Chemical Engineering Science. Vol. 49. No. 248, pp. 546S.5472.

1994
Copyright 8 1995 Elscvier Scicncs Ltd
Printed in Crreat Britain. All rights reacrwd
wo9-251X/94 S7.00 + 0.00

0009-2509(94)00383-l

THE SIMULATED MOVING BED CHEMICAL REACTOR

THOMAS N. HAYNES and HUGO S. CARAM*

Department of Chemical Engineering, Lehigh University, Bethlehem, PA 18015, U.S.A.

(Received 9 May 1994; acceptedfor publication 5 October 1994)

Abstrati-A novel method of maintaining a transient cyclic steady state in a fixed bed tubular reactor is
examined. This process, the simulatedmoving bed chemical reactor, utilizesthe differencesin the thermal
and mass velocities of the system to maintain autothermal behavior within the reactor. This system can
process low calorific fuels and low temperature feed streams, is insensitive to minor disturbances, and
distributes the load evenly over all rhe catalyst in the bed. This work presents, for the first time, theoretical
resultsconcerning the operation of these systems. Both irreversibleand reversiblereactions are considered,
and comparisons are drawn betweenthe new process and the reverse flow reactor.

INTRODUCTION Operation of the simulated moving bed chemical

It is well known that dynamic operation can improve reactor can be separated into two phases. In the first
the conversion and selectivity of chemical reactors. phase, the exit of the first reactor bed is connected
Boreskov and Matros (1983) demonstrated that ben- through a series of insulated pipes to the inlet of a
eficial transient operation can result when the gas feed second reactor bed. This configuration is shown in
direction is periodically switched in heterogeneous Fig. 1. Consequently, the exit gas of the first reactor
gas-solid tubular reactors. This system, the reverse is used to preheat the second reactor bed. When the
flow reactor, utilizes the “wrong way” phenomena temperature front passes out of the first reactor and
and thermal storage capacities of the catalyst bed to into the second reactor, the inlet to the system is
generate autothermal behavior. Flow reversal is, how- changed from the first bed to the second bed. At the
ever, not the only way to utilize these phenomena same point, the outlet of the second reactor is con-
(Matros, 1989). Matros has explored several other nected to the inlet of the first reactor. This initiates the
methods of accomplishing sustained dynamic beha- second phase of the system and is demonstrated in
vior, which include configurations he called “the Fig. 1. In phase 2, the exit gas of the second reactor
rocker”, “the torus”, “the match” and “the wings”. is now used to reheat the first bed. When the tempera-
This paper examines, in more detail, “the torus” ture front exits the second reactor, the system is
scheme. This process, which the authors call either switched back to the phase 1 configuration and the
the simulated moving bed chemical reactor or the whole cycle is repeated.
ring reactor, also uses the thermal storage capabil- The simulated moving bed chemical reactor is an
ities of the catalyst bed and the “wrong way” beha- addition to a growing class of processes that profits
vior to maintain a favorable dynamic state. from the differences in two wave velocities. In adia-
Figure 1 shows a diagram of the simulated moving batic fixed bed reactors, the differences between the
bed reactor and illustrates its mechanical operation speed of the thermal and mass waves are utilized to
for exothermic systems. To ignite the system, the generate autothermal behavior. In chromatographic
first reactor bed is preheated to a temperature where reactor systems, the differences in the speed of the
the reaction rate is significant. By introducing a cold adsorption isotherms of the reactants and products
reactant feed gas, a “wrong way” effect (Boreskov are utilized to achieve separation of the species and
and Slinko, 1965; Pinjala et al., 1988) is generated to favor equilibrium conversion. A stable cyclic
and a temperature front with a high maximum tem- steady state can be maintained by either reversing
perature develops. This temperature front moves the flow or simulating a moving bed. Thus, four dis-
down the length of the first reactor bed and would tinct areas exist, which combine one of the two meth-
exit the system if no other measures were taken. In ods to accomplish one of the two objectives. The two
this case, the reactor would extinguish. Reverse flow most researched systems are the reverse flow reactor
operation prevents extinction by reversing the feed (Matros, 1989), which uses flow reversals to generate
direction before the front exits the reactor. On the autothermal behavior, and the simulated moving bed
other hand, in the simulated moving bed reactor, chromatographic reactor (Ray et al., 1994), which
the temperature front is maintained by chasing it uses a simulated moving bed for chromatographic
around a ring of at least two reactors. separation. Recently, it was demonstrated that
reverse flow operation can be exploited for chromato-
‘To whom correspondence should be addressed. graphic separation (Agar and Ruppel, 1988). Finally,

5465
5466 THOMAS N. HAYNES and HUGO S. CARAM

Insulation Insulation

Fig. 1. Switching mechanism of simulated moving bed reactor

the present paper examines using a simulated moving Note that, although this model formulation allows for
bed for maintaining autothermal behavior. variable lengths of each section of the ring reactor, we
will limit our focus to systems to where all the reactor
beds are identical (Li = L). Parameters for this model,
MODELING which describe the processing of a possible dilute
For the purposes of this study, a reduced version of waste gas stream of volatile organic compounds, can
the homogeneous, or one-phase, model given by be found in Eigenberger and Nieken (1988). The
Eigenberger and Nieken (1988) will be used. The model parameters are reproduced in Table 1 and are
homogeneous model considers the gas and solid distinguished by a low adiabatic temperature rise of
phase to be at the same temperature. In addition, 50K.
the mass accumulation term of the mass balance is Equations (1 j-(7) are integrated numerically using
neglected, axial mass dispersion is ignored, and all a modified finite-difference method (Haynes et al.,
parameters are considered to be constant. For an 1992). The method represents the stiff linear portions
adiabatic plug flow reactor, the mass balance can be of the energy balance implicitly and the nonlinear
expressed as reaction rate explicitly. This introduces stability to
the solution, but avoids the requirement of a non-
(1) linear equation solver. The resulting set of equations
can be quickly solved for each time step by using a
where i = 1 indicates the first reactor and i = 2 indi- slightly modified tri-diagonal matrix solver.
cates the second reactor. The corresponding energy To insure ignition, a uniform high temperature
balance is profile was chosen as the initial conditions for the
6’Ti d’ Ti calculations. The reactor is now simulated until the
Pb$b x = xs - “&P&g !?Y!
aZ + E(-AI%)r(Xi, 7-f) switching time is reached, at which point the bound-
ary conditions are switched to reflect the movement of
(2) the inlet and outlet of the gas stream. If the switching
For eq. (2), the boundary conditions are the standard continues long enough, a stable cyclic steady state is
Danckwert’s boundary conditions, except that the approached, which satisfies
feed temperature to each reactor is alternated
U(z, t) = U(z, c + 2t.y) (8)
between the system feed temperature and the outlet
temperature of the other reactor. The complete set of It is convenient to define a pseudo-solids flow velo-
boundary conditions for each phase are: city of a system of N similar length beds (Li = L) as

Phase 1: Phase 2:

Xl(O,f) = 0 X1(0, t) = Xz(J52, t) (3)

x2 (0, t) = XI (G 17) X*(0,2) = 0 (4)

A- = vcpgcpg(Tl(O,r) - x dTl(O, t)
To) ~ = w~c~#I(O, t) ~ lz(L2, I)] (5)
az

A- = vq+cpg[T2(0, t) - T, (151,t)] AaT2(o,

~
a2
4
= YFP~c~~(T~(O,t) - 7-O) (6)

aq(G, 1) = o
t3Z
(7)
 The simulated moving bed chemical reactor 5467

Table 1. Parameters for the combustion of trace amounts of The cyclic steady state shows several notable char-
volatile organic compounds in waste gas streams acteristics (Matros, 1989). First, even though the gas
0.121 mol/m3 206,000 J/mol
velocity is relatively high (= 1 m/s), the propagation
G (-AH)
66,512 J/m01 0.8 speed of the temperature profile is of the order of
k” 29,732 s-’ ;, 2.06J/msK 1 mm/s. Thus, a large amount of gas can be processed
L” 0.5m PhC&lb 2.0 x lo6 J/m3 K before switching is required. Second, the maximum
T” 293 K P&G? 500 J/m3 K
temperature of the reactor is high (1167 K) despite
” 1 mjs
the fact that the feed temperature is low (293 K) and
the adiabatic temperature rise is small (50K). Third,
part of the catalyst bed is used to preheat the feed and
3 = L/t* (9) cool the effluent, which may eliminate some or all of
where the switching time, ts, is constant for all beds. the need for external heat exchange.
This term is equivalent to the discretized solids velo- An additional and important property, which has
city that is observed by a moving frame of reference not been previously mentioned, is also apparent from
fixed at the temperature front. If the pseudo-solids Fig. 2. The temperature and conversion profiles show
velocity is kept constant as the number of beds that the reaction occurs within a relatively narrow
approaches infinity, then the model will reduce to zone when compared to the overall reactor length.
that of a moving bed reactor which recycles the solids. This zone migrates throughout the entire length of
Directly related to the pseudo-solids velocity is the the reactor, which means that all of the catalyst will
net convective strength (NCS) of the system. The be utilized uniformly. This feature is not present in
NCS is a direct measure of the strength of the convec- either the steady state or reverse flow reactor. Since
tion processes in the reactor and is given by the thermal wave also propagates throughout the
entire body of the reactor, the “temperature load”
NCS = &CpsVc - &C,,bi (10) of the ring reactor will also be more evenly distribu-
The value of (NCS)A T will give the flow sensible heat ted.
between two points in the reactor which have tem-
perature difference AT. When the NCS is positive
(negative), then the heat capacity of the gas is larger Comparison with the reverseflow reactor
(smaller) than the heat capacity of the flow of solids, There are two main processes which should be
and there will be a convective flow of energy in the examined as alternatives to the ring reactor: the
positive (negative) direction. Note that by proper steady state system and the reverse flow reactor.
choice of the cycle time, and hence pseudo-solids The ring reactor, and the reverse flow reactor for
velocity, it is always possible to set NCS = 0. This is that matter, utilize the available energy of reaction
the “convectionless” case. more efficiently than the steady state system. Since
in the present case AT& = 50 K, ignition and reac-
tion could not be accomplished in a steady state sys-
IRREVERSIBLE REACTIONS tem, unless an external heat exchange system was also
An important potential application of this technol- considered along with the reactor. Therefore, a
ogy is the combustion of trace amounts of volatile simple and direct comparison against a steady state
organic compounds in waste gas streams. This sys- system is not possible with the present case.
tem can be modeled using a simple, irreversible, It is proper to compare the ring reactor with the
first-order reaction A -+ B. In this case, the reaction reverse flow reactor. Table 2 shows the results of the
rate for the model is r(X, T) = qk,eeRjRT( 1 - x). calculations. The comparison was accomplished by
For the parameters given in Table 1, the cyclic designing a “convectionless” ring reactor to achieve
steady state of the two bed “convectionless” ring 99% average conversion with a AT,, of SOK. This
reactor is shown in Fig. 2. Since the two reactors are gave a total reactor length for both beds of 1.056m
identical and at cyclic steady state, both reactors see and a switching time of 528 s. A reverse flow reactor
the same set of temperature profiles, except for a was then designed to produce an average of 99%
time delay equal to the switching time, conversion with AT,,d = 50K and a cycle time of
T1 (z, t) = Tz(z, t + tJ. A consequence of this beha- 528s. By equating the cycle time of the two pro-
vior is that, in the future, only one reactor bed need cesses, variations in the switching losses can be kept
be presented to fully explain the cyclic steady state. to a minimum. Notice that, at the beginning of phase

Table 2. Comparison between the reverse flow reactor and the ring reactor for an irreversible system

Length (m) Maximum Average Adiabatic Average Maximum load Cycle time
temperature temperature temperature conversion (mol/m3 s) (s)
(K) (K) rise (K) W)

Reverse flow reactor 0.902 1165.6 594.3 50 99 0.237 1056

Ring reactor 1.056 1175.9 583.8 50 99 0.116 1056
5468 THOMASN. HAYNES and HUGO S. CARAM

Fig. 2. Cyclic steady state of a simulated moving bed reactor.

- Simulated Moving Bed

120@
Chemical Reactor --. .
I _..*
(L = 1.056 m) / ‘.
; “;
,OOO- ._.......... f:
Reverse Flow Reactor
:’ ‘.
(L = 0.902 111) /\ ‘:,
Z

200!~,!~~,> 1,,,,,,,,,,,,,1,,,,,,,,,,,,,,,,,,,,1,,,1,,,,,

0 I). I 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Length (m)

Fig. 3. Temperature profiles in the reverse flow reactor and the ring reactor for an irreversible system. Both reactors provide
99% time-averaged conversion for a system with a AT,, = 50 K.

2 (l), the unreacted material accumulated in bed 1 (2) case. The main difference occurs because a longer tail
will be swept out of the reactor. section of the reactor temperature profile is required
Figure 3 shows a comparison of the two tempera- to keep the ring reactor stable. This is the main cause
ture profiles just prior to switching. There is a striking of the extra reactor length that is necessary for opera-
resemblance between the temperature profiles in each tion with the simulated moving bed chemical reactor
 The simulatedmoving bed chemicalreactor 5469

and may represent a significant disadvantage of using to the inlet of the system when the inlet/outlet switch
a ring reactor. occurs. This can lead to excessive energy losses and
Evaluation of the performance of the two reactors can be eliminated by allowing for a slight convective
just based on length considerations is somewhat mis- driving force. This is demonstrated in curves C and D
leading. The load of the catalyst can be defined as which show the cases where the NCS is 8 and 15.2 J/
m’s K respectively. In these cases, the small amount of
y(z) = i ‘=Y=
r[X(z, a), T(r, (r)] dcl convective flux shifts the reaction zone deeper into the
(11)
tcyc J0 reactor, which leads to lower energy losses and higher
temperatures. Too much net convection will even-
and can be important for comparison purposes. It is tually lead to serious convective heat losses and
especially important in systems where the catalyst would extinguish the system. The figure also shows
deactivation is a direct function of the load on the the case where the NCS is -4 J/m%K which means
catalyst. As can be observed from Fig. 4, the catalyst that the net convection is in the negative direction. As
load distribution for the ring reactor is nearly con- the figure shows, this quickly leads to lower tempcra-
stant except for a small end effect region. In addi- tures and poorer performance.
tion, the load on the reverse flow reactor is much The maximum temperature as a function of the
less evenly distributed, and the maximum load in pseudo-solids velocity is shown in Fig. 6. The figure
the reverse flow reactor is much higher than that of also shows that the window of operation where igni-
the ring reactor. Therefore, one would expect the tion can be maintained is relatively narrow and is
catalyst to deactivate more rapidly and more about 29s. This is in direct contrast to the reverse
unevenly in the reverse flow system than in the ring flow reactor where the available window of opera-
system. This extended catalyst life may offset the cost tion of cycle times is much larger (w 500s;
of the longer beds of the ring reactor, and will make Eigenberger and Nieken, 1988). In the reverse flow
the ring reactor more comparable to the reverse flow reactor, this large window occurs because the equiva-
reactor. Further investigation and experimental lent net convective strength is always zero and is not a
efforts are required for further evaluation of this function of cycle time. Since the window of operation
property. is narrow, operation of the ring reactor will require
precise switching. However, if the control system has
Influence of the net convection the ability to make these precise switches, then the
Figure 5 shows the influence of the pseudo-solids cycle time can be. used to affect the shape of the
velocity on the form of the cyclic steady state tem- profile, which is not as easily done in a reverse flow
perature profiles. All bed lengths for the simulation reactor.
were 0.5m each, and the cycle time was varied to
obtain the desired pseudo-solids velocity. Curve B
of the figure shows the temperature profile which REVERSIBLE REACTIONS
was presented previously in Fig. 2. This is the case The important potential applications of both the
where NCS = 0 and is the “convectionless” case. reverse flow reactor and the ring reactor are revers-
For curve B, the reaction zone is located fairly close ible exothermic reaction systems which often have low

Reverse Flow Reactor

(L = 0.902 m)

Simulated Moving Bed

Chemical Reactor
(L = 1.056 m)

Length (m)
Fig. 4. The time-averagedcatalystload in each system.
5470 THOMAS
N. HAYNESand Huoo S. CARAM

Reactor 1 Profiles

Fig. 5. Influence of the pseudo-solids velocity upon the temperature profiles of the cyclic steady state. A, NCS = -4
J/m2 s K; B, NCS = 0; C, NCS = 8 J/m2 s K, D, NCS = 15.2 J/m2 s K.

“Convectionless” ;
Cafe w

Window of
Operation
29 Seconds

n
Pseudo-Solid Velocity (mm/s)
Fig. 6. The available window of operation.

adiabatic temperature rises. This includes the oxida- versions than the reverse flow reactor for the same
tion of sulfur dioxide, the production of methanol, the length reactor. In the example chosen, the time-aver-
Claus process to produce elementary sulfur, and aged outlet conversion for the ring reactor is 67.9%
ammonia synthesis. For the sake of generality, this and for the reverse flow reactor is 66.2%. The better
work only considers the simple A +P B reaction with performance that is achieved in the ring reactor for
first-order kinetics [r(X, T) = k,qe-E’RT( 1 - x)- reversible systems can be attributed to the long tail
/c,Pae-e’RT( q/c + X)] . Hypothetical parameters section of the temperature profile. The reaction will
for this system are available (Levenspiel, 1989) and come very close to equilibrium at the maximum tem-
are presented in Table 3. perature of both systems. Then, as the temperature
The conversion and temperature profiles for a ring drops along the trailing edge of the profile, an extra
reactor and a reverse flow reactor of equal lengths are push in conversion is obtained. In the reverse flow
presented in Fig. 7. In the case of the reversible sys- reactor, the trailing edge profile is very steep and
tem, the ring reactor now begins to give higher con- the reactor rapidly becomes kinetically limited. On
 The simulated moving bed chemical reactor
I

Fig. 7. Cyclic steady states of the reverse flow and ring reactor for a reversible system.

tlet Conversion

600 700 800 900

Temperature (K)
Fig. 8. The trajectories of the reverse flow and ring reactor in the temperature conversion phase plane

Table 3. Parameters for the generic reversible reaction graph, the extra conversion is obtained. Notice that
both trajectories approximate the optimal trajectory
G 1 mol/m3 TO 293 K
0 mol/m3 1 m/s profile.
c;:
E 83,700 J/mol ;-/AH) 80,000 J/mol
167,400 J/m01 0.4
2 1.0966x lo6 1/s : 4.5 J/m s K CONCLUSIONS
k: 6.566 x 1O’Ol/s 1,328 J/m3K
PbCPb
The simulated moving bed chemical reactor is pri-
L 0.6m P.?CPR 1.0 x IO6J/m3K
marily applicable for mildly exothermic reaction sys-
tems and, therefore, can be considered an alternative
the other hand, the ring reactor shows a long smooth to the reverse flow reactor. Applications include, but
trailing edge and allows for higher conversion. The are not limited to, the combustion of volatile organic
conversion profiles in Fig. 7 show this behavior. compounds, the oxidation of sulfur dioxide, the pro-
This extra push in conversion can be better under- duction of methanol, the Claus process to produce
stood by examining the temperature conversion phase elementary sulfur, and ammonia synthesis. In gen-
plane in Fig. 8. Shown on the figure is the equilibrium eral, the reactor will favor reactions requiring high
restriction of the reaction, along with the optimal temperatures to enhance the reaction rate and lower
reactor trajectory. For both the systems, the system exit temperatures for high equilibrium outlet conver-
quickly approaches the equilibrium restriction sion. It is especially effective for low heating value
around 900 K. When the trailing edge of the tempera- mixtures, since it also provides a simple, inexpensive
ture profile is reached and the temperature drops, the and highly efficient integrated heat recovery system.
equilibrium restriction is relaxed. Under these circum- The major advantages of the simulated moving bed
stances, which occur after the “pinch point” in the reactor system are its autothermal behavior and its
5472 THOMASN. HAYNESand HUGO S. CARAM

nearly uniform catalyst utilization. It is, however, a Greek letters

very complex dynamic multibed system. There is only y time-averaged catalyst load, mol/m3s
a small window of cycle times in which the reactor can (-AH) heat of reaction, J/mol
maintain ignition. In addition, the reactor is less AT*,j (-AH)~/P,%E
stable than the reverse flow reactor and requires void fraction
longer beds in the irreversible reaction rate case. thermal axial dispersion, J/Km s
bulk density of catalyst bed, kg/m3
NOTATION density of gas stream, kg/m3
catalyst bed heat capacity, J/kg K
reactant gas heat capacity, J/kg K
concentration of reactive species, mol/m3 REFERENCES
inlet concentration of reactive species, mol/m3 Agar, D. W. and Ruppel, W., i988, Extended reactor con-
inlet concentration of product species, cept for dynamic denox design. C&m. Engng Sci. 43 (Q,
mol/m3 2073-2078.
E rate constant activation energy, J/mol Boreskov, G. K. and Matros. Y. S., 1983, Flow reversal of
reaction mixture in a fixed catalyst bed - a way to
6 reverse rate constant activation energy, J/mol increase the efficiency of chemical processes. Appi.
k, reaction rate constant pre-exponential factor, Catal. 5, 337-343.
l/s Boreskov, G. K. and Slinko, M. G., 1965, Modelling of
kr reverse rate constant pre-exponential factor, chemical reactors. Pure Awl. Chem. 10, 61 I.
Eigenberger, G. and Nieken,.c., 1988, Catalytic combustion
l/s with periodic flow reversal. Chem. Engng Sci. 43 (S),
L length of reactor, m 2109-2115.
NCS net convective strength, J/ms K Haynes, T. N., Georgakis, C. and Caram, H. S., 1992, The
reaction rate mol/m3s application of reverse flow reactors to endothermic reac-
x gas constant, 8.314 J/mol K tions. Chem. Engng Sci. 47 (9-l l), 2927-2932.
t time, s Levenspiel, O., 1989, The Chemicul Reactor Omnibook. OSU
Book Stores, Corvallis, OR.
ts?y= cycle time, s Matros. Y. S.. 1989. Catalvtic Processes Under Unsteadv-
ts switching time, s Stat; Condi;ions. &evie~, New York.
T reactor temperature, K Pinjala, V., Chen, Y. C. and Luss, D., 1988, Wrong-way
TO inlet feed temperature, K behavior of oacked-bed reactors: Imoact of thermal dis-
v gas velocity, m/s persion. AZdhE J. 34 (IO), 166%167?
^ Rav, A. K., Carr. R. W. and Aris, R.. 1994. The simulated
pseudo-solids velocity, m/s &nter&rent moving bed chromatographic reactor: A
l- l-C/C novel reactor-separator. Chem. Engng Sci. 49 (4), 469-
z length coordinate 480.