Separation of racemic bicalutamide by an optimized combination of continuous chromatography and selective crystallization

Article abstract of DOI:10.1021/op200136z

A racemic mixture of bicalutamide, a drug substance used in the treatment of prostate cancer, was separated by simulated moving bed chromatography, with the objective of maximizing throughput at reduced outlet purity. The enriched extract stream was purified further by a crystallization process exploiting a shift in the eutectic composition. The optimal purity in between both process steps was identified. The separation scheme developed was validated on a scale of 600 g, and the results were compared to those of state-of-the art and discussed. The investigated scheme revealed a wide range of promising coupling conditions while showing superior productivities and enhanced process robustness.

Full text of DOI:10.1021/op200136z

ARTICLE
pubs.acs.org/OPRD
Separation of Racemic Bicalutamide by an Optimized Combination of
Continuous Chromatography and Selective Crystallization
Henning Kaemmerer,*,†,‡,z Zoltan Horvath,‡ Ju Weon Lee,‡ Malte Kaspereit,§,# Robert Arnell,^
Martin Hedberg,^ Bj€orn Herschend,^ Matthew J. Jones,^ Kerstin Larson,^ Heike Lorenz,‡ and
Andreas Seidel-Morgensten‡
†Evonik Industries AG, 63457 Hanau, Germany
‡Physical and Chemical Foundations of Process Engineering, Max Planck Institute for Dynamics of Complex Technical Systems,
39106 Magdeburg, Germany
§Process Synthesis and Process Dynamics, Max Planck Institute for Dynamics of Complex Technical Systems, 39106 Magdeburg,
Germany
^AstraZeneca R&D S€odert€alje, Pharmaceutical Development, 15185 S€odert€alje, Sweden
S Supporting Information
b
ABSTRACT: A racemic mixture of bicalutamide, a drug substance used in the treatment of prostate cancer, was separated by
simulated moving bed chromatography, with the objective of maximizing throughput at reduced outlet purity. The enriched extract
stream was purified further by a crystallization process exploiting a shift in the eutectic composition. The optimal purity in between
both process steps was identified. The separation scheme developed was validated on a scale of 600 g, and the results were compared
to those of state-of-the art and discussed. The investigated scheme revealed a wide range of promising coupling conditions while
showing superior productivities and enhanced process robustness.
1. INTRODUCTION
(R)-Bicalutamide (compound 1, Figure 1) is the active phar-
maceutical ingredient (API) in AstraZeneca’s product CASO-
DEX. The compound is an orally active, once-a-day nonsteroidal
anti-androgen for the treatment of prostate cancer.1 CASODEX is
the racemate of bicalutamide. The manufacturing method designed
Figure 1. Structures of racemic bicalutamide 1 and the pure (R)-
during a development program within AstraZeneca for the purpose of
enantiomer 2.
manufacturing the pure (R)-enantiomer of bicalutamide (compound
2, Figure 1), utilizes SMB-HPLC for full enantioseparation, followed
investigate the temperature and solvent dependence of the
by crystallization of pure, crystalline solid from methanolꢀwater to
position of the eutectic for this racemic compound forming system.
obtain the required purity and solid-state properties of the API.
As will be shown in this article, detailed characterization24 of
An analysis of the published asymmetric syntheses of both
the crystallization behavior of bicalutamide, provides insight into
(R)- and (S)-bicalutamide leads to the conclusion that all
new possibilities in terms of process solutions for manufacturing
available methods lack industrial potential for various reasons
the desired (R)-enantiomer.
such as low overall yield or high cost of raw materials.2ꢀ10
Consequently, the racemate, which is produced using an efficient
and cost-effective chemical process (not shown), is currently the
cheapest source of the pure enantiomer.
2. DESIGN AND SETUP OF A CHROMATOGRAPHIC PRE-
ENRICHMENT STEP
Furthermore, the overall efficiency of the manufacturing
process for the active (R)-enantiomer could be further improved,
providing a method for the efficient and economic conversion of
the undesired (S)-enantiomer of bicalutamide were available.
2.1. Experimental Section I: Parameter Determination for
Chromatography. 2.1.1. Chromatographic System. An initial
screening procedure revealed good performance of the
stationary phase Chiralpak AD (Daicel Chemical Industries
Ltd.). Within this study, the adsorption behavior of the two
Since the internal AstraZeneca process design was developed to
perform full enantioseparation of the racemic material, it should
enantiomers onto the stationary phase was investigated
be possible to improve significantly the capacity of the chroma-
using an analytical column with dimensions of 250 mm ꢁ
tography process developed using process integration concepts
developed within the European Union funded Framework
Special Issue: INTENANT
Program 7-project INTENANT.11
Previous internal studies at AstraZeneca on the crystallization
of the pure enantiomer of bicalutamide did not comprehensively
Received: May 26, 2011
Published: September 19, 2011
r
2011 American Chemical Society
331
dx.doi.org/10.1021/op200136z Org. Process Res. Dev. 2012, 16, 331–342
|

Organic Process Research & Development
ARTICLE
Table 1. Column efficiency measured for (S)-bicalutamide
(1) and (R)-bicalutamide (2) as a function of flow rate
(Ver. 7.8, MathWorks, U.S.A.) with a spatial discretization of
Δz = 2DL/u (ref 15, p 49) and an approximate stability criterion
based on the Henry coefficient of the more strongly adsorbed
component (ref 15, p 152).
u0
When using the inverse method, it is important to describe the
injection profiles accurately.13 Here we describe both the front
(f) and the rear (r) of each injection profile by the following
simple approximation:
mL/min
cm/min
NTP1
NTP2
HETP1 mm
HETP2 mm
1
2
3
3
4
5
6.02
12.03
18.05
18.05
24.07
30.09
432
499
432
417
390
336
288
213
126
122
88
0.58
0.50
0.58
0.60
0.64
0.74
0.87
1.17
1.98
2.05
2.84
2.81
cmax , i
ciðz ¼ 0, 0 e t e tiÞ ¼
, i ¼ ðf, rÞ
ti ꢀ t
ð3Þ
1 þ exp
wi
89
In the above expression, cmax,i (i = f, r) is the maximum height
of an injection profile, while ti and wi are parameters determined
by least-squares fitting eq 3 to the measured front and the rear
sections of the injection profiles, determined in the absence of a
column.
Two examples of fitted injection profiles are shown in Figure 2
together with a comparison of the model employed here to a
more detailed, but numerically too inefficient, continuously
stirred-tank-reactor CSTR model of the injection system.
Adsorption was assumed to occur predominately according to
the bi-Langmuir mechanism leading to the following isotherm
equation:
4.6 mm packed with a preparative chiral stationary phase
with a particle size of 20 μm. Pure methanol was used as
mobile phase.
2.1.2. Axial Dispersion and Mass Transfer. Using the above
column, the total porosity was determined as εT = 0.622 from an
injection of a nonretained compound.
The column efficiency was assessed on the basis of apparent
stage numbers NTPi from ‘analytical’ injections using the well-
known approximation12
2
tR2, i
ðtR, iðmeasÞ ꢀ Vtube=QÞ
NTPi ¼ Lc HETPi≈5:54
¼ 5:54
3
w21=2, i
w21=2, i
b1, ici
b2, ici
qi ¼ qS1, i
þ qS2, i
with i ¼ 1, 2 ð4Þ
1 þ b1, ici
1 þ b2, ici
ð1Þ
In order to reduce the number of free parameters to be
determined, the Henry constants of both enantiomers were
determined using very small injections and were estimated as
H1 = 0.536 and H2 = 3.808. These initial slopes of the adsorption
isotherms were kept constant in the later calculations by requir-
ing Hi = qS1,i b1,i + qS2,i b2,i (i = 1, 2). Furthermore, it was assumed
that the first adsorption site is nonselective, i.e. qS1,1 = qS1,2 and
b1,1 = b1,2, and that the saturation capacity on the second chiral
where Lc and HETPi are the column length and the height-
equivalent to a theoretical plate, respectively. In eq 1, the
measured retention time tR,i(meas) was corrected by the delay
due to the dead volume of the system, Vtube/Q, where Q is the
flow rate. w1/2 denotes the peak width at half its height. Table 1
lists the column efficiencies obtained experimentally as a function
of the flow rate for the two enantiomers. A significant difference
can be observed for the NTP values of the two enantiomers. Due
to the observed differences in magnitude of the dispersive effect,
the transport-dispersive model13 was employed for the simula-
tion of the elution profiles of the enantiomers, which are later
required for determining the adsorption isotherms (see section
2.1.2). In this model different mass transfer coefficients km,i, were
adjusted for the two components, while the axial dispersion
coefficient, DL, was assumed to be equal for both components.
The axial dispersion coefficient was calculated from the Chung
and Wen correlation13 as DL = 1.21 ꢁ 10ꢀ6 m2/s. The molecular
diffusion coefficient required for these calculations was estimated
as Dm = 3.62 ꢁ 10ꢀ9 m2/s using the StokesꢀEinstein equation.14
The two mass transfer coefficients km,1, and km,2 were obtained
from the measured HETP values by fitting the latter to the
following relation:12
adsorption site is equal for both enantiomers, i.e. qS2,1 = qS2,2
.
These assumptions reduce the number of variable parameters
to three.
For the parameter determination, a set of six different experi-
ments close to the solubility limit of a racemic mixture, with
injection volumes of 200, 1000, and 2400 μL each at two
injection concentrations of 10.3 and 20.6 g/L, respectively, were
performed. The three free parameters were then determined by
simultaneously minimizing the least-squares difference between
simulated elution profiles and those recorded in the six experi-
ments. The corresponding results are shown in Figure 3. To
confirm the quality of the model parameters derived from the fit,
they were validated against independent experimental runs with
asymmetric composition as shown in Figure 4. The agreement
achieved is considered very good.
The isotherm parameters obtained from the fitting procedure
are listed in Table 2. The adsorption isotherms calculated using
these parameters are valid up to concentrations of approximately
20 g/L.
2.2. Identification of Operating Parameters for SMB Chro-
matography. 2.2.1. Theoretical Approach. As for the determina-
tion of the adsorption isotherms, the transport-dispersive model
was used to represent the chromatographic separation. Triangle
theory,16 one of the short-cut design methods based on equilibrium
theory, was used to identify initial operating parameters for SMB
chromatography. The design parameters in triangle theory are
defined by the ratio of the net flow-rate of liquid to the simulated
!
2
k00, i
2DL
u
HETPiðuÞ ¼
þ 2
, i ¼ 1, 2 ð2Þ
1 þ k00, i k0 km, i
u
0, i
where u = Q/ετA is the interstitial velocity and k00,i the retention
factor for the respective components. The coefficients obtained
were km,1 = 245.9 minꢀ1 and km,2 = 58.8 minꢀ1
.
2.1.3. Determination of Adsorption Isotherms. Adsorption iso-
therms were determined using the inverse method (peak fitting).13
This corresponds to minimizing the difference between experi-
mentally obtained and simulated elution profiles. For the simula-
tion, the transport-dispersive model was implemented in Matlab
332
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
Figure 2. Modeling of injection profiles for the chromatographic runs devoted to quantify the adsorption isotherms. The graphs reveal minor
differences between experimental data (dotted) and the two fitted models of a continuously stirred-tank-reactor CSTR (solid line) and the
approximation function as described in section 2.1.2.
Figure 3. Results of the inverse method applied to six different chromatographic experiments. Comparison of experimental data and fitted elution
profiles using a single set of parameters.
flow-rate of solid as shown below,
Four objective functions representing the goal of the corre-
sponding zones can be defined as:
QitS ꢀ εTVC
mi ¼
!
ð5Þ
MB, Rc
ð1 ꢀ εTÞVC
G1 ¼ log
G2 ¼ log
G3 ¼ log
G4 ¼ log
ð6Þ
ð7Þ
ð8Þ
ð9Þ
MMin, Rc
where εT is the total porosity of the packed column, Qi is the
volumetric flow-rate of the liquid in zone i, tS is the port
switching interval, and VC is the volume of the column. Since
the SMB process has five fundamental operating parameters
(the four flow rates, Qi, and the switching time, tS), the process
can be designed on the basis of the four expressions for the
zone flow-rate ratios mi above after fixing one parameter.
Here, a constraint on the switching time was used (see below).
The most promising SMB operating conditions that maximize
throughput while delivering the required outlet purities can
then be found by optimizing the four zone flow-rate ratios.
Each zone flow-rate ratio has to ensure a proper functioning of
the corresponding zone.
!
!
1 ꢀ PurB, Extr
1 ꢀ PurTarget
1 ꢀ YldTarget
1 ꢀ YldB, Extr
!
MMin, Rc
MA, Rc
333
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
Figure 4. Model validation against independent experiments with asymmetric injection composition. The experiments shown were not used in the
inverse method.
Table 2. Estimated adsorption model parameters
Table 3. Parameters used for the optimization of SMB
operating conditions and for the simulations
(R)-bicalutamide,
i = 2
(S)-bicalutamide,
parameters
value
parameter
[g/L]
i = 1
target purity
target yield
70ꢀ99.9%
99%
qS1,i
b1,i
43.33
43.33
[L/g]
[g/L]
[L/g]
0.01026
15.02
0.01026
15.02
qS2,i
b2,i
Tolerance
0.00609
0.22387
total mass balance
min. recycled amount
zone flow-rate ratio
objective function
0.00005
0.001
Here, Gj is the objective function, PurB,Extr and YldB,Extr are the
actual purity and yield of component B (A is the nontarget
component) at the extract port, respectively. PurTarget and YldTarget
are the target values for purity and yield, respectively. Since the mass
flux of solutes are exponentially increased or decreased by the
changes of the zone flow-rate ratios, the logarithm values were used
to linearize the relationship between the objective values and the
flow rate ratios.17 Note that the objective function Gj corresponds to
the zone logarithm error of which sign indicates the direction to
increase or decrease corresponding zone flow-rate ratio. MMin,Rc is
the minimum mass ratio of each component allowed for recycling
between zones 1 and 4 in one port switching interval. Because of the
periodic port switching in SMB operation, it is difficult to quantify
the recycled amount between zones 1 and 4. We therefore assumed
that the liquid-phase concentration profiles in zones 1 and 4 are
constant to the average concentrations of zone 1 inlet and zone 4
outlet, respectively. On the basis of these assumptions, MA,Rc and
MB,Rc can be calculated from
0.00005
0.01
weighting factors (wP/wD)
numerical method for PDE
no. of nodes
0.5/0.5
FEM with BUDS scheme
100 per column
Implicit Euler
integration method
tolerance
0.00001
routine, which uses only the gradient to obtain the next step
direction, fails to converge due to the influence of the other
zones. Convergence of the optimization routine was assured by
the proportionalꢀdifferential method, which is part of the
proportionalꢀintegralꢀderivative controller (PID) control
method17,18 used. The direction vector, D, was calculated with
B
the following equations:
cA, 1inQExtr
1
̅
T
MA, Rc
¼
ð10Þ
BI P
¼
ðG1, G2, G3, G4Þ ,
cA, FeedQFeed
SP
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
2
2
2
SP ¼ G1 þ G2 þ G3 þ G4
ð12Þ
cB, 4outQRaff
̅
MB, Rc
¼
ð11Þ
ꢀ
ꢁ
cB, FeedQFeed
T
1
ΔG1 ΔG2 ΔG3 ΔG4
BI D
¼
¼
,
,
,
,
SD Δm1 Δm2 Δm3 Δm4
where cA,1in and cB,4out are the average concentrations of component
A and B at the inlet of zone 1 and the outlet of zone 4, respectively.
Four objectives were fulfilled during all simulations of SMB
chromatography with the four design parameters (m1 to m4)
and one fixed set of operating conditions.
As mentioned above, the separation characteristic of each zone
mainly depends on its flow-rate ratio, so that each objective can
be fulfilled separately. However, if too small an absolute tolerance
for the zone flow-rate ratio is used, the iterative optimization
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ꢀ
ꢁ
ꢀ
ꢁ
ꢀ
ꢁ
ꢀ
ꢁ
2
2
2
2
ΔG1
Δm1
ΔG2
Δm2
ΔG3
Δm3
ΔG4
Δm4
SD
þ
þ
þ
ð13Þ
SD
DB ¼ ðwP BI P þ wD BI DÞ
ð14Þ
2
334
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
Figure 5. Changes of productivity and desorbent flow-rate as a function of purity of the target component and the port switching interval. Lines are
drawn as a guide to the eye.
Figure 6. Changes of productivity (open symbols) and desorbent consumption (DC, full symbols) as a function of purity of the target component at the
fixed desorbent flow-rate (QDsrb = 200 mL/min). Lines are drawn as a guide to the eye.
where BI P and BI D are, respectively, the unit vectors of the
directions calculated from the proportional and derivative gains,
wP and wD are the weighting factors of the proportional and
differential directions (wP + wD = 1), and SP and SD are the vector
sizes of the proportional and differential directions, respectively.
At the end of the iterations, the proportional and differential
vectors for the next iteration were obtained from the previous
and current objective functions and zone flow-rate ratio values.
2.2.2. Results of the Estimation Procedure. In order to separate
components with high purity and yield, zones 1 and 4 should
prevent product streams from being contaminated by the recycled,
strongly adsorbed and weakly adsorbed, components, and the
minimum mass ratio of recycled component to feed component
was limited to 0.1%, ensuring that these conditions are met.
Minimization of the loss of target component was achieved by
fixing the yield of target component to 99%. The purity of the target
component was varied from 70% to 99.9%, surveying the changes of
productivity and desorbent consumption. Table 3 summarizes the
parameters for the optimization and simulation. The entire SMB
simulation was carried out with Aspen Chromatography, Ver. 2006.5;
Aspen Technology Inc.: Burlington, MA, USA ,and the optimization
routine was implemented in Aspen as a Visual Basic script.
The operating conditions were obtained for four zone flow-rate
ratios by fixing one set of operating conditions, where the operation
constraints determine the exact conditions. The minimum port
switching interval was 0.25 min due to the mechanical limitations of
the valve system, and the desorbent flow-rate was limited to no more
than 200 mL/min in line with the pump capacity. Figure 5 shows the
change of both productivity and desorbent flow-rate, QDsrb, as a
function of target component purity and port switching interval.
Only 0.1% of the total feed amount was allowed to be recycled
through zones 1 and 4, so that the concentration profiles in zones
335
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
1
1 and 4 are relatively close to zero. This implies that the optimal
m1 and m4 values are constant to the changes of m2 and m3 values,
so that the flow-rate at the desorbent port was not changed by the
purity of the target component since the desorbent flow-rate is
decided by the difference between m1 and m4 and the fixed port
switching interval (= (1 ꢀ εT) (m1 ꢀ m4) VC/tS). Itincreasedasthe
port switching interval shortened. When the port switching interval
reached the lower limit, the flow-rate at the desorbent port was over
700 mL/min. The flow rate at the desorbent port was lower than
200 mL/min only when the port switching interval was at least 0.75
min. Consequently, the constraint on the pump flow rate is of
greater importance than the port switching interval.
Specific productivities and desorbent consumptions produ-
cing the target compound with the required purity and yield were
obtained from the optimized operating conditions. Figure 6
shows the changes in productivity and desorbent consumption
as a function of the purity of the target component. The port
switching intervals were around 0.48 min. In the low purity
region (up to 95%), productivity decreases slowly with increasing
purity, while the desorbent consumption increases. Once the
purity of the target component exceeds 95%, the desorbent
consumption increases rapidly with a concurrent rapid decline of
productivity. At 99.9% target purity, approximately 60% more
desorbent consumption is required compared to a target purity
of 95%, and 40% of productivity is lost.
xeu
¼
ð16Þ
2
ðαrÞ
1 þ
4
The Schr€oderꢀvan Laar equation29,30 can be used to
estimate the ideal solubilities of a compound in solution.
The analogous PrigogineꢀDefay equation31 allows for the
estimation of the ideal solubilities of a racemic compound.
Both equations require knowledge of the heat of fusion and
the melting point of the solute. Combining both equations
with eq 15 leads to an expression for the eutectic composi-
tion requiring only the aforementioned calorimetric proper-
ties.
1
xeu
¼
0
1
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
"
!#
u
u
t
ΔhfðR, SÞ
1
1
B
B
C
C
2
0:25 exp
ꢀ
3
3
TðfR, SÞ
B
B
B
B
B
@
C
C
C
C
C
A
R
T
2
3
!
ΔhfðRÞ
1
1
4
5
exp
ꢀ
TðfRÞ
R
T
1 þ
4
ð17Þ
Using eq 17, a shift of the eutectic composition of bicalutamide
from 98% at 273 K to close to 95% at 333 K is calculated, which
correlates well with experimental findings. This shift of the
eutectic composition is a prerequisite for the two-step crystal-
lization process adopted here, which is explained in the follow-
ing. Starting from any asymmetric solution of enantiomers
(Figure 7a, point 1, xi(nRit)ial), preferential enrichment in the liquid
phase can be achieved through solvent removal and sub-
3. COUPLING OF CHIRAL CHROMATOGRAPHY AND
SELECTIVE CRYSTALLIZATION
The thermodynamic separation of a pair of enantiomers by
crystallization requires comprehensive knowledge of the corre-
sponding solidꢀliquid phase equilibrium. The eutectic composi-
tion is of particular interest, i.e. the composition at the intersec-
tion of the solubility isotherm of the single enantiomer and the
solubility isotherm of the racemic compound. Kaspereit et al.,19,20
Gedicke et al.21 and Amanullah et al.22,23 proposed a hybrid
process coupling SMB chromatography and selective crystal-
lization. The concentration of the target enantiomer delivered by
the pre-enrichment step and used as feed for the crystallization
step must exceed the eutectic composition. The approach de-
scribed in the publications cited (denoted ‘classical’ in the follow-
ing) is compared to a two-step crystallization process, which was
described in a more recent publication,24ꢀ26 together with more
detailed solidꢀliquid phase equilibrium data for bicalutamide in
pure methanol and in aqueous solvent mixtures. The process
concept is briefly described here, allowing for the derivation of
simple mass balances for process quantification.
Solutions of enantiomers of bicalutamide in water or methanol
belong to a class of systems that exhibit a shift of the eutectic
composition, that can be estimated with a simple shortcut method.
Interactions among the heterochiral enantiomers are not pro-
nounced, and solute/solvent interactions prevail. Wang et al.27
and Klussmann et al.28 presented two equivalent shortcut methods
where the ratio αr of the solubilities of the stoichiometric compound
xl(R,S) and the single enantiomer x(l R) (eq 15) is used to express the
eutectic composition xeu in solution (eq 16). It is assumed that
xl(R,S) equals the sum of the solubilities of the enantiomers at the
stoichiometric (here: racemic) composition of the compound.
sequent thermodynamic equilibration at low temperature Tlow
.
The composition will approach point 2 at the phase boundary
between the inner two-phase region and the three-phase region
(Figure 7a, shaded), and the solution will separate into a solid
phase with racemic composition (point 3) and a liquid phase
of eutectic composition (point 4). The eutectic composition
eu,1
(R)
x
is richer in the target enantiomer than the initial
initial
(R)
composition x
at point 1. This enriched liquid phase is
separated from the solid phase and reused at a higher
temperature Thigh. Point 40 (Figure 7b) represents the initial
state for the second crystallization step. There are two options
for shifting point 40 into the outer two-phase region
(Figure 7b, shaded), in which selective crystallization of the
target enantiomer can take place. Either more solvent is
evaporated in order to concentrate the solution up to the phase
boundary with the three-phase region, or alternatively, the solubility
isotherm can be shifted upwards by controlled addition of an
antisolvent. The optimal yield is obtained at the phase boundary
between the outer two-phase and the three-phase region (point 4).
The solution separates into a liquid phase with the lower eutectic
composition xe(uR,)2 (point 5) and a solid phase consisting solely of the
target enantiomer (point 6).
eu
eu,1
eu,2
The change of eutectic composition Δx = x ꢀ x with
(R)
(R)
(R)
temperature can be relatively large (methionine/water: 8%;32,33
less for other systems27,34,35). The theoretical process yield Y(R)
,
defined as the mass of the pure product over the mass of the
excess of the more abundant enantiomer in the initial mixture,
depends both upon the difference in the eutectic composition at
the respective temperatures and the eutectic composition itself
!
xlðR, SÞ
xlðRÞ
αr ¼
ð15Þ
336
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
Figure 7. Ternary phase diagram (not to scale) representing the first (a) and the second (b) crystallization step as discussed in the text. (1) liquid phase
as obtained from the pre-enrichment step, (2) overall composition after solvent removal, (3) solid- and (4) liquid-phase compositions after equilibration,
(40) overall composition after antisolvent addition, (5) liquid-phase composition after equilibration, (6) product composition.
and can be written as:24.
step C. The mother liquor remaining after the second crystallization
step is of eutectic composition.
eu, 1
ðRÞ
eu, 2
ðRÞ
x
ꢀ x
1
Provided that the crystallization equipment is limited in size
and that the extract stream emanating from unit A has to be
processed batchwise, an internal recirculation of streams is
advantageous. More specifically, a fraction of the racemic stream
leaving port B4 can be reused and pooled with the stream leaving
A2. Furthermore, the eutectic stream leaving C2 can be recycled
completely. From the second batch onwards, the same amount of
product leaves at port C4 in each subsequent batch, whereas
the amount of racemic feed required at port A1 decreases. The
overall process yield increases significantly in this manner. The
yield involving the decreased outlet stream at port A2 can be
balanced as:
Y
¼
ð18Þ
ðRÞ
ð2 xeu, 1 ꢀ 1Þ ð1 ꢀ xðeRu,Þ2Þ
3
ðRÞ
An alternative to eq 18 more suitable for a combined process
relates the product mass to the overall amount of the target
enantiomer in the fraction, which is obtained from a pre-
enrichment step and used as feed for the crystallization process:
ð2 xinitial ꢀ 1Þ xðeRu,Þ1 ꢀ xeu, 2
3
ðRÞ
ðRÞ
tot
ðRÞ
Y
¼
ð19Þ
ð2 xeu, 1 ꢀ 1Þ ð1 ꢀ xðeRu,Þ2Þ
3
ðRÞ
Analogously, for the classical coupling scheme the theoretical
yield can be defined above the eutectic composition19,20 as:
A2, reduced
A2, reduced
m
ꢀ m
ðRÞ
ðSÞ
recycle
ðRÞ
initial
ðRÞ
Y
¼ 2 x
ꢀ 1 ¼
initial
ðRÞ
eu
ðRÞ
3
A2, reduced
A2, reduced
ðRÞ
x
ꢀ x
1 ꢀ x
1
m
þ m
tot, classical
ðRÞ
ðSÞ
Y
¼
ð20Þ
eu
ðRÞ
initial
x
ðRÞ
mC4
mA2, reduced
¼
ð23Þ
On this basis, a first estimate of process productivity for the
combined processes can be written as:
With eq 23 the productivity of consecutive batches of a coupled
process can be calculated as:
tot, classical
ðRÞ
tot, classical
ðRÞ
P
¼ Y
Pchrom
ð21Þ
ð22Þ
3
tot, recycle
ðRÞ
tot, recycle
ðRÞ
P
¼ Y
Pchrom
ð24Þ
3
tot
ðRÞ
tot
ðRÞ
3
P
¼ Y
Pchrom
The crystallization steps proposed here are entirely defined by
thermodynamic states; thus, the fastest crystallization method
available can be applied as long as no pronounced mother liquor
or impurity inclusion or other kinetically dominated phenomena
occur. For this and for practical reasons24 an antisolvent crystal-
lization was carried out, which approached thermodynamic
equilibrium within a few minutes. Apart from crystal purity no
further quality criterion was required for the product.
The proposed coupled separation scheme is illustrated in
Figure 8. A racemic feed enters unit A (SMB Chromatography)
where the extract stream becomes enriched by the target
enantiomer and leaves the unit at the port A2. Passing a mixing
node, it is concentrated and then further enriched in the first of two
crystallization steps (B). Here, a racemic solid fraction is crystallized.
The enriched liquid fraction is separated from the solids by filtration
and heated to a higher temperature. Water is added as antisolvent,
and the product is crystallized and harvested by filtration within
4. EXPERIMENTAL SECTION II: PROCESS VALIDATION
4.1. Chemicals and Equipment. The stationary phase used
for the adsorption isotherm determination described above was
also used for process validation. Four columns with dimensions
10 cm ꢁ 2.5 cm (Groupe Novasep SAS, France) were packed
with the stationary phase. Pure methanol (>99%, CVM Chemie-
Vertrieb Magdeburg GmbH & Co. KG, Sch€onebeck, Germany)
was used as the mobile phase. The columns were installed in a
preparative SMB system (CSEP 912 KNAUER SMB production
unit, four preparative pumps: K-1800, two UV-detectors:
K-2501, Knauer, Germany). The UV detectors operating at
300 nm were used to monitor both the extract and the raffinate
port. The SMB unit employed a multifunction valve for con-
necting up to 12 columns. Operating the SMB unit with fewer
columns required a redesign of the tubing, which is explained
briefly in the Supporting Information. A comparison of the
337
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
Figure 8. Setup of the proposed coupled process including recycling streams. A racemic feed mixture is pre-enriched by SMB chromatography (A) and
further purified in two subsequent crystallization steps (B) and (C). Fractional recycling of the solid phase from (B) and the entire mother liquor from
(C) allows for a decrease in the outlet stream A2 and therefore less load onto the chromatographic unit, while keeping the product stream C4 constant.
Figure 9. Simulated (open dots) and experimentally determined pro-
Figure 10. Optimized dimensionless internal flow rate ratios mi and feed
ductivities (squares) for different requirements on the extract purity.
rates for the given extract purities. For the separation task considered, the
Experimental points (open squared symbols) are shifted horizontally to
feed rate correlates to the m2-value, while all other mi-values remain constant.
lower purities compared to the corresponding theoretical points (open
dots). Full squares denote experimental results after adjusted operation
cases, where larger discrepancies between experiment and simula-
tion appeared, zone IV was not capable of entire desorbent
regeneration. Thus, traces of bicalutamide entered back into zone
I and decreased the purity at the extract port. Clearly, either the
adsorption isotherms or the transport coefficients used in the
simulation were not chosen optimally or the experimental setup
was not sufficiently well represented by the model. Nonetheless, the
model was deemed sufficient, since effective pre-enrichment, which
is the key property for process coupling, was achieved as shown by
Figure 9. Higher extract purities were obtained from a simple
extrapolation of the m-values as shown by Figure 10. The plot
shows, that all m-values except m2 are effectively constant for extract
purities up to 99.9%. The productivity of the system is therefore
governed by the m2-values and the feed flow rate only. It is generally
possible to utilize this information to obtain higher extract purities.
The experimentally determined full symbols shown in Figure 9 are
based on this rather pragmatic approach. It should be noted that
these operating points were optimized for productivity, but that the
optimization was performed within a region where the model is less
accurate. Thus, more productive operating conditions may exist in
addition to those explored here.
conditions as described in the text. The specific solvent consumption
(SC) for the corresponding operation points is also given. Lines are
drawn to guide the eye.
column packing qualities using large injections of bicalutamide
revealed very similar elution profiles for all columns. Analo-
gously, the respective porosities of the columns, determined
using an unretained solvent, were almost identical, indicating that
a good basis for SMB separation existed.
Crystallization was carried out in a 20-L jacketed vessel equipped
with a stirrer and a draft tube for agitation, temperature control, and
a vacuum pump connected to a reflux cooler suitable for solvent
removal. The distillate was weighed automatically using a balance.
The analysis of the product fractions was carried out using a
conventional analytical HPLC system consisting of a Dionex
P580 gradient pump, an ASI100 automated sample injector, a
TCC100 column oven, and an UVD340U DAD detector. The
UV-wavelength used for detection was 300 nm. The temperature
of the column was controlled at 298 K, and the sample volume
injected sample was 5 μL.
4.2. Pre-enrichment by Chromatography. An initial experi-
mental validation of the simulation study was performed for three
selected sets of operating parameters. The experimental productivity
over purity curve obtained is shown in Figure 9 together with the
simulated productivity and the corresponding solvent consumption.
Both curves are similar for low extract purities, but deviate increas-
ingly for higher purities. For a given set of parameters, the purity at
the extract port was computed as 85%, while the experimental
validation revealed an extract purity of 76%. It was found, that in
The specific solvent consumption per mass of bicalutamide in
the extract fraction increases significantly above 98% purity, as
can be seen in Figure 9. This is unfavorable, as a purity of 99.5% is
required for this compound.
Figure 11 illustrates the productivity as a function of extract
purity for a coupled process consisting of a pre-enrichment step
based upon the productivity curve determined experimen-
tally and the two-step crystallization process described above.
338
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
Table 4. Process conditions for the chromatographic
separation runs
internal dimensionless flow rates (m1, m2, m3, m4)6.713, 0.665, 1.426, 0.203
External Flow Rates
eluent
feed
23.7 mL/min
200.0 mL/min
36.5 mL/min
179.7 mL/min
raffinate
extract
Volumes
eluent
feed
256.0 L
30.3 L
47.5 L
233.9 L
Figure 11. Optimized productivities of the stand-alone chromato-
graphic process (upper thick solid line) and for coupled processes.
The classical case (see text) is denoted as a dotted line (lower right
corner), while the processes as conducted in this work are illustrated
by the thin solid line. The gain in productivity for this process
considering internal recycling of streams is reflected by the thick
dashed line.
raffinate
extract
purity raffinate
99.54 %
274.0 g
92.38 %
322.0 g
7.4 g
bicalutamide in raffinate
purity extract
bicalutamide in extract
mass lost on column, etc.
overall separation time
switching time
The dashed line corresponds to 100% product purity according
to eq 22, the solid line corresponds to the experimentally
determined line in Figure 9 (lower line). Clearly, a significant
increase in productivity can be achieved with the coupled
process, provided the pre-enrichment step is optimized for
purities below 90%. Interestingly, there is no global optimum
but rather a plateau for lower extract purities. Thus, the optimal
solution composition leaving the first and entering the second
process unit is not limited by a specific extract purity for this
particular separation problem. Taking into account the lower
solvent consumption at reduced purity requirements, the curve
suggests that operating at 70% purity may be the most promising
economically. However, solvent consumption was not an issue here,
and the purity was lowered only to the start of the productivity
plateau at 90% extract purity. The chromatographic separation of
racemic bicalutamide was split into two runs. The process para-
meters and results of the chromatographic separation are given in
Table 4.
4.3. Selective Crystallization Employing a Two-Step Pro-
cess. The chromatographic extract fraction with a volume of 237 L
and rich in the (R)-enantiomer was fed to the crystallizer in portions
and was concentrated to 8.2 L by vacuum distillation at approxi-
mately 200 mbar. The vessel was then cooled to 275 K with the aim
of establishing a high eutectic composition of ∼98% (R)-enantio-
mer in the liquid phase. However, in contrast to the laboratory-scale
experiments, no nucleation occurred within an hour of reaching the
final temperature. Thus, a negligible amount of racemic bicaluta-
mide was therefore added as seed crystals. Both a peristaltic pump
and a vacuum pump were utilized to withdraw the enriched liquid
phase, which was filtered using common HPLC frits. The remaining
solid phase was removed, analyzed, and stored. The liquid phase was
fed back to the cleaned vessel and heated up to 333 K. A defined
amount of water (8.0 kg) was then added. Water acts as an
antisolvent, reducing the solubility and forcing the solute to crystal-
lize. After equilibration the solid phase was separated from the
mother liquor by vacuum filtration. The crystals were then washed
with the mother liquor followed by chilled methanol. The solids
were then vacuum-dried. The mother liquor was fed back to the
reactor and cooled to 283 K in order to crystallize as much of
the remaining bicalutamide as possible. Analysis of the chilled
mother liquor showed that the bicalutamide content was below
the detection limit.
21:27 hh:mm
29 s
5. RESULTS OF THE INDIVIDUAL PROCESS STEPS AND
PROCESS COUPLING
The pre-enrichment by SMB chromatography was performed
over a period of 22 h and resulted in a purity of 92.38% (R)-
enantiomer at the extract port, which was slightly above the
targeted 90% purity. The raffinate was virtually free of the (R)-
enantiomer.
The results of the individual crystallization steps are summar-
ized in Figure 12. The values given in brackets denote the theoretical
values. Although a broad agreement of the experimental values with
the expected values is observed, the deviation of the solid phase (SP)
yield and purity from the first crystallization step is significant but
can be accounted for. The mother liquor level was well below the
upper end of the draft tube after solvent evaporation, which led to
poor mixing and pronounced splashing of the liquor onto the
reactor internals and the reactor wall. Evaporation of the solvent
from these drops resulted in a solid crust with the same enantiomer
ratio as the mother liquor. The purity of the solids crystallized
from the bulk solution was found to be 56% (R)-enantiomer.
Since both solid fractions were combined, the resulting total
composition was 70% (R)-enantiomer, reducing the amount
of the target enantiomer in the liquid phase and consequently
reducing the yield after the second crystallization step. The
optical purity of the product was close to 100%. It is advanta-
geous that the mother liquor in the second crystallization step
has a high purity, as this reduces the probability of the crop
purity being reduced by adhering mother liquor. Although not
investigated here, it is conceivable to force the system into the
three-phase region in order to continue preferential growth of
the target enantiomer by further decreasing the temperature.
The process conditions are rather favorable for running the
process in this manner as a large number of crystals is present
and the slope of the solubility curve (see24) provides further
scope for cooling crystallization, providing nucleation of
the distomer can be avoided, increasing the yield of the (R)-
enantiomer.
339
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
crystallization and downstream processing becomes negligible,
and the chromatographic step is clearly the time-limiting opera-
tion in the entire process scheme. However, the behavior of the
system investigated here should not be generalized, and other
products might require a more detailed investigation.
Figure 11 includes a productivity vs extract purity curve for the
classical approach to coupling a pre-enrichment step with selective
crystallization (lower right corner, dotted line). Increased produc-
tivity can be expected if the pre-enrichment results in a composition
that exceeds the eutectic composition, which is >95% in the case of
bicalutamide. The range of process conditions where this is achieved
is rather narrow for the classical coupling. By using eqs 21 and 22 it
can be shown that the productivity of the classical coupling generally
outperforms the two-step process for systems with low eutectic
compositions (Figure 11, thin solid line). For compounds exhibiting
higher eutectic compositions the two-step process is both more
efficient and more robust due to the wider range of process
conditions available. Furthermore, the two-step process has the
potential for a significant increase in the process yield, and exhibits
superior productivities compared to the classical coupling for every
eutectic composition. When a product is manufactured, generally
more than one batch will be made. As discussed above, both the
solid phase isolated in the first crystallization step and the mother
liquor remaining after the second step can be recycled in subsequent
batches, resulting in a considerable improvement in overall yield as
shown in Figure 13. Here a series of realistic and imperfect batches
identical to the one conducted (losses in the first crystallization step)
were considered. Assuming there are no process-dependent losses
of material, the second batch already allows 100% of the excess of
the more abundant enantiomer in the extract fraction to be
harvested (not shown). Higher yields in the crystallization process
also lead to a higher productivity of the coupled process, shown as
the thick dashed line in Figure 11.
Figure 12. Flowchart illustrating mass balances of the individual purification
steps.
6. SUMMARY AND CONCLUSIONS
Figure 13. Potential of yield improvements by a series of crystallization
batches identical to the one conducted. If experimental stirring issues are
avoided and product losses are assumed negligible, complete product
recovery would be possible from the second batch on.
The isolation of a single enantiomer from a pre-enriched
mixture of bicalutamide enantiomers was successfully scaled up
from the laboratory scale24 using only a few grams of material to a
scale of 600 g of racemate. The throughput of the pre-enrichment
step employing SMB chromatography was increased at the expense
of product purity in this step. The fraction leaving the extract port
was subject to two subsequent crystallization steps. Both steps
increased the chiral purity, and the final product was 99.99% pure.
Experimental issues prevented the theoretical yield of 56% (R)-
bicalutamide (with respect to the initial amount of (R)-bicalutamide
in the initial mixture. Thus, the theoretical yield after the first step
was reduced to 42%. The determined yield was actually found to be
slightly higher (45%), which might be attributed to preferential
crystallization during filtration of the crop. When compared to a
classical separation scheme (where the enantiomers are separated by
chromatography and crystallization) that merely serves to isolate the
single enantiomer, it was shown that this type of coupled process is
capable of significantly improving the productivity when isolating a
single enantiomer from a racemic solution. The productivity and
yield can be further enhanced by recycling the waste streams from
the first and second crystallization steps in subsequent batches. The
approach adopted here is superior to alternative processes for the
compound studied and is likely to be beneficial for any system with a
eutectic rich in the single enantiomer. The theoretical framework
presented can serve as a generic guideline for other chiral separation
problems with similar characteristics.
The coupling of the two processes presented here rests on a
number of assumptions. We are fully aware that a comparison of
the coupling of two processes on the basis of a dynamic model for
one step and just equilibrium considerations for any following
step is always of limited accuracy. Crystallization processes are
frequently kinetically dominated, and crystal purities can be
correlated with the gradient of the driving force applied. How-
ever, this does not apply to the compound considered in this
study, and some illustrative numbers may justify this assumption.
The accelerated processing of 600 g of racemic bicalutamide
developed here requires almost 22 h of chromatographic separa-
tion. Generating a product of 99.99% purity using the same
equipment requires a considerably longer time due to signifi-
cantly reduced productivity. Thermodynamic equilibrium was
achieved in the first crystallization step within a short period of
time after the addition of seeds. The second, antisolvent crystal-
lization was also completed rapidly. Additional time is required
for setting up the crystallizer and for product removal and
filtration. However, the majority of process steps are either
common with the downstream processing of product frac-
tions from a chromatographic unit or can be run parallel to
further chromatographic separations. The time required for
340
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
’ ASSOCIATED CONTENT
QDsrb
Q
qi
desorbent flow rate mL/min
flow rate mL/min
solid-phase concentration of component i g/L
saturation capacity of component i at the S1 sites at
saturation g/L
S
Supporting Information. This material is available free
b
of charge via the Internet at http://pubs.acs.org.
qS1,i
’ AUTHOR INFORMATION
qS2,i
adsorption of component i at the S2 sites at saturation
g/L
vectors sizes
Corresponding Author
henning.kaemmerer@evonik.com
SP,SD
tR,i
tf
retention time of component i min
retention time of the peak front min
retention time of the peak tail min
retention time of component I min
retention time segregating peak front and peak tail min
port switching interval in SMB process min
superficial velocity cm/min
Present Addresses
zEvonik Industries AG, 63457 Hanau, Germany
#Friedrich-Alexander University of Erlangen-Nuremberg, Chair
for Separation Science & Technology, 91058 Erlangen, Germany
tr
tR,i
tf/r
ts
u0
V
column volume mm3
volume of tubings, connectors, etc. mL
wP,wD weighting factors
’ ACKNOWLEDGMENT
Vtube
This work was part of the European Union FP7 Project
‘INTENANT’, funded by Grant Number FP7-NMP2-SL2008-
214129.11 The support is gratefully acknowledged. In addition,
we thank Dr. Jan von Langermann, Jacqueline Kaufmann, Martin
Uxa, and Antje Hofmann for the comprehensive support in the
experimental work.
wf
wr
w1/2,i
xl(R,S)
xl(R)
xeu
empirical factor to describe the peak front
empirical factor to describe the peak tail
width of a peak at half its height of component i min
fraction of the racemic compound
fraction of the (R)-enantiomer
eutectic composition
’ NOMENCLATURE
A
αr
YldB,Extr yield of component i at the extract port, respectively
YldTarget yield of the target component at the extract port
column cross-sectional area mm2
ratio of the solubilities of the stoichiometric compound
and the single enantiomer
z
axial locus in column mm
b1,i
bi-Langmuir model adsorption parameter, site 1, com-
ponent i L/g
bi-Langmuir model adsorption parameter, site 2, com-
ponent i L/g
average concentration of component A at the inlet of
zone 1 g/L
’ REFERENCES
(1) Wirth, M. J. Urol. 2004, 172.
b2,i
(2) Thijs, L. U.S. Pat. Appl. 2003/0073742, 2003.
(3) Tucker, H.; Chesterson, G. J. J. Med. Chem. 1988, 31, 885.
(4) Zazhi, Z. Y. G. Chin. J. Pharm. 2006, 37, 73.
(5) S€or€os, B. WO 01/00608, 2006.
cA,1in
cB,4out average concentration of component B at the outlet of
(6) Ekwuribe, N. N. WO 01/28990, 1999.
zone 4 g/L
(7) James, K. D.; Ekwuribe, N. N. Tetrahedron 2002, 58, 5905.
(8) MDL ISIS, 2.5; Available Chemicals Directory (ACD) commer-
cial catalogue database.
(9) Fujino, A. Tetrahedron Lett. 2007, 48, 979.
(10) Sugai, T.; Ikunaka, M. JP 2008007492, 2008.
(11) See: www.intenant.eu.
(12) Guiochon, G.; Felinger, A.; Shirazi, D. G.; Katti, A. M. Funda-
mentals of Preparative and Nonlinear Chromatography, 2nd ed.; Academic
Press: Boston, 2006.
(13) Chung, S. F.; Wen, C. Y. AICHhE J. 1968, 14, 857–866.
(14) Einstein, A. Ann. Phys. 1905, 17, 549–560.
(15) Antos, D. Doctoral Thesis, Oficyna Wydawnicza Politechniki
Rzszowskiej, Rzesz^ow, Poland, 2003.
c
liquid-phase concentration g/L
concentration maxima of injection profile g/L
direction vector
cmax,f
D
B
Dc
Dm
DL
εT
column diameter mm
molecular diffusion coefficient m2/s
axial dispersion coefficient m2/s
total void fraction
Gj
objective function
HETP height of a theoretical plate mm
enantiomers: 1, (S)-bicalutamide; 2, (R)-bicalutamide
i
BI P,BI D unit vectors of proportional and differential
directions
component
(16) Storti, G.; Mazzotti, M.; Morbidelli, M.; Carra, S. AIChE J.
1993, 39, 471–492.
(17) Lee, J. W.;Wankat, P. C. J. Chromatogr., A 2010, 1217, 3418–3426.
(18) Tan, K. K.; Wang, G.-G.; Hang, C. C.; Hagglund, T. Advances in
PID Control. Springer Verlag: London, 1999.
K
km,i
k00
kΔp
Lc
mass transfer coefficient of component i 1/min
retention factor
pressure drop coefficient bar s/m
column length mm
dimensionless zone flow rates in SMB process
(19) Kaspereit, M., Doctoral Thesis, Otto-von-Guericke-University,
Magdeburg, Germany, 2006.
mj
(20) Kaspereit, M.; Gedicke, K.; Zahn, V.; Mahoney, A. W.; Seidel-
Morgenstern, A. J. Chromatogr., A 2005, 1092, 43–54.
(21) Gedicke, K.; Kaspereit, M.; Beckmann, W.; Budde, U.; Lorenz,
H.; Seidel-Morgenstern, A. Chem. Eng. Res. Des. 2007, 85, 928–936.
(22) Amanullah, M.; Abel, S.; Mazzotti, M. Adsorption 2005, 11,
893–897.
(23) Amanullah, M.; Mazzotti, M. J. Chromatogr. A 2006, 1107, 36–45.
(24) Kaemmerer, H.; Jones, M. J.; Lorenz, H.; Seidel-Morgenstern, A.
Fluid Phase Equilib. 2010, 296, 192–205.
MMin,Rc minimum mass ratio of each component to be recycled
between zone 1 to zone 4
mass of component i to be recycled between zone 1
to zone 4
number of transport plates
Mi,Rc
NTP
Puri,Extr purity of component i at the extract port
PurTarget purity of target component at the extract port
p
pressure bar
341
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342

Organic Process Research & Development
ARTICLE
(25) Lorenz, H.; Kaemmerer, H.; Polenske, D.; Seidel-Morgenstern, A.
EP08163733.2, 2008.
(26) Lorenz, H.; Kaemmerer, H.; Polenske, D.; Seidel-Morgenstern, A.
WO 2010/025968 A2, 2008.
(27) Wang, Y.; LoBrutto, R.; Wenslow, R. W.; Santos, I. Org. Process
Res. Dev. 2005, 9, 670–676.
(28) Klussmann, M.; White, A. J. R.; Armstrong, A.; Blackmond,
D. G. Angew. Chem., Int. Ed. 2006, 45, 7985–7989.
(29) Mullin, J. W. Crystallization; Butterworth-Heinemann: Oxford,
1997.
(30) Tulashie, S. K.;Kaemmerer, H.;Lorenz, H.;Seidel-Morgenstern, A.
J. Chem. Eng. Data 2010, 55, 333–340.
(31) Prigogine, I.; Defay, R. Chemical Thermodynamics; Longmann:
London, UK, 1973.
(32) Polenske, D.; Lorenz, H. J. Chem. Eng. Data 2009, 54, 2277–2280.
(33) Kaemmerer, H.; Lorenz, H.; Seidel-Morgenstern, A. Chem. Ing.
Tech. 2009, 81, 1955–1965.
(34) Zhang, Y.; Ray, A.; Rohani, S. Chem. Eng. Sci. 2009, 64, 192–197.
(35) Worlitschek, J.; Bosco, M.; Huber, M.; Gramlich, V.; Mazzotti,
M. Helv. Chim. Acta 2004, 87, 279–291.
342
dx.doi.org/10.1021/op200136z |Org. Process Res. Dev. 2012, 16, 331–342