrocyclization catalyst was formed by adding
tellurium to a nonacidic zeolite [44–47]. Thus, up to Chromia and molybdena, unsupported or when
10 wt.% Te could be added into the pore structure present on a support, exhibit a very complex structure.
of NaX zeolite merely by heating a physical mix- The pioneering work of Eischens and Selwood [51]
ture of the zeolite and elemental Te. Unfortunately, in showed very clearly that chromia was spread upon
B.H. Davis / Catalysis Today 53 (1999) 443–516 453

Fig. 6. (left) Effect of molybdena concentration on the activity of activated alumina F, surface area 91 m2 /g, for the conversion of heptane to
toluene (4, 450◦ C; 䊉, 497◦ C) (from [57]). (right) Catalytic activity for heptane dehycrocyclization as a function of chromium concentration
(from [51]).

the surface of the support at low loadings. However, et al. [58] indicate that for molybdena-alumina cata-
for a catalyst with 10% chromia, three-dimensional lysts that are reduced in hydrogen to temperatures of
chromia particles existed on the support surface. 750◦ C, the skeletal isomerization of hydrocarbons is
In general, three-dimensional chromia particles are essentially by an acidic mechanism. However, when
formed more easily on a silica support than on an the reduction is at 950◦ C, the skeletal isomerization
alumina support, and this difference becomes more behaves like the catalyst was unsupported Mo metal.
pronounced at chromia loadings below about 10 wt.% On the metal, isomerization occurs via a metalloacycle
[52,53]. Chromia-alumina catalysts show evidence for intermediate giving exclusively bond shift migration.
surface acidity, even when some alkali has been in- In summary, the surface of chromia and molybdena,
corporated (e.g., [54,55]). Voltz et al. [55] concluded unsupported and supported, is complex with a range
that the differences in acidity between the oxidized of metal valence states. Under conditions normally
and reduced states were one-half of the correspond- encountered in use as a dehydrocyclization catalyst,
ing differences between the amounts of Cr6+ present both chromia and molybdena show an acidic character.
on the surface. They also indicate that potassium Any mechanistic study with these catalysts must make
hydroxide decreases the acidity of both the oxidized sure that this acidity has not partially or completely
and reduced catalysts; the potassium increases the determined the aromatic product or the isotopic label
surface oxidation of the catalyst over one that does distribution.
not contain potassium. Metallic catalysts are in general less complex than
Molybdena–alumina catalysts have similar activi- metal oxide catalysts; however, metal alloys may meet
ties for n-heptane dehydrocyclization at 1 and 20 atm or exceed the complexity of the metal oxides. The
[56]. Chromia is more active than molybdena at 1 atm characterization of metal naphtha reforming catalysts
but is essentially inactive at 20 atm pressure. The de- has been reviewed recently and will not be repeated
hydrocyclization activity increases linearly with Mo here (e.g., [59–62]).
loading to about 14 wt.% molybdena, and then be- With few exceptions, platinum is considered to be
comes constant with further increases in Mo loading in the metallic state in Pt-alumina catalysts. In most of
(Fig. 6); this indicates that molybdena is present on these catalysts, Pt is very well dispersed so that essen-
the surface up to a monolayer loading before three tially all Pt atoms are present in an exposed position.
dimension molybdena particles begin to form. Holl Thus, the chemisorption of hydrogen occurs according
454 B.H. Davis / Catalysis Today 53 (1999) 443–516

to a stochiometry of H/Pt = 1. With usage this ratio mechanism was advanced to account for this [66].
will decrease and the Pt particles will be found in the Thus, an equilibrating reaction involving C5 - and
1.0–10 nm range. As indicated above, McHenry et al. C6 -ring compounds was established and the dehydro-
[36] believed they had evidence to support the view genation of the C6 -ring compounds continuously
that Pt was not present in the metallic state but was in caused the equilibrium to shift to produce six-carbon
an ionic form that could be extracted with dilute HF ring aromatic compounds.
or acetylacetone. More recently Botman et al. (e.g., Silvestri et al. [67] converted n-heptane over cat-
[63]) have presented data that they believe support the alysts with 100 ppm H2 S added to the hydrocar-
view that Pt is present in some positive valence state. bon/hydrogen feed stream. This level of H2 S poisoned
In spite of these dissenting views, most believe that the Pt function from converting n-heptane to aromat-
platinum is present in a metallic state. ics but still allowed for dehydrogenation to the extent
The situation is more complicated when two or more that the n-heptane/n-heptenes equilibrium was main-
metals are present. There was initially much debate tained. With either a Pt-carbon catalyst or a Pt-free
over the state of tin in platinum-tin-alumina catalysts. acidic alumina catalyst, essentially no aromatics were
However, it is now generally accepted that Pt-Sn alloys produced. However, when a physical mixture of these
are formed [59,64]. It is now generally accepted that two catalysts was utilized under the same reaction
some, if not all, of the Re is in close association with Pt conditions, a large increase in the formation of ring
in Pt-Re-alumina catalysts. Furthermore, it is usually products was observed. The authors considered this
necessary to add sulfur during the ‘start-up procedure’ to demonstrate that alkane cyclization reactions can
for Pt-Re catalysts, and the small amounts of sulfur proceed by a dual functional mechanism and that the
added during this period are generally considered to acid function can effect cyclization.
be bonded to the Re atoms, and not Pt. In the case of the bifunctional catalyst, isomeriza-
The Pt in Pt-KL or BaKL zeolites is usually present tion reactions eliminate the ability to utilize tracers,
in crystals that are not larger than the pore channels. either isotopic or alkyl groups, to understand the cy-
The Pt in these catalysts is especially prone to crys- clization process. For example, it has been shown that,
tal growth in the presence of even very low (ppb) with bifunctional catalysts, the conversion of n-octane
sulfur concentrations [59–62]. Because of the lack of under normal reforming conditions (100–400 psig;
interconnection of the long channels of the zeolite, 0.68–2.7 MPa) initially leads to the formation of
the larger platinum particles effectively block entry methylheptanes, dimethylhexanes and ethylhexanes
into the channels and thereby greatly decreases the [12–14]. Subsequently, the C8 -alkane isomer mixture
catalytic activity. is converted to aromatics. Furthermore, the conver-
sion of the C8 -alkane isomer mixture to C8 -aromatics
occurs more rapidly than n-octane is converted to aro-
matics using a non-acidic, monofunctional Pt-Al2 O3
5. Hydrocarbon conversions catalyst (Fig. 7). These results indicate that the acid
function of a bifunctional catalyst must be involved
5.1. Bifunctional catalysis in the cyclization step, presumably involving a car-
bocation formed from an alkene produced by alkane
An aspect of the conversion of alkanes that has dehydrogenation with the metallic function. While of
attracted attention is the role of acidity, when present interest from the commercial viewpoint, the bifunc-
[65]. The introduction of chlorided Pt-alumina cata- tional pathway will not be considered in detail in this
lysts in the late 1940s provided incentive for much discussion. Thus, the following sections emphasize
interest in the mechanism of alkane conversion and those studies that utilize monofunctional dehydrocy-
this was the basis for the development of theories clization catalysts. These monofunctional catalysts
to account for bifunctional catalysis. In the case of are of special interest for n-hexane dehydrocycliza-
naphtha reforming, the interconversion of C5 - and tion since the bifunctional pathway does not produce
C6 -rings permitted the naphthenes to be converted aromatics in a selective manner; the formation of the
to aromatics with high selectivity and a bifunctional five-carbon ring by the acid function must involve a
B.H. Davis / Catalysis Today 53 (1999) 443–516 455

Fig. 7. The total aromatics in the liquid product vs. n-octane conversion (including conversion to iso-octane isomers) for catalysts based
upon acidic (A) and non-acidic (NA) supports. (䊊) 0.6 wt.% Pt on precipitated alumina, COP; (䊐) 1.0 wt.% Pt on UCI alumina; (N)
1.0 wt.% Pt on silica, SIL; (䊏) 1.0 wt.% Pt on non-acidic alumina, NAA (from [14]).

Table 8
structure with a high-energy primary carbocation and
Relative rates of hydrogenolysis of various C–C bonds in n-hexane
this inhibits the rate of the bifunctional pathway. with Pt catalysts
Catalyst Relative rates Reference
5.2. Hydrogenolysis C1 –C2 C2 –C3 C3 –C4

Pt single crystal (1 0 0) 1.0 2.5 5.5 [72]


5.2.1. Hydrogenolysis of alkane Pt single crystal (1 1 1) 2.3 1.0 4.4 [72]
One of the alkane reactions is hydrogenolysis [68]. Pt-SiO2 1.2 1.0 2.9 [68]
It was established that the kinetics of hydrogenolysis Pt-SiO2 -Al2 O3 1.14 1.0 2.9 [69]
of ethane was consistent with the occurrence of dehy- Pt film (1 ␮g/cm2 ) 1.0 1.3 1.8 [70]
Pt-SiO2 (1.8 nm) 1.0 1.1 3.4 [71]
drogenation to produce a surface species with carbon
having three bonds to the metal surface. This dehy-
drogenation and multiple bonding to the surface was by data representative of hydrogenolysis of n-hexane,
sufficient to weaken the carbon–carbon bond, to al- it is apparent that at low and intermediate temper-
low it to rupture to produce two C1 fragments, and the atures (100–300◦ C) similar product distributions are
two fragments to be hydrogenated to produce methane obtained by a number of investigators using a variety
(e.g., [68–71]). With higher carbon number alkanes, of catalysts (Table 8).
a selectivity is introduced since there are multiple Davis et al. [72] report that the positional selectivity
carbon–carbon bonds. For n-hexane three types of for n-hexane hydrogenolysis showed a marked depen-
carbon–carbon bond positions are available. As shown dence on the surface structure of platinum. Surface
456 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 8. Dependence of the product distribution from hydrogenolysis of n-hexane on Pt particle size (data from [70]).

faces with a high concentration of (1 0 0) microfacets close agreement with Davis et al. [72] and Myers
[(1 0 0) and (13 1 1)] exhibit a high specificity for scis- and Munns [69], Leclercq et al. found the relative
sion of internal C–C bonds. In contrast, those surfaces rates of C–C bond hydrogenolysis in n-heptane to be
composed mostly of (1 1 1) microfacets ((1 1 1), (3 3 2) C1 –C2 : C2 –C3 : C3 –C4 = 2.0 : 1.0 : 2.3.
and (10 8 7)) showed a clear preference for central and As shown by Carter et al. [11], metals like Rh
terminal C–C bond scission. The results reported by have a reaction rate for n-heptane hydrogenoly-
Davis et al. [72] differ from those who report a nearly sis that is about 100 times greater than platinum
statistical bond hydrogenolysis of n-hexane [73,74]. (Fig. 9) [79]. These reactions were conducted at low
In agreement with the results of Myers and Munns temperatures (88–300◦ C) and pressure (1 atm) with
[69], van Schaik [75], Cinneide and Gault [76], An- H2 /n–C7 H16 = 5. Under these conditions, hydrogenol-
derson and Shimoyama [70] find a preference for the ysis competes effectively with isomerization, and both
hydrogenolysis of the central C–C bond. Furthermore, reactions exceed that of dehydrocyclization (Table 9).
for conversion at 273◦ C, 100 Torr, H2 /h.c. = 10/1 in Considering all data for n-hexane hydrogenoly-
a recirculating apparatus, the position of C–C bond sis presents a picture of constrasts. First, nickel and
attack did not appear to be dependent upon the Pt rhodium show a marked difference from Pt since the
particle size, except perhaps for very large particles two former metals show a dominance of hydrogeno-
(Fig. 8). For the largest particles, terminal C–C bond lysis of the terminal C–C bond. This is apparently
hydrogenolysis may be favored but, as shown in Fig. 8, due to Ni and Rh adsorbing the carbon fragment
the central C–C bond cleavage is favored for particle so strongly that, rather than forming methane and
sizes in the range normally encountered in naphtha re- n-pentane from n-hexane, the dominant pathway is for
forming catalysts. The data in Fig. 8 show a relative the C5 -fragment to remain absorbed and for the ter-
selectivity of C1 –C2 : C2 –C3 : C3 –C4 = 1 : 1.3 : 1.8. minal C–C bond cleavage to continue until n-hexane
Leclercq et al. [77] converted 16 different alka- has been converted to six methane molecules. In ad-
nes with a variety of branching with a Pt-alumina dition, those metals that produce primarily methane
catalyst. Ignoring those reactants with a quaternary are more active for hydrogenolysis than those with
carbon, the authors favored a 1,1,3-adsorbed species statistical or center bond hydrogenolysis.
such as postulated by Anderson [78] although all In summary, Pt, among Group VIIIB metals, is
of their results could not be reconciled by a sin- relatively inactive for hydrogenolysis reactions. How-
gle mechanism. Rather, Leclercq et al. considered ever, at low temperature and pressure conditions,
their data to support the view of the participation hydrogenolysis makes a major contribution to the
of 1,2-, 1,3-, 1,4-, and 1,5-diadsorbed species. In total conversion. At the temperatures and pressures
B.H. Davis / Catalysis Today 53 (1999) 443–516 457

Fig. 9. Catalytic activities of group VIII noble metals for n-heptane hydrogenolysis. The activities are compared at a temperature of 205◦ C
at 1 atm pressure and a H2 /nC7 mole ratio of 5/1 (from [79]).

Table 9
Reactions of n-heptane on metals in the presence of hydrogena (from [43])
Metal Temperature (◦ C) Percent total conversion Percent hydrogenolysis Percent isomerization Percent dehydrocyclization
Pd 300 6.4 5.8 0.4 0.2
Rh 113 2.9 2.7 0.2 –
Ru 88 4.0 3.7 0.3 –
Pt 275 2.3 0.6 0.7 1.0
275 9.4 3.4 3.8 2.2
275 21.4 8.0 10.0 3.4
Ir 125 1.5 1.3 0.2 –
a Conditions: 1 atm, H2 /nC7 mole ratio = 5.

normally encountered in commercial reforming, hy- platinum catalysts, the data agree with the nonselec-
drogenolysis makes a less significant contribution to tive reaction pathway. Gault [84] concludes, however,
the total conversion [80]. that when all data are considered, more than two ring
opening pathways are needed to describe the data.
The use of metal films has become a popular ap-
5.2.2. Hydrogenolysis of alkylcyclopentanes proach to evaluate the activity and selectivity of a
The study of hydrogenolysis of alkylcyclopentanes metal catalyst while eliminating, or at least mini-
predates that of dehydrocyclization [81] and these mizing, the impact of support effects. By controlling
studies were continued by one of the discovers of the parameters of preparation, metal films can be
dehydrocyclization [82,83]. Gault, in his authoritative deposited so that they produce a narrow range of
review, [84] identified two extremes for hydrogenoly- particle sizes and, at the same time, the average par-
sis of methylcyclopentane as nonselective (statistical) ticle size can vary over a wide range. Unfortunately,
and selective, as outlined in Scheme 4. this approach is limited to temperatures considerably
An examination of the data collected in Table 10 below those where dehydrocyclization occurs at a
show a variety of selectivities. With highly dispersed reasonable rate; at dehydrocyclization temperatures
458 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 4.

Table 10
Products from hydrogenolysis of methylcyclopentane
Catalyst Reduction temperature (◦ C) T (◦ C) PH2 H2 /MCP Products (mol%) Reference
2MPa 3MPa Ha

PtKL 300 270 0.89 8 32(1.1)b 29(1.0)c 29 [88]


300 270 0.98 40 33(1.0) 33(1.0) 33 [88]
PtKL 450 270 0.98 40 39(0.13) 30(1.0) 30 [88]
Pt-SiO2 450 270 0.98 40 51.8(3.1) 16.6(0.54) 30.9 [88]
Pt-Al2 O3 450 270 0.96 22 66(2.9) 23(2.5) 9.3 [88]
EUROPAT1(0.17 nm) 227 227 0.99 316 39(2.0) 20(0.49) 41 [89]
247 247 0.99 316 38(2.1) 18(0.41) 44 [89]
297 297 0.99 316 38(1.5) 26(0.72) 36 [89]
Pt-SiO2 (2.2 nm) 227 227 0.99 316 46(2.3) 20(0.59) 34 [89]
247 247 0.99 316 43(2.4) 18(0.49) 39 [89]
297 297 0.99 316 39(1.5) 26(0.74) 35 [89]
Pt-Al2 O3 (2.3 nm) 227 227 0.99 316 50(2.5) 20(0.67) 30 [89]
Pt(0.48)Re(1.8)Al2 O3 550 300 0.90 9.5 54(1.5) 37(4.2) 8.8 [90]
Pt-SiO2 (1.8 nm) 300 277 0.95 18 40(2.1) 19(0.50) 38 [71]
PdNaY 350 275 0.94 16 45(1.7) 27(1.6) 17 [91]
Pt-SiO2 210–260 1 atm total – 56(1.4) 42(25) 1.7
Pt-Al2 O3 210–260 50(1.3) 40(4.0) 10 [92]
PtC 210–260 53(1.3) 42(8.4) 5 [92]
Pt-SiO2 300 250 0.95 20 42.1(2.7) 15.6(0.38) 41.3 [85]
PtNaY 300 250 0.95 20 52.9(2.1) 25.6(1.3) 19.9 [85]
Pt-SiO2 485 6.8 2 36.8(1.5) 24.1(0.62) 39.1 [86]
Pt(0.2%)-Al2 O3 250–310 ∼1 high 41(2.0) 21(0.55) 38 [87]
a 2MP = 2-methylheptane; 3MP = 3-methylpentane; H = n-hexane.
b Numbers in parentheses are 2-MP/3-MP.
c Numbers in parentheses are 3-MP/H.

sintering of the film occurs so that only very large Gault [94] converted methylcyclopentane with a se-
metal particles are present. The results from two stud- ries of Pt-alumina catalysts wherein the Pt particle size
ies are compared in Fig. 10 [73,93]. While Kramer was controlled by varying the Pt content from 0.15
and Fischbacker [93] obtain less n-hexane with the to 20 wt.%. The runs were made in a batch reactor
larger Pt particles than Anderson and Shimoyama at 315◦ C and 30 atm (3.03 MPa). The ratios of bond
[73] do, the results for the highly dispersed Pt cat- breaking (R␤/␥ and R␥/␣ from Scheme 4) are plotted
alysts are in excellent agreement and are consistent versus increasing platinum content (increasing plat-
with a nonselective hydrogenolysis reaction pathway. inum particle size) in Fig. 11. For the lower loading,
B.H. Davis / Catalysis Today 53 (1999) 443–516 459

Fig. 10. Selectivity for conversion of methylcyclopentane (left from data in [70]; right from data in [93]).

Fig. 11. Dependence of values of R1 (R␤/␥ ) and R2 (R␥/␣ ) on Pt content of catalyst (from [94]).

lower Pt particle size catalyst the ratio R␤/␥ is nearly Table 11


constant at 2.5, slightly higher than the value of 2.0 for Platinum-catalyzed ring openinga (from [86])
pure nonselective bond rupture. The other selectivity H2 partial pressure (atm) 6.8 10.2 13.6 17.0
strongly depends on Pt particle size. Ring-opening rate (␮mol/s g) 9.8 14.0 17.6 19.9
For the hydrogenolysis of n-alkanes, the rate of Hexane composition (mol%)
C–C bond rupture normally has an inverse depen- 2-methylpentane 36.8 39.6 41.0 39.6
dence on hydrogen pressure. However, the data 3-methylpentane 24.1 20.9 22.0 22.3
n-hexane 39.1 39.5 37.0 38.1
of Brandenberger et al. [86] show that the rate of
ring-opening hydrogenolysis of methylcyclopentane a 0.71 wt.% Pt/SiO2 , 485◦ C, 3.4 atm MCP.
has a positive, rather than negative, dependence on
hydrogen partial pressure (Table 11). These results carbon that is multiply bonded to a surface Pt atom.
imply that the rate limiting step does not involve de- Equally important, it is evident that hydrogenolysis is
hydrogenation of methylcyclopentane to produce a nonselective and that the selectivity for ring-opening
460 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 5.

Table 12 Burnett [97]. These authors utilized a 2 wt.%


Products from the conversion of 14 C-labeled 1,3-dimethylcyclo- Pt-SiO2 catalyst and operated at 482◦ C and 18 atm
pentane at 482◦ C and 18 atm (from [97])
(1.8 MPa) with a hydrogen/hydrocarbon ratio of 6.
Compounds Mol% They converted a mixture that contained 2 vol%
Methane + ethane 8.5 of 14 C-labeled 1,3-dimethylcyclopentane present
Propane 3.6 in a 4 : 1 n-heptane : m-xylene mixture. The 14 C
i + n-butane 3.9 content of the products permitted them to iden-
i + n-pentane 1.3 tify those that were derived from the conversion of
Methylcyclopentane 0.5
2-methylpentane 3.7
1,3-dimethylcyclopentane (Table 12). The isopentane
3-methylpentane 1.5 products agree with those calculated for a nonselec-
n-hexane 2.4 tive mechanism (Scheme 5).
2,4-dimethylpentane 10.3 Toluene accounts for only about 14% of the
2-methylhexane 22.8 toluene plus isoheptane products. Because the 2- and
3-methylhexane 22.7
n-heptane 0.0
3-methylhexane hydrogenolysis products can undergo
Toluene 8.8 cyclization to produce toluene, the ring expansion
pathway must be smaller than represented by 14%
toluene. The products from the ring opening reaction
does not depend to a measurable extent upon hy- are in excellent agreement with those expected for a
drogen partial pressure. In contrast, Levitskii and non-selective hydrogenolysis reaction and with those
Minachev [95] reported a selectivity that depends of Brandenberger et al. [86].
upon hydrogen partial pressure for a Pt-SiO2 catalyst Hardy and Davis [98] converted n-propylcyclo-
(n-hexane : 2-methylheptane : 3-methylheptane = 1.0 : pentane and 1-methyl-2-ethylcyclopentane with a
4.2 : 1.5 at 10 atm H2 and 1.0 : 1.4 : 0.8 at 50 atm). Pt-nonacidic alumina catalyst at 482◦ C and about
Lester [96] obtained data that suggest that hydro- 14 bar. The products from both C8 -isomers fit very
carbons which contained alkyl substitution and with well the products predicted for a non-selective hy-
the longest chain having 5-carbons could cyclize on drogenolysis mechanism (Scheme 6).
supported Pt-catalysts to form a cyclopentane struc- Studies were also conducted where each of the
ture which then underwent intermolecular ring expan- two alkyl- or dialkyl-cyclopentane isomers were con-
sion to produce a six-carbon ring intermediate which verted when these compounds were present at about
subsequently dehydrogenated to produce the aromat- 10–15% in n-heptane. The amount of C8 -aromatics
ics. To date, additional reports with the high selecti- and C8 -isoalkanes could be determined and the results
vity for aromatics that was obtained by Lester have showed that only 7–8% of the conversion products
not appeared. from the alkylcyclopentanes were C8 -aromatics. The
That the 5-carbon ring does not directly ex- data from these runs are compared in Table 13 to
pand to a six-carbon ring on a monofunctional the data of Csicsery and Burnet [97] as well as the
Pt catalyst operated at reforming conditions has data obtained by Huang et al. [99] and all show that
been conclusively demonstrated by Csicsery and less than 20% of the alkylcyclopentane underwent
B.H. Davis / Catalysis Today 53 (1999) 443–516 461

Scheme 6.

Table 13
Percentage of aromatics produced in the conversion of alkylcyclopentanes
Reactant Reaction conditions (Aromatics/(aromatics + Reference
hydrogenolysis products)) × 100

n-heptane labeled (2.5 vol% 1,3 482◦ C, 18 atm, 6 : 1 H2 : hydrocarbon, 13.6 [97]
[14 C]dimethyl-cyclopentane) 2% Pt-SiO2
n-heptane (15% n-propylcyclopentane) 482◦ C, 13 atm, 9 : 1 H2 : hydrocarbon, 7.1 [98]
0.6% Pt on nonacidic alumina
n-heptane (8% 1-methyl-2-ethylcyclopentane) 482◦ C, 13 atm, 9 : 1 H2 : hydrocarbon, 8.3 [98]
0.6% Pt on nonacidic alumina
n-octane (2.4% n-[14 C]propylcyclo-pentane) 482◦ C, 13 atm, 7 : 1 H2 : hydrocarbon, 10–18 [99]
2% Pt-KL

ring expansion to produce aromatics. Since some of n-hexane increases with increasing particle size; this
the iso-octanes formed by hydrogenolysis can cy- is in contrast with the trend obtained by most work-
clize by direct six-carbon ring formation to produce ers with increasing Pt particle size. As the amount of
C8 -aromatics, the actual percentage will be less than Re added to a constant loading of Pt increases, the
shown in Table 13. Thus, under reforming conditions methylcyclopentane conversion first decreases, then
the dominant reaction pathway was hydrogenolysis undergoes an increase to about five times that of Pt
of the cyclopentane ring to produce iso-octanes and alone, and appears to reach a constant activity above
not ring expansion. this level (Fig. 13). The addition of Re to Pt causes the
Yamada et al. [90] investigated the impact of the selectivity to change so that the amount of n-hexane
reduction temperature upon a series of Pt-Re-alumina decreases with a corresponding increase in both 2-
catalysts that were then used for the hydrogenolysis of and 3-methylpentane. The cyclopentane content is the
methylcyclopentane at 300◦ C and 1 atm total pressure. same for Pt, Re and the Pt-Re mixtures (Fig. 14)
For the catalyst that was reduced at a low temperature, The selectivity for the hydrogenolysis of methyl-
the three allowed hexane isomers are present in a ra- cyclopentane appears more complex for Pt-zeolite
tio that is similar to that of a nonselective mechanism. catalysts. The pore geometry of the zeolite structure
In addition, slightly more than 10% of the products limits the size of the Pt particle to about 1.0 nm; thus,
result from hydrogenolysis to produce methane and these catalysts should resemble those of highly dis-
cyclopentane (Fig. 12). As the reduction temperature persed Pt catalysts. However, Sachtler and coworkers
increases (and Pt particle size increases), the prod- reported that this was not the case (e.g., [85,100,101])
uct distribution changes. Surprisingly, the amount of for Pt-NaY zeolites. From the viewpoint of activity,
462 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 14. Effect of Re/(Pt + Re) content on selectivity (Cata-


lyst: Pt (0.48)-Re(x), reduction temp.: 550◦ C, W/F: 2.4 g h/mol,
H2 /methylcyclopentane: 9.5 mol/mol; 䊊: 2-methylpentane, 䊉,
3-methylpentane, 䊎: n-hexane, : cyclopentane) (from [90]).

Fig. 12. Effect of reduction temperature on selectiv-


ity (W/F: 2.4g h/mol, H2 /methylcyclopentane: 9.5 mol/mol 䊊:
2-methylpentane, 䊉: 3-methylpentane, 䊎: n-hexane, : cyclopen- vicinity of the alkali ions [103]. Alvarez and Resasco
tane (from [90]). [104] found that the enhanced activity was more pro-
nounced for higher Pt containing catalysts and con-
cluded that, since the Pt crystals must have the same
maximum size in the low and higher Pt containing
catalyst, the number of Pt interacting with alkali ions
must be larger in the higher Pt loaded catalyst. The Pt
loading effect was therefore considered to favor local
interactions as the cause for the augmentation of the
hydrogenolysis activity.
The difference in Pt supported on silica and KL
zeolite is illustrated in Fig. 15. While the Pt-silica
catalyst gives approximately a nonselective prod-
uct distribution, with the Pt-KL zeolite catalyst the
amount of 3-methylpentane is increased and the
amount of both 2-methylpentane and n-hexane is de-
creased. These authors compare data for their low and
higher Pt-loaded KL zeolite catalysts and for their
Pt-CabOSil together with three sets of selectivity
Fig. 13. Effect of Re content on activity (reduction temp.: data taken from the literature ((a) [88]; (b) [105]; (c)
550◦ C, W/F:2.4 g h/mol, H2 /methylcyclopentane: 9.5 mol/mol, 䊊: [106]) as a function of dispersion. The data for the
Pt (0.48)-Re(x) bimetallic catalyst, 两: Re(x) catalyst) (from [90]). Pt-CabOSil are similar to those for the three curves;
however, the data for the two zeolite catalysts are
very different. The 3-/2-methylpentane ratio is larger
the basicity of the cation appears to play a role. Long for the zeolite catalysts, and for the higher Pt loading,
range electron transfer from the alkali to the metal the 3-isomer is formed in larger amounts than the
has been proposed (e.g., [102]) while others con- 2-isomer. This is the result that Gault would attribute
sider the effect to be due to local interactions in the to selective hydrogenolysis that is associated with
B.H. Davis / Catalysis Today 53 (1999) 443–516 463

Fig. 15. 3-Methylpentane/2-methylpentane (left) and 3-methylpentane/n-hexane ratio dependence on platinum dispersion (H/M) for the
conversion of methylcyclopentane with a Pt-KL zeolite catalysts (a, from [88]; b, from [102]; c, from [103]; 䊏, Pt-CabOSil; N, low
Pt-loaded KL zeolite; 䉬, high-Pt KL zeolite) from [104]).

the latter approaching that obtained for silica and


alumina supported Pt. The situation is similar with
the 3-methylpentane/n-hexane ratio where the zeolite
samples have selectivities anticipated for much larger
Pt particles.
Moretti and Sachtler [100], because methylcy-
clopentane has a length/width ratio of about 1.7,
suggest that methylcyclopentane will align to diffuse
so that the length of the molecule is aligned with the
pore direction. If the 1.0 nm Pt particle fills the pore
cavity, exposing the metal through the cage window,
one end of the methylcyclo-pentane will contact the Pt
Fig. 16. 3-Methylpentane/2-methylpentane ((䊉) 3MP/2MP ratio) particle. In one orientation, this alignment will cause
product ratio (left axis) and H/Pt (䉬) chemisorption capacities
the C–C bond that leads to 3-methylpentane to be
(right axis) as a function of calcination temperature on the 0.6%
Pt/KLTL catalyst. (Reduction temperature of 350◦ C and methyl- attacked. However, the other alignment will cause the
cyclopentane ring opening reaction temperature of 340◦ C) (from closest approach to favor the C–C bond that produces
[107]). n-hexane to be attacked, and the data are not consis-
tent with this. Moretti and Sachtler [100] introduce
an additional geometrical effect. In one approach the
large platinum crystals; however, Alverez and Resasco methyl group is pointed to the metal and in the other
[107] show that this is not the case. They determined the tertiary hydrogen will contact the metal; since
the dispersion (H/Pt ratio) and the 3-/2-methylpentane methylcyclopentane is restricted from turning over in
selectivity for increasing calcination temperatures, the pore; only the orientation with the tertiary hydro-
followed by reduction at 350◦ C, and found that the gen approaching the metal will lead to hydrogenolysis
ratio remained high and constant up to a calcination to produce n-hexane. Thus, only about half of the
temperature greater than 377◦ C (Fig. 16). Only when methylcyclopentanes that approach the Pt crystal with
the sample was heated above 377◦ C did the disper- the methyl end of the molecule will be in the proper
sion and the 3-/2-methylpentane ratio decrease, with orientation to react, and this will cause a decrease in
464 B.H. Davis / Catalysis Today 53 (1999) 443–516

Table 14 Table 15
Comparison of products obtained from the conversion of n-octane C8 -Alkane distribution calculated for nonselective hydrogenolysis
with Pt-BaKL and Pt-KL catalysts (from [102]) of all C–C bonds in n-propylcyclopentane and the normalized
experimental distribution (from [99])
References Reference
[9,10] [99] Experimentala Calculatedb
Catalyst Pt-BaKL-866 Pt-KL a. 3-ethylhexane 51.2c 37.5
Pressure 6.9 bar 13.8 bar b. 2-, 3-, and 4-methylheptane 51.2c 37.5
Temperature 733 K 725 K c. n-octane 28.3 25.0
LHSV 6 ca. 1 d. cyclopentane 10.7 12.5
Conversion (%) 30 27 e. methylcyclopentane 6.12 12.5
C8 aromatics f. ethylcyclopentane 3.65 12.5
Ethylbenzene + ortho-xylene 88 93.5 a Average of four runs.
Meta- + paraxylene 12 6.5 b Calculated assuming statistical C–C bond hydrogenolysis.
Methylheptanes/C8 aromatic 0.19–0.28a 0.52 c These products could not be separated using the GC column
3-ethylhexane/methylheptanes 0–0.51a 0.15
needed for 14 C counting.
2-methylheptane (8)b 18c(21)b,c
3-methylheptane (34)b 26(28)b
4-methylheptane (58)b 41(51)b
3-ethylhexane 15 Considering all of the data for methylcyclopentane
a Range of values because of overlapping of 3-ethylhexane and hydrogenolysis, it appears the conversion is essentially
4-methylheptane. nonselective for platinum that is highly dispersed on
b Percentage based upon methylheptanes.
non-zeolitic supports. For the non-zeolite catalysts that
c 2-methylheptane peak corrected for coelution of toluene.
are effective as dehydrocyclization catalysts, it is con-
cluded that any contribution of C5 -ring formation will
the hydrogenolysis rate to produce n-hexane relative lead to isomerization through a nonselective mecha-
to that of 3-methylpentane. nism for hydrogenolysis of cyclopentane isomers. It
Huang et al. [99] converted n-octane that contained is puzzling that only a few authors report hydrogeno-
3 wt.% n-propyl-[1-ring-14 C]-cyclopentane with a lysis of methylcyclopentane to produce methane plus
Pt-KL zeolite catalyst at 482◦ C and about 14 bar. The cyclopentane. For Pt-zeolite catalysts hydrogeneolysis
preparation, activation and use of these Pt-zeolite of methylcyclopentane is selective but for n-propyl cy-
catalysts is difficult since many variables must be clopentane it is nonselective. There is insufficient data
controlled. Hughes et al. [9,10] reported results for to define the overall selectivity character of Pt-zeolite
a catalyst that had a long (months) life. The data catalysts.
Huang et al. [99] obtained are compared in Table 14
with those reported by Hughes et al. [9,10]; it is seen 5.2.3. Hydrogenolysis of alkylcyclohexanes
that the products at about the same conversion levels The hydrogenolysis of cyclohexane or alkylcy-
are nearly identical, both with respect to the aro- clohexane does not make a significant contribution
matic and the iso-octane isomers. Unlike much of the during their conversion under normal reforming pro-
data obtained for methylcyclopentane hydrogenoly- cesses. The results obtained for the conversion of
sis, Huang et al. found that the hydrogenolysis of 14 C-labeled methylcyclohexane show that more than

n-propylcyclopentane occurred for both the n-propyl 95% of the methylcyclohexane was converted to
side chain and for the cyclopentane ring. Since the toluene with either Pt-nonacidic alumina or Pt-acidic
14 C-label was in the ring, the identification of isotopi- alumina at 482◦ C and 400 psig [108]. Demethylation
cally labeled ethyl- and methyl-cyclopentane prod- by hydrogenolysis and/or dealkylation/transalkylation
ucts clearly show that it originated from the labeled was the most prominent side reaction, even with the
n-propylcyclopentane reactant. The experimental data Pt-acidic alumina catalyst. Cracking to C6 - products
are in reasonable agreement with the distribution ex- made a small contribution to the minor amounts of
pected for non-selective hydrogenolysis of the eight side reactions.
C–C bonds located both in the n-propyl group and Dermietzel [109] showed that methyl-labeled
the cyclopentane ring (Table 15). toluene, added to a light naphtha, does not undergo
B.H. Davis / Catalysis Today 53 (1999) 443–516 465

significant conversion (less than 3 wt.%) during re- catalysts. Thus, with Pt-nonacidic catalyst there was
forming at 520◦ C and pressures in the 0.5–2.0 MPa no need to invoke the complex bicyclic mechanisms
range. that evolved to explain the aromatic isomers produced
Iglesia et al. [110] converted n-heptane containing by dehydrocyclization of alkanes.
a small fraction of 13 C-methyl labeled toluene over a
Te/NaX zeolite catalyst at 450◦ C, 4.5 kPa hydrocar-
bons and 96 kPa hydrogen in a recirculating reactor. 6.2. Comparison of the conversion of an alkane and
The added toluene, when 26% of the n-heptane had its corresponding naphthene
been converted, still contained 99.5% of the 13 C in the
methyl position and 0.5% in the ortho-ring position. Shown in Scheme 7 are some C8 alkane iso-
It is therefore clear that one does not need to con- mers that would form 1,1,3-trimethylcyclopentane
sider secondary reactions such as hydrogenolysis of (5) by a C5 -ring closure. Of the four shown,
alkylcyclohexanes or of the aromatic products that only 2,4-dimethylhexane (4) provides a path-
are formed during dehydrocyclization with monofunc- way that could lead to any cyclopentane isomer
tional Pt catalysts. other than 1,1,3-trimethylcyclopentane. The data
in Table 16 for the chromia catalyst make it ap-
parent that 1, 2, and 3 do not yield the same C8
aromatic distribution. In addition, they do not re-
6. Ring formation: C5 - and C6 -ring forming semble the C8 aromatic distribution obtained from
mechanisms the conversion of 1,1,3-trimethylcyclopentane. The
amount of ethylbenzene and o-xylene relative to
6.1. General m-xylene which Pines and Chen [117] obtained
from 1-methyl-2-ethylcyclopentane eliminate this
One of the features that has received much atten- ring structure from making a major contribution to
tion is the size of the ring initially formed during the the reaction pathway. 1,2,3-trimethylcyclopentane
dehydrocyclization process. As noted above, much of can also be eliminated as a possible intermediate
the early work that required additional mechanisms because only o-xylene and p-xylene can be formed
in addition to a direct six-carbon ring was conducted from it by a simple ring expansion pathway. The only
with metal oxide catalysts. C5 -cyclization pathway leading to m-xylene is through
The dehydrocyclization of many alkanes, each con- 1,1,3-trimethylcyclo-pentane followed by ring expan-
taining eight or more carbons and at least one chain sion route d in compound 5 . For 2,5-dimethylhexane,
with six or more carbons, with a Pt-nonacidic alu- the trace amount of m-xylene excludes ring expansion
mina catalyst produced 90% or more of the aromatics d so that the ring expansion must follow path a in or-
predicted for the operation of only a six-carbon ring der to produce the p-xylene product. Thus, cyclization
forming mechanism [111]. Fogelberg et al. [112,113] of both 2,4- and 2,5-dimethylhexane through the com-
utilized a commercial Pt-alumina catalyst and a mon C5 ring compound, 1,1,3-trimethyl-cyclopentane,
Pt-catalyst made ‘nonacidic’ by calcining the alu- would mutually exclude all of the ring expansions
mina support at a high temperature to produce a low that lead to C8 aromatics. A direct C5 -cyclization for
surface area for the dehydrocyclization of n-octane 2,2,4-trimethylpentane followed by ring expansion to a
as well as C8 -hexane and heptane isomers. With the six-carbon ring requires expansion a but not d; further-
commercial catalyst they obtained a nearly equili- more, a would have to be at least 5 times larger than
brium C8 -aromatic composition; however, with the c + d to yield the experimental aromatic distribution.
nonacidic catalyst the aromatics were predominantly Hence, the data in Table 16 exclude C5 ring closure
(> 90%) those expected for a direct six-carbon ring for both 2,4- and 2,5-dimethylhexane. Dautzenberg
formation. and Platteeuw [118] were, to our knowledge, the first
With these nonacidic Pt-catalysts the dehydrocy- to use this reasoning to eliminate a common cyclopen-
clization of C8 –C10 alkanes produces simpler results tane ring structure being formed from several alkane
than earlier workers had obtained with the metal oxide isomers. On the other hand, the data in Table 16,
466 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 7.

Table 16
Aromatic distribution from the conversion of C8 alkanes and the corresponding naphthenes over chromia–alumina catalysts
Reactant Aromatic selectivitya Aromatic (mol%) Reference
Bz Tol EB OX MX PX

1 (2,2-DMH)b 12 13 50 8 (21)c 11 (31) 7 (18) 12 (31) [115]


– – 21 2 (2) 37 (47) 33 (42) 7 (9) [34]
2 (2,5-DMH)b 35 – 14.7 Trace Trace 2.6 (3) 82.7 (97) [115]
3 (2,2,4-TMP)b 21 – 17.6 5.3 (6.4) 77 (94) [117]
9 – 7.2 – – – 93 (100) [116]
4 (2,4-DMH)b 26 11 58 – – 27 (87) 4 (13) [115]
5 (1,1,3-TMCP)b – 6.7 – 6.7 (7) 60 (64) 27 (29) [112]
a Total conversion to aromatic/total conversion.
b DMH =dimethylhexane; TMP = trimethylpentane.
c Numbers in parenthesis are for the mol% based on C aromatics.
8

except for 2,2,4-trimethylpentane, are consistent with intermediate for dehydrocyclization over Pt as well as
direct C6 ring formation. The aromatic distribution over chromia. Again, the aromatic product data are
from 2,2,4-trimethylpentane is essentially the same consistent with direct C6 -ring formation.
as that from 2,5-dimethylhexane; it has been sug- The aromatic distributions in Table 18 (taken from
gested that a portion of the large amount of isobutenes [34]) are from nine carbon cyclohexanes that con-
formed by cracking of 2,2,4-trimethylpentane subse- tain a quaternary carbon and for the alkanes that can
quently dimerize to form 2,5-dimethylhexane which directly form these alkylcyclohexanes by a C6 -ring
then undergoes dehydrocyclization by C6 -ring closure closure. For chromia, the reaction pathway for the
to form p-xylene. The data in Table 16 are consistent three trimethylcyclohexanes is clear: it is demethy-
with this view. lation to produce the corresponding xylene, and for
Applying the above logic to the data in Tables 16 the small amount of C9 aromatics with the predomi-
and 17 allows us to eliminate a common cyclopentane nate isomer being the one expected by a 1,2-methyl
B.H. Davis / Catalysis Today 53 (1999) 443–516 467

Table 17
Aromatic distribution from the conversion of C8 alkanes and the corresponding naphthenes over Pt on nonacidic alumina catalysts (from
[34])
Reactant Aromatic selectivitya Aromatic (mol%) Reference
Tol EB OX MX PX

1 (2,2-DMH)b 52 – 22 (47)c 18 (39) 8 (16) [96]


2 (2,5-DMH)b – Trace Trace 7 93 [111]
0.6 1.6 11 87 [118]
3 (2,2,4-TMP)b 36 42 0 0 35 (60) 23 (40) [96]
1.6 7 50 42 [118]
4 (2,4-DMH)b 6 2 90 2 [111]
5 (1,1,3-TMCP)b 35 30 0 1.5 (2.1) 42 (60) 27 (38) [96]
1.8 6.4 56 36 [118]
6 (1,1-DMCH)b 60 1 (2) 10 (24) 30 (71) 1 (2) [111]
53 (2) (60) (28) (10) [118]
a Total conversion to aromatic/total conversion.
b DMH = dimethylhexane; TMP = trimethylpentane.
c Numbers in parenthesis are for the mol% based on C aromatics.
8

Table 18
Aromatic distribution from C9 alkanes and the corresponding trimethylcyclohexanes (data from [34])
Reactant Xylenesa Trimethylbenzenesa
o- m- p- 1,2,3- 1,2,4- 1,3,5-

Cr2 O3
1,1,2-TMCHb 85 – – 14 – –
1,1,3-TMCHb 1 97 – – 2 –
2,2,4-TMHb 8 27 12 7 33 1
2,6-DMHEPTb 3 70 – 4 12 4
1,1,4-TMCHb – 2 90 – 8 –
2,2,5-TMHb – 16 16 2 54 00
Pt(1)
1,1,3-TMCHb – 73 – – 10 8
2,6-DMHEPTb,c 2 49 10 18 15 2
Pt(2) or Pt(3)20
1,1,2-TMCHb 20 2 – 57 18 –
1,1,3-TMCHb – 12 – 15 60 –
2,2,4-TMHb 7 13 3 – 32 18
2,6-DMHEPTb 2 12 – 8 38 18
1,1,4-TMCHb 1 2 1- 2 85 –
2,2,5-TMHb – 6 3 8 54 2
a Total may not be 100% because of aromatics other than listed above.
b TMCH = trimethylcyclohexane; TMH = trimethylhexane; 2,6-DMHEPT = 2,6-dimethylhept-3-ene.
c Same distribution was obtained with and without added hydrogen.

migration from the geminal-dimethyl position. Note for 2,2,4-trimethylhexane would lead to the same inter-
that C6 -ring closure for 2,6-dimethylyheptene forms mediate as was formed from 2,6-dimethylheptene; the
only 1,1,3-trimethylcyclohexane. The aromatics from experimental distribution from 2,2,4-trimethylhexane
2,6-dimethylheptene closely resemble those from contained a greater proportion of C9 aromatic than
1,1,3-trimethylcyclohexane except there is slightly was produced from the corresponding naphthene.
more of the C9 aromatics than was formed from This suggests that a direct C6 -cyclization occurs with
1,1,3-trimethylcyclohexane. A direct C6 -ring closure both 2,6-dimethylheptene and 2,2,4-trimethylhexane
468 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 17. Aromatic product distribution for the conversion of 1,1,3-trimethylcyclohexane over two different catalyst preparations, Pt (1)
(treated with aqueous KOH after reduction) and Pt (3).

but that a methyl migration, prior to or during cycliza-


tion, occurs more readily from a quaternary carbon
than from a tertiary one. The data presented in Fig. 17
for the aromatization of 1,1,3-trimethylcyclohexane
demonstrates how sensitive the aromatic selectivity
of Pt-Al2 O3 -K catalysts is to even small changes in
the catalyst preparation.
Considering all of the data in this section conclu-
sively demonstrate that it is almost certain that these
alkanes cyclize to alkylcyclohexane-like structures. In Scheme 8.
those instances for alkanes with a quaternary carbon,
a triene mechanism is not possible.
The results obtained by Kilner et al. [119] are en-
lightening. These authors converted [1-14 C-2-13 C]-
6.3. Isotopic labeled compounds ethylcyclopentane with a chromia-alumina catalyst at
505◦ C and 1 atm. The aromatization product, toluene,
Under conditions where the molecule does not un- was analyzed for 13 C and 14 C content in the methyl,
dergo isomerization prior to or after cyclization, the the ring carbon-1, and the sum of the ring -2, -3, and
labeled molecule technique may be equivalent to that -4 positions. The results they obtained are summa-
of using appropriately alkyl substitution. Thus, the rized in Scheme 8.
products from the conversion of an alkyl or polyalkyl If we assume that the cyclopentane ring undergoes
alkane with one or more six-carbon or longer chain hydrogenolysis as described above, we anticipate the
can serve as a means of deducing a mechanism in just label to be distributed as shown in Scheme 9.
the same way as the use of carbon isotope labeling Pathway c leads to 3-ethylpentane which can-
with the same compound. However, with n-hexane and not undergo direct six-carbon ring formation and
n-heptane the alkyl label is not applicable and the use should not produce toluene as rapidly as the prod-
of carbon isotopes is the only approach. ucts from pathways a and b. Pathway a leads to
B.H. Davis / Catalysis Today 53 (1999) 443–516 469

Scheme 9.

Table 19
Isotope distribution expected for toluene produced from n-heptane
and 3-methylhexane Scheme 10.
Reactant Label Methyl 1 2 3 4
14 C 50 0 50 0 0
below for chromia catalysts. This is true even when
13 C 0 50 0 50 0 the chromia-alumina catalyst contains alkali as it did
in the experiments by Kilner et al. [119].
The first to use 14 C-isotope in a study of the de-
14 C 0 0 0 100 0
13 C
hydrocyclization mechanism was Wheatcroft [120].
0 0 100 0 0
Wheatcroft [120] considered Pitketkly and Steiner’s
[121] analysis of the reaction pathway for the aroma-
tization of n-heptane (Scheme 10) where k1 is the rate
[1-14 C-2-13 C]-n-heptane and pathway b leads to of dehydrogenation and k2 is the rate of cyclization. A
3-methyl-[1-14 C-2-13 C]-n-hexane; thus, the products steady state was assumed for the heptene concentra-
from pathways a and b can undergo cyclization to di- tion in Scheme 10. Wheatcroft treated Scheme 10 as a
rectly produce a six-carbon ring. In Table 19, we com- system of consecutive unimolecular reactions as well
pare the isotope distribution expected if isotopically as assuming that hydrogenation of heptene was of sec-
labeled n-heptane and 3-methylhexane underwent ondary importance. Wheatcroft solved the differential
direct six-carbon ring formation. By comparing the equations describing the system and obtained an ex-
distribution of label that was obtained experimentally pression relating the fraction of intermediate (heptene)
with the one expected if hydrogenolysis had occurred in the product to the amount of heptane conversion.
to produce labeled n-heptane and 3-methylhexane, we Wheatcroft used literature values from earlier studies
conclude that hydrogenolysis followed by cyclization [122–125] and found that a ratio of k2 /k1 = 5 provided
made little, if any, contribution to the production of a good fit to the data (Fig. 18). Wheatcroft concluded
toluene. On the other hand, the isotopic distribution that his treatment showed that the alkene is a com-
is consistent with a bifunctional mechanism in which pulsory intermediate and that the rate of cyclization is
ethylcyclopentene, formed by dehydrogenation, un- five times faster than the rate of dehydrogenation to
derwent ring expansion at an acidic site, and the the olefin.
six-carbon ring then underwent dehydrogenation to Wheatcroft used a chromia (12 wt.%)-␥- alumina
produce toluene. For this bifunctional mechanism we catalyst. He plotted the percentage of ring labeled
would expect the 14 C label to be in the methyl position toluene versus the percent toluene in the product
and 13 C to be in the 1-ring position; the experimental (Fig. 19). He considered the results to show that, at
isotopic distribution is in good agreement with the low conversions, cyclization followed the pathway in
distribution predicted for the bifunctional mechanism. Scheme 11.
We consider this data to strongly support the opera- This cyclization pathway provides exclusively ring
tion of a bifunctional mechanism that can impact the labeled toluene. 1-Heptene isomerization occurs so
isotopic distribution obtained with the data described that at high conversion the 14 C-label becomes almost
470 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 11.

Scheme 12.

equally distributed between the two terminal posi-


tions. This would permit toluene to be formed with
equal label at the methyl and the ring of toluene in
(Scheme 12).
Wheatcroft indicated that his data can be explained
by this scheme provided the rate of isomerization of
1-heptene to 2- + 3-heptenes is about equal to the rate
of cyclization. Another explanation for Wheatcroft’s
Fig. 18. Percent heptene in the product vs. percent conversion in data would be that bifunctional isomerization to
k1 k2
the product. Theoretical lines for A → B → C. Experimental points methyl hexenes occurs and the isomerization is poi-
from [120]. soned more rapidly with increasing heptene conver-
sion than the aromatization reaction is.
Wheatcroft’s mechanism requires chemisorption of
the saturated portion of 1-heptene to be much more
rapid than the adsorption of the double bonded car-
bons; studies of the competitive conversion of n-octane
and 1-octene described below indicate that this is not
the case. Another argument against such a mechanism
is that the above intermediate resembles a Grignard
reagent and intramolecular cyclization of a Grignard
with the above structure leads to the formation of a
five-carbon ring [126–128].
Mitchell [129] found only 25–28% methyl label in
toluene that was produced from 1-14 C-n-heptane with
a chromia-alumina catalyst. Mitchell proposed that
this activity could be accounted for by the following
mechanism with statistical C–C bond breaking in the
cyclobutane ring (Scheme 13).
Fig. 19. Percent ring labeled toluene vs. percent toluene in product. Mitchell did not provide a mechanism to account
Line is best fit for straight line (from [120]). for the dehydrogenation of the methyl group that
B.H. Davis / Catalysis Today 53 (1999) 443–516 471

Scheme 13.

Scheme 14.

is detached from the catalyst and for its subsequent For 1-14 C-n-octane dehydrocyclization, ethylben-
cyclization to produce a four-carbon ring, a highly zene should have 50% of the label in the ethyl group,
strained structure. However, the equal participation of and this should be in the ␤-position of the ethyl group
the above two structures provide toluene that contains if it was formed by direct cyclization to a six-carbon
25% methyl labeled toluene, as obtained in Mitchell’s ring. For chromia on acidic alumina, Pines and Chen
experiment. found 52 and 50% methyl label in the first and second
Shortly following the report by Mitchell, Pines un- samples collected; 88–92% of the label was in the
dertook extensive studies of dehydrocyclization. In ␤-position of the ethyl group. For o-xylene, 100% of
these studies, Pines attempted to define the role of the radioactivity should be in the side chain methyl
acidity of the support. He and his coworkers employed groups, and the first and second samples contained 79
three catalysts in these studies: unsupported chromia; and 90% side chain label, respectively. Thus, for the
chromia supported on an acidic alumina derived from two aromatic isomers allowed by direct six-carbon
aluminum isopropoxide hydrolysis and chromia sup- ring closure, 80% or more of the 14 C-label is in
ported on a nonacidic alumina derived from potas- the position expected for a six-carbon ring closure
sium aluminate. Pines and coworkers obtained data mechanism. With chromia on nonacidic alumina,
for 1-14 C-n-heptane [130] and 1-14 C-n-octane [117] as the activity in ethylbenzene and o-xylene was also
well as labeled methyl hexanes. For the conversion of consistent with that expected for a direct six-carbon
1-14 C-n-heptane, 50% of the initial activity should be ring formation for the second sample they collected.
located in the methyl group. For two of the catalysts, However, the first sample collected for dehydrocy-
unsupported chromia and chromia on acidic alumina, clization of 1-14 C-n-octane had a low side chain
Pines and Chen [130] found 40% or more of the label activity just as had been found for the conversion of
in the methyl position. However, the surprising result 1-14 C-n-heptane with this catalyst. Pines and cowork-
was that the methyl label for the first sample collected ers also tried to include the 14 C distribution in m-
from the conversion of 1-14 C-n-heptane with the chro- and p-xylene in their mechanism. These results led
mia on nonacidic alumina catalyst was only 17.4%; Pines and coworkers [25] to propose that various ring
this was close to the value expected if there had been size intermediates and methyl insertion isomerization
complete scrambling of the isotope to each of the po- reactions were involved in the dehydrocyclization
sitions of toluene. Pines and Chen proposed a reaction mechanism. These are demonstrated in Scheme 15 by
pathway that included the formation of cycloheptane some of the possible reaction pathways for n-octane
and its subsequent conversion to toluene (Scheme 14). and 3-methylheptane.
472 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 15.

Pines and coworkers disregarded the data show-


ing that compounds with various ring sizes, such
as methylcycloheptane, underwent aromatization at
a slower rate than the n-alkane that they would be
formed from. To account for this observation, they
concluded that the various sized rings, such as methyl-
cycloheptane, did not leave the surface but underwent
ring contraction while adsorbed on the catalyst sur-
face. This would require adsorption to be the rate
Scheme 16.
determining step although this was not advanced by
Pines and coworkers.
If we consider the 14 C-label distributions in ethyl-
benzene and o-xylene that were obtained by Pines and Greater than 90% of the 14 C in the aromatic products is
coworkers, only the early samples collected when us- in the position predicted for six-carbon ring formation.
ing the chromia on nonacidic alumina are inconsistent Feighan and Davis [132] converted [4-14 C-]-n-
with at least 80%, and usually greater than 80%, of the heptane with their version of the three catalysts used
aromatics being formed by a mechanism that involves by Pines and coworkers: unsupported chromia and
direct six-carbon ring formation. chromia supported on acidic and nonacidic aluminas.
Cannings et al. [115,131] converted 2,2-dimethyl- With this reactant, all of the 14 C should be present in
4-14 C-methyl pentane (isooctane) and 3-[14 C]-methyl- the ring position meta to the methyl group if toluene
heptane with a potassium and cerium promoted is produced by direct six-carbon ring formation. In
chromia-alumina catalyst. With 3-[14 C]-methylheptane contrast to Pines and coworkers, Feighan and Davis
the aromatics composition and 14 C distribution be- degraded the aromatic ring so that the activities in the
tween the side chain ring are given in Scheme 16. ring positions as well as the methyl group could be de-
The C8 -aromatic distribution is the one expected for termined from experimental data (Scheme 17) [133].
a mechanism with direct six-carbon ring formation. All of the results were consistent with at least 80%,
B.H. Davis / Catalysis Today 53 (1999) 443–516 473

Table 21
Conversion of labeled heptane to toluene over Pt and Pt bimetallic
catalysts
Catalyst Charge = % 14 C in Methyl position
Sample n-[1-14 C]- n-[4-14 C]- Reference

Pt-Al2 O3 -K 1 36 3.2 [135]


2 41 1.3 [135]
3 44 1.1 [135]
Pt-Al2 O3 -K – 40 – [136]
10% Pt-Al2 O3 – 50 2.5 [138]
Pt–Sn-Al2 O3 -K 1 52 0.57 [137]
2 49 0.29 [137]
Pt-Re-Al2 O3 -Aa 1 34 4.1 [137]
2 32 2.5 [137]
3 32 – [137]
a Sec-butyl amine added to the charge.

Scheme 17.

Table 20
14 C in the methyl position of toluene from the conversion of 14 C

labeled heptane over chromia catalysts


Catalyst Samplea Reference Scheme 18.
1 2 3 4
n-[1-14 C] heptane Davis also converted [135] [1-14 C-]- and [4-14 C-]-n-
Cr2 O3 44.7 47.3 47.4 [130] heptane with Pt supported on nonacidic alumina.
Cr2 O3 -Al2 O3 -A 38.7 39.8 41.7 [130]
Cr2 O3 -Al2 O3 -B 17.5 21.7 32.1 [130]
The side chain label and the label in the ring were
Cr2 O3 -Al2 O3 -B 43.0 39.0 40.0 [134] consistent with 80% or more of the toluene be-
Cr2 O3 -Al2 O3 -B 28 [129] ing formed by a direct six-carbon ring mechanism.
n-[4-14 C] heptane Nogueira and Pines [136] later reported data for their
Cr2 O3 3.7 1.8 0.6 [132] Pt-nonacidic alumina catalyst that agreed with those
Cr2 O3 -Al2 O3 -A 1.6 1.7 1.3 1.2 [132]
Cr2 O3 -Al2 O3 -B 2.6 2.8 2.7 3.0 [132]
obtained by Davis. Davis also converted [1-14 C-]- and
[4-14 C-]-n-heptane with Pt-Re-acidic alumina and
a Collected at increasing time-on-stream. Pt-Sn-nonacidic alumina [137]. For the Pt-Re catalyst,
sec-butyl amine was added to poison the acid function
and usually greater than 80%, of a mechanism that during the conversion since alkali metals are severe
involves direct six-carbon ring formation (Table 20). poisons for this catalyst. Again, the results with both
Later, Davis and Venuto [134] converted both bimetallic catalysts, Pt-Re and Pt-Sn, were consistent
[1-14 C-]- and [4-14 C-]-n-heptane with chromia on with more than 80% of the aromatics being formed
nonacidic alumina and found an isotope distribution by a direct six-carbon ring formation (Table 21).
for all samples that was consistent with 80% or greater The label distribution expected for converting
of the toluene being formed by a direct six-carbon heptane, labeled at C-1 or C-4, to toluene is given
ring formation. Unfortunately, Professor Pines did not in Scheme 18. Analysis of the toluene ring label
have a supply of his chromia on nonacidic alumina distribution from C-1 and C-4 labeled heptane can
for use in this study. distinguish between a direct C6 closure and a C5
474 B.H. Davis / Catalysis Today 53 (1999) 443–516

Table 22
or Pt catalysts. Experimental difficulties prevented us
Ring 14 C distribution in toluene obtained from the conversion of
n-4- 14 C heptane over chromia catalysts [132] from making a complete ring degradation for our Pt
studies; the data in Table 23 have been calculated so
Catalyst Sample (C2 )/(C1 + C2 + C3 ) C2 /C3
that all results are presented in a similar way. The ex-
Cr2 O3 1 0.070 0.081 perimental value falls about midway between the 0.5
2 0.023 0.15 expected for direct C6 ring closure and the 0.33 ex-
3 0.015 0.017
pected for the 1,2-dimethylcyclopentane intermediate.
Cr2 O3 -Al2 O3 -A 1 0.045 0.047
2 0.042 0.051 It will take a very careful and difficult ring label analy-
4 0.045 0.048 sis to eliminate 1,2-dimethylcyclopentane as an inter-
Cr2 O3 -Al2 O3 -B 1 0.11 0.14 mediate using carbon tracers. The C-2 and C-3 labeled
2 0.057 0.070 heptane pair will provide the same label distribution
4 0.041 0.047
in the toluene as the C-1 and C-4 labeled pair so ad-
ditional information will not be attained from use of
C-2 and C-3 labeled heptane.
closure to 1,2-dimethylcyclopetane followed by ring Amir-Ebrahimi et al. [138] converted [1-13 C]-n-
expansion (Tables 22 and 23 ). The methyl label of heptane with a 10 wt.% Pt-alumina catalyst at 380◦ C
toluene produced from [4-14 C]-heptane with all three and 3 Torr heptane/760 Torr helium. They determined
chromia catalysts is consistent with, at most, 20% of the position of the label using mass and microwave
a cycloheptane ring intermediate. The results from spectroscopy. After hydrogenation to methylcyclohex-
[1-14 C]-heptane, except for the runs with nonacidic ane, mass spectrometrical analysis showed that the
chromia by Pines and Chen [130] and by Mitchell toluene had 50% of the 13 C-label within the ring. The
[129], are also consistent with 80% or greater direct distribution of the 50% ring label was reported to be
C6 ring closure. Davis [137] obtained that similar distributed with 34% in the ortho (2-) position and
results were obtained with the [1-14 C]-heptane using 16% in the meta (3-) position. A label content for the
Pt-, Pt bimetallic, and supported and unsupported 1- and 4- position was not given nor was there an es-
chromia catalysts. timate for the lower limits of detection. To account
The toluene ring label distribution was determined for their experimental label distribution, these authors
using our modification of the Steinberg and Sixma indicated that a three-step mechanism was involved:
method (Scheme 17) [133]. The data in Tables 22 1,5-ring closure, ring opening, and 1,6-ring closure.
and 23 show that C5 -ring closure to ethylcyclopen- The authors do not explain why 1,6-ring closure does
tane followed by ring expansion to a six carbon ring not contribute in the first step.
does not make a major contribution to the aromatic A variety of techniques are now being used to
formation from [4-4 C]-n-heptane with either chromia identify surface species resulting from C–H cleavage

Table 23
Ring label distribution in toluene obtained from n-heptane over Pt catalysts
Charge Sample (C2 )/(C1 + C2 + C3 ) Reference

Experimental Calculated based on

C6 EtCPa 1,2-DMCPa
n-4-14 C- 2 0.033 0.0 0.33 0.0 [135]
3 0.065 0.0 0.33 0.0 [135]
n-4-14 C- – 0.06b 0.0 0.33 0.0 [138]
n1-13 C- 1 0.40 0.50 0.33 0.33 [135]
3 0.44 0.50 0.33 0.33 [135]
n-1-13 C- – 0.40 0.50 0.33 0.33 [138]
a EtCP = ethylcyclopentane; DMCP = dimethylcyclopentane.
b Based on a maximum of 5% 13 C at positions other than meta.
B.H. Davis / Catalysis Today 53 (1999) 443–516 475

Scheme 19.

Scheme 20.

Scheme 21.

Scheme 22.

of alkanes. These species include, among others These or similar structures will be encountered in
[139], the following schemes.
Amir-Ebrahimi et al. [138] utilized the non-selective
(Scheme 19) and selective (Scheme 20) mechanism
for alkylcyclopentane isomerization/hydrogenolysis
to explain their data:
However, Schemes 19 and 20 were inadequate
so that the authors introduced a third mechanism
(Scheme 21):
The direct ring closure, illustrated for 3-methyl-
hexane, is described as a carbene-olefin addition
(Scheme 22):
Combining these schemes, they described the three
step mechanism as (Scheme 23):
It appears that the above scheme was constructed to
fit the data that had been obtained, and was based upon
the initial reactant structure. There is no reasonable
476 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 23.

way to predict the contributions of pathways a and Table 24


13 C distribution n-heptane-1-13 C reaction products (Te/NaX)a
b in Scheme 23 nor the relative amount of the two
(from [110])
pathways in a or the three in b.
Unfortunately, the acidity of the catalysts used by Contact time (ks): 3.6 22.7
n-heptane conversion (%): 7.2 33.9
Gault and coworkers is not well defined. If we consider 13 C distribution (%)
that some modest amount of isomerization occurs in Toluene
competition to the normal 1,6-ring formation, we have Methyl- 48.1 47.7
the following possibilities (Scheme 24): Ortho- 47 46.0
The simple alkane isomerization mechanism allows Para- 1.1 1.0
Meta- 3.8 4.6
for 1,6-ring formation to produce toluene equally la-
1- 0 0.6
beled in the methyl group and in the ortho-ring po-
sition as well as the label at the meta-position (3-). 3-methylhexane
CH3 (1,6,7) positions – 100
The isomerization scheme would also provide label CH2 , CH (2,3,4,5) – 0
in the para-position (4-); however, the experimental
3-ethylpentane
value for this was apparently below the level that could
CH3 (1,5,7) – –
be detected. Even so, it is apparent that the complex CH2 , CH (2,3,4,6) – –
three-step mechanism, which must finally involve a
2-methylhexane
direct 1,6-ring closure, should be advanced only af- CH3 (1,6,7) – –
ter the possibility of a direct 1,6-ring formation plus a CH2 , CH (2,3,4,5) – –
small fraction of the simple bifunctional isomerization a From NMR measurements.
process has been eliminated.
Iglesia et al. [110] converted [1-13 C]-n-heptane
with a Te/NaX zeolite catalyst at 723 K and 100 kPa ortho-ring positions (Table 24). The position of the
(H2 : h.c. = 21). The toluene product contained greater 13 C in the methylhexane isomers that were produced

than 93% of the isotopic label in the methyl and are given in Table 24. These show that the amount of
B.H. Davis / Catalysis Today 53 (1999) 443–516 477

Scheme 24.

the isotope in the CH2 positions was too low to de-


tect; 100% of the isotope detected was in the methyl
positions. Iglesia et al. considered the aromatization
to occur through the formation of heptatrienes which
subsequently cyclize, in analogy with the hexatriene
mechanism outlined below. The additional feature
of the proposed mechanism is the formation of the
bicylcic structure from 1,3,6-heptatriene and the sub-
sequent rearrangement/hydrogenolysis of the cyclo-
propane ring to locate the isotope label in the meta
and para position (Scheme 25).
The simpler explanation of the dehydrocyclization
of a small amount of methylhexanes that result from
isomerization, as illustrated above in Scheme 24, can
equally well account for the isotopic label being in the
meta and para positions as well. In any event, the data
reported by Iglesia et al. are in excellent agreement
with the dominant (about 95%) fraction of the toluene Scheme 25.
478 B.H. Davis / Catalysis Today 53 (1999) 443–516

being formed by a six-carbon ring forming mechanism


and provide no direct data to support the hepatriene
as the intermediate in the cyclization pathway.
Considering all of the tracer studies, the only data
that are inconsistent with a direct six-carbon ring for-
mation were obtained using chromia supported on
nonacidic catalysts. The above data for C7 + alkanes
are also mostly in agreement with aromatics being
formed by a mechanism that involves direct six-carbon
ring formation. The only exception was for some of
the chromia catalysts, and these showed a distribution
that was consistent with a dual functional isomeriza-
tion, such as was found by Kilner et al. [119]. For
example, the data of Pines and Chen [117,130], for
the conversion of labeled n-heptane and n-octane with
their potassium containing chromia-alumina catalyst,
changed with increasing reaction time to become con-
Fig. 20. Ratio of naphthene/paraffin conversion with Pt-nonacidic
sistent with that of six-carbon ring formation. The hy- alumina catalyst at 300◦ C (䊉) 482◦ C (䊏) and 1 atm (data from
pothesis of some acidity for some chromia catalysts [141]).
and a contribution of some isomerization through a bi-
functional mechanism is a much simpler mechanism d(cycloalkane) kc
than resorting to the formation of all ring sizes that = (1)
d(alkane) ka
undergo contraction-expansions.
where kc and ka are the rate constants for cycloalkane
and alkane, respectively. Based upon the rates mea-
sured separately, kc /ka is expected to be much greater
7. Irreversible adsorption mechanism than one. On the other hand, if the same assumptions
are made for the conditions except that adsorption is
Early work showed that the dehydrogenation of a irreversible, the relative rate of conversion of the two
cycloalkane was much more rapid than the dehydro- reactants is:
cyclization of the corresponding alkane. For example, d(cycloalkane)
Fogelberg et al. [112] found that while ethylcyclohex- =1 (2)
d(alkane)
ane underwent complete conversion at 375◦ C, under
the same conditions only 62% n-octane conversion It was found that the reversible adsorption path-
was attained at 525◦ C. Thus, in any competitive con- way was followed for the lower temperature range
version we should expect a much greater conversion (300–400◦ C). When the conversion was carried out
rate for the cycloalkane than for the correspond- at 482◦ C the ratio was close to 1, indicating that the
ing alkane. Aromatics are apparently not strongly reaction pathway followed essentially irreversible ad-
adsorbed; for example, Archibald and Greensfelder sorption and the adsorption constants for the two re-
[140] found that the effect of the aromatic was simply actants were nearly the same (Fig. 20). This situation
that of an inert diluent for the conversion of an equiv- was found to apply to a number of competitive conver-
olume mixture of n-heptane and toluene at 490◦ C sion situations: naphthene/alkane mixture, a mixture
with a promoted chromia-alumina catalyst. of two alkanes with six or more carbons; a mixture
If it is assumed that Langmuir–Hinshelwood ad- of alkanes, one with more and one with less than 6
sorption behavior is followed in a competitive conver- carbons and a mixture with alkane/naphthene with
sion, the relative rate of conversion of an equimolar five carbon rings [141]. Thus, at 1 atm and 482◦ C,
ratio of the cycloalkane and alkane that have the same the conversion of alkanes and cycloalkanes involves
adsorption constant should be [141]: essentially irreversible adsorption of the reactant.
B.H. Davis / Catalysis Today 53 (1999) 443–516 479

Fig. 21. Protium incorporation in n-octane for conversion of n-octane-d18 /methylcyclohexane-d0 at 482◦ C with Pt-Al2 O3 (䊐), nonacidic
Al2 O3 containing 1% K (䊏), and acidic Al2 O3 (䊉) (from [142]).

Fig. 22. Deuterium incorporation into n-octane at 482◦ C using a Pt-SiO2 catalyst: (䊐) n-octane-d18 /methylcyclohexane-d0 , and (䊉)
methylcyclohexane-d14 /n-octane-d0 (from [142]).

For irreversible adsorption during the conversion was irreversible adsorption of the two reactants when
of a mixture of the alkane and cycloalkane, one alumina was used as the support.
but not the other labeled with deuterium, observ- There was little exchange between the two reac-
able exchange between the two gas phase reactants tants (Fig. 22) when a mixture of n-octane-d18 and
should not occur. When a mixture of n-octane-d18 methylcyclohexane-d0 or a mixture of n-octane-d0 and
and methylcyclohexane-d0 was converted over a methylcyclohexane-d14 was converted at 482◦ C with
Pt-alumina catalyst at 300◦ C or 482◦ C, there was a Pt-SiO2 . While there was essentially no H/D ex-
considerable exchange (Fig. 21) [142]. More super- change in the reactants, the aromatic products derived
vising, there was as much or more exchange when from the two reactants has exchanged to essentially an
this mixture was converted over acidic alumina with- equilibrium H/D composition (Fig. 23). In contrast to
out Pt, or even over a nonacidic alumina, than when the conversion at the high temperature (482◦ C), there
Pt was present. This exchange activity of the alumina is exchange to a nearly statistical distribution when
support precluded a determination of whether there mixture of n-octane-d18 and methylcyclohexane-d0 or
480 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 23. Deuterium distribution in ethylbenzene produced in methylcyclohexane/octane/D2 run with Pt-SiO2 catalyst at 482◦ C (from [142]).

Table 25
Deuterium incorporation in the starting reagents using a Pt-SiO2 catalyst (from [142])
Deuterium Content Methylcyclohexane n-octane
300◦ C 375◦ C 482◦ C 300◦ C 375◦ C 482◦ C
Aa Ba Aa Aa Ba Aa Ba Aa Aa Ba
d0 1.44 4.15 – – 94.5 15.0 1.09 78.1 98.2 74.4
d1 3.68 13.6 – – 4.74 29.2 4.49 16.4 1.53 8.39
d2 4.90 21.3 – – 0.59 28.0 10.6 4.39 0.29 4.21
d3 6.09 24.4 – – 0.16 17.3 17.1 0.97 – 3.44
d4 7.68 19.3 – – – 7.46 20.4 0.15 – 3.04
d5 7.48 12.1 0.08 – – 2.39 19.1 – – 2.35
d6 8.32 4.57 0.14 – – 0.58 13.8 – – 1.77
d7 8.64 0.59 0.24 – – – 8.06 – – 1.26
d8 8.61 – 0.43 – – – 3.56 – – 0.73
d9 9.29 – 0.54 – – – 1.31 – – 0.36
d10 8.71 – 0.93 0.16 – – 0.40 – – –
d11 8.79 – 2.01 0.59 – – – – – –
d12 8.10 – 7.02 4.93 – – – – – –
d13 5.49 – 16.9 9.54 – – – – – –
d14 2.77 – 71.7 84.76 – – – – – –
d15 – – – – –
d16 – – – – –
d17 – – – – –
d18 – – – – –
a A = mixture of methylcyclohexane-d (12%) and n-octane-d (88%); B = mixture of methylcyclohexane (56.3%) and n-octane (42.9%)
14 0
and D2 gas (8.5 ml/min).

a mixture of n-octane-d0 and methylcyclohexane-d14 temperature is employed with compounds representa-


was converted at 300◦ C (Table 25). Thus, we have the tive of naphtha reforming.
surprising result that there is more exchange in the These results indicate that the rate limiting step
reactants at the lower temperature than when a high of the dehydrocyclization reaction is adsorption that
B.H. Davis / Catalysis Today 53 (1999) 443–516 481

Table 26
Dehydrocyclization of a equimolar mixture of n-octane-d0 and
n-octane-d18 with a Pt/SiO2 Catalyst at 482◦ C and 1 atm with a
Pt-SiO2 Catalyst (from [143])
Time-on- Conversion of octane-d0 /
stream (h) Conversion (%) conversion of octane-d18
d18 d0

2 5.76 21.7 3.77


3 4.75 19.0 4.00
4 5.09 20.0 3.92
5 3.93 14.9 3.78
6 4.02 12.1 3.02
7 3.54 11.1 3.14
8 3.37 10.2 3.08

Table 27
Fig. 24. The deuterium distribution in the ethylbenzene (䊉) and
Conversion of a mixture of n-octane and methylcyclohexane with
o-xylene (䊐) produced during the dehydrocyclization of an equal
a Pt/SiO2 catalyst at 482◦ C and 1 atm without added hydrogen
molar mixture of C8 H18 and C8 D18 (the data shown is for the
(from [143])
sample collected at the third hour; the data shown are representative
Reagents (mol) Conversion of MCH/ of those for the samples collected at other times) (from [143]).
conversion of octane

MCH-d0 octane-d18 4.0 + 0.2


MCH-d0 octane-d0 1.2 + 0.2 to a kinetic isotope effect of 3.3 ± 0.6 for the con-
MCH-d14 octane-d0 0.41 + 0.05 version of the n-octane/methylcyclohexane mix-
ture. Likewise, a kinetic isotope effect was found
when methylcyclohexane-d14 was substituted for
involves breaking of the C–H or C–D bond. If the ini- methylcyclohexane-d0 ; in this instance a kinetic iso-
tial adsorption is rate limiting there should be a kinetic tope effect was calculated to be 2.9 ± 0.6. It is empha-
isotope effect. As described below, this was found to sized that these kinetic isotope effects were obtained
be the case. experimentally in a competitive conversion so that
An equimolar mixture of n-octane-d0 and n-octane- both reactants were subjected to identical reaction
d18 was converted at 482◦ C and atmospheric pres- conditions of temperature, catalyst state and hydro-
sure [143]. Samples were collected at intervals gen(deuterium) pressure and surface coverage (apart
and the kinetic isotope effect calculated from the from the surface H/D ratio).
amount of unconverted reactants. Because of an in- The aromatic products, if H/D exchange did not
verse isotope effect that is applicable in g.c. anal- occur, should consist of ethylbenzene-d0 and -d8 and
ysis, two distinct peaks, with baseline separation o-xylene-d0 and -d8 . However, the H/D isopotomer
between the two, were obtained for n-octane-d0 distribution (Fig. 24) show that there was essentially
and n-octane-d18 , allowing one to obtain the isopo- equilibration of the H/D in both aromatic products.
tomer distribution within each peak [144]. The ki- This requires that the aromatic product and/or the pre-
netic isotope effect for the conversion of n-octane-d0 cursors to remain on the surface for a sufficient time
and n-octane-d18 was 3.5 ± 0.4 (see Table 26). The for this equilibration to occur.
data for the kinetic isotope effect for the conver- If the irreversible alkane adsorption step de-
sion of the n-octane and methylcyclohexane mix- termines the aromatic product, there should be a
tures are compiled in Table 27. Whereas the ratio of large kinetic isotope effect in the formation of
the conversion of methylcyclohexane/n-octane was o-xylene relative to ethylbenzene for the conversion
1.2 for the undeuterated components, the conver- of 2,2,7,7-tetradeutero-octane [145] (Scheme 26).
sion ratio was found to be 4.0 when n-octane-d18 Thus, whereas the ratio of ethylbenzene/o-xylene
was substituted for n-octane-d0 ; this corresponds formed with undeuterated n-octane is about 1.0, the
482 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 26.

Fig. 26. Fraction of unlabeled toluene versus total toluene found


during the conversion of a mixture of [3-14 C]-n-heptane (76%) and
unlabeled methylcyclohexane (24%) at 520◦ C with a Pt-alumina
catalyst (data from [148]).

tial C–H bond that is broken in the rate-determining


alkane adsorption step.
Isagulyants et al. [146] and Kazanskii and Rozen-
gart [147] converted various cyclohexane-hexane mix-
tures in which the cyclohexane was labeled with 14 C.
Fig. 25. Percentage of 14 C-labeled benzene versus the percentage
of 14 C-labeled cyclohexane in a mixture with unlabeled n-hexane
As shown in Fig. 25, the 14 C labeled benzene was
using a pulse (䊉) and flow (X) reactor (drawn from data in directly related to the amount of 14 C labeled cyclo-
[146,147]). hexane in the charge. The slope shows that cyclo-
hexane was converted about 1.5 times as rapidly as
n-hexane; this is similar to the value we obtained with
kinetic isotope effect of about 3.5 would require this mixture using a number of catalysts [141]. Il’in
this ratio to change to 3.5 for the conversion of et al. [148] found a selectivity for the conversion of
2,2,7,7-tetradeutero-octane. The aromatic distribution a n-[3-14 C-]heptane-methylcyclohexane mixture over
obtained for the conversion of 2,2,7,7-tetradeutero- Pt-Al2 O3 -K (Fig. 26); correcting the slope for the ap-
octane with the Pt-SiO2 catalyst at 482◦ C was essen- proximate 3-fold excess of n-heptane in the reactant
tially the same as those of the undeuterated n-octane. gives a ratio of 4.4.
This result requires that the adsorbed alkane undergoes Manninger et al. [149] converted a mixture of
H/D exchange much more rapidly than cyclization so n-hexane and [14 C]-cyclohexane with a Pt KL zeo-
that the aromatic is formed without regard to the ini- lite catalyst at about 300◦ C in a recirculating reactor.
B.H. Davis / Catalysis Today 53 (1999) 443–516 483

gen partial pressure. Thus, as the hydrogen pressure


is increased to several atmospheres, it appears that the
mechanism gradually changes from irreversible to re-
versible adsorption.
The question of whether an alkene, and even di-
ene and triene (see discussion below), is an interme-
diate in the dehydrocyclization mechanism has been
debated. To define whether this is the case, a mixture
of 1-octene-d0 and n-octane-d18 was converted with
the Pt-SiO2 catalyst [151]. Several observations for
the conversion of this reaction mixture bear heavily
Fig. 27. Molar radioactivity in the cyclohexane fraction of the ef- upon the dehydrocyclization mechanism that operates.
fluent as a function of the conversion, X. Feed: mixture of 10 Torr
First, there is some H/D exchange in the small amount
n-hexane + 2 Torr [14 C]-cyclohexane in a closed-loop circulation re-
actor containing 0.8% Pt/KL catalyst. T = 300◦ C, p(H2 ) = 120 Torr of alkene that is observed in the products, indicating
(circles) and 480 Torr (triangles) (from [150]). that the added alkene undergoes adsorption/desorption
during the conversion. Secondly, there is not a de-
tectable amount of n-C8 H16 D2 or n-C8 H17 D1 ; this re-
Two hydrogen pressures were used, 16 and 64 kPa, sult shows that the adsorbed alkene does not undergo
with a 20% cyclohexane mixture (0.27 kPa cyclohex- hydrogenation to the alkane which then desorbs to the
ane and 1.33 kPa n-hexane). At the lower hydrogen gas phase. Thus, the alkene must be adsorbed in a form
pressure the cyclohexane label was constant for sev- that allows some fraction to desorb as well as in a man-
eral levels of conversion (Fig. 27); however, at the ner that subsequently leads to the aromatic products.
higher hydrogen partial pressure the relative activity Thirdly, the alkene in the alkene/alkane mixture is ad-
in cyclohexane declined with increasing conversion. sorbed and converted to aromatics much more rapidly
Menon and Paál [150] indicate that the higher hydro- than the alkane. Similar results were obtained for a re-
gen partial pressure might be able to hydrogenate the action mixture of 1-heptene-d0 and n-octane-d18 . The
alicyclic surface precursors of benzene (cyclohexene results show that the adsorption of an added alkene
or cyclohexadiene) to cyclohexane. At 300◦ C and is greatly favored over that of the alkane. Since the
atmospheric pressure, the equilibrium composition rate limiting step for alkane dehydrocyclization is the
of cyclohexane/benzene should be far (95+%) on the initial C–H bond rupture, the added alkene must en-
benzene side. It does not appear reasonable for des- ter the cyclization reaction pathway at some step that
orption of cyclohexane, following hydrogenation of follows the rate limiting step or must be converted in
cyclohexane, of cyclohexene or cyclohexadiene to be a parallel pathway.
more rapid than the subsequent dehydrogenation to Price et al. [152] found a large kinetic isotope effect
benzene. Details of the analytical procedure were not (kH /kD > 2 at 400◦ C) when deuterated hydrocarbons
given other than a gas chromatograph equipped with are reacted simultaneously (or separately) with similar
a radioactivity detector was used. A similar result undeuterated hydrocarbons. Thus, the authors found
could be obtained if a product derived from n-hexane a kinetic isotope effect of about 2.5 from the conver-
coeluted with cyclohexane and served to dilute the sion of deuterated and undeuterated cyclohexane with
cyclohexane with respect to the original activity. In a Te/NaX zeolite at 400◦ C (Fig. 28). Furthermore,
any event, these results are difficult to explain and the the authors found when a mixture of C6 H12 and
authors [150] indicate that further experiments are C6 D12 were converted in the presence of gaseous H2
necessary to elucidate this question. or D2 , the kinetic isotope effect was similar. Equally
For total pressures in the 400 psig range and 482◦ C, important, they found little intermolecular H/D ex-
the relative conversion of a cycloalkane/alkane mix- change nor was there exchange with the gaseous H2
ture favored cycloalkane by a factor that was at least 20 or D2 . Since they found that cyclohexane converted
or greater. Our results suggested that the atmospheric at the same rate in the presence of gaseous H2 or D2 ,
pressure conversion was nearly independent of hydro- they showed that there is no isotope effect with the
484 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 28. First order plots for conversion of cyclohexane-d0 (N) and cyclohexane-d12 (䊉) in a D2 in recirculation reactor at 400◦ C (40 Torr
cyclohexane, 300 Torr H2 or D2 ) (from [152]).

activation of hydrogen (deuterium) with this catalyst.


These authors concluded that the absence of exten-
sive H–D exchange and presence of a large kinetic
isotope effect strongly suggested that the slow step
in the reaction is the cleavage of one or more C–H
bonds. The authors converted undeuterated cyclohex-
ene in the presence of gaseous D2 ; they found both
cyclohexane and benzene as the products. Neither the
cyclohexane or benzene had incorporated significant
amounts of deuterium, showing that these products
were formed from cyclohexene disproportionation.
While they did not detect dienes or trienes in the
products, these authors interpreted their results using
the triene mechanism outlined below; however, they
had no direct evidence to support this mechanism. At
the same time, their results are consistent with the rate
limiting step being the initial C–H bond rupture in a Fig. 29. Comparison between n-hexane (20 Torr) hydrogenolysis
mechanism that involves irreversible adsorption of the and aromatization rates over Pt(10,8,7) at 300◦ C as a function of
reactant. hydrogen (D2 ) pressure (from [153]).
Using single crystal Pt catalysts, Davis et al.
[153] found an inverse kinetic isotope effect
(kH /kD = 0.3–0.77) for the initial rates for the hy- Iglesia et al. [110] converted a mixture of 1-heptene
drogenolysis, isomerization and C5 -cyclization of (20 mol%) and [1-13 C-]-n-heptane with a Te/NaX ze-
n-hexane and n-heptane at 1 atm and 520–640 K olite catalyst. They utilized a batch, recirculating reac-
(247–367◦ C). The isotope effect did not depend upon tor system in which the reactants and products, as they
the crystal face of the Pt single crystal. They consid- accumulate, are recirculated over the catalyst. The data
ered the inverse kinetic isotope effect to depend upon are reported in terms of toluene turnover, defined as
a combination of kinetic and thermodynamic effects. the moles of C7 converted to toluene per Te moles in
These authors found that the aromatization reaction, the catalyst, as shown in Fig. 30. When the 13 C-labeled
unlike hydrogenolysis and isomerization, exhibited toluene content in the products is plotted against reac-
the normal kinetic isotope effect (e.g., Fig. 29). tion time, the data fit a straight line that extrapolates to
B.H. Davis / Catalysis Today 53 (1999) 443–516 485

Scheme 27.

Fig. 30. Moles of labeled and unlabeled toluene formed from step. It appears that the data described above for our
1-heptene (20%)/n-heptane-1-13 C (80% mixtures) (723 K; hydro-
deuterium-labeling studies show that the C–H (D)
carbon, 4.5 kPa; hydrogen, 96 kPa) (from [110]).
bond breaking process occurs in the step leading from
the physically adsorbed species to the chemically ad-
zero at zero time. However, the 12 C-toluene extrapo- sorbed species as shown in Scheme 27 where g = gas
lates to a non-zero value at time zero. The 12 C-toluene phase, a = adsorbed phase, and p and c = physical and
concentration rapidly attains a limiting value that is chemical, respectively.
in the range of 0.4–0.5 toluene turnovers. These au- The experimental data for deuterium labeled com-
thors report that the heptenes and heptane, as deter- pounds can be explained using the above mechanism.
mined by 13 C content, had equilibrated within 2 ks First, there is very little H/D exchange in the uncon-
contact time. We believe that a more reasonable ex- verted reactants. The step leading from physisorption
planation for the experimental data at short contact to chemisorption is essentially irreversible; thus, even
times is that the 1-heptene underwent conversion at though there is extensive H/D exchange following
a much more rapid rate than the n-heptane, and the chemisorption, the H/D exchanged species do not
authors make this observation in their paper. Thus, return to the gas phase. Secondly, Scheme 27 ex-
during the initial stages of the conversion, the unla- plains the primary kinetic isotope effect because the
beled 1-heptene is consumed at a much more rapid C–H (C–D) bond breaking occurs in the irreversible
rate than the [1-13 C]-n-heptane is. From the data in chemisorption step. Thirdly, the cyclization step fol-
Fig. 30, we judge that the unlabeled 1-heptene conver- lows the chemisorption step. The experiment for
sion is essentially complete in the first 2 ks of contact the dehydrocyclization of 2,2,7,7-tetradeuteriooctane
time. This rapid depletion of the unlabeled 1-heptene showed that there is no isotope effect in the step that
would leave only 13 C-labeled heptane as a reactant af- determines whether ethylbenzene or ortho-xylene is
ter about 2 ks so that the products could only be derived formed [145]. Finally, this mechanism indicates that
from the 13 C-labeled heptane. For the latter interpre- there will be H/D exchange in the reactant as well as
tation, the data of Iglesia et al. are completely consis- in the products at higher H2 (D2 ) pressure since high
tent with our data for the conversion of 1-octene-d0 concentrations of [H] and [D] will shift the above
and n-octane-d18 . The data in Fig. 30 would be con- steps of Scheme 27 to the left. While the experiment
sistent with a mechanism where the alkene and alkane with D2 has not been conducted at high pressure to
are converted through either a common or in parallel date, the competitive conversion of methylcyclohex-
reaction pathways. ane and n-octane at H2 pressures of 200 or 400 psig
It is apparent that there is a primary isotope effect are consistent with reversible adsorption, in agree-
for the conversion of both n-octane and methylcyclo- ment with the above prediction. These experiments
hexane. These results require that the C–H (C–D) bond with deuterium have been conducted at 100 psig and
breaking process be involved in the rate-determining exchange, as postulation does occur.
486 B.H. Davis / Catalysis Today 53 (1999) 443–516

Previously, it was proposed [142] that the first step


of chemisorption is as shown in Eq. (3).
RH  R• + H• (3)
and that R•is further dehydrogenated in a second step
to form a methylene species:
R•  R : +H (4)
We do not have experimental evidence to prove this hy-
pothesis. However, the reverse of reaction (3) is widely
accepted as the final step in the olefin hydrogenation.
The kH /kD value (3.5 for n-octane, 2.9 for methyl-
cyclohexane) obtained from this study is higher than
expected if only the zero point energy of the C–H and Fig. 31. Schematic potential energy curves showing alteration of
C–D bonds are considered. According to a classical irreversible adsorption for 1 atm conditions to reversible adsorption
view of the kinetic isotope effect, the ratio of rate con- for 200 psig conditions.
stants is:
kH exp1E0
= (5) A large number of results for the proton transfer re-
kD RT action show that kH /kD at ordinary temperature (25◦ C)
where 1E0 is equal to the zero point energy differ- usually lies between 3 and 7; however, it is not uncom-
ence of C–H and C–D bonds, and has a value of mon to obtain values of kH /kD as high as 10 at 25◦ C
1.2 kcal/mol. Based on this equation, at 482◦ C the [155]. In a few instances, values as large as 23 are en-
maximum value of kH /kD is calculated to be 2.2, some- countered (e.g., [156,157]). These results indicate that
what smaller than determined experimentally. the explanation of the variable kinetic isotope effect
The above calculation for the maximum value for in terms of only the zero point energy does not ade-
kH /kD assumes that 1E is the same at 482◦ C as it is at quately account for the experimental facts. Some at-
room temperature. Considering the inharmonic nature tribute the larger values of kH /kD to a tunneling effect
of the vibration mode that introduces the isotope ef- (e.g., [158]). Schwab [159] reviewed the application
fect, this assumption may be questioned. For example, of a tunnel effect to heterogeneous catalysis as early
if the Morse potential-energy function for deuterium as 1937. While the tunnel effect can be adequately ac-
broadens more rapidly than for hydrogen as the tem- counted for in some instances (e.g., radioactive decay
perature increases, the difference in the ground state and semiconductors), the theory lacks a quantitative
energies between C–H and C–D would become larger, formulation for application in catalysis even today.
and the kinetic isotope effect would be larger than cal- The data collected in this study and in our earlier
culated above. However, this does not appear to be a competitive conversions of naphthene/alkane mixtures
likely explanation. in the pressure range of 100–400 psig [160] are consis-
The energy difference between gaseous H2 and D2 tent with the chemisorption step(s) having a hydrogen
is 1.8 kcal but the difference between H and D ad- dependence (Fig. 31). This figure has similarities with
sorbed on nickel is 2.4–2.8 kcal/mol; i.e., the heat of the one showing the change in relative activation en-
adsorption of D2 is larger than that of H2 by about ergies illustrated by the potential energy curves shown
1 kcal/mol [154]. Any difference in the energy of ad- in Fig. 32 [161]. For adsorption on a clean metal sur-
sorption to form H-Pt and D-Pt have been neglected face, the activation energy to go from the intermedi-
in the calculation of kH /kD using Eq. (5). Should there ate chemisorbed state to the final chemisorbed state
be a difference between the energy for Pt-D and Pt-H, is lower than the one required for desorption. How-
and should it be in the desired direction, 1E in Eq. (5) ever, the potential energy curves change with surface
would be larger and would provide for a larger value coverage and at high coverages (θ = 1) the situation is
of kH /kD . reversed so that conversion to the physically adsorbed
B.H. Davis / Catalysis Today 53 (1999) 443–516 487

occurred in the isobutene formed during the dehydro-


genation, but there was only a small exchange in the
isobutane reactant. These data indicate that adsorption
is essentially irreversible under these high temperature
reaction conditions. The elimination of the second
hydrogen is rapid and reversible in quasi-equilibrium,
allowing for extensive hydrogen-deuterium exchange.
There also seems to be a quasi-equilibrium between
the surface hydrogen and the gaseous deuterium
that allows a significant amount of deuterium on the
catalyst surface. The result for high-temperature de-
hydrogenation of isobutane and the high-temperature
dehydrocyclization of alkanes are therefore consistent.
In summary, there is a significant primary kinetic
isotope associated with the rate limiting step for
dehydrocyclization and for the dehydrogenation of
methylcyclohexane. The results from the competi-
Fig. 32. Schematic of potential energy diagram showing the relative
activation energies for chemisorption of intermediate and final tive conversion of n-octane and methylcyclohexane
chemisorbed states and the variation with surface coverage (from as well as the small amount of H/D exchange in the
[161]). unconverted reactants show that the reactants are ir-
reversibly adsorbed on the Pt catalyst at 482◦ C. The
state has a lower activation energy than the step lead- large value of kH /kD indicates that factor(s) in addi-
ing to the final chemisorbed state. tion to the normal primary kinetic isotope effect must
For the conversion of a mixture of n-octane and be involved in the reaction.
methylcyclohexane at 1 atm, both reactants behave as
though the potential energy curve corresponds to the
case of θ = 0 in Fig. 31. The lower activation energy 8. Triene intermediate
leading to the chemisorbed state of a carbene (Fig. 31)
has a lower activation energy than does desorption The mechanism for dehydrocyclization has been
to the gas phase for the run at even 1 atm. However, proposed by many to involve a series of dehydro-
the change in selectivity to favor the conversion of genation steps and isotopic tracer studies that involve
the naphthene at 100–400 psig (8 : 1 H2 : naphthene) is the addition of a labeled potential intermediate have
consistent with a potential energy curve where the con- provided data that are considered to support this
version of the alkyl radical to form the carbene (first scheme [166–180]. Thus, for n-octane dehydrocy-
chemisorbed state to second chemisorbed state) has clization this mechanism could be represented as
a higher activation energy than desorption does. This follows (Scheme 28).
hypothesis requires that the unconverted alkane has In the above, a and g refer to adsorbed and gas
undergone exchange with deuterium when that gas is phase species, respectively; =, ==, and === refer to
present at high pressure; unfortunately, this experiment one, two or three double bonds, respectively; CyC6 ==
has not been successfully accomplished at this date. refers to either ethyl- or 1,2-dimethyl-cyclohexadiene;
The dehydrogenation of isobutane with Pt-Sn/Al2 O3 and Ar refers to ethylbenzene or o-xylene. An ex-
also exhibits a kinetic isotope effect even in the ample of an experiment used to confirm the above
presence of hydrogen [162–165]. From the kinetic mechanism involves the conversion of a reaction mix-
analysis, the authors concluded that the rate limiting ture of 14 C-labeled hexene and unlabeled hexane, and
step is the dissociative adsorption of isobutane, and determining the 14 C content of the species involved
the isotopic exchange studies support this view. When in the reaction network in Scheme 28. However, the
deuterium was added to the feed at temperatures irreversible nature of the initial, rate-limiting step
above 550◦ C, a relatively high degree of exchange had precludes the addition of any of the intermediates
488 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 28.

shown in Scheme 28 from providing proof of the cyclization depends on hydrogen pressure so that it
sequential dehydrogenation mechanism. becomes less significant to the overall conversion at
Many results using suspected intermediates with higher pressures. To us, it appears that the KIM re-
isotopic label led to the conclusion that dehydrocy- sults are suggestive of hexatriene intermediates but a
clization occurs through cyclization of the triene in the determination of the amount that this pathway con-
gas phase [166–180]. This method has become known tributes to the aromatics formation must await further
as the ‘Kinetic Isotope Method’ (KIM). More recent experimentation.
results, especially those of Paál, Tetenyi and coworkers
[181–184] and Kazanskii and coworkers [185], sug-
gest that dehydrocyclization occurs by a consecutive 9. Conversion of alkylpentanes
stepwise dehydrogenation pathway through the alkene
and diene to the triene which desorbs to the gas phase There are ample data to show that C5 -ring formation
where it undergoes a C6 thermal cyclization. However, occurs for the low temperature range (about 300◦ C)
the KIM method is especially susceptible to ‘chemi- but, as documented above, C5 -ring formation makes
cal disguises’ [186] and this has not been recognized little or no contribution to the mechanism that leads to
by all workers. aromatic compounds at high (about 480◦ C and higher)
Gál et al. [187] reviewed the application of KIM temperatures. Furthermore, hydrogen/deuterium tracer
and Guczi and Tetenyi [188] considered the applica- studies indicate that irreversible adsorption is a domi-
tion of KIM to dehydrocyclization; thus, we will not nating feature of the higher, but not the lower, tempera-
consider this subject in detail. In their review, Gál ture region. With Pt-Sn-ZSM-5 catalysts, Dessau finds
et al. [187] state that ‘If the rate of equilibration be- that n-pentane can be converted with high selectivity
tween the gas phase and the surface is much larger (80%) to cyclopentane [189]. Thus, it is not a question
than that of the chemical reaction, we can assume that of whether cyclization to a cyclopentane ring struc-
the specific activity of the intermediates will be the ture occurs; rather, it is a matter of defining the role of
same in the gas phase as that on the catalyst surface. cyclopentanes in the formation of aromatic products.
Thus, the specific activities, as measured in the gas Much attention has also been given to the produc-
phase, can be used in the relevant KIM expression’. tion of aromatics from hydrocarbons in which the
Paál and Tetenyi [181–184], in their extensive study longest carbon chain contains five carbons; the dehy-
of dehydrocyclization of hexane–hexene mixtures, felt drocyclization of iso-octane (2,2,4-trimethylpentane)
that they had good reasons to believe that the assump- is an example of such an alkane that has been used.
tion of Langmuir–Hinshelwood kinetics could not be It is obvious that a direct C6 -ring forming mechanism
applied. In spite of this, they used the data to establish cannot apply for such a reactant. Thus, we will not
their view of a dehydrocyclization pathway involving include a further discussion of the lower temperature
a triene intermediate. conversion of alkyl pentanes.
The data of both Isagulyants et al. [146,147] and
Il’in et al. [148] are also consistent with a mechanism
that does not have adsorption–desorption equilibrium 10. Atmospheric vs. reforming pressures
of the reactants. Dautzenberg and Platteeuw [118]
concluded that cyclization of hexane occurred by (i) a
Silvestri and Smith [190] determined the effect of
platinum catalyzed six-ring closure and (ii) a thermal
hydrogen partial pressure on n-hexane dehydrocy-
six-ring closure of hexatrienes; however, the thermal
clization activity of Pt-carbon and TeNaX catalysts.
B.H. Davis / Catalysis Today 53 (1999) 443–516 489

1/2
clization rate depending on PH , as expected for an
irreversible chemisorption mechanism.
Results with n-octane enable us to make a connec-
tion between atmospheric pressure operations and the
reforming process which is carried out at pressures
in the 60–400 psig (0.41–2.72 MPa) range [191]. A
nearly equilibrium aromatic composition is obtained
with a Pt on acidic alumina catalyst operating at
400 psig (2.72 MPa) (Fig. 35). For Pt : Sn = 4 : 1 (and
for Pt) on nonacidic alumina, at 200 psig (1.36 MPa)
the aromatic products are only those allowed by direct
C6 -ring formation to produce equal amounts of ethyl-
benzene and o-xylene (Fig. 36). For a Pt : Sn = 1 : 4
nonacidic alumina catalyst, the aromatics formed at
the early time on stream were nearly the equilib-
rium composition. By the seventh hour on stream,
the aromatics were greater than 90% of the isomers
allowed by direct C6 ring closure. In addition the
o-xylene : ethylbenzene ratio was 2 : 1 (Fig. 37) just as
it was with this catalyst at atmospheric pressure. The
conversion of 3-methylheptane following the run with
n-octane produces an aromatic distribution that is es-
sentially the same as 1 atm, and are consistent with a
direct six-carbon ring formation. With these catalysts,
the traces of benzene and toluene in the liquid prod-
ucts could not have altered the aromatic composition
significantly even if they were both derived from
only one of the C8 -aromatic products. Consequently,
the aromatic selectivity is the same at atmospheric
pressure and at 200–400 psig (1.37–2.75 MPa) ; thus,
the aromatic selectivity data obtained at atmospheric
Fig. 33. Relative dehydrocyclization activities for Te/NaX (䊊)
and Cr2 O3 /Al2 O3 (䊉) catalysts as a function of hydrogen partial
pressure with monofunctional catalysts should be ap-
pressure (from [190]). plicable at reforming conditions. Unfortunately, the
superiority of the Pt-Sn catalyst in activity and cata-
lyst life was not as pronounced in the 200–400 psig
The relative rate for aromatization of both cata- (1.37–2.75 MPa) range as at atmospheric pressure.
lysts had a common dependence on hydrogen partial On the other hand, Callender et al. [192] ob-
pressure (Fig. 33). The authors developed a rate ex- tained results with Pt-SiO2 at 15 atm which, when
pression based upon a consecutive dehydrogenation extrapolated to zero residence time, suggested that
pathway that leads to heptatrienes and their subse- C5 -cyclization was responsible for at least 90%
quent cyclization that depended on the third power of the total cyclization for n-heptane. They felt it
of hydrogen. As shown in Fig. 34, the authors fit the likely, following Keulemans and Voge’s [193] re-
highest two data points to a line with a slope of -3 sults with an acidic Pt-Al2 O3 catalyst, that the
as required for the triene mechanism. However, if toluene resulted from ethylcyclopentane rather than
instead of the two highest data points, one uses the 1,2-dimethylcyclopentane. Csicsery and Burnett [97],
two lowest partial pressure points, one obtains a slope using a Pt-SiO2 catalyst at 18 atm, found that for the
of −0.71. The value for the two lowest points shows conversion of 14 C-labeled 1,3-dimethylcyclopentane,
reasonable agreement with the relative dehydrocy- ring expansion to toluene could account for only
490 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 34. Log–log plot of dehydrocyclization rate for Te/NaX catalyst as a function of hydrogen partial pressure (redrawn from [190]).

Scheme 29.

a maximum of about 10% of the total naphthene occurred on the metal function; hence, if ring ex-
conversion. Csicsery and Burnett reported that the pansion of 1,2-dimethylcyclopentane was catalyzed
conversion to aromatics was less than the amount by the metal function it should have occurred with
converted to lower weight products. Ring open- 1,3-dimethylcyclopentane during the study by Csic-
ing of 1,3-dimethylcyclopentane was nonselec- sery and Burnett. The results of Davis [135] and
tive to form 2,4-dimethylpentane, 2-methylhexane Gault [194] with n-[4-13 C] heptane eliminate ethylcy-
and 3-methylhexane; these products accounted for clopentane as a major reaction pathway intermediate
greater than 50% of the total conversion. There at atmospheric pressure. The results with n-octane at
seems to be little doubt but what hydrogenolysis 400 psig (2.75 MPa)and 482◦ C are inconsistent with
B.H. Davis / Catalysis Today 53 (1999) 443–516 491

Fig. 37. C8 -aromatic distribution from the conversion of n-octane


Fig. 35. C8 -aromatic distribution from the dehydrocycliza- (left) over a Pt : Sn = 1 : 4 (conditions same as for Fig. 36) and
tion of n-octane over Pt on acidic alumina (500 psi; 3-methylheptane following overnight flush with nitrogen (EB,
H2 : hydrocarbon = 10 : 1; temp., 482◦ C) (EB, ethylbenzene; OX, ethylbenzene; OX, o-xylene; MX, m-xylene; PX, p-xylene).
o-xylene; MX, m-xylene; PX, p-xylene) (from [191]).

400 psi (1.37 and 2.75 MPa) was insignificant com-


pared to the amount of o-xylene [191]. Hence, we
must conclude that: (a) a cyclopentane intermediate
does not contribute to the aromatics formation over the
metal function or (b) an extremely unique ring expan-
sion is operating which allows pathway to o-xylene
but excludes the pathway to m-xylene in Scheme 29.
The results from the conversion of an alkane with a
quaternary carbon are not the same at atmospheric and
reforming pressure. At atmospheric pressure methyl
migration from the quaternary carbon occurred to pro-
duce C9 -aromatics as well as C8 -aromatics resulting
from demethylation. At 400 psig (2.75 MPa) we did
not obtain C9 -aromatics from 2,2,5-trimethylhexane;
only C8 -aromatics resulting from demethylation were
Fig. 36. C8 -aromatic distribution from the conversion of n-octane obtained.
over a Pt : Sn (4 : 1) on nonacidic alumina catalyst (200 psi;
H2 : hydrocarbon = 10 : 1; temp., 482◦ C) (EB, ethylbenzene; OX,
o-xylene; MX, m-xylene; PX, p-xylene) (from [191]).
10.1. Hydrogen pressure effects

At all temperatures, surface hydrogen concentration


a cyclopentane intermediate, as outlined in following must play an important role in alkane conversion. Sin-
the Scheme 29. felt [195] reported that the rate of hydrogenolysis of
1-Ethyl-2-methylcyclopentane, which should yield ethane could be expressed as a simple power law:
at least as much m-xylene as o-xylene, must contribute r = kPEn PHm . (6)
to any C5 -cyclization mechanism for n-octane because
propylcyclopentane cannot lead directly to o-xylene Cimino et al. [68] had found this equation to apply
by ring expansion. But the amount of m-xylene ob- for an iron catalyst. Approximate values of n and
tained with Pt-Al2 O3 -K at atmospheric and 200 and m for Pt-SiO2 and Pt-Al2 O3 were 0.7 to 0.9 and
492 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 38. Dependence of the rate of hydrogenolysis on H2 pressure


at 608 ± 1 K using Pt/Al2 O3 (5): 䊊, C2 H6 ; 䊐, C3 H8 ; 4, n-C4 H10 .
PA = 0.071 atm (from [196]).

−1.7 to −1.8, respectively. Thus, the rate increases


with increasing ethane partial pressure but decreases
markedly with increasing hydrogen pressure. This
was interpreted in terms of a mechanism that involves
partial, or even complete, dehydrogenation of the ad- Fig. 39. Dependence of the rate of n-C4 C10 hydrogenolysis on H2
sorbed ethane and then the subsequent hydrogenation pressure at various temperatures using Pt/Al2 O3 (from [196]).
of the carbon residue to produce methane. In this
instance, the slow step is considered to be the break-
ing of the C–C bond. While the rates were similar modifies the initial power law to
for the two supports, the activation energies differed kPA PHx
widely and this was considered to indicate a strong r= 2 (7)
interaction between the support and platinum. Similar PA + k2 PHx+1.2
results have been reported by subsequent workers. For
example, Bond [196] summarizes work with C2 –C4 While the hydrogen dependence on the kinetics and
alkanes, showing that the rate initially increases as the mechanistic considerations for alkane hydrogenolysis
hydrogen partial pressure increases (Fig. 38), attains is now well established, they are of secondary concern
a maximum and then decreases with further increases for dehydrocyclization.
in hydrogen partial pressure. The hydrogen partial As we have shown, the nature of alkane adsorp-
pressure effect is also temperature dependent with tion changes from reversible at lower temperatures
the maximum rate shifting to higher hydrogen partial to become essentially irreversible at higher temper-
pressure with increasing temperature (Fig. 39). atures where dehydrocyclization occurs at a reason-
In Bond’s recent summary, the kinetics includes able rate. Thus, while the mechanistic understanding
Langmuir–Hinshelwood methodology for the adsorp- of the effect of hydrogen partial pressure described
tion of both the alkane (A) and hydrogen (H); this for alkane hydrogenolysis is of interest, it does not
B.H. Davis / Catalysis Today 53 (1999) 443–516 493

Fig. 41. Hydrogen coverage (monolayer) as a function of temper-


ature at 1.3, 13, 131 Pa, 1.3, 13, 26.3, 53 kPa (from [198]).

Fig. 40. Amounts of adsorbed hydrogen measured by isotopic


transient kinetic titration over 5.3 mg of EUROPT 1 reduced at
773 K in the absence of ethane (squares) or under hydrogenolysis
conditions (triangles) (from [197]).

appear possible to include it in a consideration of


the dehydrocyclization mechanism which does not in-
volve Langmuir–Hinshelwood reversible adsorption
methodology.
Martin et al. [197] used an isotopic transient kinetic
technique to measure the hydrogen surface coverage
in the presence and absence of ethane. They found Fig. 42. Isobar measured at 7 × 10−8 Torr at pressure
10−3 –10−4 Torr is the isobar shifted by about 100◦ C to higher
a similar hydrogen surface coverage for the EU-
temperatures. Metal with a heat of adsorption higher than that of
ROPT1 (Pt-SiO2 ) catalyst over the temperature range Pt, 1 shift the curves to the right (from [199]).
400–800 K whether ethane was present or absent
(Fig. 40). Furthermore, the amount of adsorbed hy-
drogen from the isotopic transient kinetic technique agreement with those reported for measurements using
agreed closely with those for data obtained using a a Pt filament and a pressure of 7 × 10−8 Torr [199]
volumetric technique with the same catalyst [198]. (Fig. 42).
For the static volumetric conditions with 0.13 kPa hy- Iglesia et al. [168] considered the kinetic coupling
drogen partial pressure, the amounts adsorbed at 227 and hydrogen surface fugacity for the conversion of
and 277◦ C was 88 and 75 ␮mol/g-cat. compared to n-hexane with Te/NaX, H-ZSM-5 and Ga/H-ZSM-5
the values of 94 and 67 ␮mol/g-cat. from the isotopic catalysts. These authors propose that the rate-limiting
transient kinetic technique. The authors found that α, hydrogen desorption steps, and the high surface hy-
the coefficient of Temkin’s law, drogen fugacities that result, control the rate and se-
θH ≈ αlogPH , (8) lectivity of dehydrogenation and related reactions on
many nonmetal surfaces. They found that competitive
was nearly constant over the range 227–527◦ C.
At reactions of 3-methylcyclohex-1-ene (about 20 mol%)
pressures in the range of 50 kPa, the surface coverage and [1-13 C]methylcyclohexane showed that the ini-
by hydrogen, θ H , is about 0.5 of a monolayer at 527◦ C tial toluene products are unlabeled and concluded
(Fig. 41). The data shown in Fig. 41 are in qualitative that methylcyclohexenes are required intermediates in
494 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 43. Rate of dehydrocyclization versus hydrogen pressure Fig. 44. Rate of dehydrocyclization versus hydrogen pres-
(T = 470◦ C, PHC = 2.2 bar, Pt/Al2 O3 , ro = rate at PH2 = 18 bar) (from sure for various hydrocarbon pressures (T = 470◦ C, PHC = 2.2 bar,
[200]). Pt/AL2 O3 , C = 1.5 wt.%, PHC (bar): N 0.5, 䊊 1, 䊉 5) (from [201]).

toluene formation. However, as indicated above, and


as found by Shi and Davis [151] for the conversion of
1-octene and deuterium-labeled n-octane, these results
are also consistent with the adsorption of the alkene
being much more rapid than that of the alkane and
that the initial products are formed exclusively from
the alkene due to adsorption kinetics, and not due to
surface mechanism considerations. Likewise, the con-
clusion reported by Iglesia et al. [168] does not appear
to agree with that of Price et al. [152] who found that
added hydrogen or deuterium did not impact the rate
of dehydrocyclization with a similar catalyst.
Rohrer et al. [200] found that the rate for n-heptane
dehydrocyclization increased with increasing hydro-
gen pressure up to a maximum value, and beyond
this value the rate fell off with increasing hydrogen
pressure (Fig. 43). The isomerization rate underwent Fig. 45. Rate of dehydrocyclization versus hydrocarbon pressure
a similar trend at 471◦ C but was less obvious at the for various hydrogen pressures (T = 470◦ C, Pt/Al2 O3 , C = 15 wt.%)
higher temperature of 527◦ C. Bournonville and Franck (from [201]).
[201] show figures from a thesis [202] that show the
same trend in n-heptane dehydrocyclization rate ver- tion at the maximum increases with carbon number
sus hydrogen pressure as was obtained by Rohrer et so that n-decane is converted about four times more
al. (Fig. 44). In addition, these authors report that the rapidly than n-hexane. Bournonville and Frank [201]
hydrogen partial pressure that corresponds to the max- consider three reaction pathways for dehydrocycliza-
imum rate shifts to higher pressure with increases in tion: metallic monofunctional path, metallic bifunc-
the partial pressure of n-heptane (Fig. 45). The rate tional path and acid bifunctional path. They consider
maximum occurs at a common hydrogen partial pres- the metallic monofunctional path takes place on the
sure as the carbon number of the n-alkane increases metallic phase only and involves a cyclohexanic in-
from 6 to 10; however the rate of dehydrocycliza- termediate which very rapidly dehydrogenates to the
B.H. Davis / Catalysis Today 53 (1999) 443–516 495

Fig. 46. The variation of n-octane conversion at atmospheric pres-


sure with increasing tin loading using a nonacidic alumina support Fig. 47. n-Octane conversion with increasing Sn/Pt ratio us-
(point at right is for Pt on SnO2 only) (from [203]). ing an acidic alumina support (482◦ C, 7.8 atm, LHSV = 1.8, and
H2 /n-octane = 3/1) (from [203]).

aromatic. In their view, the metallic bifunctional path


involves the cyclization to the cyclopentanic interme-
diate which undergoes ring expansion on the acid func-
tion before dehydrogenating. The bifunctional acid
path involves the cyclization of olefins formed over the
metal function. These authors consider cyclization to
be the probable rate determining step and that equilib-
rium between adsorption and desorption is very rapid.
At atmospheric pressure tin added to a Pt-nonacidic
alumina catalyst acts as a promoter, producing the
maximum activity at about Sn/Pt = 4 (Fig. 46) [203].
However, when tin is added to a Pt-acidic alumina cat-
alyst, for the conversion of n-octane at 482◦ C, 7.8 atm
(0.79 MPa), and H2 /n-octane = 3, tin acts as a poison
(Fig. 47). The tin, in this case, eliminates acid sites of
the support and, since the bifunctional pathway involv-
ing cyclization at the acid site is about 30 times the rate
of the monofunctional metal pathway [12], the conver-
sion of n-octane declines with loss of catalyst acidity.
The ratio methylheptanes/aromatics in the products Fig. 48. The methylheptanes/aromatics ratio in products from de-
hydrocyclization of n-octane at 482◦ C and 400 psig (2.75 MPa)
from the conversion of n-octane with Pt-nonacidic
for Pt- and Pt-Sn-alumina catalysts (from [160]).
alumina catalysts and with Pt-Sn-nonacidic alumina
catalysts depend upon the total pressure (the hydrocar-
bon/hydrogen mole ratio remained constant) (Fig. 48) the ‘metal monofunctional isomerization’ pathway,
[160]. Thus, for the high pressure naphtha reform- presumably through the hydrogenolysis of alkylcy-
ing (400 psig, 2.75 MPa), about 3–4 times as much clopentanes that are formed in competition with the
methylheptane isomers are formed as C8 -aromatics. direct six-carbon ring formation pathway. At lower
Since the support for each of the three catalysts conversion levels of n-octane, the C8 -aromatics are es-
is nonacidic alumina, the reaction pathway is pre- sentially the same as formed at atmospheric pressure
dominantly the monofunctional metal pathway. This and are only the two expected for direct six-carbon
means that the addition of tin decreases dramatically ring formation. However, the incorporation of tin
496 B.H. Davis / Catalysis Today 53 (1999) 443–516

Table 28
Aromatic products from n-octane dehydrocyclization with Pt-Sn-nonacidic alumina catalysts (from [203])
Catalyst Pressure Aromatics

Ethylbenzene o-xylene m-xylene p-xylene

Pt- 1 atm 48 50 1.2 1.0


Pt-Sn = 3 : 8 1 atm 32 59 5.0 2.4
Pt 7.8 atm 45 48 7.0 Trace
Pt-Sn = 1 : 3 7.8 atm 38 62 Trace Trace

alters the ratio of ethylbenzene : o-xylene from 1 : 1 they formed a ‘template’ for the spacings of the hy-
without tin to 1 : 2 for Pt/Sn = 3 (Table 28) [203]. drogen atoms in the cyclohexane reactant. Only those
In summary, hydrogen partial pressure plays an metals that were of the proper size and formed the
important role in naphtha reforming and in dehydro- proper crystal face to correspond to the hydrogen spac-
cyclization. In most instances, these effects have been ings in cyclohexane would be active catalysts and data
determined at low (1 atm and lower) pressures and were assimilated in the 1930s to support this view. A
are therefore difficult to extend to normal reforming doublet site may operate in some cases so that it could
conditions. Alkane hydrogenolysis decreases with be possible, in principle, to dehydrogenate n-heptane
increasing hydrogen partial pressure so that, in com- to heptenes on doublet sites even though cyclization
parison to other reactions, it makes a small relative could not be accomplished with the doublet site.
contribution during naphtha reforming. Hydrogen Balandin based his model on a surface that was
partial pressure also impacts the rate of dehydrocy- identical to the bulk metal in structure and lattice spac-
clization negatively, and this is true for catalysts with ing. However, recent results suggest that this is not the
acidic or nonacidic supports. With the acidic support, case but that, in the case of the small metal particles
cyclization occurs predominantly by an acid bifunc- in the range of tens of angstroms (ca. 5 nm for Pt),
tional reaction pathway so that increases in pressure the natural structure may be that of a five-fold sym-
(above about 5 atm) apparently decreases the rate by metry structure [205,206]. These small metal particles
decreasing the ‘pseudo steady state alkene concentra- with the icosahedral surface planes look like closely
tion’. Since similar aromatic isomer compositions are packed [111] planes. Thus, whereas normal fcc struc-
obtained at low (1 atm; 0.101 MPa) and high (27 atm; tures have [110] planes for surfaces, the icosahedral
2.73 MPa) pressures for the monofunctional metal structure provides the [111] planes with the greater
pathway, hydrogen partial pressure does not alter number of nearest neighbor bonds which lowers the
the metal catalyzed cyclization pathway. However, energy of the system. More surprising, the experimen-
increasing hydrogen partial pressure decreases the tal observations indicate that the nearest neighbor dis-
rate of dehydrocyclization so that it must function tance is smaller in the surface layer than that of the
to decrease the concentration of the reactive surface bulk material [207].
intermediate(s) in the dehydrocyclization reaction Joyner et al. [208] found n-heptane dehydrocycliza-
pathway. Unfortunately, much remains to be defined tion did not occur on the [111] crystal face of Pt and
with respect to the specific roles of hydrogen in the only occurred on the suitable stepped Pt surface with
dehydrocyclization reaction pathway(s). an appropriate atomic structure, terrace width and step
orientation which was covered with an ordered layer
of carbonaceous deposit template. If the ordered layer
11. Catalytic site failed to form on account of the presence of surface im-
purities the dehydrocyclization reaction did not occur.
The sextet model proposed by Balandin was sim- Thus, it appears that the present view of small parti-
ple and appealing [204]. The model required six va- cles and the large single crystal studies are in conflict
cant catalytic sites for cyclohexane dehydrogenation. as it concerns dehydrocyclization. On the one hand,
These six sites were required to be arranged so that small Pt crystallites are active dehydrocyclization
B.H. Davis / Catalysis Today 53 (1999) 443–516 497

catalysts and, according to [205,206], should expose can support a mechanism but cannot ‘prove’ a mech-
the (1 1 1) face; on the other hand, the (1 1 1) crys- anistic pathway. On the other hand, the mechanism
tal face of large Pt single crystals is not active for does not even have to be consistent with hypothesized
dehydrocyclization. pathways.
The results of Blakely and Somorjai [209] for At the molecular level, dehydrocyclization may be
cyclohexane show a product distribution of ethy- viewed as a chemical reaction between a catalytic site
lene : propylene : cyclohexene : benzene of 10 : 1 : 0.5 : and an alkane. A valid and complete mechanism must
1; this corresponds to a selectivity for benzene of less include a description of the catalytic site. At present,
than 20%. The hydrogenolysis reaction decreased the nature of the catalytic site is poorly defined. This
the selectivity further so that with an equal turnover is true even for the monofunctional metallic dehy-
number for dehydrogenation and hydrogenolysis at drocyclization pathway. The best descriptions of the
about 7 × 1013 kink atoms/cm2 , the dehydrocycliza- catalytic site come from studies of Pt single crystal
tion selectivity is about 10%. This is a very poor faces. While instrumental techniques, such as LEED,
dehydrogenation selectivity compared to supported allow one to accurately define the surface features,
Pt catalysts since it would be unusual to find a ben- this method suffers from two distinct disadvantages:
zene selectivity at atmospheric or reforming pres- (1) the reaction is conducted at very low pressures
sure of less than 80%. At 200 psig (137 MPa) and and (2) the results are obtained with Pt particles that
H2 : hydrocarbon 9 : 1, we found that 14 C-labeled are enormously large in comparison to those found
methylcyclohexane, added to a naphtha fraction, was in supported commercial catalysts. The low pressures
converted to aromatics with a selectivity greater than provide reaction pathways to produce surface species
97% with the nonacidic catalyst and even 91% when that are strongly dehydrogenated; this is illustrated
we used an acidic commercial catalyst [108]. The by the picture of the surface which is nearly covered
data suggest that the small particles have a selectivity by carbon or carbon-rich islands (Fig. 49) [18]. Hy-
that differs from high loading or bulk metal catalysts. drogenolysis and cracking become significant reaction
Much work remains to be done to connect the prop- pathways with large Pt particles relative to the rates
erties of small and massive Pt metal particles for both of these two reactions with very small Pt particles.
catalytic activity and selectivity. Thus, the benzene selectivity reported by Somorjai
and coworkers (e.g., [18]) is very low compared to
that obtained with a commercial Pt-alumina catalyst
12. Discussion where Pt is present in small particles. While the data
generated with the single crystal catalysts are of in-
The mechanism for alkane dehydrocyclization is terest, it is not certain how to directly extrapolate the
very complex and is composed of a multitude of indi- data to the small particles present in supported naph-
vidual steps with a variety of intermediate compounds. tha reforming catalysts.
One of the complicating features of many of the mech- The aromatic products formed from alkanes with at
anisms advanced to date has been the desire to include least one six-carbon chain are those that are predicted
all reaction products within a single mechanism. In for a mechanism that involves direct six-carbon ring
some instances, this desire has led to the formulation of formation. For many C8 –C10 alkane isomers, 85%,
a complex intermediate structure, such as a bicyclic in- and usually more than 85%, of the aromatics are those
termediate, to explain the formation of minor products. expected for direct six-carbon ring formation. For the
In discussing reaction mechanisms, one should keep dehydrocyclization of [1-14 C]- and [4-14 C]-n-heptane,
in mind that there are features which are experimen- 80%, and usually greater than 80%, of the 14 C appears
tally verifiable and hypotheses which are consistent at positions predicted for a direct six-carbon ring for-
with the experimental observation but are not directly mation. Furthermore, the distribution of 14 C within
proven by the data. The latter are analogous to kinet- the ring precludes a significant contribution by ei-
ics. A ‘correct’ mechanism must be consistent with ther alkylcyclopentane or cycloheptane intermediates.
the kinetic data but even the best kinetic data will not Similar results for dehydrocyclization of 14 C-labeled
limit the data to a single mechanism. Thus, kinetic data alkanes were obtained with both metal and metal oxide
498 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 49. In this model for the working structure and composition of platinum reforming catalysts most of the surface is continuously covered
by a strongly bound carbonaceous deposit whose structure varies from two- to three-dimensional with increasing reaction temperature.
Uncovered patches or ensembles of platinum surface sites always exist in the presence of this carbonaceous deposit. Bound breaking and
chemical rearrangement in reacting hydrocarbon molecules take place readily at these uncovered sites (from [18]).

catalysts, provided a nonacidic support was used. The ried out at low (about 300◦ C) temperatures and atmo-
only exception to this result was at early reaction times spheric pressure. Furthermore, these five-carbon rings
with a chromia-alumina catalyst that contained potas- can undergo hydrogenolysis to produce isomeric alka-
sium and/or other alkali metals and here the results are nes with the same carbon number. The hydrogenolysis
consistent with a bifunctional pathway that involves using highly dispersed Pt catalysts occurs predomi-
both the metal and acid functions. At total reaction nantly, or only, by a nonselective mechanism wherein
pressures up to 400 psig (2.75 MPa), the aromatic the C–C bonds are broken on a statistical basis. With
products are those expected for direct six-carbon ring methylcyclopentane, however, the methyl-ring C–C
formation. Furthermore, the aromatic product distri- bond does not undergo hydrogenolysis at a com-
bution when operating at high pressures is the same as parable rate to the other five C–C bonds. While a
obtained at atmospheric pressure, indicating that the few papers have reported a few percent of cyclopen-
monofunctional metal dehydrocyclization pathway tane in the products, it is much less than the 12.5%
leading to aromatics is the same at low and high pres- expected if all C–C bonds in methylcyclopentane
sures. The conclusion is that the aromatics are formed underwent hydrogenolysis at the same rate. In most
in the metal monofunctional pathway at temperatures papers reporting products from the hydrogenolysis of
of about 480◦ C and higher only by a reaction pathway methylcyclopentane, the amount of cyclopentane is
that includes direct six-carbon ring formation. insignificant, or even so small that it is not reported.
There is strong evidence to support the formation With n-propylcyclopentane, the situation appears
of a five-carbon ring when dehydrocyclization is car- to be different with each C–C bond undergoing
B.H. Davis / Catalysis Today 53 (1999) 443–516 499

Scheme 30.

hydrogenolysis at about the same rate when a Pt-KL ane structures to aromatics means that the final prod-
zeolite catalyst is used at 100 psig (0.69 MPa) and uct will be aromatics. However, the selectivity for ring
482◦ C. With alkylcyclopentanes and Pt-silica or size will not be impacted by the subsequent dehydro-
Pt-nonacidic alumina catalysts, ring enlargement to genation to aromatics unless the two ring structures
aromatic products is much slower than hydrogenolysis are themselves interchanging, as is the case with the
to produce isomeric acyclic alkane isomers. bifunctional pathway that includes an acid function.
The formation of alkylcyclopentanes and their sub- However, with the monofunctional metal catalyzed cy-
sequent hydrogenolysis provides a monofunctional clization, where the five- and six-carbon ring struc-
pathway for alkane isomerization. When operating tures are not at equilibrium or even interconverting,
at 482◦ C and 1 atm (0.101 MPa), significant amounts the selectivity can be impacted by secondary reactions:
of neither alkylcyclopentanes nor isomerized alkanes dehydrogenation of the six-carbon ring compounds to
are observed. On the other hand, the monofunctional aromatics and hydrogenolysis of the five-carbon ring
metal catalyst produces isoalkanes when operating at compounds to acyclic alkanes. Thus, with the mono-
482◦ C and 400 psig (2.75 MPa), and the amount of functional metal catalysts, dehydrocyclization to form
isoalkanes may exceed the amount of aromatics. The
simplest mechanistic pathway for monofunctional
metal catalyzed isomerization is the formation of
alkylcyclopentanes and their subsequent hydrogenol-
ysis. Thus, we have proposed that both five- and
six-carbon ring formation occurs at both 1 and 27 atm
(0.101 and 2.75 MPa) conditions. The difference be-
tween these two pressures is the fate of the five-carbon
ring compounds. At 27 atm (2.75 MPa), the dominant
pathway is hydrogenolysis to produce isomers of the
alkane charged. At 1 atm (0.101 MPa) the five-carbon
ring compounds that are formed predominantly un-
dergo dehydrogenation and further reaction to pro-
duce coke; this is supported by the fact that 20–50%
of the carbon passed to the catalyst ends up as coke
on the catalyst. Thus, the following pathways are
proposed for alkane dehydrocyclization at these two
extreme pressures (Scheme 30).
At 482◦ C, there is little difference in the free energy
Fig. 50. Initial probabilities of trapping-mediated dissociative
for the cyclization to the cyclopentane or cyclohex- chemisorption of (a) C3 H8 , (b) CH3 CD2 CH3 and (c) C3 D8 on
ane ring structures; the difference between the two Pt-(110)-(1 × 2) as a function of reciprocal surface temperature
ring structures is that dehydrogenation of cyclohex- (Ppropane = 1.7 × 10−6 Torr) (from [212]).
500 B.H. Davis / Catalysis Today 53 (1999) 443–516

either the five- or six-carbon ring compounds should For the dehydrocyclization of n-octane, the amount
be about equally favored, and the secondary reactions of ethylbenzene should be twice that of o-xylene since
will provide products that are characteristic of each of there are two pathways leading to the former aro-
the naphthene compounds formed as initial products. matic but only one leading to the xylene isomer. Ex-
McMaster and Madix [210] found a kinetic isotope perimentally, with a Pt-silica or Pt-nonacidic alumina
effect for the initial dissociation probabilities of C2 H6 catalyst about equal amounts of these two aromatics
and C2 D6 on a Pt(1 1 1) crystal surface. They utilized are formed. We advanced the view that the o-xylene
supersonic molecular beam techniques and varied the was formed in a larger amount than expected because
average translational energies. They concluded that the the cyclization pathway leading to its formation in-
kinetic isotope effect could be ascribed to the quantum volved only secondary C–H bond breakage whereas
mechanical tunneling of a hydrogen or deuterium atom the formation of ethylbenzene required breaking one
through the barrier to dissociation. secondary and one primary C–H bond. Since it re-
Weinberg [211,212] found a kinetic isotope for quires more energy to break a primary C–H bond
the dissociative adsorption of methane, ethane and than a secondary one, this could explain the formation
propane on Pt and Ir using flow or molecular beam of more o-xylene than expected from statistical con-
techniques. Briefly, a kinetic isotope effect was found siderations. However, the data for the conversion of
for C3 H8 , CH3 CD2 CH3 and C3 D8 (Fig. 50). The dif- 2,2,7,7-tetradeuteriooctane clearly show that this ex-
ferent rates for these three isopotomers show that both planation is not valid [145]. Since the kinetic isotope
primary and secondary C–H (C–D) bonds cleave un- effect for alkane conversion is about 3, the pathway
der the reaction conditions. The authors calculate that leading to o-xylene should be much slower since it in-
the difference between cleavage of primary and sec- volves the breaking of C–D bonds. Experimentally, the
ondary C–H bonds is 425 cal/mol; the pre-exponential amounts of ethylbenzene and o-xylene were the same
factor also favors cleavage of the secondary C–H for the deuterium-labeled and the unlabeled n-octane.
bond. In comparing data for different Pt faces, an Thus, the landing site cannot be the factor that deter-
implication was that the geometrical structure leads mines the aromatic selectivity. This requires that the
to greater differences in the activation energy than carbon attached to Pt rapidly change so that all adsorp-
the electronic difference for Pt and Ir with the same tion positions of n-octane become the same, or nearly
geometric structure. the same prior to the cyclization step. This means that
At this time, the surface science studies clearly de- there must be extensive H/D exchange of the reactant
fine a kinetic isotope effect in the chemisorption of prior to the cyclization step so that both cyclization
small alkanes; however, much remains to be done to pathways have the same kinetic isotope effect.
directly relate this to naphtha reforming conditions. The consecutive dehydrogenation to the -ene, -diene
The large kinetic isotope effect, the nearly equal rate and -triene followed by cyclization was advanced as
of conversion of the alkane and the alkyl-cyclohexane the reaction pathway for the direct six-carbon ring
under competitive conversion conditions, and the lack formation. As described in the triene section above,
of hydrogen/deuterium exchange in the reactants at the early work involved the addition of 14 C labeled
482◦ C and atmospheric pressure conditions are con- intermediates to the alkane. This reaction scheme, as
sistent with a reaction mechanism that includes irre- it has evolved, is illustrated by the version presented
versible adsorption of the reactants. It is realized that by Paál [213] (Scheme 31).
the use of irreversible adsorption is analogous to that Paál indicates that surface unsaturated species are
of an acid catalyst. Just as it is never possible to have true intermediates of aromatization and that those ap-
an acid site in a catalyst without a conjugate base, pearing in the gas phase are the products of surface de-
it is never possible to have a catalytic conversion in hydrogenation and desorption. The desorption should
which the adsorption of the reactant is completely be less likely with increasing unsaturation, as indicated
irreversible. Thus, there is a small amount of H/D by the shorter arrows in Scheme 31. It is true that the
exchange in the reactant under these ‘irreversible ad- gas-phase alkenes become less stable with increasing
sorption’ conditions; however, it is much less extensive unsaturation; however, the aromatic is as unsaturated
than occurs in the aromatic product. as the -triene and it readily desorbs as a product. The
B.H. Davis / Catalysis Today 53 (1999) 443–516 501

Scheme 31.

cis-triene isomer is claimed to easily cyclize but the difference between the equilibrium concentration
trans-triene isomer only cyclizes following isomeriza- of 1,3,5-octatriene and 2,4,6-octatriene. A different
tion to the cis isomer. According to Paál, the surface ethylbenzene : o-xylene ratio is obtained for the Pt
pool ‘remembers’ the original reactant structure; how- and the Pt-Sn catalysts. This shows that the catalyst
ever, this conclusion is not consistent with our results plays a dominant role in defining the primary aro-
with 2,2,7,7-tetradeuterooctane where there was no matic products and this appears possible only if the
kinetic isotope effect for the aromatic products. intermediates that determine the aromatic selectivity
If the results for the kinetic isotope effect and for reside on the catalyst surface during the cyclization
the lack of H/D exchange in the reactants with essen- step. This, together with the increasingly unfavorable
tially equilibration of H/D in the products are accepted, free energy changes as the acyclic compound products
then the initial chemiadsorption step is rate limiting. If become more unsaturated, make it unlikely that gas
this is the case, the addition of 14 C-labeled products, phase trienes could be formed as part of the cycliza-
such as the -ene, -diene, or -triene, cannot provide tion pathway, as has been proposed by some workers.
valid proof for these species as true intermediates in Alkylcyclopentane compounds cannot be formed
the overall reaction pathway. In fact, the competitive by the triene mechanism. While the concentration of
conversion of 1-octene and n-octane-d18 show that the alkylcyclopentane products are low at the high pres-
alkene is preferentially adsorbed at the double bond sure (7 atm; 0.71 MPa), there are large amounts of
and is converted at 10 or more times faster than the acyclic alkanes that could result from the hydrogenol-
alkane. Furthermore, 1-octene does not undergo hy- ysis of the cyclopentane compounds. Furthermore,
drogenation (deuteration) to return to the gas phase as n-pentane is reported to selectively produce cyclopen-
n-alkane in measurable amounts. Thus, while the re- tane as the major product with a nonacidic Pt-zeolite
sults of many early tracer studies are consistent with catalyst [189]. Several cyclization schemes have been
the mechanism as outlined by Paál, the results can- proposed to produce alkylcyclopentanes as initial de-
not be used to confirm the triene mechanism outlined hydrocyclization products; one such scheme is shown
above. in Scheme 32 [214].
The selectivity for the formation of benzene from The bonding to the catalytic site in Scheme 32 is
n-hexane is much smaller than for the formation represented to be unknown. A similar scheme has been
of ethylbenzene and o-xylene from n-octane. There advanced by Bent [215] (Scheme 33).
should be little difference in the equilibrium con- The pathway to produce a six-carbon ring can be
centration of the triene formed from n-hexane and identical to Scheme 32 except that there will be one
n-octane; likewise, there should not be a significant more carbon in the cyclization product (Scheme 34).
502 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 32.

Scheme 33.

Scheme 34.

In both Schemes 32 and 34, the initial cyclic tures in the initial cyclization step is well established.
transition state involves ring sizes that are the most It seems reasonable that the mechanism for the for-
thermodynamically stable ones and which are the mation of these two ring structures include a common
dominant structures formed in reactions used to effect pathway as shown in Scheme 31, and there is no com-
cyclization reactions during organic synthesis. The pelling experimental data to show that this is not the
formation of both the five- and six-carbon cyclic struc- case.
B.H. Davis / Catalysis Today 53 (1999) 443–516 503

the order Pt(100)-(1 × 1) > Pt(100)-(5 × 20) > Pt(111).


Upon heating hydrogen is evolved with exhaustively
dehydrogenation of the alkane. This is interesting
but not directly correlated with dehydrocyclization at
reforming conditions.
Teplyakov et al. [218] have studied the dehydro-
cyclization of 1-hexene to benzene on Cu3 Pt(111).
This alloy was chosen since earlier work had shown
that Cu modified the reactivity and selectivity of Pt
[219–221]. Teplyakov and Bent [222] concluded that
Scheme 35.
the rate-determining step in the dehydrocyclization of
1-hexene is not the dehydrogenation step. Instead, cy-
clization accompanied by the loss of two hydrogen
atoms was considered to be the rate limiting step. The
authors suggest that the common intermediate for the
dehydrocyclization of 1-hexene and 1,3,5-hexatriene
has the structure of hexa-␴-bonded triene. In the TPR
experiment the hexa-␴-bonded hexatriene intermedi-
ate converts to benzene at 132◦ C, a temperature far
lower than used for dehydrocyclization of alkanes. The
triene is bonded with C–H bonds nearly parallel to the
surface. In agreement with Paál, these authors obtained
data suggesting that the trans-triene formed carbon on
the catalyst whereas the cis-isomer formed benzene.
Vasquez and Madix [223] studied the conversion
of 1-hexene, hexadienes and hexatrienes with Pd(111)
and H(D)Pd(111). With clean and hydrogen-covered
Pd surfaces, 1-hexene hydrogenation to produce gas
phase hexane was never observed. The conjugated
hexadienes and trienes were hydrogenated to produce
1-hexene, but not n-hexane, at the low temperatures. In
this study, above 127◦ C fragmentation of the adsorbed
Fig. 51. (top) Proposed closed-packed assembly of n-octane on
Pt(100)-(1 × 1). (bottom) Proposed closed-packed assembly of
species and dehydrogenation of these fragments oc-
n-octane on Pt(100)-(5 × 2) (from [217]). curs. Structure-sensitive dehydrocyclization of the hy-
drocarbons leads to desorption-limited evolution of
benzene; cyclization could occur at temperatures as
Surface science studies are playing an increasingly low as 60◦ C.
important role in understanding alkane activation and The low selectivity for benzene at the low temper-
dehydrocyclization. An early attempt at this effort is atures with these large Pt or Pd catalysts make it im-
illustrated for various Pt faces in Scheme 35 [216]. possible to relate the results to high temperature de-
As with many initial efforts, Scheme 35 represents an hydrocyclization conditions.
oversimplified view. The structure and temperature dependence of
Examples of efforts to extend the scheme by Pi- n-hexane skeletal rearrangement was investigated near
mentel [216] have continued. Manner et al. [217] atmospheric pressure over a series of platinum single
have utilized as catalysts the (1 × 1) and (5 × 20) crystal surfaces with well-defined surface structure
surfaces of Pt(100) and Pt(111). They propose a and composition. Davis et al. [72] found that aro-
close-packed assembly of n-octane (Fig. 51). They matization of n-hexane to benzene displayed unique
found that the facility of C–H bond activation was in structure sensitivity in which the rates and selectivities
504 B.H. Davis / Catalysis Today 53 (1999) 443–516

were maximized on platinum surfaces with high


concentrations of [111] microfacets. The rates of
isomerization, C5 -cyclization, and hydrogenolysis
reactions displayed little dependence on platinum
surface structure. Selectivities for n-hexane skeletal
rearrangement varied markedly with temperature and
hydrogen pressure at near atmospheric conditions.
One situation that provides strong, but indirect, sup-
port for a common pathway to produce the five- and
six-carbon ring compounds is the conversion of acyclic
compounds that can produce six-carbon ring structures
with geminal dialkyl groups. We have carried out such
experiments with C9 -alkane isomers and compared the
products to those obtained by converting the alkylcy-
clohexane that would be formed by the cyclization of
the alkane (Scheme 36). The similarity of the products
from the alkylcyclohexane compound and the acyclic
alkane that would form the same alkylcyclohexane by
six-carbon ring formation is striking.
In addition, 2-methylheptane can undergo cycliza-
tion to produce m-xylene and 1,1-dimethylcyclohexane
by a direct six-carbon ring formation. On the other
hand, 3-methylheptane cannot form a geminal
dimethylcyclohexane structure by direct six carbon
ring formation. Conversion of 1,1-dimethylcyclohexane
with a Pt-nonacidic alumina catalyst shows that
demethylation to produce toluene is a dominant re- Scheme 36.
action pathway. When 2-methylheptane is converted
with Pt-SiO2 , Pt-nonacidic alumina or Pt-ZSM-5
zeolite catalysts, significantly more toluene is pro- as the active site to the cyclization reaction and could
duced by 2-methylheptane than is obtained from form the multiple bonding necessary to allow the re-
3-methylheptane at a similar conversion level. The moval of a hydrogen and to incorporate it within the
conversion of C9 -acyclic alkanes produces similar coordination sphere while at the same time forming
results (Scheme 36). Thus, the significant amounts the multiple bonding required to effect cyclization of
of aromatic products that would result from the for- the remaining hydrocarbon radical.
mation of geminal dialkyl cyclohexanes followed If a single Pt site is capable of the cyclization
by dealkylation by metal catalyzed hydrogenolysis scheme, it is difficult to explain the differences in
provide strong evidence that cyclization to produce aromatic selectivity between Pt and Pt-Sn catalysts.
aromatics involves an adsorbed cyclohexane structure It has been proposed that complex reactions, such as
rather than an adsorbed triene structure that is formed hydrogenolysis, are possible only on sites comprised
prior to the cyclization step. of multiple atoms. By forming an alloy, such as PtSn,
Mechanisms such as those in Scheme 22 involve the Pt atoms become isolated and surrounded by Sn
only one transition metal atom, and in this case a Pt atoms, and these isolated atoms cannot effect hy-
atom. There are examples in homogeneous transition drogenolysis as readily as the multiple Pt atom sites.
metal catalysis where multiple bonding to a central Either the widely accepted cyclization and isomeriza-
transition metal occurs that would be as demanding tion mechanism that involves a single transition metal
as required for the cyclization step. Thus, in principle, atom must be abandoned or it must be accepted that
it is concluded that a single Pt atom could function an ensemble of transition metal atoms is not required
B.H. Davis / Catalysis Today 53 (1999) 443–516 505

for cyclization and hydrogenolysis reactions. Accept- activity are due to the deposition of carbon on the cat-
ing the view that a single Pt metal site can effect the alyst, a reaction that was not included in the model.
cyclization step, then the different catalytic properties As has been demonstrated [141,142], the conversion
of Pt and PtSn, and other Pt alloys, must be due to of the alkane at the lower temperature is essentially
electronic, and not geometric, effects. In making this by reversible adsorption whereas at the higher tem-
assertion, it is realized that geometric and electronic perature (365◦ C), irreversible adsorption should at
effects are not independent; rather, the assertion is least be approached. Thus, while the conversion of
made in analogy with acidity of the acid–base pair n-hexane provides a good first effort, the use of large
where one of the conjugate effects is dominant. Thus, Pt crystals and the low temperatures in these studies
it is viewed that the changes in activity and aromatic suggest that the data differ from those expected for
selectivity are impacted more strongly by the mod- the normal naphtha reforming conditions.
ification of the electronic factor of alloying than the Dadyburjor and Ruckenstein [225] utilized the path-
changes in the geometry of the catalytic site. ways shown in Scheme 38 and data for the conver-
sion of 2-methylpentane-2-13 C [226] in their analysis
of the role of multiple landing areas for adsorption.
12.1. Modeling The authors accounted for three types of landing sites
in which: (1) two surface sites are tied up in form-
A trend in catalysis today is the development of ing two carbon-surface bonds, leading to hydrogeno-
models to aid in understanding the reaction mech- lysis products, (2) three surface sites are involved in
anisms. One of these involves a catalyst design bonding with the adsorbate by forming two nonad-
approach based on a kinetic model of the reac- jacent carbon-surface bonds, leading to isomerization
tion/catalyst system to be used in conjunction with products, and (3) five surface sites are used to form
experimental studies. In this approach, the perfor- a cyclic intermediate with one carbon-surface bond,
mance of the catalyst is simulated by automated leading to 13 C scrambled products. Adsorbed hydro-
modeling of proposed chemical reaction mechanisms, gen is considered to be dissociated with each hydro-
and the results of this modeling is used in defining gen being bonded to a single surface site. The relative
the experimental work. The kinetic model is refined amounts of these adsorbed species determine the selec-
using the experimental data. The authors indicate that tivity for the reaction. Using their model, the authors
“. . . modeling in the absence of experimental data were able to obtain agreement between the experi-
may be a mathematical exercise, while experimen- mental data and the model. Initially we proposed that
tation in the absence of a kinetic model may lack the selectivity for the aromatic products obtained from
direction. . . [224]”. For the conversion of n-hexane, the conversion of n-octane was determined by the at-
a reaction scheme utilized 17 elementary reversible tachment of n-octane to the landing site [35]. Since
steps, 12 adsorbed species, and the formation of six two secondary C–H bonds would need to be broken
stable reaction products (Scheme 37). In spite of the to form o-xylene and a secondary and primary C–H
complexity of the reactions in Scheme 37, the au- bond would need to be broken in forming ethylben-
thors caution that “It should be remembered that this zene, we proposed that the higher concentration of
sequence of steps is used for illustrative purposes o-xylene was due to breaking the weaker secondary
only.” They utilized the data generated by Davis et al. C–H bonds. However, subsequent results at 482◦ C us-
[72] for the conversion of n-hexane using the Pt(111) ing 2,2,7,7-tetradeuteriooctane showed that C–H bond
single-crystal platinum catalyst operated in a batch breaking with subsequent H/D exchange was so rapid
reactor at 300 and 365◦ C. The data and the model that the initial adsorption site could not determine the
both show that methylcyclopentane and hydrogenoly- aromatic product selectivity [145]. Thus, it appears
sis products are dominant. The experimental data, but that it is not the landing site but rather the stability
not the model, show the selectivity changes toward of the different adsorbate-catalyst structures that will
benzene and hydrogenolysis products at the higher determine the product selectivity for aromatics forma-
of the two temperatures that were used. Dumesic et tion. Presumably this would also apply to the struc-
al. [224] suggest that the selectivity and decreased tures shown in Scheme 38.
506 B.H. Davis / Catalysis Today 53 (1999) 443–516

Scheme 37.

Zeigarnik and Valdés-Pérez [227] utilized results of any single labeling experiment. First, the
MECHEM, a computer aid for elucidating reaction reaction pathways shown in Scheme 39 are rather
mechanims, to predict the location of an isotope in the limited. For example, it appears more likely that ring
labeled products from the conversion of n-heptane. expansion/contraction between five- and six-carbon
They utilized the four reaction pathways shown in rings for pathway 3, Scheme 39 is more likely than
Scheme 39 in their calculations and obtained the re- the formation of the bicyclic ring structure that is
sults shown in Table 29. They concluded that it was shown in Scheme 39. Pines and coworkers tried to
not possible to distinguish the mechanisms from the establish that pathway 4, Scheme 39, was reasonable
B.H. Davis / Catalysis Today 53 (1999) 443–516 507

Scheme 38.

Table 29
Predicted positions of isotopic labels in products of n-heptane dehydrocyclizationa
Hypothetical mechanism (see Scheme 1) Initial position of the label in n-hepatane
[1-14 C]-n-heptane [2-14 C]-n-heptane [3-14 C]-n-heptane [4-14 C]-n-heptane

1,2

4 All (five) positions All (five) positions All (five) positions All (five) positions
a Solid circles stand for carbon labels.

by converting cycloheptane under the same condi- a much wider range of products. Thus, they could
tions that they used for n-heptane. They found that only explain the results if the adsorption of cyclohep-
the intermediate, cycloheptane, was converted more tane was much slower than that of n-heptane, and we
slowly than n-heptane and, at the same time, provided consider this to be unlikely.
508 B.H. Davis / Catalysis Today 53 (1999) 443–516

by which Pt produces significant amounts of aromat-


ics (e.g., [111,112]).
While there has been discussion as to whether the
conversion of alkanes to aromatics follows the clas-
sical bifunctional mechanism involving both metal
and acid sites, more work needs to be directed to-
ward answering this question. For example, Shum et
al. [228] used a mechanical mixture of a non-acidic
platinum catalyst and acidic alumina, and concluded
that the metal function controls dehydrocyclization.
Sivasanker and Padalkar [229] investigated the de-
hydrocyclization of C6 through C8 alkanes using
Pt-alumina catalysts; however, these authors carried
out their reactions at atmospheric pressure where cat-
alyst aging was rapid. In spite of this, these authors
concluded that the metal function was the critical
parameter with n-hexane and n-heptane; however,
Scheme 39.
both metal function and acid sites could be important
for n-octane. Ako and Susu [230] also investigated
the dehydrocyclization of n-octane with monofunc-
In summary, the modeling of alkane dehydrocy- tional and bifunctional Pt/Al2 O3 catalysts; however,
clization provides stimulating intellectual exercises the total pressure was only 1.8 atm. Callender et al.
but the complexity of the mechanism does not allow [192] expressed the view that Pt catalyzed cyclization
one to include a sufficient number of pathways for through C5 -ring closure and that ring expansion led
the model to provide results that improve upon the to aromatics. They reported that alumina had activity
judicious choice of isotopically labeled reactants. for cyclization of the alkenes produced from dehy-
drogenation but that for dispersed Pt, the metal cat-
12.2. Bifunctional catalysis alyzed cyclization was more rapid. Henningsen [231]
reported that the predominant pathway for the con-
Apart from PtKL zeolite, the commercial naphtha version of n-heptane to toluene is with a bifunctional
reforming catalysts are bifunctional. The foregoing has mechanism, and that the acid catalyzed pathway was
concentrated upon the monofunctional pathway since more rapid.
it is not possible to utilize either the reaction products Dautzenberg and Platteeuw [118] followed the con-
nor the position of isotopic label to understand the re- version of n-hexane with Pt on a non-acidic alumina
action pathway using bifunctional catalysts. Thus, the catalyst. They concluded that dehydrocyclization of
bifunctional aspects of dehydrocyclization are consid- n-hexane to benzene proceeds along two different re-
ered briefly in the following. action paths: (a) Pt-catalyzed ring closure whose con-
Some authors, e.g., [96], were of the view that the tribution to total rate of aromatization is proportional
platinum function alone could catalyze aromatics for- to total Pt surface regardless of metal particle size and
mation by cyclization to a five carbon ring followed (b) thermal six-ring closure of hexatrienes formed by
by ring expansion to a six carbon ring and subsequent dehydrogenation from n-hexanes over the Pt function.
dehydrogenation to aromatics. Others contended that They argued for a monofunctional reaction pathway
direct six-carbon ring formation was the only pathway for Pt that involves both five- and six-ring formation:
B.H. Davis / Catalysis Today 53 (1999) 443–516 509

Fig. 52. The total aromatics in the liquid product vs. C8 -alkane
conversion for catalysts based upon acidic (A) and non-acidic
(NA) supports. (䊊) 0.6 wt.% Pt on precipitated alumina, COP;
Fig. 53. The ratio iso-octanes/aromatics vs. n-octane conversion.
(䊐) 1.0 wt.% Pt on UCI alumina; (N) 1.0 wt.% Pt on silica, SIL;
(䊊) 0.6 wt.% Pt on precipitated alumina, COP; (䊐) 1.0 wt.% Pt
(䊏) 1.0 wt.% Pt on non-acidic alumina, NAA (from [12]).
on UCI alumina; (N) 1.0 wt.% Pt on silica, SIL; (䊏) 1.0 wt.% Pt
on non-acidic alumina, NAA (from [12]).
The conversion of n-octane with monofunctional
and bifunctional catalysts has been investigated
at 100 psig (0.69 MPa) and 482◦ C with a hydro- The aromatic distributions clearly indicate that the
gen/hydrocarbon molar ratio of 8/1. The aromatics in Pt on silica or non-acidic alumina catalysts do not have
the liquid product as a function of n-octane conversion acidity. Even so, iso-octane isomers are formed more
show that isomerization of n-octane to iso-octanes rapidly than aromatics are produced for both of these
is more rapid that aromatics formation for the bi- nonacidic catalysts. Thus, even for these non-acidic
functional catalyst (Fig. 7). Furthermore, the slope of supports, an isomerization pathway must operate. As
the two lines in Fig. 52 show that the bifunctional shown in Fig. 53, C8 -iso-alkanes are produced about
pathway is more rapid that the Pt monofunctional six times faster than aromatics at the 10% n-octane
pathway. In addition, the data in Fig. 7 show that conversion level; at lower n-octane conversion lev-
the conversion of n-octane with the monofunctional els of the line in Fig. 53 suggests that this ratio of
catalyst produces aromatics with only a 6–10% selec- i-C8 /aromatics approaches 8 or higher. For this reason,
tivity; the monofunctional mechanism at reforming the rates in Fig. 7 should be viewed as the rate of Pt
pressures leads primarily to isomerization as well as catalyzed conversion of n-octane, and the selectivity
some cracking products. This selectivity difference is for producing an aromatic product is only about 10%
more apparent in Fig. 52 which shows the aromatics of the total n-octane conversion. Thus, the Pt function
selectivity versus total C8 conversion. Under the re- catalyzes four types of reactions: (1) aromatization by
action conditions utilized for n-octane conversion, the direct six carbon ring formation; (2) isomerization by
alkane products from the conversion of n-propyl- and cyclization to a five-carbon ring structure followed by
1-methyl-2-ethyl-cyclopentane are those predicted hydrogenolysis of the alkylcyclopentane intermediate;
for hydrogenolysis of the cyclopentane ring on a (3) conversion of n-octane to C1 through C7 hydrocar-
statistical, i.e., nonselective basis. bons by hydrogenolysis; and (4) the dehydrogenation
510 B.H. Davis / Catalysis Today 53 (1999) 443–516

of alkanes and naphthenes. Thus, for the Pt catalytic functional pathway dominates in producing aromatic
function the following reaction network applies: products over that of the monofunctional Pt pathway.

In the above scheme, an alkylcyclohexane structure The monofunctional pathway can be represented as
is shown although there is no data in this paper to (Scheme 40):
eliminate the structure being the one obtained from The bifunctional pathway (Scheme 41) for the
cyclization of an octatriene. However, the cyclization production of aromatics may be represented as: (Pt,
to the alkylcyclopentane could not occur through such platinum; A = support acidity; B represents the series
a triene mechanism. conversion involving the three steps Pt, A, Pt).
The formation of more iso-octanes than aromatics In summary, Pt monofunctional and bifunctional
with the single functional Pt catalyst shows that the catalysts produce a common set of products apart from
pathway to produce aromatics by ring expansion of the C8 -aromatic and i-C8 -alkane distributions at low
the cyclopentane ring to a cyclohexane structure does conversions. The difference between the two types of
not occur at a significant rate. This conclusion is in catalysts therefore is essentially that of the rate and se-
conflict with some early claims (e.g., [96]); however, lectivity of producing aromatics versus C1 – C7 prod-
the conclusion agrees with studies with 14 C labeled ucts. Overall, the bifunctional pathway provides both
cyclopentane [97,99]. the most rapid and the most selective pathway for the
In a recent review [232] it is stated that “It has formation of aromatics from n-alkanes. The bifunc-
been shown that the isomerization and hydrocracking tional pathway provides a nearly equilibrium distribu-
reactions require the catalyst to have two separate and tion of methyl heptanes whereas the monofunctional
distinct functions – a hydrogenation–dehydrogenation platinum pathway produces predominantly the iso-
function and an acidity function”. This requirement is mers allowed from hydrogenolysis of alkylcyclopen-
clearly not necessary; isomerization of the n-alkane tanes, primary products formed by the five-carbon ring
can, and does, occur over the metal function through cyclization of the n-alkane. The nonacidic Pt catalyst
C5 -ring formation and subsequent hydrogenolysis, and produces aromatics by direct six-carbon ring forma-
this occurs at a rate that is faster than the formation of tion. The acidic catalyst produces essentially a mixture
aromatics. Likewise, hydrocracking occurs to a greater of C8 -aromatics that is near the thermodynamic equi-
extent with a nonacidic supported catalyst than with librium mixture, presumably because of isomerization
the acid supported catalyst. reactions that proceed metal catalyzed cyclization and
Another feature of the Pt-acidic catalyst is that it by the acid catalyzed cyclization.
produces aromatics at least 10 times the rate of the
Pt-non-acidic catalyst. At the same time, the selec-
tivity for aromatics production from n-octane for the 13. Conclusions
Pt-acidic catalyst is about five times greater than for
the Pt-non-acidic catalyst. With the acidic catalyst, the Dehydrocyclization is a complex reaction with the
bifunctional pathway produces aromatics more rapidly potential for many kinetic disguises. In spite of this,
and more selectively than by the Pt function only there is a vast body of experimental data that support
pathway. Thus, for the Pt on an acidic support the bi- specific mechanistic features.
B.H. Davis / Catalysis Today 53 (1999) 443–516 511

Scheme 40.

Scheme 41.

For the naphtha reforming catalyst, a bifunctional For the monofunctional metal catalyzed pathway,
pathway operates with cyclization at the acid site is the aromatics are produced by a cyclization that di-
more rapid than the monofunctional metal catalyzed rectly produces a six-carbon ring. Cyclization to other
cyclization pathway. size rings followed by ring contraction/expansion does
512 B.H. Davis / Catalysis Today 53 (1999) 443–516

Fig. 54. Schematic reaction coordinate for the conversion of an alkane and H/D exchange at low (l) and high (h) temperature (I, R•
II, Ri; Ed , desorption activation energy; P, precursor physically adsorbed state; Er , activation energy for rate determining step; Ex )l,h
activation energy for low, high temperature desorption of H/D exchange in adsorbed species; Ede )l and Ede )h, activation energy for second
dehydrogenation at low and high temperature, respectively; and Ex,2 )h, activation energy for exchange in second step) (from [143]).

not materially contribute to the production of aromat- lack of significant H/D exchange in the reactants, in-
ics with either the metal or metal oxide catalysts. cluding cyclohexane structures when fed under com-
With the monofunctional pathway, both five- and petitive reaction conditions with n-alkanes, confirm
six-carbon rings are formed as initial products. The this conclusion. Likewise, the nearly equilibrium H/D
six-carbon ring pathway leads directly to aromatic composition of the aromatic products show that once
products. For conversions at normal reforming con- adsorbed, H/D exchange occurs much more rapidly
ditions (60–400 psig (0.4–2.75 MPa)), the five-carbon than aromatics formation (Fig. 54). Furthermore, the
ring undergoes nonselective hydrogenolysis reactions experiments with deuterated reactants show that aro-
to produce acyclic alkane products and these prod- matics desorption is not the rate limiting step.
ucts themselves may subsequently undergo cyclization The mechanism changes from irreversible adsorp-
(Scheme 30, path b, c). At atmospheric conditions the tion at atmospheric pressure to one that is essentially
five-carbon ring compounds undergo coking/cracking a reversible one at 100 psig (0.69 MPa) or higher. This
reactions, and thereby contribute to the rapid aging dependence upon hydrogen pressure is expected for
of the catalyst but not to the production of significant a rate limiting step that involves the initial rupture of
amounts of aromatics (Scheme 30, path b, d). the C–H bond.
The predominant monofunctional metal catalyzed
cyclization step does not occur through a triene inter-
mediate. Similar products are obtained form a gem- Acknowledgements
inal dimethylcyclohexane reactant and from alkanes
that can form this cyclohexane structure by direct The author’s dehydrocyclization studies have been
six-carbon ring formation. Since the acyclic alkanes impacted by many previous investigators. One of them
that contain geminal dimethyl substitution cannot form is the person responsible for this special issue, Dr. John
the triene structure, they must cyclize through some Sinfelt. Initially this influence was indirectly through
other mechanism. his publications dating from the early 1960s and con-
The dominant dehydrocyclization pathway at at- tinuing to the present. Of equal importance has been
mospheric pressure occurs by a mechanism that in- the interesting discussions with John concerning the
volves essentially irreversible adsorption with the ini- history of naphtha reforming as well as catalysis in
tial chemisorption being the rate controlling step. The general.
B.H. Davis / Catalysis Today 53 (1999) 443–516 513

References [34] B.H. Davis, J. Catal. 23 (1971) 340.


[35] B.H. Davis, J. Catal. 42 (1976) 238.
[1] B.H. Davis, Appl. Catal. 115 (1994) N27. [36] K.W. McHenry, R.J. Bertolacini, H.M. Brennan, J.L. Wilson,
[2] B.A. Kazansky, A.F. Plate, Ber. 69 (1936) 1862. H.S. Seelig, Actes Congr. Int. Catal., 2nd, 1960 2 (1961)
[3] A.V. Grosse, U.S. Patent 2,124,566, Applied 30 September 2293.
1936. [37] H.E. Kluksdahl, R.J. Houston, J. Phys. Chem. 65 (1961)
[4] A.V. Grosse, J.C. Morrell, W.J. Mattox, Ind. Eng. Chem. 32 1469.
(1940) 528. [38] M.F.L. Johnson, C.D. Keith, J. Phys. Chem. 67 (1963) 200.
[5] B.L. Moldvasky, H.D. Kamisher, Compt. Rend. Acad. Sci. [39] B.H. Davis, J. Catal. 23 (1971) 355.
U.R.S.S. 1 (1936) 355. [40] B.H. Davis, J. Catal. 42 (1976) 247.
[6] V.I. Karshev, M.G. Severyanova, A.N. Siova, Khimixva [41] T.R. Hughes, R.L. Jacobson, K.R. Gibson, L.G. Schormack,
Tverdogo Topliva Chem. Solid Fuel 8 (1936) 559. J.R. McCabe, Proc. Am. Pet. Inst. Refin. Dept. 45 (1976)
[7] J. Turkevich, Heterogeneous catalysis, selected American 875.
histories, in: B.H. Davis, W.P. Hettinger Jr. (Eds.), ACS [42] J.H. Sinfelt, J. Catal. 29 (1973) 308.
Symp. Series, vol. 222, 1982, p. 463. [43] J.H. Sinfelt, Bimetallic Catalysts, Wiley, New York, 1983.
[8] V. Haensel, M.J. Sterba, Progress in petroleum technology, [44] J.N. Miale, P.B. Weisz, J. Catal. 20 (1971) 288.
ACS Adv. Chem. Series 5 (1951) 60–75. [45] W.H. Lang, R.J. Mikovsky, A.J. Silvestri, J. Catal. 20 (1971)
[9] T.R. Hughes, W.C. Buss, P.W. Tamon, R.L. Jacobson, in: 293.
Y. Murakami, A. Iijima, J.W. Ward (Eds.), Proc. 7th Int. [46] R.J. Mikovsky, A.J. Silvestri, E. Dempsy, D.H. Olsen, J
Zeolite Conf., Elsevier, Amsterdam, 1986, pp. 725–732 Catal. 22 (1971) 371.
[10] T. R. Hughes, R. Jacobson, P. Tamm, in: J. Ward (Ed.), [47] D.H. Olson, R.J. Mikovsky, G.F. Shipman, E. Dempsy, J.
Catalysis, Elsevier, Amsterdam, 1998, pp. 317–333. Catal. 24 (1972) 161.
[11] J.L. Carter, J.A. Cusumano, J.H. Sinfelt, J. Catal. 20 (1971) [48] J.R. Bernard, Proc. 5th Int. Congr. Zeolites, Heyden, London,
223. 1980, p. 686.
[12] R. Srinivasan, D. Sparks, B.H. Davis, J. Mol. Catal. 88 [49] S.J. Tauster, J.J. Steger, J. Catal. 125 (1990) 387.
(1994) 325. [50] R.J. Davis, E.G. Derouane, J. Catal. 132 (1991) 269.
[13] R. Srinivasan, B.H. Davis, J. Mol. Catal. 88 (1994) 343. [51] R.P. Eischens, P.W. Selwood, J. Am. Chem. Soc. 70 (1948)
[14] D. Sparks, R. Srinivasan, B.H. Davis, J. Mol. Catal. 88 2271.
(1994) 359. [52] N.E. Fouad, H. Knözinger, H.M. Ismail, M.I. Zaki, Zeit. für
[15] M. Orchin, L. Reggel, R.A. Friedel, E.O. Woolfolk, Physik. Chem. 173 (1991) 201.
Aromatic Cyclodehydrogenation, U.S. Bureau of Mines, [53] N.E. Fouad, H. Knözinger, Zeit. für Physik. Chem. 171
Technical paper 728, 1948. (1991) 75.
[16] A.H. Steiner, in: P.H. Emmett (Ed.), Catalysis, vol. 4, [54] K. Wlodzimierz, Bull. Polish Acad. Sci. Chem. 32 (1984)
Reinhold, New York, 1957, p 529. 501.
[17] C. Hanch, Chem. Revs. 52 (1953) 353. [55] S.E. Voltz, A.E. Hirshler, A. Smith, Abstract, ACS meeting,
[18] G.A. Somorjai, S.M. Davis, Plat. Metals Rev. 27 (1983) 54. Phys. Inorg. Chem. Sect., p. 64.
[19] P.S. Skell, J.J. Havel, M.J. McGlinchey, Accounts of Chem. [56] B.S. Greensfleder, R.C. Archibald, D.L. Fuller, Chem. Eng.
Res. 6 (1973) 97. Progress 43 (1947) 561.
[20] K.A. Biesiada, C.T. Koch, P.B. Shevlin, J. Chem. Am. Soc. [57] A.S. Russell, R.J. Stokes, Ind. Eng. Chem. 38 (1946) 1071.
102 (1980) 2098. [58] Y. Holl, F. Garin, G. Maire, A. Muller, P.A. Engelhard, J.
[21] A. Sárkány, J. Catal. 89 (1984) 14. Grosmangin, J. Catal. 104 (1987) 225.
[22] R.J. Kokes, A.L. Dent, Advan. Catal. 22 (1972) 1. [59] B.H. Davis, G.J. Antos, in: G.J. Antos, A.M. Aitani,
[23] G.H. Twigg, Trans. Faraday Soc. 35 (1939) 1006. J.M. Parera (Eds.), Catalytic Naphtha Reforming: Science,
[24] A. Farkas, L. Farkas, E.K. Rideal, Proc. Roy. Soc. London Technology, Marcel Dekker, New York, 1995, pp. 113–180
146 (1934) 630. [60] G.B. McVicker, J.L. Kas, J.J. Ziemak, W.E. Gates, J.L.
[25] H. Pines, C.T. Goetschel, J. Org. Chem. 30 (1965) 3530. Robbins, M.M.J. Treacy, S.B. Rice, T.H. Vanderspurt, V.R.
[26] R.C. Pikethly, H. Steiner, Trans Faraday Soc. 35 (1939) 979. Cross, A.K. Ghosh, J. Catal. 139 (1993) 48.
[27] K.M. Sharan, Catal. Rev. Sci. Eng. 26 (1984) 141. [61] M. Vaarkamp, J.T. Miller, F.S. Modica, G.S. Lane, D.C.
[28] J. Happel, Isotopic Assessment of Heterogeneous Catalysis, Koningsberger, J. Catal. 138 (1992) 675.
Academic Press, New York, 1986. [62] G. Jacobs, C.L. Padro, D.E. Resasco, J. Catal. 179 (1998)
[29] G.A. Somorjai, Accounts Chem. Res. 9 (1976) 248. 43.
[30] W.P. Hettinger Jr., personal communication. [63] M.J.P. Botman, L.Q. She, J.Y. Zhang, W.L. Driessen, V.
[31] R.L. Jacobson, H.E. Kluksdahl, C.S. McCoy, R.W. Davis, Ponec, J. Catal. 103 (1987) 280.
Am. Petrol. Inst. Ref. Div. Mtg. Chicago, May 1969. [64] R. Srinivasan, B.H. Davis, Platinum Metals Rev. 36 (1992)
[32] H. Pines, W.O. Haag, J. Am. Chem. Soc. 82 (1960) 2471. 151.
[33] V. Haensel, IV Simposio Iberoamericano di Catalysis, [65] J. Turkevich, H. Young Jr., J. Am. Chem. Soc. 63 (1941)
Mexico City, Plenary Lecture, 1974. 519.
514 B.H. Davis / Catalysis Today 53 (1999) 443–516

[66] G.A. Mills, H. Heinemann, T.H. Milliken, A.G. Oblad, Ind. [102] C. Besoukhanova, J. Guidot, D. Barthomeuf, J. Chem. Soc.
Eng. Chem. 45 (1953) 134. Faraday Trans. I 77 (1981) 1595.
[67] A.J. Silvestri, P.A. Naro, R.L. Smith, J. Catal. 14 (1969) [103] V. Bonaci-Koutecky, J. Koutecky, P. Fantucci, V. Ponec, J.
3860. Catal. 111 (1988) 409.
[68] A. Cimino, M. Boudart, H.S. Taylor, J. Phys. Chem. 58 [104] W.E. Alvarez, D.E. Resasco, Catal. Lett. 8 (1991) 53.
(1954) 796. [105] R. Kramer, H. Zueg, J. Catal. 85 (1984) 530.
[69] C.G. Myers, G.W. Munns Jr., Ind. Eng. Chem. 50 (1958) [106] H. Glassl, K. Hayek, R. Kramer, J. Catal. 68 (1981) 397.
1727. [107] W.E. Alvarez, D.E. Resasco, J. Catal. 164 (1996) 467.
[70] J. R. Anderson, Y. Shimoyama, in: W. Hightower (Ed.), 5th [108] B.H. Davis, J. Catal. 29 (1973) 395.
Int. Congr. Catal. J., Miami Beach, FL, 20–26 August 1972, [109] J. Dermietzel, Proc. 5th Intern. Symp. Heterogeneous
North-Holland, Amsterdam, 1973, pp. 695–715. Catalysis, Varn, 1983, Part I, pp. 452–458.
[71] M. Chow, G.B. McVicker, J. Catal. 112 (1988) 290. [110] E. Iglesia, J. Baumgartner, G.L. Price, K.D. Rose, J.L.
[72] S.M. Davis, F. Zaera, G. Somorjai, J. Catal. 85 (1984) 206. Robins, J. Catal. 125 (1990) 95.
[73] E. Santacessaria, D. Gelosa, S. Carra, I. Adami, Ind. Eng. [111] B.H. Davis, P.B. Venuto, J. Catal. 15 (1969) 363.
Chem. Prod. Res. Dev. 17 (1978) 68. [112] L.-G. Fogelberg, R. Gore, B. Rånby, Acta Chem.
[74] Z. Karpinski, T. Koscielski, J. Catal. 63 (1980) 313. Scandinavica 21 (1967) 2041.
[75] J.R.H. van Schaik, R.P. Dessing, V. Ponec, J. Catal. 38 [113] L.-G. Fogelberg, R. Gore, B. Rånby, Acta Chem.
(1975) 273. Scandinavica 21 (1967) 2050.
[76] A.D. Cinneide, F.G. Gault, J. Catal. 37 (1975) 311. [114] H. Pines, G.-T. Goetschl, J.W. Dembinski, J. Org. Chem.
[77] G. Leclercq, L. Leclercq, R. Maurel, J. Catal. 50 (1977) 87. 30 (1965) 3540.
[78] J.R. Anderson, Advan. Catal. 23 (1973) 91. [115] F.R. Cannings, A. Fisher, J.F. Ford, P.D. Holmes, R.S. Smith,
[79] J.H. Sinfelt, Advan. Catal. 23 (1973) 91. Proc. Conf. Radioisotopes Int. AEC, Copenhagen, 1960, vol.
[80] B.H. Davis, Proceedings, 8th Int. Cong. Catal., vol. 2, 1984, 3, p. 205.
pp. 469–480. [116] H. Pines, S.M. Csicsery, J. Catal. 1 (1962) 313.
[81] N.D. Zelinskii, B. Kazanskii, A.F. Plate, Chem. Ber. B66
[117] H. Pines, C.T. Chen, 2nd Congr. Int. Catal. Paris 1960 (1961)
(1933) 1415.
367.
[82] B.A. Kazanskii, T.S. Bulanova, Izv. Akad. Nauk. SSSR Otd.
[118] F.M. Dautzenberg, J.C. Platteeuw, J. Catal. 19 (1970) 41.
Khim. Nauk, 1947, pp. 29–40.
[119] A.E.H. Kilner, H.S. Truner, R.J. Warne, Radioisotope
[83] B.A. Kazanskii, Usp. Khim. 17 (1948) 655.
Conference, 1954, Proc. 2nd Conf. Oxford, 19–23 July 1954,
[84] F.G. Gault, Adv. Catal. 30 (1981) 1.
vol. 3, Physical Sci., Industrial Applications, Butterworths,
[85] M. Chow, S.H. Park, W.M.H. Sachtler, Appl. Catal. 19
London, 1954, p. 23.
(1985) 349.
[120] R.W. Wheatcroft, Ph.D. Thesis, University of California,
[86] S.G. Brandenberger, W.L. Callender, W.K. Meerbott, J.
1949.
Catal. 42 (1976) 282.
[121] R.C. Pitketkly, H. Steiner, Trans. Faraday Soc. 35 (1939)
[87] G. Maire, G. Plouidy, J.C. Prudhomme, F.G. Gault, J. Catal.
979.
4 (1965) 556.
[122] H.S. Taylor, J. Turkevich, Trans. Faraday Soc. 35 (1939)
[88] M. Vaarkamp, P. Dijkstra, J. van Grondelle, J.T. Miller, F.S.
921.
Modica, D.C. Koningsberger, R.A. van Santen, J. Catal. 151
(1995) 330. [123] H. Hoog, J. Verheus, F. Zuiderweg, Trans. Faraday Soc. 35
[89] H. Zuegg, R. Kramer, Appl. Catal. 9 (1984) 263. (1939) 993.
[90] M. Yamada, E. Yamamoto, A. Amano, J. Japan Petrol. Inst. [124] W. Mattox, J. Am. Chem. Soc. 66 (1944) 2059.
25 (1982) 112. [125] H.S. Taylor, H. Fehrer, J. Am. Chem. Soc. 63 (1941) 1387.
[91] X. Bai, W. Sachtler, J. Catal. 129 (1991) 121. [126] T.C. Rees, Ph.D. Dissertation, Pennsylvania State University,
[92] I.V. Gostunskaya, K. Ch’ing-fêng, B. A. Kazanskii, Izv. 1966.
Akad. Nauk SSR, Ser. Khim. (1964) 1073. [127] H.M. Walborsky, Accounts Chem. Res. 23 (1990) 286.
[93] R. Kramer, M. Fischbacher, J. Mol. Catal. 51 (1989) 247. [128] M. Julia, Accounts Chem. Res. 4 (1971) 386.
[94] F.G. Gault, Compt Rendu 245 (1957) 1620. [129] J.J. Mitchell, J. Am. Chem. Soc. 80 (1958) 5848.
[95] I.I. Levitskii, Kh.M. Minachev, in: F. Márta, D. Kalló (Eds.), [130] H. Pines, C.T. Chen, J. Org. Chem. 26 (1961) 1057.
Mechanisms of Hydrocarbon Reactions: a Symposium, [131] F.R. Cannings, A. Fisher, J.F. Ford, P.D. Holmes, R.S. Smith,
Elsevier, Amsterdam, 1975, pp. 81–95. Chem. Ind. (1960) 228.
[96] G.R. Lester, J. Catal. 13 (1969) 187. [132] J.A. Feighan, B.H. Davis, J. Catalysis 4 (1965) 594.
[97] S.M. Csicsery, R.L. Burnett, J. Catal. 8 (1967) 74. [133] H. Steinberg, F.L.J. Sixma, Rec. Trav. Chim. 79 (1960) 679.
[98] R.H. Hardy, B.H. Davis, Acta Chim. Hungar. 124 (1987) [134] B.H. Davis, P.B. Venuto, J. Organic Chem. 36 (1971) 337.
269. [135] B.H. Davis, J. Catal. 29 (1973) 398.
[99] C.-S. Huang, D.E. Sparks, H.A. Dabbagh, B.H. Davis, J. [136] L. Nogueira, H. Pines, J. Catal. 70 (1981) 404.
Catal. 134 (1992) 269. [137] B.H. Davis, J. Catal. 46 (1977) 348.
[100] G. Moretti, W.M.H. Sachtler, J. Catal. 115 (1989) 205. [138] V. Amir-Ebrahimi, A. Chaplin, P. Parayre, F.G. Gault,
[101] G. Moretti, W.M.H. Sachtler, J. Catal. 116 (1989) 361. Nouveau J. Chem. 4 (1980) 431.
B.H. Davis / Catalysis Today 53 (1999) 443–516 515

[139] J.M. Cogen, W.F. Maier, J. Am. Chem. Soc. 108 (1986) [167] O.V. Bragin, V. Tovmasyan, G. Isagulyants, A. Greish, L.
7752. Kovalenko, A. Liberman, Izv. Akad. Nauk SSSR, Ser. Khim.
[140] R.C. Archibald, B.J. Greensfelder, Ind. Eng. Chem. 37 (1977) 2041.
(1945) 356. [168] E. Iglesia, J. Baumgartener, G. Price, J. Catal. 134 (1992)
[141] B.H. Davis, J. Catal. 23 (1971) 365. 549.
[142] B. Shi, B.H. Davis, J. Catal. 147 (1994) 38. [169] G.V. Isagylyanta, M.I. Rozengart, B.A. Kazanskii, Yu. I.
[143] B. Shi, B.H. Davis, in: J.W. Hightower, W.N. Delgass, E. Derbentsev, Yu. G. Dubinskii, Dokl. Akad. Naud SSSR 191
Iglesia, A.T. Bell (Eds.), 11th Int. Congr. Catal., Elsevier, (1970) 600.
Amsterdam, 1996, pp. 1145–1154. [170] G.V. Isagulyants, A.A. Greish, L.I. Kovalenko, N.N.
[144] B. Shi, B.H. Davis, J. Chromatography A 654 (1993) 319. Rozhdestvenskaya, M.I. Rozengart, Izv. Akad. Nauk SSSR
[145] B. Shi, B.H. Davis, J. Catal. 162 (1996) 134. Ser. Khim. (1980) 864.
[171] G.V. Isagylyanta, M. Rozengart, V.G. Bryukhanov, Izv.
[146] G.V. Isagulyants, M.I. Rozengart, B.A. Kazanskii, Yu. I.
Akad. Nauk SSSR Ser. Khim. (1980) 870.
Derbentsev, Yu. G. Dubinskii, Dokl. Akad. Nauk SSSR 191
[172] I.V. Kalechits, Kinet. Katal. 8 (1967) 1114.
(1970) 600.
[173] B.A. Kazanskii, G.V. Isagulyants, M.I. Rozengart, Yu.G.
[147] B.A. Kazanskii, M.I. Rozengart, Dokl. Adak. Nauk SSSR
Dubinsky, L.I. Kovalenko, 7th Int. Congr. Catal., paper 92,
197 (1971) 1085.
1972.
[148] V.F. Il’in, G.V. Isagulyants, M.I. Rozengart, Yu. N. Osov, [174] Z. Paál, P. Tetenyi, Acta Chim. Acad. Sci. Hung. 54 (1967)
Neft. 15 (1975) 808. 175.
[149] I. Manninger, X.L. Xu, P. Tétényi, Z. Paál, Appl. Catal. 51 [175] G.V. Isagulyants, A.A. Balandin, Kinet. Katal. 2 (1961) 737.
(1989) L7. [176] M.I. Rozengart, E. Mortikov, B.A. Kazanskii, Dokl. Akad.
[150] P.G. Menon, Z. Paál, Ind. Eng. Chem. Res. 36 (1997) 3282. Nauk SSSR 166 (1966) 619.
[151] B. Shi, B.H. Davis, J. Catal. 168 (1997) 129. [177] Z. Paál, J. Catal. 105 (1987) 540.
[152] G.L. Price, Z.R. Ismagilov, J.W. Hightower, in: T. Seiyama, [178] A.A. Balandin, G.V. Isagulyants, Yu. I. Derbentsev, Kinet.
K. Tanabe (Eds.), Proc 7th Int. Catal. Congr., Elsevier, Katal. 2 (1961) 741.
Amsterdam, 1981, pp. 708–723. [179] G.V. Isagulyants, Yu.I. Derbentsev, Russ. Chem. Rev. 38
[153] S.M. Davis, W.D. Gillespie, G.A. Somorjai, J. Catal. 83 (1969) 714.
(1983) 131. [180] B.A. Kazanskii, in: F. Márta, D. Kalló (Eds.),
[154] A. Ozaki, Isotopic Studies of Heterogeneous Catalysis, Symp. Mechanisms of Hydrocarbon Reactions, Elsevier,
Academic Press, New York, 1977, pp. 175–177. Amsterdam, 1975, pp. 15–48.
[155] R.P. Bell, The Proton in Chemistry, 2nd ed., Cornell [181] Z. Paál, P. Tetenyi, Acta Chim. Acad. Sci. Hung. 53 (1967)
University Press, Ithaca. NY, 1973, pp. 250–296. 193.
[156] E.S. Lewis, L.H. Funderburk, J. Am. Chem. Soc. 89 (1967) [182] Z. Paál, P. Tetenyi, Acta Chim. Acad. Sci. Hung. 54 (1967)
2322. 175.
[157] E.S. Lewis, J.K. Robinson, J. Am. Chem. Soc. 90 (1968) [183] Z. Paál, P. Tetenyi, Acta Chim. Acad. Sci. Hung. 55 (1968)
4337. 273.
[158] R.P. Bell, Acid–Basic Catalysis, Oxford University Press, [184] Z. Paál, P. Tetenyi, Acta Chim. Acad. Sci. Hung. 58 (1968)
Oxford, 1941, pp. 143–207. 105.
[159] G.M. Schwab, Catalysis, van Nostrand, Princeton. NJ, 1937, [185] B.A. Kazanskii, G.V. Isagulyants, M.I. Rozengart, L.I.
pp. 260–262. Kovalenko, Yu.G. Dubinsky, 5th Int. Congr. Catal. Miami
1972 (1973) 1277.
[160] B.H. Davis, ACS Preprints, Div. Petrol. Chem. 28 (1983)
[186] J. Wei, Ind. Eng. Chem. 58 (1966) 38.
420.
[187] D. Gál, E. Danózy, I. Nemes, T. Vidóczy, P. Hajdu, Annals
[161] D.O. Hayward, B.M.W. Trapnell, Chemisorption, 2nd ed.,
N.Y. Acad. Sci. 213 (1973) 51.
Butterworths, London, 1964, p. 133.
[188] L. Guczi, P. Tétényi, Annals N.Y. Acad. Sci. 213 (1973)
[162] L.C. Loc, N.A. Gaidai, S.L. Kiperman, Proc. 9th Int. Catal. 173.
Congr., vol. 3, 1988, p. 1261. [189] R.M. Dessau, U.S. Patent 5,283,385.
[163] L.C. Loc, N.A. Gkaidai, B.S. Gudkov, M.M. Kostyukovskii, [190] A.J. Silvestri, R.L. Smith, J. Catal. 29 (1973) 316.
S.J. Kiperman, S.B. Kogan, Kinetika i Kataliz 27 (1986) [191] B.H. Davis, J. Catal. 42 (1976) 376.
1365. [192] W.L. Callender, S.G. Brandenberger, W.K. Meerbott, Proc.
[164] L.C. Loc, N.A. Gaidai, B.S. Gudkov, M.M. Kostyukovsloo, Int. Congr. Catal. 5th 1972 (1973) 1265.
S.J. Kiperman, N.M. Podkletnova, S.B. Cogan, N.R. Bursian, [193] A.I.M. Keulemans, H.H. Voge, J. Phys. Chem. 59 (1959)
Kinetika i Kataliz 27 (1986) 1371. 476.
[165] D.E. Resasco, G.L. Haller, Catalysis, a Specialist Periodical [194] F.G. Gault, private communication.
Report, vol. 11, Royal Soc. Chem., Cambridge, 1994. [195] J.H. Sinfelt, J. Phys. Chem. 68 (1964) 344.
[166] Z. Paál, in: G.J. Antos, A.M. Aitani, J.M. Parera (Eds.), [196] G.C. Bond, Ind. Eng. Chem. Res. 36 (1997) 3173.
Catalytic Naphtha Reforming, Marcel Dekker, New York, [197] G.A. Martin, R. Dutartre, S. Yuan, C. Marquez-Alvarez, C.
1995, pp. 19–44. Mirodator, J. Catal. 117 (1998) 105.
516 B.H. Davis / Catalysis Today 53 (1999) 443–516

[198] A. Crucq, L. Degols, G. Lienard, A. Frennet, Acta Chim [216] G. Pimentel, Chemtech 16 (1986) 536.
Acad. Sci. Hungary 111 (1982) 547. [217] W.L. Manner, G.S. Girolami, R.G. Nuzzo, Langmuir 14
[199] E.H. van Broekhoven, V. Ponec, Prog. Surf. Sci. 19 (1985) (1998) 1716.
351. [218] A.V. Teplyakov, A.B. Gurevich, E.R. Garland, B.E. Bent,
[200] J.C. Rohrer, H. Hurwitz, J.H. Sinfelt, J. Phys. Chem. 65 J.G. Chen, Langmuir 14 (1998) 1337.
(1961) 1458. [219] H.C. deJongste, F.J. Kuijers, V. Ponec, 5th Int. Cong. Catal.
[201] J.-P. Bournonville, J.P. Franck, in: Z. Paál, P.G. Menon 2 (1977) 915.
(Eds.), Hydrogen Effects in Catalysis: Fundamentals and [220] H.D. deJongste, V. Ponec, J. Catal. 63 (1980) 389.
Practical Applications, Elsevier, Amsterdam, 1988, pp. [221] H.C. deJongste, V. Ponec, F.G. Gault, J. Catal. 63 (1980)
653–677. 395.
[202] C. Alvarez-Herrera, Thesis, Poitiers, 1977. [222] A.V. Teplyakov, B.E. Bent, J. Phys. Chem. B 101 (1997)
[203] B.H. Davis, in: L. Guczi, F. Solymosi and P. Tétényi (Eds.), 9052.
Proc. 10th Int. Cong. Catal., Elsevier, Amsterdam, 1993, p. [223] N. Vasquez Jr., R.J. Madix, J. Catal. 178 (1998) 234.
889. [224] J.A. Dumesic, B.A. Milligan, L.A. Greppi, V.R. Balse, K.T.
[204] A.A. Balandin, in: A.A. Balandin (Ed.), Scientific Selection Sarnowski, C.E. Beall, T. Kataoka, D.F. Rudd, Ind. Eng.
of Catalysts, Translated by Israel Program for Scientific Chem. Res. 26 (1987) 1399.
Translation, Jerusalem, 1968. [225] D.B. Dadyburjor, E. Ruckenstein, J. Phys. Chem. 85 (1981)
[205] A.S. Ogawa, S. Ino, J. Crystal Growth 13/14 (1972) 48. 3396.
[206] S. Ino, J. Phys. Soc. Japan 26 (1969) 1559. [226] A. Chambellan, J.M. Dartigues, C. Corolleur, F.G. Gault,
[207] J.J. Burton, Cat. Rev. Sci. Eng. 9 (1974) 209. Nouv. J. Chim. 1 (1977) 41.
[208] R.W. Joyner, B. Lang, G.A. Somorjai, J. Catal. 27 (1972) [227] A.V. Zeigarnik, R.E. Valdés-Pérez, J. Computational Chem.
405. 19 (1998) 741.
[209] D.W. Blakely, G.A. Somorjai, J. Catal. 42 (1976) 181. [228] V.K. Shum, J.B. Butt, W.H.M. Sachtler, Appl. Catal. 11
[210] M.C. McMaster, R.J. Madix, Surf. Sci. 294 (1993) 420. (1984) 151.
[211] W.H. Weinberg, J. Vac. Sci. Technol. A 10 (1992) [229] S. Sivasanker, S.R. Padalkar, Appl. Catal. 39 (1988) 123.
2271. [230] C.T. Ako, A.A. Susu, J. Chem. Tech. Biotechnol. 36 (1986)
[212] W.H. Weinberg, Langmuir 9 (1993) 655. 519.
[213] Z. Paál, discussion at end of reference 203. [231] J. Henningsen, discussion at end of reference 14.
[214] V. Ponec, Advan. Catal. 32 (1983) 149. [232] R. D. Srivastava, Heterogeneous Catalytic Science, CRC
[215] B.E. Bent, Chem. Rev. 96 (1996) 1361. Press, Boca Raton, 1988, p. 126.

You might also like