Você está na página 1de 7

Chemical Engineering Science 56 (2001) 6081–6087

www.elsevier.com/locate/ces

Dynamics of a mixed slurry reactor for the three-phase


methanol synthesis
)
Marko Setinca
, Janez Levecb; ∗
a Laboratory for Catalysis and Chemical Reaction Engineering, National Institute of Chemistry, P.O. Box 3430, SI-1001 Ljubljana, Slovenia
b Department of Chemical Engineering, University of Ljubljana, P.O. Box 537, SI-1001 Ljubljana, Slovenia

Abstract
A dynamic model for the liquid-phase methanol synthesis with the Langmuir–Hinshelwood kinetic formulation is presented for
a well-mixed slurry reactor. Responses to a step-change in the reactor feed composition were studied in a temperature range of

200 –240 C and total pressures between 34 and 41 bar. A commercial Cu=ZnO=Al2 O3 and squalane were used as a catalyst and
inert liquid phase, respectively. All transport, kinetic, and adsorption parameters, excepting the equilibrium adsorption constants
for water and methanol, were evaluated in separate experiments. The two parameters were obtained by 8tting the model predictions
to the experimental responses and were found to obey the van’t Ho: relation. A good agreement between experimental and
predicted results found in a rather wide range of operating conditions validates the Langmuir–Hinshelwood kinetic formulation as
an appropriate tool for describing the processes on catalyst surface. ? 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Methanol synthesis; Kinetics; Slurry reactor; Catalytic hydrogenation; Dynamic experiments

1. Introduction tions on dynamic behavior of the gas-phase methanol


synthesis (e.g. Ali & Goodwin, 1997), no single paper in
In the late 1970s the liquid-phase methanol synthesis the open literature has appeared on the dynamics of the
was introduced in order to utilize CO-rich syngas derived liquid-phase methanol synthesis. The information on the
from coal gasi8cation. The liquid-phase is used to absorb three-phase slurry dynamics itself is also meager. First
the heat released during the reaction and thus prevents investigations had employed transient experiments only
the potential formation of hot spots due to overheating to evaluate mass transfer rates for adsorption in aqueous
and consequently the catalyst deactivation. In general, a slurries of activated carbon (Furusawa & Smith, 1973;
slurry and three-phase upward ?ow 8xed bed reactors are Niiyama & Smith, 1976). Sulfur dioxide oxidation in car-
preferred for fast and highly exothermic reactions since bon slurries was studied later by Komiyama and Smith
they both provide better catalyst utilization and tempera- (1975), whereas Ramachandran and Smith (1977) pre-
ture control. However, Kodra and Levec (1991) showed sented a systematical analysis of the dynamic behavior
that there is almost no di:erence between slurry bubble of three-phase systems. A rather complete work on the
column and trickle bed reactors. slurry reactor dynamics has been reported by Fu, McCoy,
The objective of this study was to investigate the dy- and Smith (1989) using oxidation of aqueous sul8te so-
namic behavior of a well-mixed slurry reactor for carrying lution catalyzed by activated carbon as a model reaction.
out the liquid-phase methanol synthesis. It was of a partic- Recently, Oinas and Haario (1994) presented a general
ular interest to see whether the Langmuir–Hinshelwood set of transient equations and solved them numerically
kinetic law, which o:ers a deeper insight on the indi- for simultaneous reactions and mass transfer in di:erent
vidual steps of the reactions taking place on the cata- slurry systems.
lyst surface, is appropriate tool for describing transient
behavior of the system. While there are many publica-
2. Model formulation
∗ Corresponding author. Tel.: +386-1-476-0280; fax: +386-1-
425-9244. Hydrogenation of carbon monoxide and carbon diox-
E-mail address: janez.levec@ki.si (J. Levec). ide takes place on the surface of solid catalyst. Therefore,

0009-2509/01/$ - see front matter ? 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 1 ) 0 0 2 1 2 - 3
6082 0
M. Setinc, J. Levec / Chemical Engineering Science 56 (2001) 6081–6087

the simultaneous transport of reacting components across kinetics for the methanol and water formation on the cat-
the gas–liquid and the liquid–solid interfaces occur be- alyst surface are given by the following expressions of
fore reactants reach and react on the catalyst surface. In the Langmuir–Hinshelwood type:
setting the model equations the following assumptions EMe cCO2 (cH2 − cH2 ;eq )
were made: rMe = AMe e(− RT ) EW
; (5)
(1 + AW e( RT ) cH2 O )2
• The reactor operates as a well-mixed system at isother-
EH2 O (cCO2 − cCO2 ;eq )
mal and isobaric conditions: in batch mode for the
rH2 O = AH2 O e(− RT ) : (6)
slurry and continuous for the gas-phase. The hydrody- EW
(1 + AW e( RT ) cH2 O)
namic and transport properties are constant within the
slurry. Eqs. (5) and (6) were developed from the steady state
• The gas-phase above the liquid is also well mixed and experiments, whereas the driving-force and adsorption
its properties are the same as those in the gas-slurry terms in the above equations were formulated by means of
mixture. The gas-phase behaves as an ideal gas. )
the transient experiments (Setinc )
& Levec, 1999; Setinc,
• The liquid and solid phases form a pseudo-homo- 2001). Based on the stoichiometry of reactions (Eq. (1))
geneous slurry phase whose properties can be pre- one can write the following relations between the rates
dicted by those of pure liquid and solid phases. of the individual components
• The gas-side mass transport resistances for all gaseous rH2 = − 2rMe − rH2 O ; (7)
components are negligible.
• No mass transport limitation exists within the catalyst rCO = − rMe + rH2 O ; (8)
particles.
• The following reactions take place on the catalyst sur- rCO2 = − rH2 O ; (9)
face:
which close the equation system. Eqs. (2) – (4) are subject
CO + 2H2 ⇔ CH3 OH; (1a) to the following initial conditions
CG; i = [CG; i ]ss ; CL; i = [CL; i ]ss ; CS; i = [CS; i ]ss ;
CO2 + H2 ⇔ CO + H2 O; (1b)
for t = 0: (10)
CO + 3H2 ⇔ CH3 OH: (1c) In Eq. (10) [Ci ]ss represents the steady state values, which
According to the assumptions, the following equations had been attained in the system before the step change
govern the mass transport of the component i (i.e. in the feed composition was introduced. The steady state
CO; CO2 ; H2 , Me, H2 O, argon): concentration values were obtained from the model equa-
tions written for the steady state conditions, dCi =dt = 0;
dCG; i and solved by the Newton–Raphson method.
VG = IN CG; i; 0 − OUT CG; i
dt Eqs. (2) – (10) constitute the dynamic model of the
  slurry reactor system for the liquid-phase methanol syn-
CG; i
− (ka)L; i − CL; i VL ; (2) thesis. It should be stressed at this point that all trans-
He∗i
port, adsorption, and kinetic parameters were obtained
in the bulk gas-phase, where VG represents the total vol- by separate experiments, except the adsorption equilib-
ume of the gas-phase in the reactor (the one dispersed in rium constants for methanol and water, which were left
slurry and the one above the liquid level), and as two adjustable parameters. However, the model was
  solved by a Berkeley Madonna code, which employs the
dCL; i CG; i
VL = (ka)L; i − CL; i VL ; 4th order Runge–Kutta method. For example, in order to
dt He∗i
 
optimize one of the adsorption equilibrium constant, 3 to
VL 4 h was needed on a Pentium II 300 MHz PC.
− (ka)S; i CL; i − CS; i VL (3)
Ka; i mcat
in the bulk liquid-phase. On the catalyst surface the fol- 3. Experimental
lowing equation applies
  3.1. Slurry reactor dynamics
dCS; i VL
VL = (ka)S; i CL; i − CS; i VL − mcat ri ;
dt Ka; i mcat
The experimental setup used in this investigation con-
(4)
sists of a 300 ml stainless reactor=autoclave equipped
in which Ka; i stands for the adsorption equilibrium con- with a turbine type impeller and temperature control unit
stant for the component i and ri denotes the rate of re- (EZE-Seal, Autoclave Engineers) and a reaction system
actions taking place at the catalyst surface. The intrinsic module (CDS 804, Autoclave Engineers). The mixture
0
M. Setinc, J. Levec / Chemical Engineering Science 56 (2001) 6081–6087 6083

Fig. 1. Combined dispersion e:ect of inlet and outlet tubing of the Fig. 2. Typical results of carbon dioxide transient experiment for
experimental setup. evaluation of mass transfer coeScient, solubility, and equilibrium
constant.

of gaseous reactants (CO, CO2 , and H2 ) and inert gas


argon was metered (Brooks 5859 MFC) and fed through depends on its operating pressure. Thus, subtracting the
the system module into the reactor. The system module contribution of the outlet tubing and QM from the com-
also takes care of the reactor pressure control (Tescom, bined e:ect reactor one can quantitatively determine the
26-1723) and the reactor ePuent handling, including actual step-change function at the reactor inlet.
phase separation and vapor-phase sampling. More de-
tails about the experimental setup, analysis of the reactor 3.2. Auxiliary data
ePuents, and the catalyst used, one can 8nd in a paper
)
of Setinc and Levec (1999). The experiments were per- Due to a very small catalyst particle size (40 m) the

formed in a temperature range of 200 –240 C and total liquid–solid mass transfer coeScients were calculated by
pressures from 34 to 41 bar. Mole fractions were from setting the Sherwood number equals two. The gas–liquid
0.6 – 0.8, 0.1– 0.2, and 0.1– 0.2 for H2 , CO, and CO2 , mass transfer coeScients for CO, CO2 ; H2 , and argon
respectively, whereas the gas ?ow rate through the re- were obtained by a technique which employs pursuing
actor was 1:0 l=min. The reactor was 8lled with 150 ml the time dependent pressure drop due to the mass trans-
of squalane and 4 –40 g of commercial Cu=ZnO=Al2 O3 fer under isohoric conditions (Ledakowicz, Nettelhof, &
catalyst (40 m). The dynamic runs were carried out at Decker, 1984). The same experiments were also used to
the impeller speed of 1000 rpm. evaluate the equilibrium solubilities (Henry’s constants)
Responses to the step-changes in the concentration of after the equilibrium in a closed system was attained
feed components were measured by a quadrupole mass (dp=dt = 0). Repeating the same type of experiments in
spectrometer-QM (Leybold PGA 100). A signal was the presence of the solid catalyst enabled us addition-
recorded and analyzed by a PC via RS232. The QM was ally to determine the adsorption equilibrium constants for
connected to the module by a 5 m long stainless-steel each of the components. More speci8cally, they were ob-
capillary tubing, which ensures appropriate pressure re- tained from the di:erence between the equilibrium pres-
duction. In order to account for possible departure from sures attained with and without the catalyst. A typical
the plug-?ow behavior in the inlet and outlet reactor tub- time dependent dimensionless pressure drops for carbon
ing, separate dynamic experiments were performed with dioxide is shown in Fig. 2. As one can see, there is quite
these conduits. A combined e:ect from both conduits a large equilibrium pressure di:erence between the two
is depicted in Fig. 1. As one can see, there was some types of experiments: the di:erence is attributed to the
degree of backmixing in the conduits, which was quite adsorption on the solid catalyst. It should be emphasized
well described by means of series of well-mixed cells that these experiments were carried out in the same au-
(represented by the solid curves). The deviation shown toclave as those for the slurry dynamics.
arose mainly from the inlet reactor tubing with a 4-way Typical experimental mass transfer results at di:erent
valve used for introducing the step-change in the reac- operating temperatures are illustrated in Fig. 3 for hydro-
tor feed (about 75%). The number of cells in the inlet gen as a function of the impeller speed. By means of the
reactor tubing was found to be dependent on the linear physical properties for squalane (Cybulski, Edvinsson,
velocity in that conduit, whereas the contribution from Irandoust, & Andersson, 1993), the data were correlated
the outlet side arrives mainly from the QM and therefore )
in a conventional manner (Setinc, 2001).
6084 0
M. Setinc, J. Levec / Chemical Engineering Science 56 (2001) 6081–6087

Fig. 3. Liquid-side mass transfer coeScient for hydrogen as a function


of impeller speed at di:erent temperatures.
Fig. 4. Equilibrium adsorption constants for carbon monoxide, carbon
dioxide and hydrogen as a function of temperature.
Table 1
Henry’s law parameter values for squalane
Table 2
A E Kinetic parameters for commercial Cu=ZnO=Al2 O3 catalyst and
(bar m3 =mol) (kJ=mol) squalane system
H2 107 3707 Parameter Value
CO 168 721
CO2 373 −6120 AMe (l=mol)2 (mol=s
kgkat ) 9:15 × 1013
Me 149a −17235a EMe (J=mol K) 8:225 × 104
H2 O 330a −8633a AW (l=mol)(mol=s kgkat ) 1:32 × 102
Ar 385.2 2980.6 EW (J=mol K) 9:550 × 103
a Graaf,
AH2 O (l=mol) 3:91 × 107
Smit, Stamhuis and Beenackers (1992). EH2 O (J=mol K) 5:464 × 104

The Henry’s law constants were calculated according


to the relation ing dispersion. Thus, a typical response to a CO2 cut-o:
(exchanged for argon) in the feed stream is illustrated in
Hei = Ai eEi =RT ; (11) Fig. 5: top, for the reactants; bottom, for the products
in which the parameters Ai and Ei were obtained experi- (methanol and water). The reactor was disturbed after
mentally (except for water and methanol) and are given 1000 s of steady operation (t = 0). Due to the decrease in
in Table 1. CO2 concentration, the production of methanol and wa-
The adsorption equilibrium constants for the reaction ter was expectedly reduced. When after 2100 s CO2 was
system investigated are depicted in Fig. 4 as a function of switched on again, the methanol concentration reached a
temperature. It should be noted that the values presented higher value as it was at the previous steady state opera-
are about one-half of those reported for the conventional tion. As one can see, the rate of methanol and water pro-
gas-phase methanol synthesis (Skrzypek, Sloczynski, & duction increases proportionally to the CO2 concentra-
Ledakowicz, 1994). One may speculate that lower values tion but the rate for water is slower (compare the slopes).
are due to the presence of the liquid-phase. Therefore during this period less water was produced, and
Parameters in Eqs. (5) and (6) were obtained by because water hinders the methanol production due to
the steady state kinetic experiments and are provided in the adsorption on the catalyst active sites, more methanol
)
Table 2 (Setinc )
and Levec, 1999; Setinc, 2001). was produced than at the preceded steady state condi-
tions. However, after a while the original steady state was
achieved in both water and methanol concentrations. It
4. Results and discussion can also be seen from Fig. 5 that there is a reasonable
agreement between the model predictions and the exper-
It has to be underlined that the time-dependent con- imental results. As said previously, the only adjustable
centrations shown in Fig. 5 and Figs. 7–10 are the actual parameters in the model were the equilibrium adsorp-
experimental responses of the system, whereas the model tion constants for water and methanol. Thus, by 8tting
predictions account for the inlet as well as the outlet tub- the experimental transient concentration pro8les at three
0
M. Setinc, J. Levec / Chemical Engineering Science 56 (2001) 6081–6087 6085

Fig. 5. Responses of slurry system to cut-o: carbon dioxide in the



feed stream (200 C and 34 bar experiment). After 2100 s CO2 was Fig. 7. Responses of slurry system to cut-o: all reactant components

switched on again. Top: reactant concentration pro8les. Bottom: prod- in the feed stream (200 C and 34 bar experiment). After 2100 s all
uct concentration pro8les. reactants were switched on again. Top: reactant concentration pro8les.
Bottom: product concentration pro8les.

& Ledakowicz, 1994). Lower values found in this work


may be attributed to the presence of the liquid-phase.
In Fig. 7 the responses are shown when all three re-
actants (CO, CO2 , and H2 ) in the feed stream were ex-
changed for argon. Since all reactants were taken out of
the reaction system, it is not surprisingly that the rate of
methanol and water production reached the zero value.
When after 2100 s the feed composition was brought back
to the initial state, some time latter the reactor outlet com-
position also reached its initial steady state value. Again,
one can see how well the dynamic model predicts the
slurry reactor behavior. In the model predictions of Fig.
7 (as well as on the proceeding 8gures) the adsorption
constants for methanol and water were taken from Fig. 6.
Fig. 8 represents the results obtained when CO was ex-
changed for the inert argon. A substantial increase of the
Fig. 6. Equilibrium adsorption constants for methanol and water ob- water production rate is obviously seen. This is a conse-
tained by 8tting the model predictions to the experimental responses. quence of shifting the equilibrium in the WGSR towards
water and CO, which means more water is produced. Af-
ter the initial conditions in the feed were reestablished,
◦ ◦ ◦
operating temperatures (200 C, 225 C and 240 C) one the water concentration in the outlet stream comes back
can construct the van’t Ho: plot (Fig. 6). The adsorp- to its initial steady state value. Because the reduction of
tion enthalpies evaluated from this plot are found to be methanol production was not signi8cant at all, and be-
32 and 13 kJ=mol for methanol and water, respectively. cause CO via WGSR is also producing slowly, one can
For the gas-phase process the adsorption enthalpies of conclude that methanol mainly derives from the hydro-
46 –79 kJ=mol and 20 –67 kJ=mol for methanol and wa- genation of CO2 . This 8nding con8rms the results of
ter, respectively, were reported (Skrzypek, Sloczynski, Graaf, Winelman, Stamhuis, and Beenackers (1988).
6086 0
M. Setinc, J. Levec / Chemical Engineering Science 56 (2001) 6081–6087

Fig. 8. Responses of slurry system to cut-o: carbon monoxide in



the feed stream (200 C and 34 bar experiment). After 3000 s CO Fig. 9. Responses of slurry system to cut-o: carbon monoxide and

was switched on again. Top: reactant concentration pro8les. Bottom: carbon dioxide in the feed stream (200 C and 34 bar experiment).
product concentration pro8les. After 3500 s CO and CO2 were switched on again. Top: reactant
concentration pro8les. Bottom: product concentration pro8les.

A response to simultaneous cut-o: of both carbon-yield


compounds, CO and CO2 , is depicted in Fig. 9. As seen,
the rate of methanol production decreases a bit more
slowly as in the case when all reactant components were
exchanged for argon (Fig. 7). Due to higher solubility of
CO and CO2 with respect to that of hydrogen, the hy-
drogenation process continues with the absorbed and ad-
sorbed CO and CO2 in the slurry. The model predictions
shown here are also in good agreement with the experi-
mental responses.
A comparison between the predicted concentration pro-
8les and those obtained experimentally at higher tem-

perature and pressure (240 C and 41 bar) is illustrated
in Fig. 10. A good agreement shown in Fig. 10 is quite
convincing to conclude that it is possible to adequately
describe the slurry reactor dynamics by employing the
Langmuir–Hinshelwood kinetic formulation.

5. Conclusions

A thorough study presented here undoubtedly demon-


strates that it is possible to reasonably well describe the
dynamic behavior of a slurry reactor for the liquid-phase
methanol synthesis if the Langmuir–Hinshelwood kinetic
Fig. 10. Responses of slurry system to cut-o: carbon dioxide in
formulation were used to describe the elementary steps ◦
the feed stream (240 C and 41 bar experiment). After 3050 s CO2
on the catalyst surface. In this work all parameters were was switched on again. Top: reactant concentration pro8les. Bottom:
evaluated by separate experiments but the equilibrium ad- product concentration pro8les.
0
M. Setinc, J. Levec / Chemical Engineering Science 56 (2001) 6081–6087 6087

sorption constants for methanol and water, which were References


obtained by 8tting the dynamic model predictions to the
experimental responses. Due to a good agreement found Ali, S. H., & Goodwin Jr., J. G. (1997). Analysis of sequential
between the experimental results and predictions and due reactions in bifunctional catalyst system using isotopic transient
kinetics: methanol synthesis on Pd=SiO2 + Al2 O3 . Journal of
to the kinetics employed, one can conclude that water Catalysis, 171(1), 333–338.
plays an important role also in the liquid-phase methanol Cybulski, A., Edvinsson, R., Irandoust, S., & Andersson, B. (1993).
synthesis and further, the methanol production is propor- Liquid phase methanol synthesis: modeling of a monolithic reactor.
tional to CO2 concentration and not to CO. However, Chemical Engineering Science, 48(20), 3463–3478.
it has turned out that the values of equilibrium adsorp- Fu, C. C., McCoy, B. J., & Smith, J. M. (1989). Slurry reactors
studies of homogeneous and heterogeneous reactions. A.I.Ch.E.
tion constants for all reaction components (CO, CO2 , H2 , Journal, 35(2), 259–266.
H2 O, and CH3 OH) are about one-half of those known for Furusawa, T., & Smith, J. M. (1973). Fluid-particle and intraparticle
the conventional gas-phase methanol synthesis. One may mass transport rates in slurries. Industrial and Engineering
speculate that this is due to the presence of squalane. Chemistry Fundamentals, 12(2), 197–212.
Graaf, G. H., Smit, H. J., Stamhuis, E. J., & Beenackers,
A. A. C. M. (1992). Gas–liquid solubilities of the methanol
synthesis components in various solvents. Journal of Chemical
Notation Engineering Data, 37, 146–158.
Graaf, G. H., Winelman, J. G., Stamhius, E. J., & Beenackers, A.
C molar concentration, mol=l A. C. M. (1988). Kinetics of the three phase methanol synthesis.
Chemical Engineering Science, 43(8), 2161–2168.
He∗ Henry’s constant, dimensionless Kodra, D., & Levec, J. (1991). Liquid phase methanol synthesis:
Ka equilibrium adsorption constant, l=g comparison between trickle-bed and bubble column slurry reactors.
(ka) mass transfer coeScient, 1=s Chemical Engineering Science, 46(9), 2339–2350.
mcat mass of catalyst, g Komiyama, H., & Smith, J. M. (1975). Sulfur dioxide oxidation in
r reaction rate, mol=g s slurries of activated carbon. A.I.Ch.E. Journal, 21(4), 664–676.
Ledakowicz, S., Nettelho:, H., & Decker, W.-D. (1984). Gas–liquid
t time, s mass transfer data in stirred autoclave reactor. Industrial and
VG volume of gas-phase, l Engineering Chemistry Fundamentals, 23(4), 510–512.
VL volume of liquid-phase, l Niiyama, H., & Smith, J. M. (1976). Adsorption of nitric oxide in
aqueous slurries of activated carbon: transport rates by moment
Greek letters analysis of dynamic data. A.I.Ch.E. Journal, 22(6), 961–970.
Oinas, P., & Haario, H. (1994). Transient models for multiphase
slurry reactors. Catalysis Today, 20(3), 525–540.
IN & OUT inlet & outlet volumetric ?ow rate, l=s Ramachandran, P. A., & Smith, J. M. (1977). Dynamics of three-phase
slurry reactors. Chemical Engineering Science, 32(8), 873–880.
Subscripts Skrzypek, J., Sloczynski, J., & Ledakowicz, S. (1994). Methanol
synthesis. Warszawa: Polish scienti8c publisher.
)
Setinc, M., & Levec, J. (1999). On the kinetics of liquid-phase
i ith component (CO, CO2 , H2 , H2 O; methanol synthesis over commercial Cu=ZnO=Al2 O3 catalyst.
Me = methanol, argon) Chemical Engineering Science, 54(15 –16), 3577–3586.
G gas-phase )
Setinc, M. (2001). Liquid-phase methanol Synthesis: Dynamics of a
L liquid-phase slurry reactor. Ph.D. Thesis, University of Ljubljana, Ljubljana.
S catalyst surface

Acknowledgements

The authors wish to thank the Slovenian Ministry of


Science and Technology for the 8nancial support under
Grant No. J2-0783-0104-00 and P0-0521-0104.

Você também pode gostar