Vous êtes sur la page 1sur 8

4116 Ind. Eng. Chem. Res.

2007, 46, 4116-4123

Methanol Conversion to Light Olefins over SAPO-34: Reaction Network and


Deactivation Kinetics

D. Chen,*,† A. Grønvold,‡,§ K. Moljord,‡,⊥ and A. Holmen†


Department of Chemical Engineering, Norwegian UniVersity of Science and Technology (NTNU),
N-7034 Trondheim, and SINTEF Applied Chemistry, N-7034 Trondheim, Norway

The kinetics of the reactions involved in the conversion of methanol to light olefins over SAPO-34, including
deactivation caused by coke deposition, has been studied in an oscillating microbalance reactor between 673
and 823 K, space velocities from 50 to 2000 g/gcat,h and methanol partial pressures from 7 to 83 kPa. The
proposed reaction network involves dimethyl ether as an unstable primary product, all the olefins formed in
parallel as secondary products, and the paraffins formed from further reactions of olefins as stable tertiary
products. The selectivity to ethene increased with increasing coke content and temperature. A kinetic model
including the deactivating effect due to coke deposition has been developed to properly simulate the changes
in activity and selectivity with the coke content. A linear dependency between the coke content and the
reaction rate gave the best representation of the experimental data.

Introduction developed by Bos and Tromp.17 The conversion and the


selectivities were studied as a function of the coke content of
From synthesis gas, which can be obtained by gasification, the catalyst. An exponential dependency was found to give the
partial oxidation, or steam reforming from raw materials, such best representation of the effect of coke on the reaction rate.
as biomass, coal and natural gas, methanol is synthesized using
Unfortunately, direct measurements of coke formation in situ
a proven technology. The catalytic conversion of methanol to
are not possible in a fixed-bed reactor. The determination of
olefins (MTO) was originally an intermediate step in Mobil’s
coke in this case must be done either by burning off the coke
process to convert methanol to synthetic gasoline using H-
in the reactor or by taking the catalyst out of the reactor for
ZSM-5 as catalyst.1 However, interest has recently shifted
weighing. Such measurements of coke content in a fixed-bed
toward the MTO process following the increased demand for
olefins.2 As a consequence of the discovery of aluminophosphate reactor may, therefore, introduce additional errors, and it is also
molecular sieves, especially SAPO-34,3 it is possible to a time-consuming process. The parameters in the model were,
selectively produce ethene and propene at high methanol thus, estimated from a limited amount of data. Gayubo et al.18
conversions, due to the narrow pores (0.43 nm) extending in performed a kinetic study of MTO in SAPO-34-based catalysts
three dimensions and the mild acidity of SAPO-34.4 However, in a fixed-bed reactor, in which the catalysts were prepared by
fast deactivation of the catalyst due to coke formation has been agglomerating the SAPO-34 (25 wt %) with bentonite (30 wt
reported.5-7 %), using fused alumina as the inert charge (45 wt %). The
Most kinetic studies on methanol conversion have been made MTO reaction was studied as a function of time on-stream, and
on HZSM-5 catalysts, and a number of simplified kinetic models the initial reaction rates at different conditions were obtained
have been developed. The complexity of the model varies by extrapolating ration rates to zero time. The authors have
according to the degree of lumping proposed.8-12 A very detailed proposed a reaction network of MTO taking into account four
analysis has been performed by Mihail et al.13-14 in which 33 individual steps for the production of ethene, propene, butanes,
reactions were used. Single-event kinetic modeling of MTO on and other hydrocarbons (pentenes + paraffins). A kinetic model
H-ZSM-5 has been performed by Park and Froment15-16 on the for the reaction network was developed; however, kinetics of
basis of a detailed mechanistic description of the MTO reaction. coke deposition and deactivation during MTO were not included.
Sedran et al.11 tested several alternative kinetic models for Alwahabi and Froment19 have recently developed a single-event
conversion of methanol to hydrocarbons, including an expo- kinetic model of the MTO reaction on SAPO-34 based on the
nential activity function to account for the deactivation of the experimental data in a fixed-bed reactor. The deactivation caused
catalyst. A common characteristic of the kinetic models is the by coke formation was modeled by fitting the conversion
autocatalytic effect in the reaction network. However, to our changes with time on-stream in the fixed-bed reactor. However,
knowledge, autocatalytic behavior has not been reported for experimental data of coke formation and the deactivating effects
conversion of methanol to light olefins over SAPO-34. of the coke formed are missing.
A kinetic model for MTO over SAPO-34 based on results An oscillating microbalance reactor (TEOM) provides an
obtained in a plug flow fixed-bed reactor at 723 K has been opportunity to achieve a better quantitative understanding of
the activity and the selectivity changes with the coke forma-
* To whom correspondence should be addressed. Tel: +47 73593149. tion.20 The mechanism and kinetics of the coking process have
E-mail: chen@nt.ntnu.no. been investigated in detail in our previous work.21 The present

NTNU. work will focus on the selectivity to olefins and its changes

SINTEF Applied Chemistry.
§
Present address: Hydro Oil & Energy Research Center, Box 2561, with coke formation. A reaction network is proposed, and a
N-3907 Porsgrunn, Norway. kinetic model for this reaction network is developed, including
⊥ the deactivation due to coke deposition.
Present address: Statoil, Postuttak, N-7005 Trondheim, Norway.

10.1021/ie0610748 CCC: $37.00 © 2007 American Chemical Society


Published on Web 12/10/2006
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 4117

Experimental Section the OPE of ethene formation is, thus, relatively large. From the
OPEs, the types of products can be identified.
The setup of the TEOM reactor, the catalysts, and the The OPE curves for the yield of C2-C6 are almost straight
experimental methods are identical to those described in our lines, indicating that the selectivities to olefins are almost
previous work.22 A more detailed description of the TEOM constant on the fresh catalyst, regardless of the conversion. In
reactor can be found in a recent review.23 Calcined SAPO-34 other words, the gas-phase olefins formed over SAPO-34 are
with a unit cell composition of (Si2.88Al18P15.12)O72 was ob- quite stable, and the secondary reactions of olefins do not seem
tained from SINTEF Applied Chemistry. SAPO-34 catalysts to be very important. It has to be noted that the linear OPE
are typically cubic crystals with sizes of ∼2 µm. The SAPO- curves do not pass through zero conversion, which is a
34 particles (52-140 mesh) were dried in situ at 773 K in characteristic of secondary products. The hydrocarbons, includ-
flowing He for more than 3 h. The MTO reaction was performed ing ethene, propene, C4, C5, and C6, can thus be considered as
at WHSV ranging from 57 to 2558 g (g of cat)-1 h-1, methanol stable secondary products formed in parallel from DME at a
partial pressure ranging from 7.2 to 83 kPa, and temperatures methanol conversion less than 100%. Light paraffins (ethane
between 673 and 823 K. The runs with different space velocities and propane) are formed only at high conversions, meaning that
were carried out at 698 K and a methanol partial pressure of paraffins are stable tertiary products. The OPE curve for coke
7.2 kPa to obtain a relatively low coking rate. The runs with went through 0 at an almost identical conversion as for the
different methanol partial pressures were also performed at olefins (10%), and it increased with increasing conversion. Coke
698 K. The space velocity was adjusted to ensure an identical may, therefore, be considered as a secondary plus stable tertiary
conversion when the partial pressure of methanol was increased. product. The detailed mechanism of coke deposition has been
For high methanol partial pressures, very high space velocities discussed previously.21 Reaction intermediates such as carbe-
were required as a result of very rapid coke formation. nium ions inside the pores of SAPO-34 were considered as the
The conversion and selectivities were calculated on a CH2 major coke precursors.
basis (CH3OH f CH2 + H2O). The isomers of C4, C5, and C6 Equilibrium between olefins formed during methanol to
were lumped according to their respective carbon numbers, and gasoline over ZSM-5 is achieved.25 This is different for SAPO-
the selectivities of C1-C6 hydrocarbons were calculated by 34, in which the olefin distribution is far from equilibrium.26
normalization exclusive coke. The yield-conversion plots were Additionally, equilibrium cannot explain the constant olefin
presented on a weight basis, and the formation of DME, distribution at different space velocities, since the partial pressure
hydrocarbons (C1-C6), and water was included. The water of olefins (related to the level of oxygenates conversion) has a
fraction was calculated from the mass balance. Due to the fast profound influence on the equilibrium between olefins.25 Instead,
coke deposition on the catalyst, pulse experiments were used we believe that the constant olefin distribution is caused by the
to study the changes in activity and selectivity with coke content. low reactivity of olefin and the lower adsorption capacity for
Pulses of 3-min duration were used for PmeOH < 30 kPa, whereas olefins relative to methanol.27 The low olefin reactivity over
1-min pulses were used at high temperatures (773-823 K). It SAPO-34 has been demonstrated from the observation that the
has been shown previously that the conversion and selectivity reaction rate of propene conversion was almost 100× lower
were not affected by the pulse size.21 than the MTO reaction.27 The low reactivity of olefins was also
observed by Dahl et al.28 during ethanol and propanol conversion
over SAPO-34. Salehirad and Anderson29 reported that ethene
Reaction Network conversion was faster on the methylated SAPO-34 than on the
bare catalyst, but the reaction rate of propene was still much
As discussed previously,20 the yield-conversion plot is a
lower than for methanol.
powerful tool for distinguishing the type of product (stable or
It should be noted that the conversion of methanol was
unstable, primary or secondary) and the type of deactivation
<100% at the reaction conditions used in this work. The
(selective or nonselective). Selective deactivation has been
presence of methanol, DME, and water might to some extent
illustrated for the MTO reaction over SAPO-34. However, only
contribute to the lower reactivity of olefins through competitive
a simple reaction model was used previously,20 in which all
adsorption between polar molecules (methanol, DME, and
the hydrocarbons were lumped together; dimethyl ether (DME)
water) and less polar olefins. Water is known to suppress the
was treated as a primary product; and the olefins, as secondary
conversion of olefins and coke deposition.30,31
products. The present work differentiates the individual hydro-
Figure 1 suggests methane to be a stable primary plus
carbons according to product types and determines the effect
secondary product. Methane was suggested in the literature29,31,32
of coke formation on the distribution of olefins.
as a primary product formed directly from methanol. A
In Figure 1, the product yields at different space velocities mechanism with surface methoxy species as intermediates has
are plotted against the conversion at 698 K and a methanol been suggested to explain the formation of methane:
partial pressure of 7.2 kPa. Initial conversions are obtained by
varying the space velocity, and the different symbols in Figure
1 illustrate different space velocities. For a certain space velocity,
the conversion decreased with increasing coke content. The solid
lines enclosing such loops are OPEs (optimum performance
envelopes20,24), which are obtained from approximately the first
pulse (TOS ) 2 min). Although this treatment is not very
rigorous due to deactivation during the first pulse, it is expected
not to lead to large errors with respect to the selectivity plot in
this case, since the selectivity change with coke content is small
at low coke contents (as shown later in Figure 5). However, it
should be noted that the ethene selectivity changes significantly Another possible path for methane formation has been also
at very low coke content, and the uncertainty in determining reported29,33 (reaction 2):
4118 Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007

Figure 1. Yields versus methanol conversion during MTO over SAPO-34 at PMeOH ) 7.2 kPa and 698 K. Symbols with lines represent experimental results
at different WHSV (g (g of cat)-1 h-1): 384 (O), 253 (×), 114 (+), 82 (∆), 57 (]). Solid lines: OPE curves obtained from data of the first pulse.

where the hydride source is a large molecule deposited on the


surface of the catalyst, referred to as coke. Both mechanisms
may contribute to the methane formation. However, it is difficult Figure 2. Reaction network for MTO over SAPO-34.
to discuss the mechanism in more detail, since the yields of
methane in the present work are very low, and a relatively large mechanism,26 regardless of the nature of the “hydrocarbon pool”,
error in the analysis is expected. which is still in debate.
The reaction network given in Figure 2 is based on the
discussion above. All the hydrocarbons are formed mainly as Selectivities During MTO on SAPO-34
parallel products from DME. The methanol molecule might also
contribute to the olefin formation as a methylating agent inserted Effect of Space Velocity. All the OPE lines for the
into the surface intermediates, yielding a chain growth. However, hydrocarbons in Figure 1 are nearly linear, indicating that the
it is difficult to determine the degree of olefin formation from selectivities are independent of conversion (or space velocity).
DME and directly from methanol, so Figure 2 illustrates only This is in good agreement with the results of Liang et al.,5 but
the possible pathways for olefin formation from DME. The it is different from the observation of Cai et al.34 and Bos et
reaction network can be explained by the hydrocarbon pool al.17 These authors observed that the selectivity to ethene
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 4119

Figure 3. Ethene selectivity with coke content at 698 K but different partial Figure 5. Selectivity versus the coke content at 698 K, WHSV ) 57 g
pressures. ], 15 Kpa, 384 g (g of cat)-1 h-1; O, 30 KPa and 768 g (g of (g of cat)-1 h-1and a methanol partial pressure of 7.2 kPa. 9, Methane; b,
cat)-1 h-1; ∆, 60 KPa and 1538 g (g of cat)-1 h-1; 0, 83 kPa and 2558 g ethene; [, propene; 2, butene; ×, C5; +, C6.
(g of cat)-1 h-1.

Figure 6. Relative selectivities versus coke content at 698 K, WHSV )


57 g (g of cat)-1 h-1 and a methanol partial pressure of 7.2 kPa. b, Ethene;
Figure 4. Selectivity with temperature during the MTO reaction on SAPO- [, propene; 2, butene; ×, C5; +, C6.
34 at 7.2 kPa methanol partial pressure, 385 g (g of cat)-1 h-1and a coke
content of 4.6 wt %. [, ethene; 9, propene; b, C4; ×, C5; 2, methane.
catalyst containing coke were lower than on the fresh catalyst,
decreased with increasing space velocity. One important dif- whereas the opposite trend was observed for ethene, as shown
ference is the higher space velocity (57-384 h-1) used in the in Figure 1. This is an example of selective deactivation,20 in
present work than in the studies of Cai et al.34 and Bos et al.17. which each reaction step has its own deactivation rate.
As discussed above, the secondary reactions of olefins are not The change in selectivity of different hydrocarbons with the
important at the conversions studied in the present work, in coke content is shown in Figure 5 for WHSV ) 57 g (g of
which the primary olefin selectivity during MTO over SAPO- cat)-1 h-1. Figure 6 shows variations in relative selectivities at
34 is dominant. However, further reactions involving olefins the same conditions as in Figure 5, where the initial selectivity
are more important at the low space velocities of 2-10 g (g of was obtained by extrapolation to zero coke content from Figure
cat)-1 h-1, as in the study of Cai et al.34 This might explain the 5. These results clearly show that the selectivity to ethene
difference in the space velocity dependency on the selectivity increases with increasing coke content. The following decrease
to olefins. in selectivity of the olefins with coke content was found.
Effect of Partial Pressure. Experiments at identical conver-
sion were achieved by adjusting both the partial pressure of meth- C6 > C5 > C4, C3
anol and the space velocity. Product selectivities were found to
be independent of the partial pressure at constant coke content. Changes in the product selectivities for a complex reaction due
An example of the changes in the selectivity to ethene with coke to coke formation on a zeolite-type catalyst can be caused by
content at different partial pressures is illustrated in Figure 3. changes in conversion, acidity (including density and strength
Effect of Temperature. The effect of temperature on the distribution), and shape selectivity.20 The linear OPE pattern in
product (hydrocarbon) selectivities at a constant coke content the yield-conversion plots for all the hydrocarbons (Figure 1)
of 4.6 wt % is presented in Figure 4. The selectivity to ethene shows constant selectivities at different conversions on the fresh
and methane increases while the selectivity of propene decreases catalyst. In other words, the change in the conversion level itself
with temperature. C4 and C5 selectivities are only slightly due to the coke deposition should not influence the selectivity
influenced. As shown previously, the coke selectivity increases to the hydrocarbons. The heat of adsorption as function of the
significantly with temperature.21 site coverage of methanol measured by methanol adsorption on
Effect of Coke Deposition. Figure 1 reveals that the selec- SAPO-34 indicated a uniform distribution of acid strength,27
tivities changed with the degree of deactivation due to coke depo- which is in good agreement with the result obtained from
sition. At a given conversion, the yields of C3-C6 olefins on a ammonia TPD experiments.35 In addition, the number of active
4120 Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007

sites per cage was estimated to be 0.8,36 indicating that a maxi-


mum of 80% of the cages are active for reaction. Each cage
can be considered as a microreactor, and coke deposition reduces
the number of microreactors. The number of active sites in each
accessible microreactor is close to 1; therefore, it is not expected
that the change in the number of total active sites due to coking
influences the product distribution. The change in acidic
properties is, therefore, unlikely to be the main reason, and shape
selectivity is a more likely reason for the selectivity change.
The MTO reaction on differently sized crystals has demon-
strated the importance of the shape selectivity on the selectivity
changes due to coke deposition.37 The results showed almost
no effect of crystal size on the selectivity, but a significant effect
of coke deposition. Therefore, the transition-state shape selectiv-
ity was considered to be dominant at relatively high coke
contents. The formation of larger molecules via a larger-sized
reaction intermediate was significantly suppressed by the
reduced free space in the cavities by coke deposition, hence,
enhancing the ethene formation. The much lower ethene
selectivity at the low coke content (Figure 3) can be explained
by a weak or no effect of the transition-state shape selectivity.37
However, a possible effect of the changes in the reaction
mechanism with coke content cannot be excluded.

Deactivation Due to Coke Deposition


The coke deposition and deactivation of the MTO reaction
at 698 K and PMeOH ) 7.2 kPa for different space velocities
have been discussed in detail in our previous work.21 Deactiva-
tion of zeolites due to coke deposition is normally very
complicated, since intracrystalline diffusion often influences both
activity and selectivity. Coke deposition can reduce the number
of active sites by site coverage or by pore blockage. Coke Figure 7. Changes in conversion of oxygenates with the time on-stream
deposition can also increase the effective diffusion length and, (A) and the coke content (B) at 425 °C (O), 500 °C (4), and 550 °C (0).
thus, decrease the effective diffusivity and the effectiveness Partial pressure of methane of 14 kPa, WHSV ) 263 g (g of cat)-1 h-1
factor. It has been demonstrated that intracrystalline diffusion
plays a key role in determining the activity and the coke in the present work. Coke deposition can significantly change
deposition during MTO on SAPO-34 with the crystal size used the intracrystalline diffusion resistance, increasing the complex-
in the present work.27,38-40 The role of the intracrystalline ity in modeling of the formation and consumption of DME.
diffusion in MTO has recently been reviewed by Ruthven41 and Therefore, methanol and DME are lumped together as oxygen-
Chen et al.;23 however, no attempts have been made to ates in our model. The kinetic models for coke deposition and
distinguish between different causes of deactivation in the the conversion of oxygenates were developed in our previous
modeling in the present work. work,21 in which the conversion of oxygenates was treated as a
Our previous work21 has shown that coke deposition increases first-order reaction. In the present work, the model is explored
significantly with increasing temperature. Figure 7A shows that to predict the selectivity change with coke formation. The
the conversion of the oxygenates decreases more significantly TEOM reactor is treated as an ideal isothermal plug flow reactor
with time on-stream at the higher temperature; however, Figure containing 5-10 mg of catalyst. The integral model for
7B indicates a rather similar deactivating effect of coke methanol conversion is used in the present work, but an
molecules formed during MTO at different temperatures. It approximation has been made to assume uniform distribution
means that the rapid deactivation of oxygenates at high of coke in the catalyst bed. The experimental coke contents were
temperatures is a main result of the rapid coke deposition. used to fit the parameters in the deactivation functions. In the
model, all mole fractions are calculated on a dry basis, that is,
A small difference in the initial conversion of oxygenates
on the basis of CH2 equivalents.
with temperature can be observed at lower coke content (Figure
The reactor models can be expressed as
7B). This is mainly due to the high conversions, in which the
conversion is not very sensitive to the change in the activity at b
dx
such conditions. In addition, it can partly be a result of diffusion )b
r (3)
d(W/FMeOH)
limitation. It is well-known that diffusion limitations reduce the
observed activation energy. The kinetic model for the reaction network presented in Figure
2 is expressed as follows,
Kinetic Model of the MTO Reaction Network
ri ) k0i φiy6P0 for i ) 1-5
It was found that it is difficult to model the formation and
consumption of DME properly. It is expected that the formation 5
and conversion of DME are influenced by intracrystalline
diffusion, due to the large crystals (∼2 µm) of SAPO-34 used
r6 ) ( ∑
i)1
k0i φi)y6P0 (5)
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 4121

Table 1. Reaction Rate Constants, k0i ; Deactivation Rate Constants, ri, at Different Temperatures and Parameters for Arrhenius Equation ki0 )
Ai exp(-Ei/RT)
T, °C
400 425 500 550 Ai (kmol/gcat, kPa, h) E (kJ/mol)
k01 0.22 ( 1 × 10-2 0.25 ( 1 × 10-2 0.55 ( 5 × 10-2 0.755 ( 6 × 10-2 7210 38.4
k02 0.35 ( 1 × 10-2 0.31 ( 1 × 10-2 0.67 ( 5 × 10-2 0.76 ( 6 × 10-2 40 27.0
k03 0.13 ( 1 × 10-2 0.11 ( 1 × 10-2 0.23 ( 5 × 10-2 0.28 ( 5 × 10-2 15 26.9
k04 0.038 ( 2 × 10-2 0.035 ( 2 × 10-3 0.087 ( 1 × 10-2 0.104 ( 2 × 10-2 17 49.8
k05 0.008 ( 9 × 10-3 0.011 ( 6 × 10-3 0.017 ( 1 × 10-2 0.030 ( 9 × 10-3 5 32.4
k07 0 0.006 ( 5 × 10-3 0.020 ( 1 × 10-2 0.028 ( 9 × 10-2 181 59.6
R1 0.038 ( 1 × 10-3 0.049 ( 3 × 10-4 0.054 ( 3 × 10-3 0.063 ( 1 × 10-3
R2 0.041 ( 1 × 10-3 0.052 ( 3 × 10-4 0.059 ( 2 × 10-3 0.066 ( 1 × 10-3
R3 0.040 ( 2 × 10-3 0.052 ( 8 × 10-4 0.054 ( 4 × 10-3 0.058 ( 3 × 10-3
R4 0.050 ( 8 × 10-3 0.060 ( 6 × 10-4 0.060 ( 1 × 10-2 0.062 ( 6 × 10-3
R5 0.115 ( 4 × 10-2 0.114 ( 3 × 10-2 0.059 ( 1 × 10-2 0.065 ( 1 × 10-2
R7 0.066 ( 3 × 10-2 0.066 ( 2 × 10-2 0.057 ( 3 × 10-2 0.072 ( 2 × 10-2

where i ) [ethene, propene, butenes (C4), C5, C6, oxygenates, individual parameters was tested by t-test, and the standard
ethane + propane]. The formation of paraffins (ethane and deviation of each parameter was also calculated in a 95%
propane) was treated as a secondary reaction of all hydrocarbons. confidence interval.42 All the parameters in the model have been
The rate of formation is described by eq 6, estimated separately at different temperatures. Reaction and
deactivation rate constants together with their standard deviations
r7 ) k07φ7(1 - y6)P0 (6) are listed in Table 1. Statistical analysis indicated that the overall
regression is meaningful.
where k0i is the initial rate constant, φi is the deactivation The empirical deactivation rate constant R follows the order
function, P0 is the initial methanol partial pressure, and yi is of the molecular size. Larger molecules have larger changes in
the mole fraction on a CH2 basis. selectivity with coke content and, thus, a higher deactivation
All rate constants are assumed to depend on the catalyst coke rate. This may point out the importance of the effect of
content, which is taken into account by the deactivation transition-state shape selectivity on the product selectivity.37 The
functions, φi. Different deactivation functions42,43 have been tried standard deviation for R5 and R7 is relatively large. It is mainly
to describe the change in reaction rates with coke content. The due to the relatively large experimental error for the analysis,
linear function (eq 7) was found to give the best fit to the since the mole fractions of C6, ethane, and propane are very
experimental data. low in most of the experiments.
Figure 8 shows an Arrhenius plot of the initial kinetic rate
B
φ)1-R
bC (7) constants for the formation of the various products. The rate
constants fitted Arrhenius law well. The estimated apparent
C is the weight percent of coke on the catalyst (gcoke/gcat %). It activation energies and pre-exponential factors are presented
should be noted that the experimental relationship between the in Table 1.
deactivation and the coke content is not perfectly linear, as Figure 9 shows the comparison between experimental and
shown in Figure 7B. It follows a curve of a decrease in predicted mole fractions for each component. The rate constants
effectiveness factor with increasing Thiele modulus. A rigorous were calculated from the Arrhenius equation using the param-
modeling, including intracrystalline diffusion and changes with eters presented in Table 1. The model fitted the experimental
coke deposition, will be necessary in the future to perfectly results generally well at all temperatures, except for C6. The
simulate the deactivation function of this system. reason for the deviation in predicting the C6 mole fraction is at
As mentioned above, selective deactivation for the MTO least partly due to the relatively large experimental error in the
reaction over SAPO-34 was found. The different reaction steps GC analysis, since the mole fraction of C6 is very low, especially
have different deactivation rates, that is, different Ri values. The at relatively high coke contents. The same argument also holds
effect of coke on the selectivity, that is, the ratio of the reaction for methane. The mole fraction of methane is very low at the
rates, can now be modeled by using different values for the conditions used in the present work. In addition, a very small
empirical Ri constants. fraction of methane was founded in a few blank experiments,
A fourth-order Rung-Kutta method was used to integrate possibly caused by decomposition of methanol on the metal
the ordinary differential equations. The estimation of the
parameters of the kinetic model has been carried out by the
nonlinear least-squares routine in MATLAB using the Leven-
berg-Marquardt method. The optimum function is given in eq
8,
n m
S) ∑ ∑ wij[(yiPR) - (yiEXP)]
i)1 j)1
(8)

where index i indicates the component considered; index j,


the kinetic run; yiPR and yiEXP are the predicted and experimental
molar fraction for component i; and wij is a weighting factor.
After the parameters was estimated, an extensive statistical
analysis was performed. The significance of the overall regres- Figure 8. Arrhenius plot for reaction rate constants. 9, Ethene; 2, propene;
sion was tested by means of an F-test. The significance of the ×, butenes; [, C5; b, C6.
4122 Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007

the induction period depended on the crystal size, temperature,


and concentration of oxygenates.37 For SAPO-34 used in the
present work, the effect of intracrystalline diffusion cannot be
avoided,27 and thus, the induction period is very short. For
example, the induction period measured at 698 K and 30 kPa
methanol partial pressure was about 15-30 s. Hence, ignoring
the induction period should not create a large error.

Conclusions
A reaction network for the MTO reaction has been estab-
lished, DME is a nonstable primary product, all the hydrocar-
bons are formed in parallel from oxygenates, and ethane and
propane are considered to be stable tertiary products. A relatively
simple kinetic model, in which methanol and DME were
lumped, was developed to describe the change in selectivities
with the coke content. Catalyst deactivation was well-described
by a linear relation between the coke content and the reaction
rate. The experimental data fitted the model generally well in
the temperature range of 673-823 K. The secondary reactions
of olefins were not included in this kinetic model, because their
reactivity was low.
The distribution of olefins was not affected by the partial
pressure of methanol and the space velocity, but increasing
temperature and coke deposition increased the ethene selectivity
significantly during MTO over SAPO-34. The effect of coke
on the selectivities is proposed to be a result of transition-state
shape selectivity, favoring the formation of smaller molecules
as the void volume in the cavities is reduced by coke.

Acknowledgment
The support of this work by the Norwegian Research Council
and Norsk Hydro ASA is gratefully acknowledged.

Figure 9. Parity plot for mole fraction of olefins. 698 K, methanol partial Nomenclature
pressure of 7.2 kPa and WHSV. 0, 385; ∆, 113; / , 82; ], 57 g (g of
cat)-1 h-1; 773 K, 12 kPa and 266 g (g of cat)-1 h-1, 2; 823 K, 8 kPa bar, Ai ) pre-exponential factor for the reaction rate, k0i , kmol (g of
and 268 g (g of cat)-1 h-1, [; 673 K, 7.2 kPa and 385 g (g of cat)-1 h-1, cat, kPa,h)-1
b. C ) weight percent of coke on the catalyst (g of coke/100 g of
cat)
part in the setup. The reproducibility of the methane mole E ) activation energy, kJ/mol
fraction is relatively poor. It is expected that the experimental FMeOH ) molar flow rate of methanol (mol/h)
uncertainty is larger at low conversions, such as high coke i ) 1, 2, 3, 4, 5, 6, 7 represents ethene, propene, butenes (C4),
contents and low temperatures. Therefore, kinetic modeling of C5, C6, oxygenates, and ethane + propane, respectively.
methane formation is not reported in the present work. j ) number of kinetic run
The apparent activation energies for the olefin formation are k0i ) initial rate constant for the formation of component i,
relatively low, as shown in Table 1. As discussed previously,27,37 kmol (g of cat, kPa,h)-1
this is due in part to the intracrystalline diffusion limitation. P0: initial methanol partial pressure, kPa
The intrinsic activation energy can be estimated by accounting b
r ) [r1, r2, r3, r4, r5, r6, r7]; matrix of reaction rate, kmol (g of
for the effect of diffusion.42 If we assume that MTO was cat, h)-1
controlled completely by intracrystalline diffusion, the activation ri ) rate of formation of I, kmol (g of cat,h)-1
energy should be 2× the activation energy measured, namely, S ) objective function
76 kJ/mol. The apparent activation energy for ethene formation W ) catalyst loading, g of cat
is then expected to be between 38 and 76 kJ/mol. However, for wij ) weighting factor.
zeolite-catalyzed reactions, the apparent activation energy also b
x ) [x1, x2, x3, x4, x5, x6, x7]; matrix of conversion
involves the heat of adsorption.43 The real intrinsic activation xi ) conversion to i
energy for the surface reaction is then estimated43 to be between yi ) mole fraction of component i on a CH2 basis
68 and 106 kJ/mol, taking into account the measured adsorption yiEXP ) experimental molar fraction for component i
heat of 30 kJ/mol for methanol.27 The presence of intracrystalline yiPR ) predicted molar fraction for component i
diffusion limitation makes it difficult to compare the activation Ri ) empirical deactivation constant for the reaction corre-
energy directly with the literature data. sponding the formation of i defined by eq 7
In addition, it is worth mentioning that the estimated initial φi ) deactivation function for the formation of component i
rate constant is only an approximation, and the induction period
during which conversion increased with time on-stream or coke Literature Cited
content is ignored in the present work. Induction periods have (1) Chang, C. D. Methanol Conversion to Light Olefins. Cat. ReV. Sci.
been observed during MTO over SAPO-34, and the length of Eng. 1984, 26, 323.
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 4123

(2) Vora, B. V.; Marker, T. L.; Barge, P. T.; Fullerton, H. E.; Nilsen, (25) Chu, C. T. W.; Chang, C. D. Methanol Conversion to Olefins over
H. R.; Kvisle, S.; Fuglerud, T. Economic Route for Natural Gas Conversion ZSM-5. II. Olefin Distribution. J. Catal. 1984, 86, 297-300.
to Ethylene and Propene. Stud. Surf. Sci. Catal. 1997, 107, 87. (26) Dahl, I. M.; Kolboe, S. On the Reaction Mechanism for Hydro-
(3) Lok, B. M.; Messina, C. A.; Patton, R. L.; Gajek, R. T.; Cannan, T. carbon Formation from Methanol over SAPO-34. 1. Isotopic Labelling
R.; Flanigen, E. M. Silicoaluminophosphate Molecular Sieves: Another Studies of the Co-reaction of Ethene and Methanol. J. Catal. 1994, 149,
New Class of Microprous Crystalline Inorganic Solids. J. Am. Chem. Soc. 458.
1984, 106, 6092. (27) Chen, D.; Moljord, K.; Rebo, H. P.; Holmen, A. Methanol
(4) Froment, G. F.; Dehertog, W. J. H.; Marchi, A. J. Zeolite Catalysis Conversion to Light Olefins. I. Sorption, Diffusion and Catalytic Reaction.
in the Conversion of Methanol into Olefins. Catalysis 1992, 9, 1. Ind. Eng. Chem. Res. 1999, 38, 4242.
(5) Liang, J.; Li, H.; Zhao, S.; Guo, W.; Wang, R.; Ying, M. (28) Dahl, M. I.; Wendelbo, R.; Andersen, A.; Akporiaye, D.; Mostad,
Characteristics and Performance of SAPO-34 Catalyst for Methanol-to- H.; Fuglerud, T. The effect of crystallite size on the activity and selectivity
Olefin Conversion. Appl. Catal. 1990, 64, 31. of the reaction of ethanol and 2-propanol over SAPO-34. Microporous
(6) Marchi, A. J.; Froment, G. F. Catalytic Conversion of Methanol to Mesoporous Mater. 1999, 29, 159.
Light Alkenes on SAPO-34 Molecular Sieves. Appl. Catal. 1991, 71, 139. (29) Salehirad, F.; Anderson, M. W. Solid-State 13C MAS NMR Study
(7) Grønvold, A. G.; Moljord, K.; Dypvik, T.; Holmen, A. Conversion of Methanol-to-Hydrocarbon Chemistry over H-SAPO-34. J. Catal. 1996,
of Methanol to Lower Alkenes on Molecular Sieve Type Catalysts. Stud. 164, 301.
Surf. Sci. Catal. 1994, 81, 399. (30) Marchi, A. J.; Froment, G. F. Catalytic Conversion of Methanol to
(8) Chen, N. Y.; Reagan, W. J. Evidence of autocatalysis in Methanol Light Alkenes on SAPO-34 Molecular Sieves. Appl. Catal. 1991, 71, 139.
to Hydrocarbon Reactions over Zeolite Catalysts. J. Catal. 1979, 59, 123. (31) Munson, E. J.; Kheir, A. A.; Lazo, N. D.; Haw, J. F. In Situ Solid-
(9) Chang, C. D. A Kinetic Model for Methanol Conversion to State NMR Study of Methanol-to-Gasoline Chemistry in Zeolite HZSM-5.
Hydrocarbons. Chem. Eng. Sci. 1980, 35, 619. J. Phys. Chem. 1992, 96, 7740.
(10) Ono, Y; Mori, T. Mechanism of Methanol Conversion into (32) Hutching, G. J.; Hunter, R. Hydrocarbon Formation from Methanol
Hydrocarbons over ZSM-5 Zeolites. J. Chem. Soc. Faraday Trans. 1 1981, and Dimethylether: A Review of the Experimental Observations Concerning
77, 2209. the Mechanism of Formation of the Primary Products. Catal. Today 1990,
(11) Sedran, U.; Mahay, A.; de Lasa, H. I., Modelling Methanol to 6, 279.
Hydrocarbons: Alternative Kinetic Models. Chem. Eng. J. 1990, 45, 33- (33) Schulz, H.; Barth, D.; Siwei, Z. Deactivation of HZSM-5 Zeolite
42. during Methanol Conversion: Kinetic Probing of Pore-Architecture and
(12) Benito, P. L.; Gayubo, A. G.; Aguayo, A. T.; Olazar, M.; Bilbao, Acidic Properties. In Catalyst DeactiVation, 1991: Proceedings of the 5th
J. Concentration-Dependent Kinetic Model for Catalyst Deactivation in the International Symposium, Evanston, IL, June 24-26, 1991; Bartholomew,
MTG process. Ind. Eng. Chem. Res. 1996, 35, 81. C. H., Butt, J. B., Eds.; Amsterdam: Elsevier Science Publisher B.V., 1991.
(13) Mihail, R.; Straja, S.; Maria, G.; Musca, G.; Pop, G. Kinetic Model
(34) Cai, G.; Liu, Z.; Shi, R.; He, C.; Lixin, Y.; Sun, C.; Chang, Y.
for Methanol Conversion to Olefins, Ind. Eng. Chem. Process Des. DeV.
Light Alkenes from Syngas via Dimethy Ether. Appl. Catal., A. 1995, 125,
1983, 22, 532.
29.
(14) Mihail, R.; Straja, S.; Maria, G.; Musca, G.; Pop, G. A Kinetic
(35) Nawaz, S.; Kolboe, S.; Kvisle, S.; Lillerund, K. P., Stöcker, M.;
Model for Methanol Conversion to Hydrocarbons. Chem. Eng. Sci. 1983,
Øren, H. M. Selectivity and Deactivation Profiles of Zeolite Type Materials
38, 1581.
in the MTO Process. Stud. Surf. Sci. Catal. 1991, 61, 421.
(15) Park, T. Y.; Froment, G. F. Kinetic Modeling of the Methanol to
Olefins Process. 1. Model Formulation. Ind. Eng. Chem. Res. 2001, 40, (36) Grønvold, A. G. Conversion of Methanol to Lower Alkenes over
4172. Molecular Sieve Type Catalysts. Ph.D. Thesis, The Norwegian Institute of
(16) Park, T. Y.; Froment, G. F. Kinetic Modeling of the Methanol to Technology, Trondheim, Norway, 1994.
Olefins Process. 2. Experimental Results, Model Discrimination and (37) Chen, D.; Moljord, K.; Fuglerud, T.; Holmen, A. The Effect of
Parameter Estimation. Ind. Eng. Chem. Res. 2001, 40, 4187. Crystal Size of SAPO-34 on the Selectivity and Deactivation of the MTO
(17) Bos, R. A. N.; Tromp, P. J. J.; Akse, H. K. Conversion of Methanol Reaction. Microporous Mesoporous Mater. 1999, 29, 191.
to Lower Olefins. Kinetic Modelling, Reactor Simulation and Selection. (38) Chen, D.; Rebo, H. P.; Moljord, K.; Holmen, A. Dimethyl Ether
Ind. Eng. Chem. Res. 1995, 34, 3808. Conversion to Light Olefins. Deactivation Due to Coke Deposition. Stud.
(18) Gayubo, A. G.; Aguayo, A. T.; Sánchez del Campo, A. E.; Tarrı́o, Surf. Sci. Catal. 1998, 119, 521.
A. M.; Bilbao, J. Kinetic Modeling of Methanol Transformation into Olefins (39) Chen, D.; Rebo, H. P.; Moljord, K.; Holmen, A. Diffusion and
on a SAPO-34 Catalyst. Ind. Eng. Chem. Res. 2000, 39, 292. Deactivation during Methanol Conversion over SAPO-34sA Percolation
(19) Alwahabi, S. M.; Froment, G. F. Single Event Kinetic Modelling Approach. Chem. Eng. Sci. 1999, 54, 3465.
of the Methanol-to-Olefins Process on SAPO-34. Ind. Eng. Chem. Res. 2004, (40) Chen, D.; Moljord, K.; Holmen. A. Effect of Crystal Size of SAPO-
43, 5098. 34 on Methanol Conversion to Light Olefins. Stud. Surf. Sci. Catal. 2000,
(20) Chen, D.; Rebo, H. P.; Moljord, K.; Holmen, A. Influence of Coke 130 C, 2651.
Deposition on Selectivity in Zeolite Catalysis. Ind. Eng. Chem. Res. 1997, (41) Ruthven, D. M. The Technological Impact of Diffusion in Nanpores.
36, 3473. Diffus. Fundam. 2005, 2 77.1.
(21) Chen, D.; Moljord, K.; Rebo, H. P.; Holmen, A. Methanol (42) Froment, G. B.; Bischoff, K. B. Chemical Reactor Analysis and
Conversion to Light Olefins. II. Kinetic Modelling of Coke Formation. Design, 2nd ed.; John Wiley & Sons: New York, 1990.
Microporous Mesoporous Mater. 2000, 121, 35-36. (43) Kapteijn, F.; Moulijn, J. A.; van Diepen, A. E.; Kreutzer, M. T.
(22) Chen, D.; Grønvold, A.; Rebo, H. P.; Moljord, K.; Holmen, A. Catalysis Engineering, Delft University of Technology, 2001.
Catalyst Deactivation Studied by Conventional and Oscillating Microbalance (44) Palekar, M. G.; Rajadhyaksha, R. A. Sorption Accompanied by
Reactors. Appl. Catal. 1996, 137, L1. Chemical Reaction on Zeolites. Catal. ReV. Sci. Eng. 1986, 28, 371.
(23) Chen, D.; Bjørgum, E.; Christensen, O. K.; Holmen, A. Charac-
terization of Catalysts under Working Conditions with an Oscillating ReceiVed for reView August 15, 2006
Microbalance Catalytic Reactor. AdV. Catal.; In press. ReVised manuscript receiVed October 19, 2006
(24) Campbell, D. R.; Wojciechowski, B. W. Selectivity of Aging Accepted October 20, 2006
Catalyst in Static, Moving and Fluidized Bed Reactors. Can. J. Chem. Eng.
1970, 48, 224. IE0610748

Vous aimerez peut-être aussi