Professional Documents
Culture Documents
Lode 2003
Lode 2003
Simulated Moing Bed reactors (SMBR) combine chemical reaction and adsorptie
separation within one single continuous and countercurrent unit. This integration
promises substantial improements in process performance, especially when applied to
equilibrium-limited reactions inoling such heat-sensitie products as fine chemicals
and pharmaceuticals. In this work, the interplay among the releant process design
parameters (dimensionless ratios of the fluid and solid flow rates, and the Damkohler
numbers for each section of the unit) is inestigated. For this, an analytical solution of
differential mass-balance equations for the corresponding true countercurrent process
(TCC), using as a model system the reaction A| BqC with each species exhibiting
linear adsorption behaior, was deeloped. Based on this solution, criteria were deried
for the optimum process design with respect to productiity and solent consumption.
Comparing these results with numerical simulations of an SMBR unit shows that the
TCC model does not apply to SMBR units with a finite number of columns per section,
that is, units of practical releance, because the two units exhibit different residence time
distributions and, hence, lead to different degrees of conersion.
Introduction
Reactive simulated-moving-bed processes SMBR. are hy- assembly, and the inlet and outlet streams are delivered to
brid operations combining chemical reaction and adsorptive the unit in such a way that the unit appears to be divided in
separation within one single apparatus. This process integra- four different sections comprising two columns each. The
tion allows for improvements in terms of chemical conversion countercurrent movement of the solid phase is simulated by
in the case of equilibrium limited reactions. Here, the separa- periodically shifting the inlet and outlet ports by one column
tion of the reaction products immediately after their forma- length into the direction of the liquid flow.
tion effectively suppresses the backward reaction, and com- As the reactant is fed to the unit in between sections 2 and
plete conversion of the limiting reactant can be achieved upon 3, chemical reaction is triggered and the product species are
one single pass through the reactor. SMB reactors thus rep- formed. The product species with the lower affinity to the
resent an alternative to the industrially more established re- solid phase C is less retained and transported by the fluid
active distillation processes in applications requiring benign flow toward the raffinate outlet, while the more retained
operating conditions, as is often the case for fine chemicals product B is strongly adsorbed and transported into sections
or pharmaceuticals. 2 and 1 by the movement of the solid phase. If the operating
A typical configuration of a reactive SMB process is shown conditions are properly selected, the reactant A is com-
in Figure 1 for the case of a monomolecular reaction, A| B pletely consumed within sections 2 and 3, and, therefore, is
qC, taking place in an inert solvent, S. Here, a total of eight absent in either outlet stream. Within section 1, the fluid flow
fixed beds is interconnected in order to form a closed-loop rate is adjusted so that the strongly retained product is des-
orbed from the solid phase, allowing for the recycling of pure
adsorbent into section 4. Within section 4, on the other hand,
Correspondence concerning this article should be addressed to M. Morbidelli. fluid flow rates are low enough to allow for the adsorption of
Current address of M. Mazzotti: ETH Swiss Federal Institute of Technology Zurich,
Verfahrenstechnik, Sonneggstrasse 3, CH-8092 Zurich,
Institut Fur Switzerland. the weakly retained product, and thus for the complete re-
qieq s Hi c i , 3.
HB ) HA ) HC 4.
c i , j s c ia, j , y j s 0, xjG0
qi , j s qib, j , y j s1, x j G 0. 5.
dx j U q 1y U . Hi
s s i , j . 12.
dyj 1y p . m j y Hi .
p p
c ia, j y 1q qia, j dc A, j Da j f c j .
mj q
1y p / 1y p . Hi / dyj
sy
1y p . m j y HA .
13.
s m j y Hi . c i , j 0q . s m j y Hi . c i0, j 6. dc B, j Da j f c j .
s 14.
p p dyj 1y p . m j y HB .
c ib, j y qib, j
mj q
1y p / 1q
1y p . Hi / dcC, j Da j f c j .
s 15.
s m j y Hi . c i , j 1y . s m j y Hi . c 1i , j . 7. dyj 1y p . m j y HC .
Within each section, the concentration profiles are continu- The boundary conditions required for the integration of
ous and given by the partial differential equations just de- Eqs. 13 to 15 need to be derived from an analysis of the
scribed Eq. 2.. slopes of the characteristics. In particular, the sign of i, j de-
In order to find a solution to the system of partial differen- fines the direction within the single TCC section in which
tial equations the method of characteristics is applied here. information is propagated, and, accordingly, the location of
Substituting the adsorption isotherm Eq. 3. into the differ- the boundary discontinuity. In the case of the characteristics
ential mass-balance equation for each species Eq. 2. yields exhibiting positive slopes, that is, i, j ) 0, the information
given by the boundary condition on the left end of the sec-
ci , j ci , j tion c i,a j is propagated into the domain of interest, hence,
w U q 1y U . Hi x q 1y p . m j y Hi . c i,0 j s c i,a j , and accordingly a boundary discontinuity occurs at
xj yj the right end of the section, that is, c 1i, j / c i,b j . The reverse
situation is obtained for i, j - 0, where the situation inside
s i Da j f c j . . 8. the section is determined by the boundary condition at y j s1,
hence, c 1i, j s qi,b jrHi , and the boundary discontinuity is lo-
The corresponding characteristic equations are constituted by cated at the left end of the section, that is, qi,a j / Hi c i,0 j .
the following system of ordinary differential equations In the case of the process under consideration here, the
sign of i, j for each species within one section of the unit is
dx j determined only by the difference between the flow rate ratio
s U q 1y U . Hi 9. m j and the species Henry coefficient Hi . Based on this infor-
ds
mation, only it is possible to derive criteria on the flow rate
dyj ratios within the different sections of the unit to fulfill the
s 1y p . m j y Hi . 10.
ds process specifications. As mentioned earlier, the goal of the
process is to achieve a complete separation of the product
dc i , j
s i Da j f c j . , 11. species, with the more retained product B being collected in
ds the extract stream, and the less retained one C being with-
drawn with the raffinate, and to reach a specified conversion
where s is the coordinate along the characteristic. degree of the reactant A.
By dividing Eqs. 9 and 10, it is seen that, within the physi- The necessary condition for the separation of the two
cal plane x j , y j ., the characteristic curves are parallel straight products is that the more retained product B, and the less
m1 ) HB 16.
HB ) m 3 ) m 2 ) HC 17.
HC ) m 4 . 18.
u Ec AE q u R c AR thus yielding
X s1y
u AF c AF
Da2
s1y
m1 y m 2 . c AE q m 3 y m 4 . c AR
. 22 .
X s1yexp y
1y p . HA y m 2 . / . 28.
m 3 y m 2 . c AF
The resulting expression is completely equivalent to Eq. 25.
Depending on the value of the flow rate ratios relative to the m ) H )m .
3 A 2 Since in this case A,2 - 0 - A,3 , it fol-
Henry coefficient of the reactant, three cases have to be dis- lows that c AF s c A,3
a b
s q A,2rHA , and that both sections are uti-
tinguished: lized for the reaction. Integrating Eq. 13 for js 2 and js 3
m )m ) H . Here, the concentration of A at the left yields
3 2 A
0
end of section 2 is equal to zero, that is, c A,2 s 0, and an
integration of Eq. 13 shows that the reactant is absent in all Da2 1y y 2 .
of section 2. A reaction occurs only within the third section,
and the corresponding concentration profile of A can be ob-
c A,2 sexp y
1y p . HA y m 2 . / c AF 29.
with js 3
thus leading to
m3 y m2
0
c A,3 s c AF 23. HA y m 2 Da2
m 3 y HA
m3 y m2 Da3 y 3
X s1y
m3 y m2
exp y
1y p . HA y m 2 . /
c A,3 s
m 3 y HA
exp y
1y p . m 3 y HA . / c AF . 24 .
y
m 3 y HA
m3 y m2
exp y
Da3
1y p . m 3 y HA . / . 31 .
u Fc AF X m 3 y m 2 . Xk R c AF
PRs s . 33.
L2 q L3 Da2 q Da3
dcC, j m j y HA
Substituting the result into Eq. 31 yields Eqs. 37 and 38, as sy s j . 48.
expected. These apply for given values of Da2 and Da3. dc A, j m j y HC
Within the complete separation region, the maximum value
of m 3 y m 2 . is achieved at the vertex where m 2 s HC and
By integrating Eqs. 47 and 48, expressions relating the varia-
m 3 s HB . At this point, the required conversion degree X
tions in the concentration of the product species, B and C,
can be attained by tuning the Damkohler number; the follow-
within one section to the corresponding variation in the con-
ing values are obtained from Eqs. 37 and 38
centration of the reactant, A, are obtained
2 w 1, j c A0 , j q w 2 , j y'y
yln
2 w 1, j c A0 , j q w 2 , j q'y /
Da j
sy - 0 . , 53 .
1y p . m j y HA .
The case s 0 is not considered here, since it is nongeneric,
that is, it is obtained through a very special combination of
values of the physical parameters that makes little sense to
insist on.
Each of these equations comprises an analytical relation
between the liquid-phase concentrations of the reactant at
0
the inlet and at the outlet of section j, that is, between c A, j
1
and c A, j . Since one of these two quantities is known from the
boundary conditions, the remaining one has to be found by
numerically solving either Eq. 52 or Eq. 53. It is worth noting
that, depending on the values of the flow rate ratios and of Figure 6. Lines of 90% conversion for the reversible re-
the Damkohler numbers, both functional forms of the solu- action A| B H C under conditions of con-
tion can indeed occur.
stant Damkohler numbers, Da 2 s Da 3 s1, as
The shape of the isoconversion lines within the m 2 ,m 3 .- compared to an irreversible reaction A B H
parameter plane shall be discussed using a model system C (solid line); K eq s10 (dotted line); K eq s1
characterized by the same adsorptive properties as in the case (dashed line); K eq s 0.1 (dash-dotted line);
of the irreversible reaction described earlier, that is, HA s1, Henry coefficients: HA s1; HB s1.5; HC s 0.5.
HB s1.5, HC s 0.5, and Da2 s Da3 s1. Complete separation triangle with subregions represented by
thin dashed lines.
The isoconversion curves corresponding to a conversion
degree of X s 0.90 are drawn in Figure 6 for different values
of the reaction equilibrium constant, namely K eq s10 dotted
line., K eq s1.0 dashed line., and K eq s 0.1 dash-dotted line. simulated moving-bed process for a purely separative system,
in addition to the irreversible reaction case solid line.. Also, this is not necessarily the case for a reactive process. In par-
for the sake of clarity the complete separation triangle HC - ticular, the performance of a reactive unit, as compared to
m 2 - m 3 - HB is reported. the purely separative one, depends upon the residence time
It can be readily observed that the smaller the K eq value, of the reactant within the different sections of the unit. Ac-
the smaller the region where 90% conversion or more can be cordingly, in the following the residence time distribution
achieved with respect to the irreversible reaction. Close to RTD. for an SMB system with discrete shifting of the
the diagonal, the isoconversion lines for finite K eq values are columns is compared to that of the corresponding TCC pro-
very similar to the ones for the irreversible reaction case. This cess. The results obtained from the RTD analysis are as-
can be understood by considering that under very dilute con- sessed by comparison with detailed numerical simulations of
ditions corresponding to m 3 m 2 , that is, u F 0, the sec- a simulated moving-bed reactor.
ond-order backward reaction is reduced more than the first- The synthesis of methyl acetate from methanol and acetic
order forward reaction. On the other hand, in the neighbor- acid, catalyzed by a sulfonated ion-exchange resin, under infi-
hood of the optimal point, the reversible case can achieve nite dilution conditions in methanol was chosen for these
only a worse performance than the irreversible case for the simulations. The sorption isotherms described previously can
given values of the Damkohler numbers. Compared to the be approximated under these conditions by linear isotherms,
irreversible reaction, the maximum productivity at 90% con- with Henry coefficients HA s 0.063, HW s 2.60, and HE s
version is reduced by about 12% in the case of K eq s10, but 0.038 for acetic acid, water, and methyl acetate, respectively
by 38% and by 71% in the case of K eq s1 and 0.1, respec- Lode et al., 2001.. Furthermore, because methanol is pre-
tively. sent in large excess, the esterification reaction is of zeroth
order in methanol and can be approximated by the
Comparison of TCCR with SMBR monomolecular reaction, AW q E, where the equilibrium
limitation is disregarded in order to obtain an analytical solu-
In the first part of this work, the performance of a true tion of the model equations. A pseudo-first-order expression
countercurrent reactive adsorption process characterized by for the reaction rate is adopted, which accounts for reaction
linear adsorption isotherms, infinite mass-transfer kinetics, in both the fluid and the solid phase
and negligible axial dispersion has been analyzed in order to
achieve a deeper understanding of how operating parameters
control process performance. r s k R c A s U k L c A q 1y U . k S HA c A s U rL q 1y U . rS
While the behavior of the true countercurrent process has 54.
been shown to be completely equivalent to the corresponding
m j 1y U . q U Lcol
A, j s , js 2, 3 60.
w HA1y U . q U x tU
1y 2 L col
y2
Lcol
for 0 F - 3
Lcol
A,3 tU / A ,3 / A ,3
Lcol Lcol tX3 y U
t y LcolrA ,3 . Lcol
y3 F - 3,1
E . s~
A,3 t
U A ,3 / q U
t 3 ,1 y3L colrA ,3 .
for 3
A ,3
62.
tU y tX3 3 Lcol Lcol
for 3,2 F F q
tU 3 LcolrA ,3 q LcolrA ,2 y 3 ,2 . A ,3 A ,2
3 Lcol Lcol
0 for ) q .
A,3 A ,2
Conclusions
To describe the behavior of a true countercurrent reactive
chromatography process in terms of reactant conversion and
product separation, the concentration profiles inside the unit
have been derived analytically using an equilibrium theory
model. Based on this solution, expressions for the reactant
conversion as a function of the flow rate ratios as well as of
the Damkohler numbers within the central sections of the
unit have been derived for both an irreversible and a re-
versible reaction, where A gives B and C and the adsorption
equilibria are linear. In the case of an irreversible reaction,
explicit design criteria have been derived for the Damkohler
Figure 9. Residence-time distribution for the case of m2 numbers, as well as for the flow rate ratios, in order to obtain
s1.10 and m3 s3.00, switch time 13.33 min. optimum process performance in terms of productivity and
solvent consumption at a constant degree of conversion, that
is, for specified product purities.
Note here that the constraint of Eq. 61 is more restrictive Although valid only under specific conditions, the solution
than m 3 ) m 2 ) HA , which only requires each individual term of the TCC model that has been developed indicated the
on the lefthand side of this inequality to be smaller than tU. possibility of using the true countercurrent reactor model as
For the case of m 3 s 3.00 and m 2 s1.10, the correspond- an approximation for the simulated moving-bed reactor. This
ing RTD is shown in Figure 9. The two Dirac delta functions equivalence between TCC and SMB is widely used for nonre-
originate from tracer molecules that have never entered sec- active, purely separative processes, even when the number of
tion 2, and the distance between the two peaks is given by columns is relatively small. However, a clear disagreement
the residence time of reactant A in one single column of between the analytical TCC model results in the presence of
section 3. The two rectangular regions at higher residence reactions, and the ones obtained from the numerical simula-
times, on the other hand, refer to molecules experiencing a tion of the SMBR was found for units with a finite number of
smaller fluid velocity within section 2 as compared to the one columns per section. This inequivalence is due to the differ-
in section 3. ence in the residence-time distributions in the two units. This
In order to calculate the conversion in the SMBR unit, the conclusion has been confirmed by comparing the results of
integral in Eq. 59 has to be evaluated. Here, attention has to the SMBR model given with those obtained by first calculat-
be paid to the fact that the batch-reactor conversion has to ing the residence-time distribution of the SMBR and then
be derived for a heterogeneous system. In particular, the cor- applying the model of complete segregation to calculate con-
responding material-balance equation is given by version in the case of linear reaction kinetics. Of course, in
the case where the number of columns in the SMBR is very
dc A large, the switch time in the SMBR is very small and the
w U q 1y U . HA x sy k R c A sy U k L c A RTDs of the SMBR and the TCCR become very similar.
d
However, this is a situation of little interest in practical appli-
y 1y U . HA k S c A , 63. cations. Therefore, it can be concluded that, contrary to what
is commonly done for separative systems, the analogy with
leading to the following expression for the conversion as a the corresponding reactive TCC unit cannot be used when
function of time dealing with SMBRs of practical relevance. In these cases,
the results obtained through simplified or detailed TCCR
k R models can be used as initial estimates to be refined using a
X batch s1yexp y
U q 1y U . HA / 64. detailed SMBR model.
Acknowledgments
The conversion values computed for the operating points
along a line of unitary slope in the m 2 ,m 3 . plane are shown Financial support of Roche AG, Basel, Switzerland, is gratefully
acknowledged.
by the broken line in Figure 7. It is seen that the conversion
values computed from the RTD model and through the de-
tailed SMBR model coincide within the region of flow rate Notation
ratios considered. The same conclusion is reached when con- c i, j sliquid-phase concentration of species i within section j
sidering the data shown in Figure 8. In particular, the in-
Da j sDamkohler number in section j
of the total amount of species A fed. The set of molecules eluted directly to the raffinate com-
In principle, a fraction of these molecules is eluted to the prises all molecules fed during the interval tX2 F tX - tX3 , with
raffinate within the second switch interval, tU F tX - 2 tU , while the critical feeding time tX3 given by
the remaining ones experience one or more additional port
switches, depending on the propagation velocity of species A
1 Lcol
in section 3, A,3 , and thus depending on the flow rate ratio tU s w Lcol y tU y tX3 . A ,3 x q2 . A9.
in this section. None of these molecules is shifted into section A,2 A ,3
2.
For the sake of clarity, the analysis is here restricted to the Each of these molecules has to pass through three columns,
case where all molecules of this subset are eluted during the with a part of one column being transversed at a reduced
second switch interval. This requires the molecules to travel velocity. The residence times are therefore uniformly dis-
at least a distance of two column lengths during one com- tributed between the one for the molecules being fed at tX s
plete switch cycle, and the required propagation velocity is tX2 , which are shifted to section 2, but leave the section imme-
thus lower bounded by diately after the port movement, and the ones being fed at
tX s tX3 , which are able to elute to the raffinate just prior to
Lcol the second port shifting. Accordingly
A,3 ) 2 . A6.
tU
3 Lcol 3 Lcol 2 Lcol A ,2
F F q tU y s 3,1 . A10.
Under these conditions, each of the molecules experiences
exactly one switch of the ports, and therefore has to travel a A,3 A ,3 A ,3 / 1y
A ,3 /
distance of 2q1. Lcol before being eluted with the raffinate
stream, leading to a residence time of The remaining molecules proceed only to the second column
of section 3, and hence experience an additional switch of the
3 Lcol columns before eluting to the raffinate. The residence times
2s . A7. are therefore uniformly distributed between
A,3
Finally, subset 3 comprises all the molecules that are 4 Lcol 2 Lcol A ,2 3 Lcol Lcol
3,2 s q tU y F F
present within the first column of section 3 at the end of the
feed interval, that is, at tX s tU , and are thus moved into sec-
A,3 A ,3 / 1y
A ,3 / A ,3
q
A ,2
.