You are on page 1of 9

Applied Catalysis A: General 398 (2011) 18–26

Contents lists available at ScienceDirect

Applied Catalysis A: General


journal homepage: www.elsevier.com/locate/apcata

Kinetics of propane dehydrogenation over Pt–Sn/Al2 O3 catalyst


Qing Li a , Zhijun Sui a , Xinggui Zhou a,∗ , De Chen b,∗
a
State Key Laboratory of Chemical Engineering, East China University of Science and Technology, Shanghai 200237, China
b
Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), Sem SŁlandsvei 4, N-7491 Trondheim, Norway

a r t i c l e i n f o a b s t r a c t

Article history: Langmuir–Hinshelwood kinetic models for catalytic propane dehydrogenation on Pt -Sn/Al2 O3 are pro-
Received 8 October 2010 posed based on the reaction mechanisms that take into account the one step or two-step dehydrogenation
Received in revised form 25 January 2011 and are evaluated by fitting the experiments. When taking into consideration the fitting accuracy, the
Accepted 26 January 2011
number of fitting parameters and the rigorousness of the reaction mechanism, the kinetics of propane
Available online 3 February 2011
dehydrogenation based on the assumptions of step-wise dehydrogenation, the first step dehydrogenation
as the rate-determining step, and low surface coverage of C3 H7 outperforms other kinetic models. The
Keywords:
change of catalyst activity with time-on-stream is related to the rate of coke formation that depends on
Propane dehydrogenation
Kinetics
the partial pressures of propene and hydrogen. Finally a complete kinetic model, which includes propane
Coke formation dehydrogenation and cracking and catalyst deactivation because of coking, is developed, which is found
Pt–Sn/Al2 O3 catalyst to agree well with the changes of propane conversions with time-on-stream under different operating
conditions. Direct measurement of coke formation is avoided by the approach employed in this study for
the prediction of catalyst deactivation by coke deposition.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction Up to now, investigation on the reaction of propane dehy-


drogenation has been mainly focused on improving the catalytic
Nowadays the world demand of propene is fast increas- performance. Many efforts have been devoted to improving the
ing [1–3] because of the heavy increase in the demand of activity and selectivity by adding alkali metal, alkaline earth metal,
polypropylene, acrylonitrile, propene oxide, cumene, phenol, iso- transition metal and also Ga and Sn in the catalyst [4–9], under-
propylic alcohol and many other propene derivatives. The propene standing the mechanism of coke formation on the catalyst that
obtained as a co-product from steam crackers or FCC units causes catalyst deactivation, and building reliable reaction kinet-
does not satisfy the growing demand. Consequently, there is ics for process design and optimization [10–13]. Compared with
a great interest in developing alternative routes for propene the investigation on the mechanism of propane dehydrogenation
production. Propane dehydrogenation is one of the techniques and coke formation and the study to improve the activity and sta-
for on-purpose propene production, and several propane dehy- bility of the catalyst, much less efforts have been made to establish
drogenation processes, such as Oleflex and Catofin, have been complete and reliable kinetic models under industrial operating
commercialized. conditions and taking into consideration side reactions and cata-
Propane dehydrogenation is equilibrium limited and highly lyst deactivation. These models are essential for optimization of
endothermic, which is usually carried out at 500–600 ◦ C and atmo- the reactor and the process to increase the propene yield and
spheric pressure on platinum or chromium catalysts. However, reduce the deactivation. Moreover, most of the kinetics reported
the dehydrogenation reaction is also accompanied with cracking in the literature are not complete because the catalyst deacti-
and coking reactions at such conditions. While cracking decreases vation is not considered, or the activity of the catalyst is not
the propene yield, coking deactivates catalyst. As a result, peri- related to the rate of coke formation [14–17]. Among the few
odic regeneration of the catalyst is required in industrial operation. complete kinetics published, Larsson et al. [18] combined the kinet-
Developing dehydrogenation catalysts with high activity and selec- ics of propane dehydrogenation, which was a simple power law
tivity as well as good stability against coke formation remains a model, with the kinetics of coke formation. Thereafter, Lobera et al.
challenge. [19] employed the generalized monolayer–multilayer coke growth
model (MMCGM) [20,21] to describe the kinetics of coking on Pt–Sn
catalyst. Gascón et al. [22] used a similar approach to establish the
kinetics for Cr2 O3 catalyst. Moreover, in previous publications on
∗ Corresponding authors. Tel.: +86 21 64253509; fax: +86 21 64253528. propane dehydrogenation, only the temperature was considered to
E-mail addresses: xgzhou@ecust.edu.cn (X. Zhou), chen@nt.ntnu.no (D. Chen). be responsible for catalyst deactivation. However, the gas compo-

0926-860X/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2011.01.039
Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26 19

2. Kinetic model
Nomenclature
2.1. Propane dehydrogenation
a activity of the catalyst
Ea1 activation energy of propane dehydrogenation The mechanisms of dehydrogenation on several catalysts have
(kJ/mol) been studied in a number of investigations. Jackson et al. [16]
Ea2 activation energy of propane cracking (kJ/mol) studied the dehydrogenation of propane over a chromium cata-
Eac activation energy of coking reaction (kJ/mol) lyst using isotopic labeling techniques coupled to transient flow
F flowrate of outlet gas (in Eqs. (11) and (12)) methods and identified the rate determining step as the elimina-
(m3 /min) tion of ␤-hydrogen from propyl radical. Biloen et al. [23] studied
F0 flowrate of feed gas (in Eqs. (11) and (12)) (m3 /min) the catalytic dehydrogenation of propane to propene over platinum
k1 , k2 forward reaction rate constants (mol/(min kg), in and platinum–gold alloys, and also concluded that the subse-
models (M1), (M1-1), (M2), and (M3)) quent conversion of the propyl radical into ␲-bonded propene
k10 , k20 pre exponential factors of k1 and k2 via ␤-hydrogen elimination was the rate-determining step, while
k−1 , k−2 reverse reaction rate constants (mol/(min kg)) the desorption of ␲-bonded propene had a comparatively low
ka1 , ka2 , ka3 rate constants of adsorption activation energy. Suzuki and Kaneko [24] investigated the dehy-
ka−1 , ka−2 , ka−3 rate constants of desorption drogenation of propane over chromium–alumina–potassium oxide
kc coking reaction rate constant catalysts and proposed that the rate determining step of the reac-
kd rate constant in deactivation function tion was the dehydrogenation of adsorbed propane.
K reaction equilibrium constant of propane dehydro- Recently we [25] studied the adsorption of the possible inter-
genation mediates and the pathways of propane dissociation on Pt(1 1 1)
K reaction equilibrium constant of the elimination of by density functional theory. The results revealed that propane
the first hydrogen atom is physically adsorbed on Pt because the propane molecule has
K reaction equilibrium constant of the elimination of no unpaired electron and cannot bind with Pt atom. Trapping-
the second hydrogen atom mediated dissociative chemisorption indicated that the gas phase
K2 , K3 adsorption equilibrium constants (atm−1 ) molecule is trapped in the potential field of the surface in the
K20 , K30 pre-exponential factors of K2 and K3 form of alkane, which accommodates to the temperature of the
l the number of experimental points (in Eq. (13)) surface [26]. The physically adsorbed molecule may then either
M1, M1-1, M2, M2-1, M2-2, M2-3, M3, models for the rate of desorb or dissociate. Propene is chemically adsorbed on Pt by bind-
dehydrogenation of propane ing two surface Pt atoms through its unsaturated C–C bond. More
M4 model for propane cracking interestingly, it was found that surface hydrogen has a remarkable
M5 model for deactivation influences on the overall reactions by promoting propene desorp-
MSC model selection criteria tion and hindering propene dehydrogenation. However, it is still
p the number of parameters (in Eq. (13)) difficult to establish the reaction kinetics for a complex heteroge-
Pi (i = C3 H8 , C3 H6 and H2 ) partial pressure of specie “i” (atm) neous catalyst reaction such as propane dehydrogenation based
r1 , r2 , r3 , r4 reaction rate of propane dehydrogenation solely on theoretical calculations due to the material and pressure
rc coking rate gaps between theoretical calculations and experiments under prac-
s selectivity to propene tical operating conditions. Therefore, in this study the conventional
S vacant site Langmuir–Hinshelwood (LH) approach is employed to establish
SSR sum of squared residuals the kinetic models by assuming different reaction mechanisms and
x conversion of propane rate-determining steps.
yi (i = Ar, CH4 , C2 H4 , C2 H6 , C3 H6 , C3 H8 and H2 ) the molar ratio In the first mechanism, direct dehydrogenation of gas phase
of specie “i” in the outlet gas (in Eqs. (11) and (12)) molecule to propene on Pt surface, hydrogen dissociative adsorp-
yi0 (i = Ar, CH4 , C2 H4 , C2 H6 , C3 H6 , C3 H8 and H2 ) the molar tion, and propene desorption are involved.
ratio of specie “i” in the feed gas (in Eqs. (11) and
(12)) k1
Ȳcal the model-calculated values (in Eq. (13)) C3 H8 + 3S  C3 H6 S + 2HS (I1)
k−1
Ȳobs the weighted mean of the experimental observa-
tions (in Eq. (13))
H2 adsorption heat of propene (kJ/mol)
ka−3 ka3
2HS  H2 + 2S K3 = (I2)
H3 adsorption heat of hydrogen (kJ/mol) ka3 ka−3

ka−2 ka2
C3 H6 S  C3 H6 + S K2 = (I3)
ka2 ka−2

sitions, including the partial pressures of propene and hydrogen, In fact, this is a simple mechanism that lumps up all the possible
also have significant influences on the rate of coke formation and dehydrogenation steps. By assuming the surface reaction (I3) as the
therefore catalyst deactivation. rate-determining step, the rate of propane conversion is:
For dynamic simulation and optimization of the process of
propane dehydrogenation, a transient reaction kinetics, which k1 (PC3 H8 − PC3 H6 PH2 /K)
r1 = 3
(M1)
takes all the important reactions into consideration is required. In (1 + K2 PC3 H6 + K30.5 PH
0.5
)
2
this study, a kinetic model for propane dehydrogenation includ-
ing propane cracking and catalyst deactivation over Pt–Sn/Al2 O3 where [27]:
catalyst is presented. Moreover, the rate of catalyst deactivation is
related not only to temperature, but also to gas composition, i.e., 15934 148728
K = exp(16.858 − + )atm (1)
C3 H6 and H2 concentrations. T T2
20 Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26

If the hydrogen adsorption is weak, the rate expression (M1) can two mechanisms, i.e., deep cracking of propane that leads to graphi-
be simplified to (M1-1): tized carbon, and polymerization, condensation etc., that lead to
polyaromatics [29–32]. These two mechanisms are represented by
k1 (PC3 H8 − PC3 H6 PH2 /K)
r2 = (M1-1)
1 + K2 PC3 H6 C3 H8 → 3C + 4H2 (3)

In another mechanism, both propene and hydrogen adsorption nC3 H6 → 3nCH0.5 + 2.25nH2 (4)
are considered but dehydrogenation is assumed to proceed step by
step, i.e. The deep propane cracking Eq. (3) and the propene polymeriza-
tion Eq. (4) can be assumed to be first order with respect to propane
k1 k1 and propene. As H2 has definite influences on the coking rate, it
C3 H8 + 2S  C3 H7 S + HS K = (II1)
k−1 k−1 should be included in the coking rate equation. Our DFT study
showed H2 inhibits coke formation mainly by promoting propene
k
2 k2 desorption and decreasing the surface coverage of propene, which
C3 H7 S + S  C3 H6 S + HS K  =  (II2) lowers the rate of deep cracking or polymerization of propene.
k k−2
−2
Based on our DFT analysis, propene can be considered as the pre-
ka−2
C3 H6 S  C3 H6 + S (II3) cursor for deep cracking and polymerization, and the influence of
ka2 hydrogen on the coking rate is assumed the same for both deep
ka−3 cracking and for polymerization. Therefore, the coking rate on fresh
2HS  H2 + 2S (II4) catalyst is defined as:
ka3
kc PC3 H6
If the first C–H bond cleavage of propane (II1) is the rate- rc0 = (5)
1 + K3 PH2
determining step, the rate expression can be derived as:
k1 (PC3 H8 − PC3 H6 PH2 /K) where PC3 H6 and PH2 are the partial pressures of propene and hydro-
r3 = 2
(M2) gen. On partially coked catalyst, the coking rate is
(1 + K2 PC3 H6 + K30.5 PH
0.5
+ K2 PC3 H6 K30.5 PH
0.5
/K  )
2 2
rc = ac rc0 (6)
If the second step with second C–H bond cleavage of propane
(II2) is the rate-limiting step, the rate expression is: where ac is the activity of the catalyst for coke formation.
As usual, the rate of propane dehydrogenation is written as
k2 K  (PC3 H8 − PC3 H6 PH2 /K)
r4 = 2
(M3) r = ar0 (7)
K30.5 PH
0.5
(1 + K2 PC3 H6 + K30.5 PH
0.5
+ PC3 H8 K  /(K30.5 PH
0.5
))
2 2 2
where r0 is the rate on fresh catalyst, and a is the activity of the
By assuming negligible surface coverage of C3 H7 , or C3 H6 as the
catalyst. The change of catalyst activity with time generally takes
only dominating surface species, or H atom as the only dominat-
the form [33]
ing surface species, the rate expression can be further simplified to
(M2-1), (M2-2) and (M2-3), respectively. da d
− = d (pi, T )a (8)
dt
k1 (PC3 H8 − PC3 H6 PH2 /K)
r3−1 = 2
(M2-1) where d is the deactivation order and d the “deactivation func-
(1 + K2 PC3 H6 + K30.5 PH
0.5
)
2 tion”, usually expressed in a LHHW form [34–40]. By assuming that
k1 (PC3 H8 − PC3 H6 PH2 /K) coke formation takes place on the same site for propane dehydro-
r3−2 = (M2-2) genation, and only one site is involved in the rate determining steps
(1 + K2 PC3 H6 )2 for both propane dehydrogenation and coke formation, we have a
k1 (PC3 H8 − PC3 H6 PH2 /K) first order catalyst deactivation and the same deactivation rate for
r3−3 = 2
(M2-3) propane dehydrogenation and coke formation, i.e.
(1 + K30.5 PH
0.5
)
2
Cm − Cc
ac = a = (9)
Cm
2.2. Propane cracking
where Cc is the concentration of the active sites occupied by the
In preliminary propane dehydrogenation experiments on the coke and Cm is the total concentration of the active sites. Therefore,
Pt–Sn catalyst, only methane and ethene were detected as cracking Eq. (6) is rewritten as
products, which is in good agreement with the results reported by
dCc Cm − Cc
Wu et al. [28]. Therefore in addition to propane dehydrogenation, = rc0 (10)
dt Cm
the side reaction considered is the cracking of propane to methane
and ethylene, i.e. Replacing Cc with Cm (1 − a) yields

C3 H8 ↔ CH4 + C2 H4 (2) da rc0 kd PC3 H6 kc


− = a= a kd = (M5)
dt Cm 1 + K3 PH2 Cm
The cracking rate is assumed to be first order with respect to
propane,
3. Experimental
rcracking = k2 PC3 H8 (M4)
3.1. Catalyst preparation
2.3. Coke formation and catalyst deactivation
The Pt–Sn/Al2 O3 catalyst used for kinetic study was prepared by
Coke on the catalyst could form from several routes, such as deep incipient impregnation of Al2 O3 (Pural 200) with H2 PtCl6 and SnCl4
cracking of propane, polymerization of olefins (mainly propene) solutions, respectively. The catalysts were calcinated at 530 ◦ C and
and condensation of large molecules. The coke is mainly formed by treated with steam to remove chlorine.
Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26 21

0.4 Table 1
Blank analysis.
O - W=0.10 g
0.35 Temperature (◦ C) Conversion of Selectivity to
C3 H8 (%) C3 H6 (%)
0.3 550 0.6 n.d.
Conversion of Propane

575 0.9 n.d.


0.25 600 1.4 2.33

n.d.: Not detected.


0.2

0.15 gas concentration (Agilent 4890D). The propane conversion and the
selectivity to propene were calculated, respectively, by
0.1
F0 yi0 − Fyi
T=600oC x= (11)
0.05 F0 yi0

0 Fyi − F0 yi0
0 0.2 0.4 0.6 0.8 1 1.2 s= (12)
F0 yC3 H8 0 − FyC3 H8
Diameter of catalyst particle (mm)

Fig. 1. Influences of particle size on pore mass diffusion (600 ◦ C, 65.0 kPa Ar; 35.0 kPa where F0 and F are the inlet and outlet molar flow rate. yi0 repre-
C3 H8 , 80.0 ml/min total flow). sents the molar ratio of specie “i” in the feed gas and yi represents
the molar ratio of specie “i” (i = Ar, CH4 , C2 H4 , C2 H6 , C3 H6 , C3 H8 and
3.2. Catalyst characterization H2 ) in the outlet gas.

The catalyst was characterized by N2 physisorption on ASAP


3.4. Blank test
2010 (Micromeritics, USA) at −196 ◦ C after out-gassing the samples
for at least 5 h at 190 ◦ C and 1 mm Hg vacuum. The specific surface
In order to check whether the wall of the reactor and inert pack-
areas were calculated with BET equation, and the pore volumes and
ing (mainly quartz particles) catalyze the reaction, several blank
pore size distributions were determined from the N2 desorption
tests were carried out at different temperatures with a flow rate
isotherms by using the BJH method. Prior to the kinetic experi-
of 80.0 ml/min and the results are shown in Table 1. A very small
ment, the catalyst was ground and sieved to particle sizes between
amount of propane was converted, which was possibly due to ther-
0.1 and 0.15 mm.
mal cracking. The even smaller amount of C3 H6 formed at 600 ◦ C
might also come from thermal cracking. In this study, however,
3.3. Experimental set-up C3 H6 is considered to be formed only by catalytic dehydrogena-
tion. The formation of propene from propane by thermal cracking
The tubular reactor for the kinetic experiments was a stainless or catalyzed by the wall of the reactor or by the inert packing can
tube with an inner diameter of 6 mm, the temperature of which be neglected. To elucidate whether the reactor has an influence
was maintained by an electrical furnace jacketed outside the reac- on the conversion of propene, both propene and hydrogen with
tor. Inside the catalyst packing inserted was a thermocouple for equal molar fractions were fed into the reactor at 600 ◦ C. About
indication of the real temperature for reaction. Silica sands as inert 1.0% of the propene is converted to methane, ethene and propane
packing were used for a uniform flow distribution. An online gas and, based on the carbon balance, 0.1% of the propene is estimated
chromatography (Agilent 4890D) was used to measure the outlet to be converted into coke.

0.7 3.5. Kinetic experiments


T=600oC

0.6 The kinetic experiments were carried out in the fixed bed reactor
at temperatures ranging from 530 to 600 ◦ C and under 100 kPa. The
total feed flow rate was kept 80.0 ml/min for all the runs. The partial
Conversion of Propane

0.5
pressures of propane, propene and hydrogen were varied by adjust-
ing balance flow. In each experiment 100 mg catalyst was used.
0.4
After being loaded in a tubular reactor, the catalyst was reduced
in situ in a flow of H2 (10 ml/min) for 100 min at 500 ◦ C.
0.3

0.2
Table 2
Catalyst characteristics.
0.1 Δ - W=0.20 g
Shape Pellet
ο - W=0.10 g
Diameter 1.0 mm
0
0 50 100 150 Surface area 56.6 m2 /g
Pore volume 0.25 m3 /g
W/F (g.min/mol)
Average pore diameter 18.5 nm
Major components (wt.%)
Fig. 2. Influences of space velocity on film mass diffusion (600 ◦ C, 65.0 kPa Ar;
Pt 0.5%
35.0 kPa C3 H8 , for W = 0.10 g, F is varied from 22.8 ml/min to 89.6 ml/min; for
Sn 1.5%
W = 0.20 g, F is varied from 39.5 ml/min to 155.6 ml/min).
22 Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26

0.30

Propane conversion
(a) T-1 T-2 T-3 Model prediction

0.20

0.10

0 10 20 30 40 50 60 70 80 90
Time on stream (min)

0.50
Propane conversion

(b)
0.40 Propane-1 Propane-2 propane-3 Model prediction

0.30

0.20

0.10

0 10 20 30 40 50 60 70 80 90
Time on stream (min)

0.35
Propane conversion

(c)
Propene-1 Propene-2 Model prediction
0.25

0.15

0.05

0 10 20 30 40 50 60 70 80 90
Time on stream (min)

0.20
Propane conversion

(d)
Hydrogen-1 Hydrogen-2 Model prediction

0.10

0 10 20 30 40 50 60 70 80 90
Time on stream (min)
Fig. 3. Propane conversions at different conditions: experimental (symbols) and predicted (lines). (a) T-1: 530 ◦ C, T-2: 550 ◦ C, T-3: 600 ◦ C (C3 H8 , 35.0 kPa, C3 H6 : 4.0 kPa H2 :
6.0 kPa, Ar, balance), (b) propane-1: 11.0 kPa, propane-2: 35.0 kPa, propane-3: 49.0 kPa (575 ◦ C, 0.0 kPa H2 , balance Ar), (c) propene-1: 0.0 kPa, propene-2: 10.0 kPa (575 ◦ C,
35.0 kPa C3 H8 , 7.0 kPa H2 , balance Ar), (d) hydrogen-1: 12.0 kPa, hydrogen-2: 22.0 kPa (575 ◦ C, 35.0 kPa C3 H8 , 7.0 kPa C3 H6 , balance Ar).

4. Results and discussion 4.2. Effects of mass and heat transports on the reactions

4.1. Characteristics of catalyst Catalysts with four different averaged diameters were used to
investigate the influences of internal mass diffusion. The experi-
Table 2 summarizes the characteristics of the catalyst. ments were carried out at 600 ◦ C, the highest temperature in the
Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26 23

interested temperature region. The W/F was kept 30 g min/mol con- 40.0
stant. Fig. 1 shows that the initial propane conversions of propane
did not change remarkably when the size of the catalyst was smaller

Forward reaction rate, 10-7mol/(g.s)


than 0.3 mm. Therefore, catalyst with an averaged size of 0.15 mm
was employed in the experiments to eliminate the influences of 30.0
pore diffusion.
The catalyst with an averaged size of 0.15 mm was used to study
the influences of the effects of external mass and heat transport on
the reactions. Experiments with 0.10 g and 0.20 g catalyst, respec- 20.0
tively, were carried out. The flow rate was varied to change the
gas velocity through the catalyst bed. Fig. 2 shows the changes of
initial propane conversions of propane with W/F ratios. The con-
version at any space velocity should be independent of the linear 10.0
velocity through the bed in the absence of interphase transport
limitations [17]. As seen in Fig. 2, for the same W/F ratio, the conver-
sion will be different when the W/F ratio is large (>50 g min/mol),
which is because of the large difference in the linear velocity 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
and the external mass and heat transport. When the W/F ratio is Propane partial pressure, kPa
smaller than 50 g min/mol, the influence of external transport can
be neglected. In our kinetic experiments, the W/F ratios were less Fig. 4. Dehydrogenation rates at different propane partial pressures (T = 550 ◦ C,
than 30 g min/mol to eliminate the influences of transport limita- hydrogen partial pressure in feed: 8.0 kPa, argon: balance).
tion.
fied models (M2-1), (M2-2) and (M2-3) will certainly not able to fit
4.3. Propane conversions at different conditions the experiments as accurately as (M2), but they have less param-
eters and some of them may outperform (M2) if the number of
The effects of temperature, propane, propene and hydrogen parameters is also an account. To compare the different kinetics
pressures on propane dehydrogenation are shown in Fig. 3(a–d), compromising the fitting accuracy and number of fitting parame-
respectively. Propane conversion increases remarkably with tem- ters, the criteria suggested by Akaike [41] was used,
perature (Fig. 3(a)) as expected. The conversion decreases with
⎡ ⎤

l
2
time on stream and more rapidly at higher temperatures owing to ⎢ (Yobsj − Ȳobs ) ⎥
faster catalyst deactivation. The propane conversion decreases with ⎢ ⎥
⎢ j=1 ⎥ 2p
increasing propene concentration (Fig. 3(c)), but is only slightly MSC = ln ⎢ ⎥− l (13)
⎢ l ⎥
influenced by hydrogen concentration (Fig. 3(d)). It indicates a ⎣ (Y −Y ) 2⎦
obsj calj
strong adsorption of propene, which occupies most of the surfaces.
j=1
High propene concentration decreases the free site for adsorption of
the reactant molecules, and thus lowers the reaction rate. Although This criterion takes into consideration the span of independent
no significant effect of hydrogen on propane dehydrogenation is data set and the number of experiments points for parameter esti-
observed, the deactivation rate is lower at higher hydrogen con- mation and has been normalized to be independent of the scale of
centration. It indicates an inhibition of coke formation by hydrogen. data. For the same data set for parameter fitting, the MSC value
The deactivation rate depends more strongly on propene (Fig. 3(c)) depends only on the fitting accuracy and the number of fitting
than on propane (Fig. 3(b)). It implies that propene instead of parameters. The MSC value is therefore a good indication of the
propane is possibly the coking precursor. usefulness and appropriateness of the model because it is a good
balance between under-fitting and over-fitting. The best model is
5. Model discrimination and parameter estimation the one with the largest MSC value.

Referring to the rate Eqs. (M1)–(M3) corresponding to the 30.0


different reaction mechanisms or different assumption of rate
determining step, one can see that if the adsorption terms and the
Forward reaction rate, 10-7 mol/(g.s)

reversible reaction are negligible, (M1) and (M2) can be reduced to


one rate equation which is the first order with respect to propane
and zero order to hydrogen, while (M3) can be reduced to a rate 20.0
equation which is the first order with respect to propane and −0.5
to hydrogen. Therefore it is possible to discriminate (M3) from the
other kinetic models based on the determination of the reaction
order of hydrogen. For this reason, some experiments were carried
out at very low propane and hydrogen partial pressures (smaller 10.0
than 10 kPa) and a high space velocity to guarantee negligible
adsorption and reversible reaction. The rates of propane consump-
tion shown in Figs. 4 and 5 indicate that the reaction is indeed first
order to propane but zero order to hydrogen. Therefore, the kinetic
experiments clearly exclude (M3) from the mechanisms. As it will
4.0 6.0 8.0 10.0 12.0
also be shown later, this model cannot fit experimental data well.
Hydrogen partial pressure, kPa
(M1) and (M2) are supposed to be better than (M1-1) because
both of them have taken into account of hydrogen adsorption, Fig. 5. Dehydrogenation rates at different hydrogen partial pressures (T = 550 ◦ C,
which is important in propane dehydrogenation [25]. The simpli- propane partial pressure in feed: 1.5 kPa, argon: balance).
24 Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26

Table 3 Table 4
MSC values for different models. Parameters for kinetic model.

Model MSC SSR Parameter Parameters Value and unit (confidence degree, 95%)
number
k1 k10 = 1.4e−1 ± 2.5e−2 (mmol/(s g atm))
(M1) 4.63 10.84e−4 3 Ea1 = 44.7 ± 16.9 (kJ/mol)
(M1-1) 4.55 12.53e−4 2 k2 k20 = 1.0e−3 ± 4.8e−5 (mmol/(s g atm))
(M2) 4.60 10.49e−4 4 Ea2 = 104.8 ± 9.9 (kJ/mol)
(M3) 4.04 18.21e−4 4 K2 K20 = 8.0 ± 3.8 (atm−1 )
(M2-1) 4.62 10.90e−4 3 H2 = 94.3 ± 37.9 (kJ/mol)
(M2-2) 4.53 12.74e−4 2 K3 K30 = 1.0 ± 0.2 (atm−1 )
(M2-3) 4.45 19.18e−4 2 H3 = 238.9 ± 58.7 (kJ/mol)
kd fc0 = 7.4e−4 ± 4.9e−4 (mmol/(s g atm))
Eac = 127.8 ± 55.3 (kJ/mol)
By assuming plug-flow, the reactor models with different reac-
tion kinetics and rate constants are integrated by Runge–Kutta
method to give the outlet concentrations under different operating also adsorbs relatively strongly at high hydrogen pressures, which
conditions. To simplify the discrimination of the kinetic models, the reduces significantly not only the rate of propane dehydrogena-
coking rate equations and catalyst deactivation is temporarily not tion but also the rates of coke formation and catalyst deactivation.
considered, and for each run of the experiments with a series of Although (M2-1) has a fitting residual slightly higher than (M2) and
time dependent concentration profiles, only the first analysis data (M1), it is considered as the best kinetic model for propane dehy-
(i.e., the propene and methane concentrations), which indicate the drogenation because the underline reaction mechanism is reliable,
initial rates of dehydrogenation and cracking, are used for the data the rate equations are simple, and the predictions are acceptable.
fitting. In the fitting, the reaction constants (k1 , k2 , fc , K1 , K2 , K3 ) are The model (M3) cannot well describe the experimental results,
expressed as the reparametrized form of Arrhenius equations (Eqs. which is in accordance with the above discussions that the sec-
(14) and (15)) and a reference temperature Tm (565 ◦ C) is employed. ond dehydrogenation, namely propyl dehydrogenation could not
be the rate-determining step. However, this is in disagreement with
Ea 1 1 the results of Biloen et al. [23]. The contraversial results were also
k = k0 exp − − (14)
R T Tm observed on Cr catalysts [16,24]. This implies that the rate deter-
H 1 1 mining step may be dependent on the structure of the catalyst, e.g.
K = K0 exp − (15) the size and surface orientation. A systematic study is necessary in
R T Tm
the future to test this hypothese.
The parameters are determined by the least-square method to The kinetic parameters in dehydrogenation (M2-1), cracking
minimize the SSR, (M4), and deactivation (M5), are then estimated by fitting the
changes of propene and methane concentrations with time on

l
SSR = (Yobsj − Ycalj )2 (16) stream (about 400 data points). It is noted here that, because of the
small conversion of propane, the activity of the catalyst is assumed
j=1
to be independent of the axial position. The kinetic parameters in
where Y is the outlet concentration and i represents propane, (M2-1) and (M4) obtained by fitting to initial experimental data are
propene, hydrogen, methane and ethene. used the initial values for the model fitting. No significant change
Listed in Table 3 are the MSC and SSR values for all the seven in parameter values is found by re-fitting all the data. Table 4 sum-
kinetic models. (M1) emerges as the best one according to the marizes the values of all the parameters of the reaction kinetics.
MSC value; however, it involves the lumping of the elementary The predictions made by the established reaction kinetics (M2-
steps. (M2) has the least SSR value, and is the best model to fit the 1) are shown in Fig. 3 (dashed line), displaying a good agreement
experiments; however, it has more parameters than (M1). (M1-1), with the experimental data.
as expected, is worse than (M1) and (M2) since it has not taken Tables 5 and 6 summarize the reported activation energies
the adsorption of hydrogen into account. Among the three models of propane dehydrogenation and also the adsorption energies of
derived from (M2), (M2-2) and (M2-3) have large fitting resid- propane, propene and hydrogen on platinum determined by exper-
uals, indicating the assumption of negligible coverage of either iments and DFT calculation. The activation energy for propane
propene or hydrogen is not appropriate. It indicates that hydrogen dehydrogenation varies from 30 to 170 kJ/mol, while it is deter-

Table 5
Activation/adsorption energies determined by experiments.

Dehydrogenation Model Catalyst Activation/adsorption Refs.


energies (kJ/mol)

−r = kfor PC3 H8 − krev PC3 H6 PH Pt–Sn/Al2 O3 Eafor = 34.8 ± 19.6 [18]
2
(0.54, 1.53 wt.% Earev = -89.5 ± 42.3
Pt and Sn)
k1 (PC3 H8 − (PC3 H6 PH2 /Keq ))
−r = a Pt–Sn–K/Al2 O3 Ea1 = 34.57 ± 9.13 [19]
1 + (PC3 H6 /KC3 H6 )
(0.05, 0.14 and HC3 H6 =
0.10 wt.% Pt, Sn 85.82 ± 22.46
and K)
Power law Pt/Al2 O3 Ea = 121 ± 19.6 [23]
Pt–Au/Al2 O3
k1 (KC3 H8 PC3 H8 − KC3 H6 KH2 PC3 H6 PH2 /KP )
−r = 2
Pt–Sn/Al2 O3 Ea1 = 169.7 [36]
(1 + KC3 H8 PC3 H8 + KC3 H6 PC3 H6 + KH2 PH2 ) (0.15, 0.15 wt.% HC3 H8 = 48.07
Pt and Sn) HC3 H6 = 32.02
HH2 = 30.68
Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26 25

Table 6 different operating conditions are well predicted by the deac-


Adsorption energies determined by DFT.
tivation model. The strategy for kinetic modeling employed in
Species Crystal surface Adsorption Refs. the present work is to directly integrate the coking formation
energy (kJ/mol) kinetic model into the deactivation model, making it possible
Propane Pt(1 1 1) 5.8 [25] to avoid measurement of coke content in situ. It makes the
Propene Pt(1 1 1) 38.5–48.2 [43] modeling relatively simple but still keeps a good predictive capa-
Pt(1 1 1) 89.7 [25] bility.

mined to be 44.7 kJ/mol in the present work, which agrees with the Acknowledgements
results reported by Lobera et al. [19] and Gascón et al. [22]. How-
ever, high activation energies were obtained by Biloen et al. [23] and This work is supported by NSFC (No. 20736011), PCSIRT
Chen et al. [42]. Although there was no mention about the Pt parti- (IRT0721), 111 Project (B08021), and the Opening Project of State
cle sizes in these references, we still regard that the differences of Key Laboratory of Chemical Engineering (SKL-ChE-08C05).
the activation energies are resulted by the differences in the particle
sizes of the Pt catalysts. It is believed that the surface is mostly ter-
References
minated by steps and edges on smaller particles, while the surface
is mostly terminated by terraces on larger particles [44,45]. In addi- [1] D. Akporiaye, S.F. Jensen, U. Olsbye, F. Rohr, E. Rytter, M. Ronnekleiv, A.I. Spjelka-
tion, reactions proceeding on stepped surfaces tend to have smaller vik, Ind. Eng. Chem. Res. 40 (2001) 4741–4748.
activation energies than those proceeding on terrace surfaces [2] J. Cosyns, J.-A. Chodorge, D. Commereuc, B. Torck, Hydrocarbon Process. 77
(1998) 61.
[46]. [3] M. Santhosh Kumar, D. Chen, J.C. Walmsley, A. Holmen, Catal. Commun. 9
The experimental adsorption energies of propene reported (2008) 747–750.
by Chen et al. [42] and Lobera et al. [19] were 32.02 and [4] C. Yu, Q. Ge, H. Xu, W. Li, Appl. Catal. A: Gen. 315 (2006) 58–67.
[5] L. Bai, Y. Zhou, Y. Zhang, H. Liu, X. Sheng, Ind. Eng. Chem. Res. 48 (2009)
85.82 kJ/mol, respectively, and the difference was mainly due
9885–9891.
to the different catalysts employed in the experiments. The [6] Y. Zhang, Y. Zhou, A. Qiu, Y. Wang, Y. Xu, P. Wu, Acta Phys.-Chim. Sin. 22 (2006)
calculated adsorption energy of propene was in the range of 672–678.
38.5–89.7 kJ/mol [25,43], and the differences among these adsorp- [7] Y. Zhang, Y. Zhou, H. Liu, Y. Wang, Y. Xu, P. Wu, Appl. Catal. A: Gen. 333 (2007)
202–210.
tion energies of propene were mainly because of the different [8] C. Yu, H. Xu, Q. Ge, W. Li, J. Mol. Catal. A: Chem. 266 (2007) 80–87.
models employed for calculation. The adsorption energy of propene [9] E.L. Jablonski, A.A. Castro, O.A. Scelza, S.R. de Miguel, Appl. Catal. A: Gen. 183
determined in our study is 94.3 kJ/mol, which is reasonable (1999) 189–198.
[10] M. Larsson, M. Hultén, E.A. Blekkan, B. Andersson, J. Catal. 164 (1996) 44–53.
according to the results of both experimental and DFT calcula- [11] K. Takehira, Y. Ohishi, T. Shishido, T. Kawabata, K. Takaki, Q. Zhang, Y. Wang, J.
tion. Catal. 224 (2004) 404–416.
However, the adsorption energy of hydrogen (238.9 kJ/mol) in [12] S.D. Rossi, G. Ferraris, S. Fremiotti, A. Cimino, V. Indovina, Appl. Catal. A: Gen.
81 (1992) 113–132.
our model is higher than that reported by Chen et al. [42], indicat- [13] H.P. Rebo, E.A. Blekkan, L. Bednárová, A. Holmen, Stud. Surf. Sci. Catal. (Elsevier)
ing the adsorption of hydrogen is stronger in our study. It is also 126 (1999) 333–340.
probably contributed by the corners and steps on smaller Pt parti- [14] F. Ashmawy, C. Mchuliffe, J. Chem. Soc., Faraday Trans. 80 (1984) 221.
[15] M. van Sint Annaland, J.A.M. Kuipers, W.P.M. van Swaaij, Catal. Today 66 (2001)
cle size [47,48]. On Pt with silica-supported, the heats of hydrogen 427–436.
adsorption determined by microcalorimetric measurement [49,50] [16] D.S. Jackson, J. Grenfell, I.M. Matheson, G. Webb, in: G.F. Froment, K.C. Waugh
were 95 and 93 kJ/mol at 300 K and 403 K, respectively. The ini- (Eds.) Stud. Surf. Sci. Catal. (Elsevier) (1999) 149–155.
[17] J. Le Page, Applied Heterogeneous Catalysis: Design, Manufacture, Use of Solid
tial differential heats were essentially the same for the 1.2 wt.%
Catalysts, Editions Technip, 1987.
Pt, 6:1 Pt/Sn and 6:1 Pt/Sn catalysts, while for the 1:3 Pt/Sn cat- [18] M. Larsson, N. Henriksson, B. Andersson, Appl. Catal. A: Gen. 166 (1998) 9–19.
alyst, the initial heat was approximately 75 kJ/mol [50]. It seems [19] M.P. Lobera, C. Téllez, J. Herguido, M. Menéndez, Appl. Catal. A: Gen. 349 (2008)
that the present model overestimated the adsorption heat of hydro- 156–164.
[20] I.S. Nam, J.R. Kittrell, Ind. Eng. Chem. Process Des. Dev. 23 (1984) 237–242.
gen. [21] J. Corella, A. Monzón, Ann. Quim. 84 (1988) 205.
[22] J. Gascón, C. Téllez, J. Herguido, M. Menéndez, Appl. Catal. A: Gen. 248 (2003)
105–116.
6. Conclusion [23] P. Biloen, F. Dautzenberg, W. Sachtler, J. Catal. 50 (1977) 77–86.
[24] I. Suzuki, Y. Kaneko, J. Catal. 47 (1977) 239–248.
In this study, propane dehydrogenation experiments are [25] M. Yang, Y. Zhu, C. Fan, Z. Sui, D. Chen, X. Zhou, J. Mol. Catal. A: Chem. 321 (2010)
42–49.
carried out on Pt–Sn catalyst at different temperatures and [26] W. Weinberg, Y. Sun, Science 253 (1991) 542–545.
with different inlet propane, propene and hydrogen concen- [27] M. van Sint Annaland, H.A.R. Scholts, J.A.M. Kuipers, W.P.M. vanSwaaij, Chem.
trations. A reaction kinetic model of propane dehydrogenation Eng. Sci. 57 (2002) 855–872.
[28] W. Wu, Chem. React. Eng. Technol. 25 (2009) 41–45.
including cracking and deactivation is established following the [29] P. Praserthdam, N. Grisdanurak, W. Yuangsawatdikul, Chem. Eng. J. 77 (2000)
LH approach. The first C–H cleavage in propane molecule is 215–219.
identified as the rate-determining step. The underline reac- [30] A. Brito, R. Arvelo, R. Villarroel, F.J. Garcia, M.T. Garcia, Chem. Eng. Sci. 51 (1996)
4385–4391.
tion mechanism of propane dehydrogenation in the kinetic [31] S.K. Sahoo, P.V.C. Rao, D. Rajeshwer, K.R. Krishnamurthy, I.D. Singh, Appl. Catal.
model agrees with experiments and quantum chemistry calcula- A: Gen. 244 (2003) 311–321.
tions. [32] P. Magnoux, H.S. Cerqueira, M. Guisnet, Appl. Catal. A: Gen. 235 (2002) 93–99.
[33] S. Szépe, O. Levenspiel, Catalyst deactivation, in: Proceedings of the Fourth
The kinetic model of coke formation taking into account the
European Symposium on Chemical Reactive Engineering, Pergamon Press,
effects of propene and hydrogen is integrated into the deacti- Brussels, NY, 1971, pp. 165–276.
vation model. The propene is identified as the main precursor [34] C. Chu, Ind. Eng. Chem. Fundam. 7 (1968) 509–514.
[35] E.E. Wolf, E.E. Petersen, J. Catal. 47 (1977) 28–32.
for the coke formation and hydrogen suppressed coke forma-
[36] J. Corella, J.M. Asua, J. Bilbao, Can. J. Chem. Eng. 59 (1981) 647–648.
tion and thus deactivation. The model describes well the kinetics [37] J. Corella, J.M. Asua, Ind. Eng. Chem. Proc. Des. Dev. 21 (1982) 55–61.
of catalyst deactivation, which is related to the rate of coke [38] R.D. Srivastava, A.K. Guha, J. Catal. 91 (1985) 254–262.
formation and takes into consideration the facts that propene [39] J. Corella, A. Monzón, J.B. Butt, R.P.L. Absil, J. Catal. 100 (1986) 149–157.
[40] A. Monzón, E. Romeo, A. Borgna, Chem. Eng. J. 94 (2003) 19–28.
promotes while hydrogen depresses coke formation, is proposed. [41] M. Akaike, Math. Sci. 14 (1976) 5–9.
The changes of propane conversion with time-on-stream under [42] G.W. Chen, Y.R. Yang, Chem. React. Eng. Technol. 14 (1998) 130–137.
26 Q. Li et al. / Applied Catalysis A: General 398 (2011) 18–26

[43] A. Valcárcel, J.M. Ricart, A. Clotet, A. Markovits, C. Minot, F. Illas, Surf. Sci. 519 [47] J. Rostrup-Nielsen, J.N. rskov, Top. Catal. 40 (2006) 45–48.
(2002) 250–258. [48] J.K. Nørskov, T. Bligaard, F. Abild-Pedersen, I. Chorkendorff, C.H. Christensen,
[44] R. Rioux, H. Song, J. Hoefelmeyer, P. Yang, G. Somorjai, J. Phys. Chem. B 109 Chem. Soc. Rev. 37 (2008) 2163–2171.
(2005) 2192–2202. [49] M.A. Natal-Santiago, S.G. Podkolzin, R.D. Cortright, J.A. Dumesic, Catal. Lett. 45
[45] R. Van Hardeveld, F. Hartog, Surf. Sci. 15 (1969) 189–230. (1997) 155–163.
[46] Z.P. Liu, P. Hu, J. Am. Chem. Soc. 125 (2003) 1958–1967. [50] R.D. Cortright, J.A. Dumesic, J. Catal. 148 (1994) 771–778.

You might also like