Dehydrogenation of Light Alkanes over Zeolites

Article abstract of DOI:10.1006/jcat.1997.1860

The conversion of light paraffins to olefins and the secondary reactions of the unsaturated compounds were investigated on H-ZSM5 and H-Y zeolites between 733 and 823 K. Steady state-and transient response-isotope tracing studies revealed that two mechanisms of dehydrogenation are operative. The main pathway is represented by monomolecular, protolytic dehydrogenation. This reaction contributes most to steady state olefin production. Additionally, at the initial stages of the reaction, extra framework aluminum moieties are speculated to participate in high dehydrogenation activity. This pathway is blocked at later stages of the reaction by product (hydrogen) inhibition. The intrinsic rates of protolytic dehydrogenation and olefin desorption range in the same order of magnitude. At high protolytic dehydrogenation rates, olefin desorption represents the rate determining step. Depending on the process conditions, olefins undergo secondary cracking, oligomerization, or isomerization. The latter proceeds via intramolecular rearrangement, possibly via a cyclopropylcarbenium ion at high temperatures and low conversions. At reaction conditions where bimolecular cracking prevails, isomerization is concluded to occur via secondary cracking of di- or oligomers.

Full text of DOI:10.1006/jcat.1997.1860

JOURNAL OF CATALYSIS 172, 127–136 (1997)
ARTICLE NO. CA971860
Dehydrogenation of Light Alkanes over Zeolites
Thomas F. Narbeshuber, Axel Brait, Kulathuyier Seshan, and Johannes A. Lercher
Christian Doppler Laboratory for Heterogeneous Catalysis, Department of Chemical Technology, University of Twente,
P.O. Box 217, 7500 AE Enschede, The Netherlands
Received March 17, 1997; revised August 4, 1997; accepted August 5, 1997
dergo double-bond-, cis/trans-, and skeletal isomerization,
The conversion of light paraffins to olefins and the secondary (iii) di- and/or oligomerize, or crack further into smaller
reactions of the unsaturated compounds were investigated on olefins and carbenium ions. The extent to which these pro-
H-ZSM5 and H-Y zeolites between 733 and 823 K. Steady state-
and transient response-isotope tracing studies revealed that two
mechanisms of dehydrogenation are operative. The main pathway
is represented by monomolecular, protolytic dehydrogenation. This
reaction contributes most to steady state olefin production. Addi-
reactant pressures, high conversions, and low temperatures
tionally, at the initial stages of the reaction, extra framework alu-
minum moieties are speculated to participate in high dehydrogena-
tion activity. This pathway is blocked at later stages of the reaction
by product (hydrogen) inhibition. The intrinsic rates of protolytic
cesses occur depends on the acidity of the catalyst and on
the shape and concentration of the olefins. Several groups
including ourselves showed that di- and oligomerization re-
quires high concentrations of the olefins (2, 16–18). High
favor these processes. At low olefin concentrations, desorp-
tion is preferred over the bimolecular reaction.
The isomerization of olefins was studied in a number
dehydrogenation and olefin desorption range in the same order of of publications during the past (18–35). Double-bond and
magnitude. At high protolytic dehydrogenation rates, olefin des- cis-trans isomerization, as well as the skeletal rearrange-
orption represents the rate determining step. Depending on the ment of these substances, were in the focus of interest (e.g.,
process conditions, olefins undergo secondary cracking, oligomer-
(
18, 19, 33)). Due to the increasing demand for i-butene as a
ization, or isomerization. The latter proceeds via intramolecular
rearrangement, possibly via a cyclopropylcarbenium ion at high
temperatures and low conversions. At reaction conditions where
bimolecular cracking prevails, isomerization is concluded to occur
reactant in the synthesis of methyl-tert-butyl-ether (MTBE)
32–35) isomerization of n-butene to i-butene was recently
(
discussed in several articles. Two reaction pathways appear
to describe the observations of kinetic experiments, i.e., (i) a
monomolecular rearrangement and (ii) an isomerization
via a di- or oligomer species, which selectively crack. It is
currently believed that carbenium ions with more than four
carbon atoms might isomerize via the formation of a cyclo-
propyl intermediate (31). The current argument against the
monomolecular rearrangement of the butylcarbenium ion
is based on the fact that, assuming a methyl-cyclopropyl-
carbenium ion as intermediate, a primary carbenium ion
has to be formed to yield i-butene. Houzvicka and Ponec
via secondary cracking of di- or oligomers.
�c 1997 Academic Press
INTRODUCTION
Monomolecular catalytic cracking was shown to proceed
via protolysis of the substrate and subsequent decompo-
sition of the so formed carbonium ion (1–3). Similarly,
monomolecular dehydrogenation can be visualized to pro-
ceed via protolysis followed by the release of hydrogen and
the formation of a carbenium ion/alkoxy group (4, 5). We
discussed the principal scheme of protolytic dehydrogena-
tion recently (1, 2, 5), using the conversion of n-butane
as an example. However, the mechanism(s) of alkane de-
hydrogenation and the nature of the catalytically active
sites are still controversial (3–15). Brenner and Emmett
(
32) concluded that isomerization occurs via a monomolec-
ular pathway on Brønsted acid sites.
In the present communication we focus on the high tem-
perature (733–823 K) dehydrogenation of light paraffins
and subsequent reactions of the olefins, mainly isomeriza-
tion. Two mechanisms of dehydrogenation and isomeriza-
tion will be discussed with respect to the results obtained
(
7) postulated acidic dehydrogenation sites and excluded
an earlier opinion about metal impurities to cause dehy-
drogenation. McVicker (8) suggested that strong electron
deficient sites are present that effect a radical decomposi-
tion of the substrate, a view accepted by Bizreh and Gates
13
from steady state- as well as C- and D-tracing studies. The
rate determining steps of both routes will be evaluated.
EXPERIMENTAL
(
11). Marczewski (12), and Beyer (15) held Lewis acid sites
responsible.
The catalytic conversion of light hydrocarbons (propane,
Carbenium ions, once formed via dehydrogenation on n-butane, i-butane, n-pentane, i-pentane, and n-hexane)
zeolites, may (i) desorb as olefins from the catalyst, (ii) un- was investigated on H-ZSM5 and H-Y zeolites. The
127
0021-9517/97 $25.00
Copyright �c 1997 by Academic Press
All rights of reproduction in any form reserved.

1
28
NARBESHUBER ET AL.
TABLE 1
Physicochemical Characterization (Values Taken from Ref. (36))
TABLE 2
Rates of Cracking and Dehydrogenation at 773 K
9
[
mol/g s mbar] � 10
HY-M
Cracking Dehydrogenation Cracking Dehydrogenation
HY-P
HY-M
H-ZSM5
H-ZSM5
Altotal (EDAX) [mol/g]
Altetrahedral ( Si-NMR) [mol/g]
Acid sites NH3 adsorption
2.7 � 10�
3
3.7 � 10
2.1 � 10
1.2 � 10
� 3
� 3
� 3
4.4 � 10
� 4
2
9
� 3
4
� 4
2.0 � 10
4.1 � 10
Rates
6.6 � 10�
4.2 � 10
� 4
Propane
0,54
4,20
41,0
n.d.
0,53
2,90
30,8
n.d.
1,1
7,3
37,5
126
0,6
2,8
12,3
32
=
Brønsted acid sites [mol/g]
Acid sites pyridine adsorption
Brønsted acid sites [mol/g]
6.0 � 10� 4
4.2 � 10� 4
n-Butane
n-Pentane
n-Hexane
—
=
1
Hads-NH3 [kJ/mol]
140
1.04
142
0.49
147
—
EFAL (IR �OH = 3617 cm 1)
�
[arbitrary units]
monomolecular dehydrogenation. It was clearly observed
that the apparent energies of activation for dehydrogena-
tion increased with increasing molecular weight of the feed
on HY-M and H-ZSM5, while those for cracking decreased.
The addition of the heats of adsorption of the correspond-
ing hydrocarbon to the apparent energy of activation for
cracking yielded a constant true energy of activation indi-
cating that the energy which is required to be overcome
to cleave a carbon–carbon bond does not depend on the
chain length of the hydrocarbon. In contrast, the true en-
ergy of activation for dehydrogenation increased, because
both, heats of adsorption and apparent energies of activa-
tion, increased with increasing carbon chain length. This
suggests different mechanisms or different rate determin-
ing steps for dehydrogenation and cracking. The reaction
orders for cracking and dehydrogenation were unity for all
hydrocarbons studied.
physicochemical characteristics of the samples were taken
from Ref. (36). They are compiled in Table 1. H-ZSM5 had
a Si/Al ratio of 35 and a particle size of approximately 1 �m.
Infrared spectroscopy showed two characteristic hydroxyl
�
1
adsorption bands at 3610 and 3745 cm . They were as-
signed to Brønsted acid sites and terminal hydroxyl groups,
respectively. Lewis acid sites were not detected. Iron impu-
rities were smaller than 500 ppm in the H-ZSM5 sample.
HY-P(arent) was a dealuminated faujasite, obtained
from Degussa. This sample contained large amounts of ex-
tra framework aluminum. HY-M(odified) was obtained by
leaching HY-P with 0.1 N ammoniumhydroxide solution to
remove large fractions of the extralattice aluminum species.
Steady state and transient response experiments were
carried out in a plug flow reactor using a quartz glass tube
with an inner diameter of 4 mm (for a detailed description
see Ref. (2)). The reactions were investigated between 733
and 823 K. Propane or n-butane were used as 10 or 2%
The Conversion of Branched Hydrocarbons
A detailed comparison between linear and the branched
mixture in He 5.0, respectively. Pentane and hexane were hydrocarbons was already given earlier (2). The conversion
added to the carrier gas via a syringe pump. Partial pres- of i-butane on H-ZSM5 yielded methane, propene, and the
sures of the reactants ranged from 10 Pa to 10 kPa. The butene isomers as primary products. A comparison of the
catalyst weight was varied from 0.002 to 0.05 g. Flow rates selectivities of cracking and dehydrogenation for n-butane
were adjusted from 10 to 100 ml/min to keep the conversion and i-butane is depicted in Fig. 1. The selectivity to dehy-
below 5% , so that the reactor operated under differential drogenation was higher for i-butane than for n-butane.
conditions.
The product selectivities for the conversion of i-pentane
and n-pentane at 773 K are added in Fig. 1. Methane
RESULTS
TABLE 3
Rates of Reaction, Energies of Activation,
and Reaction Orders
Apparent Energies of Activation [kJ/mol] for Cracking and
Dehydrogenation of Propane to n-Pentane (n-Hexane)
The conversion of light hydrocarbons was carried out be-
tween 733 and 823 K on H-Y and H-ZSM5 zeolites. A de-
tailed interpretation of the steady state results was already
given earlier (1, 2, 37) and will only be summarized here.
HY-M
H-ZSM5
EA
Cracking Dehydrogenation Cracking Dehydrogenation
The rates of cracking and dehydrogenation at 773 K are Propane
165
140
135
n.d.
65
140
150
n.d.
155
135
120
105
95
115
150
160
n-Butane
compiled in Table 2. The corresponding energies of activa-
n-Pentane
tion are compiled in Table 3. The observed product distri-
n-Hexane
butions could be explained by monomolecular cracking and

DEHYDROGENATION OF LIGHT ALKANES
129
TABLE 5
Conversion of n- and i-Pentane on H-ZSM5 at 773 K; Rates
9
of Product Formation � 10 [mol/g s mbar]; EA [kJ/mol]
R. no.
Reaction
n-Pentane
i-Pentane
=
4
C5-1
C5H12 → C1 + C
C5H12 → C2 + C
C5H12 → C3 + C
9.4
21
7.2
12.2
9
17
6.6
0
21
12.5
=
C5-2
C5-3
C5-4
C5-5
3
=
2
=
C5H12 → H2 + C
5
=
=
=
C
→ C + C
5
2
3
rCracking
rDehydrog.
37.6
12.2
23.6
21
EA Cracking
EA Methane
EA Ethane
115
115
115
150
150
155
130
150
FIG. 1. Selectivities to cracking and dehydrogenation for n-butane,
i-butane, n-pentane, and i-pentane on H-ZSM5 at 773 K.
EA Dehydrog.
and butenes, ethane and propene, propane and ethane,
and pentenes were the primary products in the conver-
sion of n-pentane. The conversion of i-pentane yielded the formation of methane and butenes and (ii) the formation of
same primary products as n-pentane with the exception of ethane and propenes, denoted as C5-1 and C5-2 in Table 5.
propane.
Note that in the conversion of i-pentane the step producing
As observed for the conversion of the butanes, the se- butenes showed an apparent energy of activation which was
lectivity to dehydrogenation was higher for i- than for 25 kJ/mol higher than the step producing propenes, suggest-
n-pentane.
ing that the rate determining step for C5-1 is the desorption
Tables 4 and 5 reflect the rates of reaction and the en- of the olefins while for C5-2 it should still be protolysis of
ergies of activation for the monomolecular cracking and the C–C bond.
dehydrogenation processes. The energies of activation for
n-butane and i-butane show a similar trend. The EA for de-
hydrogenation was somewhat smaller than the value for
cracking. In contrast, the comparison of the energies of ac-
H/D ISOTOPE EFFECTS
Figure 2 shows the rates of cracking and dehydrogena-
tivation for n- and i-pentane showed a distinctly different tion asa function ofthe response time for an n-(H/D)butane
behavior. With i-pentane the energy of activation for crack- transient experiment (for a detailed interpretation see Ref.
ing was much higher than that for n-pentane, while the en- (38)). The rate of n-(H )butane cracking was 1.7 times
10
ergy of activation for dehydrogenation was the same for higher than that of n-(D )butane indicating an appreciable
10
both hydrocarbons. These data suggest a change in the rate isotope effect. Note that the rate of cracking decreased in
determining steps of both processes. Looking closer to the parallel to the decreasing proton concentration on the cata-
energies of activation for cracking and breaking down the lyst surface, indicating that the O–H (O–D) bond cleavage,
overall value to EA Methane and EA Ethane, a clear distinction
can be made between the two cracking processes, i.e., (i) the
TABLE 4
Conversion of n- and i-Butane on H-ZSM5 at 773 K; Rates
9
of Product Formation � 10 [mol/g s mbar]; EA [kJ/mol]
R. no.
Reaction
n-Butane
i-Butane
=
3
C4-1
C4H10 → C1 + C
C4H10 → C2 + C
3.7
3.6
2.7
0.1
4.5
0
4.0
0.1
=
C4-2
C4-3
C4-4
2
=
C4H10 → H2 + C
4
=
=
C
→ 2 C
4
2
rCracking
rDehydrog.
7.3
2.8
4.5
4.1
FIG. 2. Rates of cracking and dehydrogenation as a function of
response time upon a steady state isotope transient (n-H10-butane → n-D10-
butane at t = 0 s; H-ZSM-5, pButane = 4 kPa; 773 K).
EA Cracking
EA Dehydrog.
135
115
125
100

1
30
NARBESHUBER ET AL.
i.e., the protonation of the alkane (deprotonation of the
zeolite) is most important for the rate determining step.
The results were confirmed by a step-up transient exper-
iment (i.e., increasing the partial pressure of n-(H10)butane
from 0 to 40 mbar) using a catalyst, which only contained
OD groups. The initial rate of cracking (surface covered
with D, feed n-(H10)butane) was again 1.7 times lower than
the rate obtained after the surface deuterium atoms were
fully exchanged with the protons from the molecules.
Because the ratios of the rates were found to be identical
for the reactions, i.e.,
�
�
r(C4H10 + H )
r(C4H10 + H )
=
1.7 =
,
[1]
�
�
FIG. 3. Pressure transient response experiment. n-Butane (incr.
→ 4 kPa), 803 K, 32 mg H-ZSM5, 30 ml/min. Response functions of
r(C4D10 + D )
r(C4H10 + D )
0
cracking and dehydrogenation products.
we conclude that it is primarily the cleavage of the OH bond
of the zeolite or the proton addition (i.e., a kinetic isotope
effect) which is responsible for the difference in the rates.
In contrast to cracking, isotopic labeling of the feed
or exchanging the surface protons with deuterium atoms
hardly influenced the rate of dehydrogenation. Only a
slightly lower rate was observed (factor 0.9–0.95) with n-
values. The steady state selectivities of cracking, primary
and secondary dehydrogenation (formation of butenes and
1
,3-butadiene) were 83.8, 16, and 0.2% , respectively.
It can be seen that the reactant (n-butane) and the crack-
ing products reach the steady state shortly after the re-
sponse function ofthe tracer gaskrypton and did not further
depend on time on stream. The dehydrogenation response
had a maximum at very short times on stream followed by a
decrease to its final steady state value (the maximum rate at
very short times on stream will be denoted as overshoot in
the text.), which was reached after approximately 50 s in this
experiment. This effect was even more pronounced for 1,3-
butadiene, i.e., the initial butadiene production exceeded
the steady state rate by a factor of 5, while the maximum
value of the butene isomers was two times the steady state
concentration. The response curve of 1,3-butadiene arrived
at the steady state after 350 s.
(
D10)butane. This suggests that protonation of the feed is
not part of the rate determining step in dehydrogenation.
It is speculated that the rate determining step for dehydro-
genation should be close to product desorption.
In order to facilitate the discussion, let us briefly summa-
rize the major facts observed so far:
(
i) The energies of activation for dehydrogenation in-
crease with increasing molecular weight within a homolo-
gous row of paraffins. This is in clear contrast to the constant
true energy of activation obtained for protolytic cracking
indicating a change in the rate determining step of the
reaction.
The dependence of the overshoot of the dehydrogena-
tion products on changes in hydrodynamic residence time
is shown in Fig. 4. When the residence time was increased
(
ii) Comparing normal- and iso-paraffins, we observed
a sudden increase in the energy of activation for cracking
when the rate of olefin production exceeds a certain limit.
Much more butenes are produced from i-pentane than from
n-pentane; thus, olefin desorption becomes the slow step.
(
iii) A distinct H/D isotope effect was observed for pro-
tolytic cracking, but not for dehydrogenation of n-butane
indicating again that protonation of the alkane is not rate
determining for dehydrogenation.
Pressure Transient Response Experiments
Pressure transient response experimentswere carried out
with n-butane in order to study the activity behavior for
very short times on stream. Figure 3 shows the dependence
of cracking and dehydrogenation as a function of time on-
stream at 803 K and a volumetric flow rate of 30 ml/min;
3
2 mg of H-ZSM5 were used without quartz dilution. The
FIG. 4. Dependence of the transient response function as a function
of residence time. Both step up transients: n-butane (incr. 0 → 4 kPa),
conversion at the final steady state was 3.5% . For reasons of
better comparison, the response curves of the products (i.e.,
8
03 K, 32 mg H-ZSM5, 0.53 g s/ml (closed symbols), 2.11 g s/ml (open
concentration vstime) were normalized to their steady state symbols), conversions: 3.5% , 14% .

DEHYDROGENATION OF LIGHT ALKANES
131
from 0.53 g s/ml to 2.11 g s/ml by decreasing the volumet-
ric flow rate from 60 ml/min to 15 ml/min, using the same
amount of catalyst, the intensities of the maxima of the re-
sponse curves decreased. The change in residence time had
no influence on the response curve of cracking. However,
the initialslope ofthe response curveswere shifted to higher
times on stream with decreasing space velocity. Note again,
that the transient response functions shown here were nor-
malized to their steady state values, so that the increase
in conversion from 3.5 to 14% after the decrease in space
velocity is not seen.
Hydrogen could not be monitored with the GC-MS-FID
system. Thus, an online massspectrometer wasused to mon-
itor the hydrogen response on-line at the exit of the reactor.
The results obtained for H-ZSM5 are shown in Fig. 5. After
increasing the n-butane pressure from 0 to 4 kPa the hydro-
gen production immediately reached the steady state value,
in clear contrast to the formation ofthe butenes. The reverse
FIG. 6. Pressure transient response experiment. n-Butane (incr.
→ 4 → 0 → 4 → 0 kPa), 803 K, 32 mg H-ZSM5, 30 ml/min. Response
0
functions of dehydrogenation. Regeneration between the two step-up
transients: 5 min in He at reaction temperature.
transient, i.e., switching from n-butane containing to pure can be clearly seen that 5 min were enough to free approxi-
helium, showed an immediate cease of hydrogen produc- mately 30% of the surface, as estimated from the integrated
tion, together with a sudden decrease in the concentrations areas of the butenes during the overshoot. The initial rate
�
8
of tracer gas and feed. Thus, the excess hydrogen produced of hydrogen desorption was estimated to be 10 mol/g s.
from dehydrogenation during the overshoot must remain
Besides H-ZSM5, pressure transient response experi-
on the catalyst and it is supposed to desorb very slowly. A mentswere carried out with HY-P and HY-M. For both cata-
close inspection of the curve in Fig. 5 shows indeed that lysts a high initial dehydrogenation activity was observed
the concentration of hydrogen after the step down did not which ceased with time on stream (Fig. 7) and reached
reach the baseline completely. This suggests that after a steady state. The major difference between the two sam-
certain time (in the absence of the reaction) the surface ples lies in this initial activity to dehydrogenation. It was far
should be free of hydrogen and the overshoot might be ob- higher on the parent (HY-P) than on the modified (HY-M)
served again. Therefore, the step up transient was repeated sample. Also the decay of the dehydrogenation activity was
after a regeneration period of 5 min between the transients. significantly slower (approx. 30 min) than the decay ob-
During this regeneration period the temperature was kept served with HY-M (approx. 5 min). This suggests that extra
constant and He was passed over the catalyst. The results of framework speciesplaya major role in the dehydrogenation
this transient response experiment are depicted in Fig. 6. It of light alkanes at short times on stream. For a detailed
FIG. 5. Pressure transient response experiment. n-Butane (0 → 4 → 0 kPa), 803 K, 32 mg H-ZSM5, 30 ml/min. Response functions of H2 and
n-butane.

1
32
NARBESHUBER ET AL.
discussion of the results obtained with the faujasite sam-
ples see Ref. (37).
Thus, H-ZSM5 and HY-M reached the steady state at
short times on stream, while HY-P showed a distinctly dif-
ferent behavior. The initial activity to dehydrogenation was
much higher and the decrease to a constant steady state
value much longer compared to the other samples indicat-
ing a higher capability of storing hydrogen at the surface.
Cofeeding n-Butane and i-Butene
In order to determine the influence of olefins on the
rates and product selectivities ofn-butane conversion, small
amounts of i-butene were added to the n-butane feed. The
quantities added to the feed stream were increased linearily
as the conversion was increased, and were in the range of
FIG. 8. The influence of the addition of i-butene upon conversion of
n-butane at 803 K.
the amount produced via dehydrogenation ofn-butane. The surface processes involved in olefin (trans-)formation. The
selectivities of the products are compiled in Fig. 8 for the equilibrium selectivities were maintained during cofeeding
cracking of n-butane on H-ZSM5 at 803 K with and without ofpure i-butene indicatingthat isomerization ofthe butenes
addition of i-butene to the feed, respectively. The selectivity is fast compared to desorption.
to the butene isomers increased from 15% to approximately
At higher conversion levels, the formation of butenes
2% upon addition of i-butene to the feed. This level rep- increased drastically. At this level of conversion, also the
2
resents the amount of i-butene cofed.
selectivity of the cracking products changed. The rate of
It wasobserved that the butene isomerswere alwaysclose production of ethene and propene was higher than that of
to thermodynamic equilibrium (1, 2, 18) suggesting that methane and ethane, indicating secondary processes, e.g.,
isomerization of the olefins is much faster than all other the formation of an octene intermediate through dimeriza-
tion followed by �-cracking. This behavior suggests a criti-
cal lower limit in the gas phase concentration of the butenes
before secondary reactions set in, i.e., we infer that a certain
butene concentration is required to start di- or oligomer-
ization reactions. This interpretation would further suggest
that the isomerization of butenes is a monomolecular pro-
cess at low concentrations and does not require the forma-
tion of di- or oligomerization, while at high concentrations
di- and oligomerization sets in.
13C-TRACING STUDIES
Theoretically, methane and propene aswellasethane and
ethene contain one 13C-atom each, if cracking of 1,4- C2-
n-butane exclusively proceed through protolysis. This leads
13
13
to overall C selectivities of 100% (methane), 50% (ethane
and ethene), and 33% (propene), respectively. Butenes, re-
sultingfrom protolyticdehydrogenation should contain two
1
3
13
C atoms each corresponding to a C selectivity of 50% .
13
The rates of 1,4- C2-n-butane conversion on H-ZSM5 be-
12
tween 733 and 823 K exactly equaled the conversion of C4-
n-butane. Thus, an isotope effect or a preferential cleav-
age of a certain C–C bond was not observed. The isotope
selectivities
X
13Cx 12
Cy Hz
IS
13Cx 12
=
Cy Hz
XCx+y Hz
FIG. 7. Pressure transient response experiment of n-butane on HY-P
and HY-M at 773 K (inc. pButane 0 → 4 kPa). Closed and open symbols
refer to the parent and the modified faujasite, respectively.
for the products and the reactant upon conversion of
n-1,4- C2-butane on H-ZSM5 at 803 K are shown in Figs. 9
13

DEHYDROGENATION OF LIGHT ALKANES
133
1
3
13
FIG. 9. Conversion of n-1,4- C-butane on H-ZSM5 at 803 K. Isotope
FIG. 11. Conversion of n-1,4- C-butane on H-ZSM5 at 803 K. Iso-
selectivities of products and reactant; 5% overall conversion.
tope selectivities of products and reactant; 50% overall conversion.
depicted in Fig. 14. It might be clearly understood from
these results that carbon scrambling does not occur under
the reaction conditions applied. Note that, nevertheless, the
butenes were found to match their thermodynamic equilib-
rium distribution, suggesting, that isomerization occurs via
a monomolecular process under the differential operating
conditions employed.
and 10. At 803 K and a conversion of 5% hardly any car-
bon scrambling was detected in the products. Methane and
ethane contained exactly 100 and 50% 13C, respectively.
Ethene, propene, and the butene isomers showed small
deviations from the 13C distribution expected from pure
monomolecular cracking. The extent of carbon scrambling,
however, increased with the increasing carbon chain length
of the olefins. These results clearly demonstrate the in-
creased reactivity of the olefins with the catalyst surface
DISCUSSION
(
see also (38)). Figures 11–13 show the influence of conver-
The Rate Determining Step
sion and temperature on the extent of carbon scrambling.
With decreasing temperature and increasing conversion the
The experiments carried out so far clearly indicate that
extent of carbon scrambling in the olefins increases. This is the rate determining steps of cracking and dehydrogenation
explained with an increased surface concentration of these are different. A first indication was given from the apparent
products and hence an increased contribution of bimole- energies of activation for the conversion of normal alkanes.
cular reactions. Note that the 13C distributions in methane, In contrast to the C–C bond cleavage, where a constant true
ethane, and n-butane did not deviate from those expected energy of activation was observed, dehydrogenation of the
for protolytic cracking. This suggests that once methane alkanes showed an increase in the energies of activation.
and ethane are formed, they do not undergo any further Note that the interaction of higher olefins with an acid cata-
reactions and desorb immediately.
lyst surface is stronger than that of smaller alkenes. Thus,
The 13C isotope selectivities for reactants and products if desorption were rate determining, a trend similar to the
12
13
upon cofeeding n- C4-butane and n-1,4- C2-butane are described would be expected.
1
3
13
FIG. 10. Conversion of n-1,4- C-butane on H-ZSM5 at 803 K. Iso-
FIG. 12. Conversion of n-1,4- C-butane on H-ZSM5. Isotope selec-
tope selectivities of butene isomers; 5% overall conversion.
tivities of butene isomers as a function of temperature.

1
34
NARBESHUBER ET AL.
be in the same order of magnitude. If all reactions pre-
ceding the rate determining step are fast, compared to the
latter, quasi-equilibrium is established for these reactions.
This requires that re-hydrogenation (of butenes) occurs
nearly as fast as dehydrogenation (of butane). Co-feeding
n-(H10)butane and n-(D10)butane (38), however, showed
that the deuterium distribution in butane was far different
from that in the butenes suggesting that dehydrogenation/
hydrogenation equilibrium was not established. Moreover,
13
an appreciable extent of C scrambling would be expected
13
in n-butane at high conversions of n-1,4- C-butane, be-
cause a high extent of 13C scrambling was observed in the
butenes at these conversions. Because none of these condi-
tions were observed, it must be concluded that the rates of
dehydrogenation and desorption are similar.
1
3
FIG. 13. Conversion of n-1,4- C-butane on H-ZSM5. Isotope selec-
tivities of ethane as a function of temperature.
Although protolysis was concluded to represent the slow
step in cracking of n-alkanes, it strongly depends on the rate
of olefin formation via cracking whether protonation of the
feed or desorption of the olefins limits the overall process
for iso-paraffins.
A second argument for different rate limiting steps of
cracking and dehydrogenation can be drawn from the com-
parison of the energies of activation observed for linear
and branched paraffins. The conversion of i-pentane yields
mainly methane and butene isomers as primary crack-
ing products, while C2- and C3- paraffins and olefins are
the main products from n-pentane cracking. Thus, it ap-
pears from the comparison of the energies of activation
that the rate constant of C4- production becomes higher
than the rate constant of olefin desorption in the case of
i-pentane and the energy of activation for the production
of methane (and butenes) does no longer reflect the process
of protolysis.
The third argument for the difference in the rate deter-
mining steps arises from the presence/absence of an isotope
effect in n-(H/D)butane conversion. While a large isotope
effect was observed for cracking, a change in the rates has
not been detected when replacing butane(H) with deuter-
ated butane.
A Second Dehydrogenation Pathway
The transient response studies suggest two regimes for
dehydrogenation of alkanes, (i) a high initial activity to de-
hydrogenation at short times on stream, and (ii) a much
lower steady state activity to dehydrogenation. In contrast,
cracking reached the steady state following the tracer gas
and did not further depend on time on stream. It was
shown earlier that protolytic cracking exclusively proceeds
on Brønsted acid sites (1, 2). Because the cracking activity
did not change with time on stream, deactivation of these
sites can be excluded to occur. It might, thus be speculated
that dehydrogenation, proceeds additionally at a second
site. This pathway must, thus, be blocked after short times
on stream.
The decrease of the high initial dehydrogenation activ-
ity was much slower on the parent faujasite, than on all
other materials (HY-M, H-ZSM5). Because this material
contained a large amount of extra framework aluminum,
we speculate that these moieties are involved in the active
sites for the second dehydrogenation pathway.
Thus, we believe that protolysis limits the rate of crack-
ing while olefin desorption controls the rate of dehydro-
genation. However, 13C and H/D tracer experiments sug-
gest that the rates of dehydrogenation and desorption must
From these results it appears that blocking of the second
pathway, which is only active for dehydrogenation, causes
the steep decrease in dehydrogenation activity after short
times on stream. On-line monitoring of hydrogen after a
pressure step up transient, however, showed that hydrogen
reached the steady state value, together with the cracking
products and the tracer gas. Thus, at short times on-stream
less hydrogen than olefins desorb from the catalyst sur-
face, indicating that hydrogen is trapped on the catalyst.
The “step-down” transient response experiment, followed
by a repetition of the “step-up” transient showed that the
blocking is reversible and that these sites are freed slowly
by desorbing hydrogen.
1
3
12
FIG. 14. Conversion of n-1,4- C-butane and n- C-butane on H-
ZSM5 at 803 K. Isotope selectivities of reactant and products.

DEHYDROGENATION OF LIGHT ALKANES
135
Based on the reports of Iglesia, Boudart, and Temkin controlled pathway would be zero order with respect to the
39–42) we propose the following model to explain this be- reactant.
(
havior. The initial dehydrogenation rate (on the additional
site, not on the Brønsted acid sites) is high. Dehydrogena-
tion proceeds by stepwise removal of hydrogen atoms, in
Isomerization
The concentrations of the butene- and the pentene-
contrast to protolytic dehydrogenation on Brønsted acid isomers were found to be in thermodynamic equilibrium,
sites. We speculate that hydrogen adatoms are stored on the independent whether they were formed via cracking or via
surface and their recombinative desorption is slow. Thus, dehydrogenation. This indicates that isomerization (double
as the reaction proceeds, the surface is subsequently filled bond-, cis/trans-, and skeletal-) is faster than desorption.
with hydrogen species. This ensemble of hydrogen species
Two mechanisms of olefin isomerization were proposed
at the catalyst surface was denoted as hydrogen pool and in literature, i.e., an intra- and an intermolecular rearrange-
it corresponds to high hydrogen surface fugacities (virtual ment. The monomolecular pathway was often rejected (es-
pressures). Because the recombinative desorption of hy- pecially for the butenes) on the basis of the argument that
drogen is rate determining, all preceding steps must be in rearrangement of the butenes would require the forma-
quasi equilibrium. This implies a zero order with respect to tion of a primary carbenium ion to form a cyclopropyl-
intermediate, which is not very likely to occur. The 12C/ C
13
the reactant at steady state for the second pathway.
Because the gas phase concentration of desorbing hydro- cofeeding experiment clearly showed that carbon scram-
gen immediately reached steady state, this reaction mech- bling occurred to a minor extent in the butenes. It was,
anism only explains a part of the catalytic activity for dehy- therefore, concluded that isomerization must occur via a
drogenation, i.e. the contribution of the second active site. monomolecular rearrangement under differential operat-
If the overall formation of olefins was limited by hydro- ing conditions. At high levels of conversion and lower tem-
gen desorption, a gentle initial slope of the appearance of peratures (both factors increase the surface coverage); i.e.,
hydrogen in the gas phase should have been observed.
under conditions where bimolecular cracking prevails, the
The dependence of the observed overshoot on the hy- extent of carbon scrambling also increased. Thus, we pro-
drodynamic residence time, i.e., a lower maximum concen- pose that under these conditions oligomer (dimer) forma-
tration of the butenes at higher values of hydrodynamic tion followed by �-cracking is the preferred pathway in
residence time (�) relative to steady state, can be explained olefin isomerization.
by our model with respect to the behavior of a differen-
tial plug flow reactor. Using Kobayashi’s transient response
data (Kobayashi (43) reported an overshoot behavior of
the N2 response for the decomposition of N2O on MnO2
and explained it by the rate limiting desorption of oxygen
from the surface) for the decomposition of N2O on MnO2,
Bennett (44) discussed the possible types of overshoot be-
havior with a variation in hydrodynamic residence time (�).
It was demonstrated that for high residence times the over-
shoot could not be observed, while at very low residence
times the overshoot peak becomes a cusp. This model of
Kobayashi is in good qualitative agreement with our ex-
perimental results, if we assume, that hydrogen disposal is
the rate determining step of the second dehydrogenation
pathway.
Thus, it is concluded that dehydrogenation proceeds via
two routes, i.e., (i) protolytic dehydrogenation on Brønsted
acid sites (reaches immediately the steady state, together
with cracking) and (ii) dehydrogenation on a different kind
of sites (possibly species affiliated to extra framework alu-
minum).
CONCLUSIONS
Steady state-, transient response-, and isotope-tracing
studies revealed two mechanisms of dehydrogenation. The
main pathway is represented by monomolecular, protolytic
dehydrogenation. This reaction contributes most to steady
state olefin production. The rate determining step for this
pathway is represented by olefin desorption from the cata-
lyst surface. Additionally, at initial stages of the reaction,
extraframework aluminum moieties are concluded to play
a significant role for dehydrogenation. Under those con-
ditions, dehydrogenation in this second reaction pathway
proceeds via stepwise hydrogen removal. The hydrogen
species recombine and desorb very slowly and quench, thus,
the rate of this process at steady state (product = hydrogen
inhibition). Depending on the process conditions, olefins
undergo secondary cracking, oligomerization, or isomeriza-
tion. The latter proceeds via intramolecular rearrangement
at low conversions, possibly via a cyclopropylcarbenium
ion. At reaction conditionswhere bimolecular crackingpre-
vails, isomerization is concluded to occur via secondary
cracking of di- or oligomers.
At steady state, the orders of dehydrogenation were de-
termined to be unity with respect to the reactant for all
hydrocarbons studied. This indicates that the additional de-
hydrogenation mechanism does not contribute significantly
at steady state. If it contributed, we would expect a lower
ACKNOWLEDGMENT
Financial support by the Christian Doppler Laboratories for Hetero-
overall reaction order, because the hydrogen desorption geneous Catalysis is gratefully acknowledged.

1
36
NARBESHUBER ET AL.
REFERENCES
23. Lucchesi, P. J., Baeder, D. L., and Longwell, P. J., J. Am. Chem. Soc.
1, 3235 (1959).
. Narbeshuber, T. F., Vinek, H., and Lercher, J. A., J. Catal. 157, 388 24. Jacobs, P. A., Declerk, L. J., Vandamme, L. J., and Uytterhoeven, J. B.,
1995). J. Chem. Soc., Faraday Trans. 71(7), 1545 (1975).
. Narbeshuber, T. F., Ph.D. thesis, University of Twente, Enschede, NL, 25. Cross, N. E., Kemball, C., and Leach, H. F., Adv. Chem. Ser. 102, 389
994. (1971).
8
1
(
2
3
1
. Haag, W. O., and Dessau, R. M., in “Proceedings 8th International 26. Hightower, J. W., and Hall, W. K., J. Phys. Chem. 71, 1014 (1964).
Congress on Catalysis, Berlin, 1984,” Vol. II, p. 305, Dechema, Frank- 27. Hightower, J. W., Gerberich, H. R., and Hall, W. K., J. Catal. 7, 57
furt am Main.
(1967).
4
5
. Krannila, H., Haag, W. O., and Gates, B. C., J. Catal. 135, 115 (1992).
. Lercher, J., van Santen, R. A., and Vinek, H., Catal. Lett. 27, 91
28. Hightower, J. W., and Hall, W. K., Amer. Inst. Chem. Eng. Symp. Ser.
63, 122 (1967).
(
1994).
29. Goldwasser, J., Engelhardt, J., and Hall, W. K., J. Catal. 71, 381 (1981).
30. Goldwasser, J., Engelhardt, J., and Hall, W. K., J. Catal. 70, 275 (1981).
31. Brouwer, D. M., and Hogeveen, H., Prog. Phys. Org. Chem. 9, 179
(1972).
6
7
8
. Bandiera, J., and Taarit, Y. B., Appl. Catal. 76, 199 (1991).
. Brenner, A., and Emmett, P. H., J. Catal. 75, 410 (1982).
. McVicker, G. B., Kramer, G. M., and Ziemiak, J. J., J. Catal. 83, 286
(
1983).
32. Hou zˇ vi cˇ ka, J., and Ponec, V., Ind. Eng. Chem. Res. 36, 1424 (1997).
33. Maxwell, I., and Naber, J. E., Catal. Lett. 12, 105 (1992).
9
. Kramer, G. M., and McVicker, G. B., J. Catal. 115, 608 (1989).
1
0. Hall, W. K., Lombardo, E. A., and Engelhardt, J., J. Catal. 115, 611 34. Mooiweer, H. H., de Jong, K. P., Kraushaar-Czarnetzki, B., Stork,
(
1989).
W. H. J., and Krutzen, B. C. H., Stud. Surf. Sci. Catal. 84(C), 2327
1
1
1
1
1. Bizreh, Y. W., and Gates, B. C., J. Catal. 88, 240 (1984).
2. Marczewski, M., J. Chem. Soc. Faraday Trans. I 82, 1687 (1986).
3. Abbot, J., and Wojciechowski, B. W., J. Catal. 113, 353 (1988).
4. Gianetto, G., Sansare, S., and Guisnet, M., J. Chem. Soc. Chem. Com- 36. Stockenhuber, M., Ph.D. thesis, Technical University of Vienna,
mun., 1302 (1986). Vienna, Austria, 1994.
5. Beyer, H. K., Horvath, J., and Reti, F., React. Kinet. Catal. Lett. 14, 395 37. Narbeshuber, T. F., Brait, A., Seshan, K., and Lercher, J. A., Appl.
(1994).
35. O’Young, C., Xu, W., Simon, M., and Suib, S., Stud. Surf. Sci. Catal.
84(C), 1671 (1994).
1
(
1980).
Catal. A: General 146, 119 (1996).
1
1
1
6. Lombardo, E. A., and Hall, K. W., J. Catal. 112, 565 (1988).
7. Abbot, J., and Woiciechowski, B. W., J. Catal. 107, 451 (1987).
38. Narbeshuber, T. F., Stockenhuber, M., Brait, A., Seshan, K., and
Lercher, J. A., J. Catal. 160, 183 (1996).
8. Abbot, J., and Wojciechowski, B. W., Can. J. Chem. Eng. 63, 818 39. Temkin, M. I., and Pyzhev, V., Acta Physicochim. 12, 327 (1940).
1985). 40. Boudart, M., J. Phys. Chem. 87, 2786 (1983).
9. Ko, A. N., and Wojciechowski, B. W., Int. J. Chem. Kin. 15, 1249 (1983). 41. Boudart, M., Ind. Eng. Chem. Fund. 25, 70 (1986).
(
1
2
2
2
0. Abbot, J., and Wojchiechowski, B. W., J. Catal. 90, 270 (1984).
1. West, P. B., Haller, G. L., and Burwell, R. L., J. Catal. 29, 486 (1973).
2. Gerberich, H. R., and Hall, W. K., J. Catal. 5, 99 (1966).
42. Iglesia, E., Baumgartner, J. E., and Price, G. L., J. Catal. 134, 549(1992).
43. Kobayashi, M., and Kobayashi, H., J. Chem. Eng. Jpn. 6(5), 438 (1973).
44. Bennett, C. O., Catal. Rev. Sci. Eng. 13(2), 121 (1976).