Resolution of a diasteromeric salt of citalopram by multistage crystallization

Article abstract of DOI:10.1021/cg300166g

Multistage crystallization has been used for resolution of the racemic mixture of citalopram to obtain pure S-citalopram. The racemate was converted to a diasteromeric salt pair using (+)-O,O′-di-p-toluoyl-d-tartaric acid, (+)DTT, as a resolving agent. The obtained salts involved solid solutions almost within the entire range of the crystalline phase composition, which excluded the possibility of a single-stage resolution. Therefore, a multistep crystallization technique was employed that allowed enrichment of the racemate with the desired diastereomer in the liquid phase. Moreover, a procedure for restoring the 1:1 composition of the depleted solid phase has been developed. The procedure was based on formation of a diasteromeric salt pair with the opposite enantiomer of the resolving agent, i.e., (-)DTT.

Full text of DOI:10.1021/cg300166g

Article
pubs.acs.org/crystal
Resolution of a Diasteromeric Salt of Citalopram by Multistage
Crystallization
Maciej Balawejder,† Dawid Kiwala,§ Heike Lorenz,‡ Andreas Seidel-Morgenstern,‡ Wojciech Piatkowski,§
and Dorota Antos*,§
†Chair of Chemistry and Food Toxicology, University of Rzeszow, Rzeszow, Poland
‡Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany
§Department of Chemical and Process Engineering, Rzeszow University of Technology, Rzeszow, Poland
ABSTRACT: Multistage crystallization has been used for resolution of the racemic mixture of citalopram to
obtain pure S-citalopram. The racemate was converted to a diasteromeric salt pair using (+)-O,O′-di-p-toluoyl-D-
tartaric acid, (+)DTT, as a resolving agent. The obtained salts involved solid solutions almost within the entire
range of the crystalline phase composition, which excluded the possibility of a single-stage resolution. Therefore,
a multistep crystallization technique was employed that allowed enrichment of the racemate with the desired
diastereomer in the liquid phase. Moreover, a procedure for restoring the 1:1 composition of the depleted solid
phase has been developed. The procedure was based on formation of a diasteromeric salt pair with the opposite
enantiomer of the resolving agent, i.e., (−)DTT.
ideal situation large difference in the solubility of diastereomers
is achieved.15 However, formation of double salts, for which
resolution of the racemate in unfeasible, or solid solutions,
where multistage crystallization is necessary to obtain
substantially diasteromerically pure salts, can also be
observed.15,16
The solid solution type of the crystalline phase behavior has
been exhibited by a system of diastereomeric salts of citalopram
with the chiral resolving agent (−)-O,O′-di-p-toluoyl-L-tartaric
acid, (−)DTT, as well as with its opposite enantiomer,
(+)DTT.17,18
Citalopram is an antidepressant drug belonging to the class
of selective serotonin reuptake inhibitors. It is used not only to
treat depression associated with mood disorders but also in the
treatment of body dysmorphic disorder and anxiety.19
Citalopram is the common name for (R,S)-1-[3-
(dimethylamino)propyl]-1-(4-fluorophenyl)-1,3-dihydroiso-
benzofuran-5-carbonitrile as a racemic bicyclic phthalane
derivative with chemical formula C20H21FN2O.20 It is supplied
in the form of the hydrobromide salt . The compound has a
single stereocenter, which binds a 4-fluori(o)phenyl group and
an N,N-dimethyl-3-aminopropyl group. Due to this chirality,
the molecule exists in two enantiomeric forms. The therapeutic
activity of citalopram resides in the S-enantiomer, commercially
available with the tradename escitalopram (supplied in the form
of the oxalate salt), due to its higher affinity to the human
INTRODUCTION
■
In the past decades considerable emphasis has been placed on
the use of single enantiomers versus racemic mixtures in many
areas of chemistry and biochemistry.1−4 This was triggered by
increasing awareness of the difference in the biological activity
exhibited by individual enantiomers of racemic drugs.5,6 Two
main approaches are available to produce a single enantiomer
involving either resolution of racemates or stereoselective
synthesis from achiral material.7 However, manufacturing
racemates is usually more economical than stereoselective
synthesis of enantiomers,8 particularly if efficient separation
techniques can be employed allowing high yield and optical
purity as well as recycling of unwanted enantiomer.6,9 There are
a number of separation methods that can be used for optical
resolution of racemates, which exploit either the enantiose-
lective behavior of biological systems, e.g., kinetic resolutions
using enzymes, or the differences in the molecular recognition
of enantiomers by auxiliary chiral agents,6 e.g., chiral
chromatography, chiral membranes, and enantioselective
crystallization.7
In the so-called classical resolution method the enantiomers
are converted to a diasteromeric salt pair by reaction with a
single-enantiomer resolving agent, i.e., forming a pair of
diastereomers that possess the same chemical formula but
have different physical properties.10 The diastereomers are then
separated using crystallization, taking advantage of the differing
solubility of the two components.11
This method is often considered the most straightforward,
economical, and easiest to perform on a large scale.12−14
However, discovery of an efficient resolution still often relies
upon screening of a large number of resolving agents. In an
Received: February 4, 2012
Revised: March 26, 2012
Published: April 4, 2012
© 2012 American Chemical Society
2557
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
Figure 1. Illustration of the ternary phase diagrams for diasteromeric crystallization: (a) ideal solution; (1 and 2) boundary composition of the initial
racemic mixture for which pure diasteromeric salt can be crystallized (e.g., S,R′ salt, where R′ is the R enantiomer of the resolution agent), (dashed
lines) corresponding boundary conodes; (b) double salt formation; (c and d) formation of a partial or complete solid solution, respectively (1) initial
racemate, (2 and 3) equilibrium compositions in the liquid and solid phase, (dashed lines) demonstration of conodes.
serotonin transporter, while the R-enantiomer is much less
potent.21
recovery of the 1:1 composition in the solid phase by
diasteromeric crystallization using (−)DTT.
As mentioned above, the diasteromeric salts of citalopram
and (−)DTT as well as (+)DTT involved solid solutions in the
crystalline phase. For this reason the resolution of these salts by
single-stage diasteromeric crystallization was found to be
unfeasible;17,18 it resulted in certain enrichment of 1:1 mixtures,
which, however, would be required to be further increased to
obtain the pure salt of the desired S-diastereomer.
2. CHARACTERIZATION OF THE PHASE BEHAVIOR
The design of diasteromeric crystallization has to be proceeded
by identification of the solid-phase behavior of the salt pair
system. The most accurate method is based on determination
of the ternary phase diagrams representing SLE data for
mixtures of the two salts and a solvent. Demonstrations of the
SLE diagrams of diasteromeric salts are presented in Figure 1.
In the case of simple eutectic resolution (Figure 1a), the pure
salt of target enantiomer is obtained in the solid phase.
Deviations from the simple resolution are exemplified in Figure
1b−d; they involve partial or complete solid solution in the
crystalline phase (Figure 1b,c) or double salt formation, where
resolution of racemate is not possible (Figure 1d). In the
former case certain enrichment of the initial racemic mixture
with the desired compound can be achieved in either the solid
or the liquid phase depending on the nature of the solid
solution, known as the Roozeboom type.23 To achieve pure
diastereomer multistage crystallization has to be used.22,24
Apart from determination of the ternary phase diagram,
which is time consuming, other auxiliary methods can be used
for evaluation of the phase behavior such as X-ray powder
diffraction (XRPD). Here, alternation of the powder patterns
with a change of the solid-phase composition can be analyzed.
Another technique allowing fast determination of the solid-
phase behavior of the diasteromeric system is differential
In this study to separate mixtures of diasteromeric salts R,S-
citalopram·(+)DTT multistage crystallization has been used.
The principles of the procedure for multistage crystallization of
solid solution forming systems have been developed in our
recent study,21 where resolution of enantiomeric mixtures of
the oxalate salt of racemic citalopram was performed. In this
case, however, the separation process could not be initialized
with racemic mixtures; proper enrichment of racemate with the
desired S-enantiomer had to be preliminary ensured to achieve
an acceptable yield and enantiomeric purity of the product.
In the case of diasteromeric salt R,S-citalopram·(+)DTT the
initial racemic mixtures can be directly processed, where
enrichment of the liquid phase with the desired diastereomer is
accompanied with its depletion in the solid solution. Because a
1:1 composition of the salt mixtures is required to initialize the
resolution, the solid phase with an excess of unwanted R-
citalopram·(+)DTT (R-diastereomer) cannot be directly
recycled to the process. Therefore, in this study the procedure
for the multistage crystallization process was combined with
2558
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
scanning calorimetry (DSC). However, applicability of the
method is not general; decomposition of diasteromeric salts
during heating may be a limiting factor.15
Sol3, see Figure 2), while the remaining part is delivered into
the inlet of the next stage for further enrichment or is recovered
as a product, P, in the last stage.
To increase the yield of operation the exhausted extract
streams of each crystallization stage can be recycled into the
inlets of previous stages (n − 1) (see Sol′2, Sol′3, in Figure 2),
excluding the first one, from which it is withdrawn as the
fraction of waste. A graphic illustration of the countercurrent
three-stage process using the ternary solubility diagram is
presented in Figure 3.
3. COUNTERCURRENT MULTISTAGE
CRYSTALLIZATION FOR SYSTEMS EXHIBITING
SOLID SOLUTIONS
As mentioned above, resolution of solid solution forming
systems requires multistage crystallization, in which enrichment
of the salt mixtures with the desired compound can be
upgraded in either the solid or the liquid phase.
Details regarding the crystallization process involving a solid
solution in the enantiomeric system are given in our previous
study.22 The concept of a countercurrent multistage crystal-
lization process is quoted below for the type of systems for
which enrichment is upgraded in the liquid phase. The same
behavior was exhibited by R,S-citalopram·(+)DTT.
The idea of three-stage countercurrent crystallization is
illustrated in Figure 2.
Figure 3. Graphic illustration of the three-stage crystallization process
on the ternary solubility diagram: (solid lines) conodes, (dashed lines)
mixing lines, (dotted lines) evaporation lines, (M) mixing points for
each stage Rn =Fn+1 + Recn +Soln, Fn+1 = Pn.
To realize the first run the recycled solid streams (Rec) with
proper enrichment has to be delivered. However, this
investment is on one-off basis; after accomplishing the first
run of the multistage operation the crystallization unit can
operate self-sufficiently without external delivery, excluding the
feed transported to the first stage.
Such a unit can operate in the continuous as well as in the
cyclic mode. In the former case the stages are directly
connected, while in the latter one the units consists of several
single self-sufficient stages between which the raffinate streams
are transported, whereas the extract streams (En, n > 1) can be
stored and used as the feed stream for the (n − 1)th stage at
any time. Flexibility of the cyclic mode facilitates experimental
realization of the process. The batch process can be realized
isocratically, i.e., at the same temperature and solvent
composition in each stage.
Figure 2. Flowsheet of the three-stage countercurrent crystallization
process.
Design of the process is based on solving equations of the
mass balance for each single stage (see Figures 2 and 3)
(F + E′n+1) + Recn = Mn
(1)
(2)
n
Each stage has two inlets, i.e., feed (F) and recycled stream
(Rec), and two outlets, raffinate (R) and extract (E). The
compositions of raffinate and extract streams are correlated by
the SLE relationship.
The exhausted solid solution, extract, has reduced enrich-
ment compared to the composition of the feed stream, while
the enriched liquid solution, raffinate, has upgraded enrich-
ment. A fraction of the enriched raffinate stream is recycled into
the same stage after evaporation of the solvent (e.g., Sol1, Sol2,
Rn + En = Mn
(F + E′n+1)·y + Recn·xi,Rec = Mn·y
n
i,F
n
i,Mn
(3)
(4)
n
Rn·y + En·xi,E = Mn·y
i,Mn
i,Rn
n
Note that the composition of each Fn and recycled stream E′n+1
is the same.
2559
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
Figure 4. Synthesis route of R,S-citalopram(+)·DTT: (a) reaction scheme of neutralization; (b) reaction scheme for formation of a diasteromeric salt
pair.
The set of eqs 1−4 is coupled with the mass balance of the
solid stream recycled into the same stage after evaporation
expressed as
dictory changes of the performance indicators, i.e., product
purity and yield. An increase of enrichment of the exhausted
stream results in an upgrade of the product purity, which is
counterbalanced by a reduction of the yield, and vice versa. The
choice of xS,E1 and the number of stages should be preceded by
economic evaluations.
A similar procedure can be designed for solid solution
systems, for which enrichment is upgraded in the solid phase.
In this case, F, P, and E are the solid solutions while Rec and R
are liquid.22
Rn = Fn+1 + Recn + Soln
(5)
Rn·y = Fn+1·y + Recn·x
i,Recn
i,Rn
i,Rn
(6)
and mass balance of the liquid stream recycled into the n − 1
stage after dissolving
E′n+1 = En+1 + Sol′n+1
(7)
4. EXPERIMENTAL SECTION
E′n+1·y = En+1·xi,E
i,F
n+1
(8)
n
4.1. Chemicals and Equipment. The following chemicals were
used in this study: racemic citalopram hydrobromide, (R,S)-1-[3-
(dimethylamino)propyl]-1-(4- fluorophenyl)-1,3-dihydroisobenzofur-
an-5 carbonitrile hydrobromide with purity Pu > 99% (Jubilant
Organosys Ltd.), escitalopram oxalate, (S)-1-[3-(dimethylamino)-
propyl]-1-(4-fluorophenyl)-1,3-dihydroisobenzofuran-5-carbonitrile
oxalate, Pu > 98% (Yick-Vic Chemical & Pharmaceuticals Ltd.),
(+)-O,O′-di-p-toluoyl-D-tartaric acid, (+)DTT, Pu ≥ 98% (Fluka),
(−)-O,O′-di-p-toluoyl-L-tartaric acid, (−)DTT, Pu = 97% (Sigma-
Aldrich), acetonitrile for HPLC (VWR Prolabo), and methanol for
HPLC (Fisher Chemical).
The laboratory equipment used in this work included the following:
HPLC chromatograph (Agilent 1200 Series), thermostat (Lauda
Ecoline RE-104), balance (Mettler Toledo), diffractometer PAN-
alytical X’Pert Pro (PANalytical GmbH, Germany).
4.2. Synthesis Routes. 4.2.1. Synthesis of R,S-Citalopram·(+)-
DTT. The resolving agent was selected based on publications
concerning resolution of racemic citalopram by diasteromeric
salt crystallization,17,18 i.e., (+)DTT was used.
Conversion of racemic citalopram hydrobromide into R,S-
citalopram·(+)DTT salt was realized in the following two steps:
conversion of racemic citalopram hydrobromide into racemic
citalopram free base, and subsequent conversion of racemic citalopram
free base into R,S-citalopram (+)DTT salt (see scheme in Figure 4).
Synthesis routes were similar to those proposed in recently published
articles17,18 with a modification lying in the replacement of the solvent
used for generation of the free base, i.e., toluene was replaced with
dichloromethane.
where Fn is the feed (liquid) delivered to the n stage, Recn is the
recycled stream (solid), Rn is the raffinate (liquid), En is the
extract (solid), E′n is the extract recycled into the previous stage
after dissolving (liquid), Soln is the solvent, Mn is the mixing
point, yi is the mass fraction in the liquid solution, xi is the mass
fraction in the solid solution, and i denotes the component of
the mixture, e.g., R- or S-diastereomer. After each stage, n, the
process can be interrupted. Then, Fn+1 denotes the product
stream, Pn,
Additionally, the SLE relationship is described by functional
dependencies in the following form
y = f(xS)
(9)
i
If the mass of the feed stream and its composition yR,F, yS,F are
known, the system running in batch mode has one degree of
freedom, i.e., only one process variable is available to alter the
process performance.22 Note that the composition of the
recycled extract streams En, n > 1 is restricted, enrichment of
these streams is set the same as the feed in the previous stage. It
is convenient to select the composition (i.e., enrichment) of the
exhausted extract stream withdrawn from the first stage, xR,E1, as
the process variable. When the value of xR,E1 is chosen then yi,R
is determined by the SLE relationship and the remaining
streams and their compositions, E, R, Rec, M, P, Sol, xi,Rec, yi,M,
are calculated by solving eqs 1−8. Manipulation of the
The procedures were realized as described below.
Conversion of Racemic Citalopram Hydrobromide into Racemic
Citalopram Free Base. A 20 g amount of citalopram hydrobromide
composition of the exhausted stream, xS,E , results in contra-
1
2560
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
Figure 5. Microscopic photos of R,S-citalopram·(+)DTT: (a) 48.3% of S-diastereomer and (b) 99.57% of S-diastereomer.
was suspended in 100 mL of water and 100 mL of dichloromethane. A
1.3 M solution of sodium hydroxide was slowly added until pH = 9.8
was achieved. This solution was intensively stirred for 45 min, and
then the phases were separated. The aqueous phase was extracted
using 15 mL of dichloromethane, while the organic phase was
extracted using 15 mL of Milli-Q water. The organic phases were
combined and dried for 2 h with anhydrous magnesium sulfate. After
filtration of the drying agent, the filtrate was concentrated using an
evaporator to obtain a consistency of honey. The residue of the solvent
was removed by a vacuum drier (t = 45 °C, p = 0.4 bar, time = 24 h).
A reaction yield of 90% was obtained.
Conversion of Racemic Citalopram Free Base into R,S-
Citalopram·(+)DTT Salt in Ethanol. A 10 g amount of racemic
citalopram free base was dissolved in 90 mL of ethanol by heating at
30 °C in the water bath. To the mass obtained 12.4 g of (+)DTT acid
was slowly added and heated until complete dissolution was obtained.
The diluted mixture was left for 6 h to crystallize at room temperature.
Afterward, the solid product was separated by vacuum filtration and
dried using a vacuum drier (t = 45 °C, p = 0.4 bar, time = 48 h). A
reaction yield of ca. 90% was achieved.
quently, conversion of R,S-citalopram free base into R,S-citalopram·(-
)DTT salt.
Conversion of R,S-Citalopram·(+)DTT into R,S-Citalopram Free
Base. A 10 g amount of R,S-citalopram·(+)DTT containing 48.2% of
S-diastereomer was suspended in 25 mL of Milli-Q water and 16.6 mL
of dichloromethane. A detailed description of this procedure has been
already given in section 4.2.1. The reaction yield was 83%.
Conversion of R,S-Citalopram Free Base into R,S-Citalopram·(−)-
DTT Salt. A 0.5 g amount of R,S-citalopram free base and 0.62 g of
(−)DTT acid were diluted in a defined volume of acetonitrile/
methanol (1:0.08 v/v).
The obtained mass was heated at reflux for 40 min. Then, the
solution was poured into the closed tube. The tube was inserted in a
double-walled thermostatted vessel, and the solution was electro-
magnetically stirred for about 72 h at 25 °C in order to establish the
equilibrium. After that time, a two-phase system was obtained and the
solid product was separated by vacuum filtration and dried using a
vacuum drier (t = 45 °C, p = 0.4 bar, time = 24 h). Both the solid and
the liquid phases were analyzed using HPLC (see below section 4.3).
The reaction yield was ca. 90%.
Because of the concrete value for the solubility of citalopram salts in
acetonitrile−methanol mixtures certain amounts of both diastereomers
could be exchanged between the phases, and after establishing the
crystallization equilibrium the liquid phase was enriched with R-
diastereomer while the 1:1 composition of the solid phase was
restored.
Due to the very low solubility of salts of R,S-citalopram·(+)DTT in
ethanol almost the entire amount of the salt mixture was transferred
into the crystalline phase, whereas soluble postsynthesis impurities
remained in the liquid phase.
This solid product was used to prepare the feed and solid solutions
for the crystallization process (see section 4.6).
4.3. HPLC Analysis. All chromatographic analyses were carried out
using a Chirobiotic V column supplied from Astec Co. (particle size 5
μm, column dimensions 250 × 4.6 mm). The chiral selector used in
this column is macrocyclic antibiotic vancomycin. The mobile phase
(eluent) composition was as follows: methanol:triethylamine:anhy-
drous acetic acid (99.9:0.06:0.055 v/v). Measurements have been
performed at a flow rate of 1 mL/min at 25 °C with an injection
volume of 5 μL. The UV signal was recorded at a wavelength of 240
nm.
Samples for HPLC analysis were prepared by dissolution in the
eluent. The samples of the dried solid solution were prepared by
dissolving up to a concentration of 1 g/L, whereas liquid solution was
sampled in an amount of 0.5 mL and diluted 10−100 times. The
concentration of both the liquid and the solid solutions was
determined by detector calibration.
4.4. SLE Measurements of R,S-Citalopram·(+)DTT. 4.4.1. SLE
Experiments. Solubility experiments were carried out for
mixtures with different diasteromeric enrichment, de. Samples
were prepared using the required composition of S- and R,S-
citalopram·(+)DTT (synthetized as described in section 4.2.1)
and suspended in 4 mL of acetonitrile/methanol solvent
(1:0.08 v/v) in closed glass vials. All samples were heated up to
60 °C in a water bath until complete dissolution. Next, these
vials were inserted in a double-walled thermostatted vessel, and
solutions were electromagnetically stirred for about 72 h at 25
°C in order to establish the liquid−solid equilibrium.
4.2.2. Synthesis of S-Citalopram·(+)DTT. Synthesis was carried out
in two steps: conversion of escitalopram oxalate into S-citalopram free
base and next conversion of S-citalopram free base into S-
citalopram·(+)-DTT.
Conversion of Escitalopram Oxalate into S-Citalopram Free
Base. The procedure was the same as that in section 4.2.1 but instead
of citalopram hydrobromide escitalopram oxalate was used. The
neutralization time was shorter and lasted 30 min.
Conversion of S-Citalopram Free Base into S-Citalopram·(+)DTT
Salt in Ethanol. A 10 g amount of S-citalopram free base was dissolved
in 30 mL of ethanol by heating at 30 °C in the water bath. To the mass
obtained 12.4 g of (+)DTT acid was slowly added and heated until
complete dissolution. The diluted mixture was left for 24 h to
crystallize at room temperature. After this time, the solid product was
separated by vacuum filtration and the residue of solvents was
removed using a vacuum drier (t = 45 °C, p = 0.4 bar, time = 48 h).
The reaction yield obtained was ca. 70%.
Analogously to R,S-citalopram·(+)DTT, the solubility of S-
citalopram·(+)DTT (S-diastereomer) in ethanol was low and the
target product was almost entirely transferred into the crystalline
phase.
This solid product was used to prepare the solid solutions for the
crystallization process (see section 4.6).
4.2.3. Synthesis of R,S-Citalopram·(−)DTT Salt. Synthesis of R,S-
citalopram·(−)DTT salt was performed in two steps: conversion of
R,S-citalopram·(+)DTT into R,S-citalopram free base, and subse-
2561
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
After establishing equilibrium samples of the saturated solution
were centrifuged (t = 25 °C, V = 1000 rpm, time = 2 min) and
separated using a syringe with a filter. Solid phases were washed using
cold ethanol, separated by vacuum filtration, and dried in a vacuum
drier (t = 45 °C, p = 0.4 bar, time = 24 h). Solid and liquid phases
collected during the solubility experiments were analyzed by HPLC.
The SLE equilibrium was also identified in very close vicinity to a
1:1 composition of salt mixtures.
For that purpose, four samples with the 0.2 g of R,S-citalopram·(+)-
DTT with composition 1:1 were prepared by dissolving in 2, 3, 4, and
5 mL of acetonitrile/methanol solvent (1:0.08 v/v) to obtain a
different concentration of supersaturated solution, which corre-
sponded to various locations of the mixing point on the 1:1 line in
the ternary solubility diagram. Each of these mixing points belonged to
different conodes.
solvent was selected on the basis of solubility screening in
different solvents. Use of single-component solvents such as
ethanol, isopropanol, and water, resulted in very poor solubility
of citalopram, whereas in methanol the solubility of R,S-
citalopram·(+)DTT was very high; hence, pure methanol was
excluded as well. Binary mixtures methanol−acetonitrile
assured proper selectivity of the separation at relatively good
solubility.17,18
The results of the SLE measurements in the presence of
acetonitrile−methanol mixture are summarized in Table 1.
Table 1. Results of the SLE Measurements for R,S-
Citalopram·(+)DTT with Different de
Further procedure was the same as described above.
SLE measurements were repeated twice; results were found to be
reproducible.
before crystallization
liquid phase
after crystallization
liquid phase
deS [%] xS [g/g] deS [%] yR [g/g] yS [g/g]
solid phase
4.4.2. Additional Observations. During the investigation of SLE of
R,S-citalopram·(+)DTT, different crystal sizes were observed depend-
ing on the ratio of diastereomers in the samples. The solid phase with
a composition of diasteromeric salts close to 1:1 was characterized by
small, poorly shaped crystals (Figure 5a), whereas the solid phase with
an excess of S-diastereomer had larger but broken crystals (Figure 5b).
Moreover, mixtures with a composition close to racemate
crystallized much easier than pure S-diastereomer. These difficulties
in crystallization of the S-diastereomer might be caused by the large
width of the metastable zone for this diastereomer, which probably
originated from its crystalline structure.
4.5. Solid-Phase Analysis by XRPD. Samples of the solid phase
obtained after the SLE measurements for R,S-citalopram·(+)DTT (i.e.,
after phase separation, washing, and drying, as described in section
4.4) were subjected to analysis by X-ray powder diffraction (XRPD)
using a PANalytical X’Pert Pro diffractometer (PANalytical GmbH,
Germany). The radiation source was Cu Kα. Samples were measured
on Si holders and recorded in a 2-theta range of 3−40° with a step size
of 0.017 and counting time of 50 s for each step.
4.6. Multistage Crystallization. The method involved preparing
samples of known compositions in 50 mL closed plastic vessels. The
mixture containing a defined amount of solid S- and R,S-
citalopram·(+)DTT obtained according to the procedures described
above in section 4.2.1 was dissolved in acetonitrile/methanol (1:0.08
v/v). Such a feed stream was mixed with a proper mass of crystals
enriched with S-diastereomer. Subsequently, this sample was heated up
to 60 °C in a water bath until complete dissolution. Then the tube was
inserted in a double-walled thermostatted vessel, and the solution was
electromagnetically stirred (300 rpm) in order to establish the liquid−
solid equilibrium (t = 25 °C, time = 72 h). Afterward, the sample of
saturated solution was centrifuged (t = 25 °C, V = 1000 rpm, time = 2
min) and separated using a syringe with a filter. The solid phase was
washed using cold ethanol, separated by vacuum filtration, and dried in
a vacuum drier (t = 45 °C, p = 0.4 bar, time = 24 h). The liquid
fraction, which was enriched with S-diastereomer, was weighed and
used as the feed stream for the next stage. The composition of both
the solid and the liquid phases was determined by HPLC (see section
4.3). The same procedure was applied for the second and third stage of
crystallization. The fraction of raffinate from the second stage was
delivered into the third step as the feed stream, while the raffinate
received after the third stage was the final product.
deS [%]
0
−3.34
−1.40
−1.08
1.00
0.483
0.493
0.495
0.505
0.507
0.512
0.517
0.546
0.558
0.617
0.719
0.996
7.68
29.2
0.0078
0.0077
0.0053
0.0040
0.0032
0.0028
0.0031
0.0030
0.0025
0.0017
0.0008
0.0070
0.0090
0.0141
0.0178
0.0223
0.0263
0.0314
0.0371
0.0816
0.0899
0.153
9.80
19.6
24.4
39.2
49.0
58.8
68.6
78.4
88.2
93.1
98.0
53.7
69.7
78.1
83.7
84.8
93.8
94.6
97.8
99.5
97.8
1.46
2.50
3.38
9.18
11.5
23.3
43.9
0.306
99.1
0.627
The SLE data were marked on the ternary diagram and
linked by tie lines (conodes). The obtained phase diagram is
illustrated in Figure 6. As it can be seen, the data relate only to
the right half of the diagram, because the measurements were
started with racemate and conducted toward pure S-
diastereomer. Conodes were concentrated within the compo-
sition range corresponding to the three-stage crystallization
experiment, i.e., up to 93% of deS in the liquid phase.
The solubility diagram shows that pure S-diastereomer
dissolves much better in the solvent compared to the racemic
mixture. Furthermore, based on the shape of the ternary phase
diagram, it was confirmed that R,S-citalopram·(+)DTT forms
the solid solution almost in the entire range of composition of
the salt mixture. Moreover, the solubility curve indicates the
solubility maximum for pure S-diastereomer, for which
enrichment with the target component can be upgraded in
the liquid phase. However, it was observed that the conode in
the very vicinity of pure S-diastereomer changed the slope
toward to the S-corner, indicating the possibility of the presence
of the eutectic composition and of a narrow range for
crystallization of pure S-diastereomer (see the phase diagram
depicted in Figure 6).
The solubility measurements are also presented in a xy
coordinate system (see Figure 7), which demonstrates the
distribution of S-diastereomer between the solid and the liquid
phase without solvent. As it can be observed, the S-
diastereomer is significantly enriched in the liquid phase. This
phenomenon was utilized in this study for separation of 1:1
mixtures of salts by multistage diasteromeric crystallization.
For process design the equilibrium composition of the liquid
and solid phases have to be correlated. To determine proper
concentration dependencies, the SLE data points presented in
5. RESULTS AND DISCUSSION
5.1. Ternary Phase Diagram of R,S-Citalopram·(+)DTT.
To determine the ternary phase diagram the solid−liquid
equilibria for the salt pairs of R,S-citalopram·(+)DTT with
different composition was quantified. As mentioned above,
sample mixtures were prepared by mixing proper amounts of S-
and R,S-citalopram·(+)DTT obtained by synthesis described in
section 4.2. As a solvent the mixture of acetonitrile and
methanol (1:0.08 v/v) was used. The composition of the
2562
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
Figure 6. Ternary solubility diagram of citalopram·(+)DTT with different de in acetonitrile/methanol at 25 °C.
y = −0.0583xS2 + 1.3139xS − 0.6222
(10)
(11)
S
y = 0.0618xS2 − 0.0976xS + 0.0382
R
where yS and yR are the mass fraction of the S- and R-
diastereomer in the liquid solution, respectively, xS is the mass
fraction of the S-diastereomer in the solid solution, and the
coefficient of determination R2 = 0.95 and 0.92 for eqs 10 and
11, respectively.
Additionally, the SLE data in the system under study were
acquired for the sample composition in the vicinity of the 1:1
composition of the diasteromeric salts. For this purpose the
supersaturated solutions of 1:1 R,S-citalopram·(+)DTT differ-
ing in concentration were subjected to crystallization. The
obtained equilibrium data are presented in Table 2. It can be
Figure 7. Distribution of S-diastereomer between the solid and the
liquid phase at 25 °C.
Table 2. SLE Measurements in the Vicinity of 1:1
Composition of Diasteromeric Salts at 25°C
before crystallization
liquid phase
after crystallization
liquid phase
Figure 8 were correlated by two empirical polynomial
functions. To increase the accuracy of the process predictions
the approximation related the data within the range selected for
the crystallization experiment (i.e., upper limit was xS < 0.55).
The SLE relationships are given as follows
solid phase
yS = yR (1:1) [g/g] deS [%] xS [g/g] deS [%] yR [g/g] yS [g/g]
0.0578
0.0400
0.0309
0.0246
−2.16
−2.74
−3.34
−5.28
0.489
0.486
0.483
0.473
1.30
2.16
7.68
7.72
0.0082
0.0088
0.0078
0.0080
0.0085
0.0092
0.0090
0.0093
seen that higher enrichment of S-diastereomer in the liquid
phase can be obtained by a decrease of concentration of
supersaturated solutions. This indicated that the conodes in the
vicinity of the 1:1 line in the ternary phase diagram differ in the
slope. Those additional equilibrium data were accounted for in
the SLE dependencies (see Figure 8).
5.2. XRPD Patterns of R,S-Citalopram·(+)DTT. Due to
decomposition of diasteromeric salts at the melting temper-
ature the calorimetric measurements could not be used to
identify the structure of the crystalline phase. For this purpose,
XRPD solid state analysis was solely used. Samples of R,S-
citalopram·(+)DTT salts with different de were subjected to
analysis. As can be observed in Figure 9, each pattern has
similar reflections with the exception of the sample containing a
high excess of S-diastereomer (99.57% S). Its XRPD pattern
Figure 8. SLE relationship for R,S-citalopram·(+)DTT at 25 °C.
2563
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
which means that no adjustment of stream composition during
operation was made.
The experiment was performed according to the procedure
described in section 4.6. The experimental data obtained are
compared to the theoretical predictions in Table 3. Product
purity and yield were used to assess the results of the
performed process. The purity was measured using the HPLC
system, while the yield of the desired S-diastereomer (YS) was
calculated as follows
mR·y
− mRec ·y
S,Recn
S,Rn
n
YS =
·100%
mF ·y
n
S,F
(12)
n
where mR is the mass of raffinate, mRec is the mass of the
n
n
Figure 9. XRPD patterns of R,S-citalopram·(+)DTT with different de.
recycled stream, and mF is the mass of the feed stream.
n
On the basis of the comparison contained in Table 3 it can
be concluded that the process was predictable. The observed
discrepancies between theoretical and experimental masses of
streams can be attributed to the very small scale of the process.
Mass losses during experiment caused errors between predicted
and experimental mass of the products. It could also result from
small inaccuracies in the SLE measurements.
reveals the existence of another crystal lattice as a new fraction.
Thus, in this pattern pure S-diastereomer and saturated mixed
crystals can be clearly identified (see additional peak shown by
the arrow in Figure 9). It can be concluded that this is
associated with the distinct crystallization area and the existence
of a mixture close to the eutectic composition, which was
identified during the SLE measurements. In the remaining
cases, the absence of new phases was confirmed. Additionally,
the substrates, i.e., S- and R,S-citalopram·(+)DTT used for
preparation of samples for the SLE measurements, were
crystallized from ethanol and also subjected to XRPD analysis.
They indicated the same structure of the crystal lattice.
On this basis, R,S-citalopram·(+)DTT was defined as a solid
solution forming system over almost the entire range of
composition.
5.3. Multistage Crystallization. The knowledge of SLE
was further exploited for designing the multistage crystallization
experiment. As mentioned in section 4.6, the design was based
on the fact that a de increase could be generated in the liquid
phase. The process was realized based on the concept of the
countercurrent multistage crystallization, which was explained
above in section 3. It should be noted that only the first run of
the three-stage process was performed with the goal of
verification of the process feasibility and reproducibility of
theoretical simulations. This means that the recycled streams
consisted of mixtures of the solid phase with proper de
obtained by addition of pure S-diastereomer to diasteromeric
mixtures of salts R,S-citalopram·(+)DTT with 1:1 composition.
Raffinate after each stage was directly delivered into the next
stage as the feed stream without partial recycling. Nevertheless,
because the process is realized in a batch mode, its extension by
including recycling is straightforward.
As mentioned above, product purity could be increased at
the expense of a lower yield by manipulation of the
composition of extract withdrawn from the first stage (xS,E ).
1
The optimal choice of xS,E should be done on the basis of
1
economic evaluations.
After accomplishing all stages, the exhausted crystalline phase
had reduced de compared to the composition of the feed
stream whereas the liquid phase was enriched with the desired
S-diastereomer. Thus, ca. 96% of S-diastereomer (deS = 90.06%,
prediction; 92.5%, experiment) in the liquid solution was
obtained after the third stage of crystallization and ca. 44% in
the exhausted extract stream of the first stage. The total yield
for the experiment performed was 7%. However, if the extract
streams withdrawn from the all stages are recycled the yield
increases markedly. For instance, single recycling of the extract
stream from the second stage can upgrade the yield with a
factor 1.26, double recycling 1.34, and when the extract from
the first stage is also recycled 1.67. Nevertheless, in the ideal
case (no mass loses), when semicontinuous processing is
interrupted, 100% of the processed mixture can be recovered
from all stages.
It can be observed that the first step of crystallization leads to
production crystals with an excess of R-diastereomer, which
cannot be directly recycled into the process.
Recovery of the 1:1 composition in the solid phase by simple
recrystallization of depleted solutions of R,S-citalopram·(+)-
DTT was unsuccessful; therefore, for this purpose a new
procedure has been developed.
5.4. Recovery of the Racemic Composition in the
Solid Phase. To increase the yield of the multistage
crystallization process, recovery of the 1:1 composition of the
diasteromeric salt was performed for extract E1 that was
enriched with undesired R-diastereomer. Recovery of the 1:1
composition from the solid solution containing 48% of S-
diastereomer (deS = −3.34) was realized in two steps. The first
step was conversion of R,S-citalopram·(+)DTT into R,S-
citalopram free base, and the second one was conversion of
R,S-citalopram free base into R,S-citalopram·(−)DTT. The
description of the two steps was explained in section 4.2.3.
For the experiment the mass and composition of the feed
stream (F1) for the first step was set as follows: m = 25 g, yS,F
=
1
yR,F = 0.02. Also, the composition of the exhausted extract
1
(xS,E ) had to be specified. Thus, for this process xS,E = 0.48 was
1
1
fixed. When the value of xS,E was set the composition of both
1
diastereomers in raffinate (y , y ) was determined by the
S,R1 R,R1
equations of the SLE relationships (eqs 10 and 11). Next, the
mass of remaining streams, i.e., En, Rn, Recn, Mn, and their
composition were calculated by solving the mass balance
equations listed in section 3.
The diasteromeric enrichment, de, of the extract streams, En,
was set the same as that of feed, Fn−1: deE2 = deF1, deE3 = deF2,
2564
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
Table 3. Experimental and Theoretical Data for the Three-Stage Crystallization
(a) data for the first stage
theoretical and experimental inlets
feed, F1
recycled stream, Rec1
a
Vsolv [mL]
mass [g]
25
yR,F [g/g]
yS,F [g/g]
deS [%]
mass [g]
0.105
xR,Rec [g/g]
xS,Rec [g/g]
deS [%]
7.66
1
1
1
1
30.7
0.0200
0.0200
0
0.462
0.538
theoretical outlets
raffinate, R1
extract, E1
Vsolv [mL]
mass [g]
24.4
yR,R [g/g]
yS,R [g/g]
deS [%]
7.66
mass [g]
0.693
xR,E [g/g]
xS,E [g/g]
deS [%]
YS [%]
1
1
1
1
30.7
0.0078
0.0091
0.5174
0.483
−3.40
33.1
experimental outlets
raffinate, R1
extract, E1
Vsolv [mL]
mass [g]
23.1
yR,R [g/g]
yS,R [g/g]
deS [%]
9.40
mass [g]
0.542
xR,E [g/g]
xS,E [g/g]
deS [%]
YS [%]
1
1
1
1
28.9
0.0099
0.0119
0.557
0.443
−11.4
43.6
(b) data for the second stage
theoretical inlets
feed, F2 (R1)
recycled stream, Rec2
Vsolv [mL]
mass [g]
23.1
yR,F [g/g]
yS,F [g/g]
deS [%]
mass [g]
xR,Rec [g/g]
xS,Rec [g/g]
deS [%]
64.8
2
2
2
2
29.1
0.0078
0.0091
7.66
0.537
0.176
0.824
experimental inlets
feed, F2 (R1)
recycled stream, Rec2
Vsolv [mL]
mass [g]
23.1
yR,F [g/g]
yS,F [g/g]
deS [%]
mass [g]
0.537
xR,Rec [g/g]
xS,Rec [g/g]
deS [%]
64.8
2
2
2
2
28.9
0.0099
0.0119
9.40
0.176
0.824
theoretical outlets
raffinate, R2
extract, E2
Vsolv [mL]
mass [g]
23.283
yR,R [g/g]
yS,R [g/g]
deS [%]
64.80
mass [g]
0.344
xR,E [g/g]
xS,E [g/g]
deS [%]
YS [%]
2
2
2
2
29.1
0.0044
0.0207
0.500
0.500
0
14.3
experimental outlets
raffinate, R2
extract, E2
Vsolv [mL]
mass [g]
22.3
yR,R [g/g]
yS,R [g/g]
deS [%]
65.0
mass [g]
0.361
xR,E [g/g]
xS,E [g/g]
deS [%]
YS [%]
2
2
2
2
27.8
0.0050
0.0235
0.509
0.491
−1.88
29.7
(c) data for the third stage
theoretical inlets
feed, F3 (R2)
recycled stream, Rec3
Vsolv [mL]
mass [g]
22.3
yR,F [g/g]
yS,F [g/g]
deS [%]
mass [g]
xR,Rec [g/g]
xS,Rec [g/g]
deS [%]
90.1
3
3
3
3
27.8
0.0044
0.0207
64.8
1.29
0.0497
0.950
experimental inlets
feed, F3 (R2)
recycled stream, Rec3
Vsolv [mL]
mass [g]
22.3
yR,F [g/g]
yS,F [g/g]
deS [%]
mass [g]
1.29
xR,Rec [g/g]
xS,Rec [g/g]
deS [%]
90.1
3
3
3
3
27.8
0.0050
0.0235
65.0
0.0497
0.950
theoretical outlets
product, P (R3)
extract, E3
Vsolv [mL]
mass [g]
23.432
yR,P [g/g]
0.0036
yS,P [g/g]
deS [%]
mass [g]
0.171
xR,E [g/g]
xS,E [g/g]
deS [%]
YS [%]
3
3
27.8
0.0682
90.1
0.462
0.538
7.66
79.7
experimental outlets
product, P (R3)
extract, E3
Vsolv [mL]
mass [g]
21.9
yR,P [g/g]
yS,P [g/g]
deS [%]
92.5
mass [g]
0.282
xR,E [g/g]
xS,E [g/g]
deS [%]
8.78
YS [%]
3
3
26.1
0.0027
0.0686
0.456
0.544
52.5
a
Molar mass M = 710.
2565
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566

Crystal Growth & Design
Article
The obtained equilibrium data for diasteromeric salts R,S-
citalopram·(−)DTT are summarized in Table 4.
ACKNOWLEDGMENTS
■
The financial support of Polish Ministry of Science and Higher
Education within the grant no. DPN/N135/NIEMCY/2010 is
greatly acknowledged.
Table 4. Result of Conversion of R,S- Citalopram Free Base
into R,S-Citalopram·(−)DTT
REFERENCES
(1) Cayen, M. N. Chirality 1991, 3, 94−98.
(2) Somagoni, J.; Reddy, S.; Koorelli, S.; Manda, S.; Yamsani, M. R.
Pharm. Anal. Acta 2011, 2, 2−4.
■
substrate
product
solid phase
R,S- citalopram free base
R,S-citalopram (−)DTT
(3) Sheldon, R. A. Chirotechnology: Industrial Synthesis of Optically
Active Compounds; Marcel Dekker Inc.: New York, 1993.
(4) Collins, A. N.; Sheldrake, G. N.; Crosby, J. Chirality in Industry II:
Developments in the Manufacture and Applications of Optically Active
Compounds; John Wiley & Sons: Chichester, 1997.
(5) Maier, N; Franco, P.; Lindner, W. J. Chromatogr., A 2001, 906,
3−33.
(6) Karamertzanis, P. G.; Anandamanoharan, P. R.; Fernandes, P.;
Cains, P. W.; Vickers, M.; Tocher, D. A.; Florence, A. J.; Price, S. L. J.
Phys. Chem. B 2007, 111, 5326−5336.
(7) Li., Z.; Grant, D. J. Pharm. Sci. 1997, 86 (10), 1439−1447.
(8) Gu, C.-H.; Grant, D. J. W. Physical Properties and Crystal
Structures of Chiral Drugs. In Handbook of Experimental Pharmacology:
Stereochemical Aspects of Drug Action and Disposition; Springer: Berlin,
2003.
yR [g/g] yS [g/g] deS [%] xR [g/g] xS [g/g] deS [%] yield [%]
0.130
0.104
0.086
0.120
0.096
0.080
−3.6
−3.6
−3.6
0.500
0.500
0.498
0.500
0.500
0.502
0
90.2
89.3
86.6
0
0.40
The key point of the operation was to use (−)DTT acid and
choose the appropriate concentration of R,S-citalopram free
base (depleted with S-isomer) in the solvent. It is evident that
the decrease of the concentration in the initial solution leads to
a higher enrichment of the solid phase. Thus, starting with a
nonequimolar mixture of citalopram free base, the racemic
composition may be reproduced or even a small enrichment of
the S-diastereomer can be achieved using (−)DTT.
(9) Rekoske, J. E. AIChE J. 2001, 47 (1), 2−5.
(10) Lorenz, H.; Sheehan, P.; Seidel-Morgenstern, A. J. Chromatogr.
A 2001, 908, 201−214.
6. CONCLUSIONS
In this study multistage crystallization was used to isolate the
desired S-diastereomer of citalopram from racemic mixture.
The process design was proceeded by measurement of the
solid−liquid equilibria of diasteromeric salts R,S-
citalopram·(+)DTT in an acetonitrile/methanol solvent
system. The determined ternary phase diagram indicated that
R,S-citalopram·(+)DTT involved solid solutions in the
crystalline phase almost in the entire composition range.
Because pure S-diastereomer had a higher solubility than
racemate, diasteromeric excess was upgraded in the liquid
phase. This phenomenon was utilized in the realization of the
crystallization process.
On the basis of the results of the SLE measurements a three-
stage crystallization experiment was designed and successfully
conducted. Nevertheless, in the first crystallization stage the
enrichment of the liquid phase with the desired S-diastereomer
was accompanied by its depletion in the solid phase. Therefore,
to increase the process yield a procedure for effective recovery
of the 1:1 composition of diasteromeric salts in the solid phase
was developed.
The results of the experiment indicated that multistage
crystallization was an efficient way to receive pure S-
diastereomer from the racemic mixture. The total yield for
the experiment performed was 7%. However, recycling of
extract streams can improve the yield. Nevertheless, when the
operation is interrupted or accomplished, 100% of the
processed mixture can be recovered from all stages.
The advantage of diasteromeric crystallization over an
“achiral approach” (i.e., without addition of the resolving
agent) was the possibility of resolution of 1:1 mixtures of salts.
In the achiral approach only mixtures enriched with the desired
enantiomer could be resolved.22
(11) Leusen, F.; Noordik, J.; Karfunkel, H. Racemate resolution via
crystallization of diastereomeric salts: Thermodynamic Considerations and
Molecular Mechanic Calculations; Pergamon Press: New York, 1993;
Vol. 49 (24), pp 5377−5396.
(12) Collet, A. Enantiomer 1999, 4, 157−172.
(13) Muller, S.; Afraz, M. C.; Gelder, R.; Ariaans, G. J. A.; Kaptein,
̈
B.; Broxterman, Q. B.; Bruggink, A. Eur. J. Org. Chem. 2005, 6, 1082−
1096.
(14) Ferreira, F. C.; Ghazali, N. F.; Cocchini, U.; Livingston, A. G.
Tetrahedron 2006, 17, 1337−1348.
(15) Kaptein, B.; Vries, T. R.; Nieuwenhuijzen, J. W.; Kellogg, R. M.;
Grimbergen, R. F. P.; Broxterman, Q. B. New Developments in
Crystallization-Induced Resolution. In Handbook of Chiral Chemicals;
Ager, D., Ed.; CRC Press: Boca Raton, FL, 2005.
(16) Sistla, V. S.; von Langermann, J.; Lorenz, H.; Seidel-
Morgenstern, A. Cryst. Growth Des. 2011, 11, 3761−3768.
(17) Dancer, R. J.; de Diego, H. L. Org. Process Res. Dev. 2009, 13 (1),
23−33.
(18) Dancer, R. J.; de Diego, H. L. Org. Process Res. Dev. 2009, 13 (1),
38−43.
(19) Parker, N.; Brown, C. Ann.Pharmacother. 2000, 34 (6), 761−
771.
(20) Aronson, J. Meyler’s Side Effects of Psychiatric Drug; Elsevier:
Amsterdam, 2009.
(21) Schatzberg, A.; Nemeroff, C. The American psychiatric publishing
textbook of psychopharmacology; American Psychiatric Publishing, Inc.:
Washington, D.C., 2009.
(22) Balawejder, M.; Galan, K.; Elsner, M.; Seidel-Morgenstern, A.;
Piatkowski, W.; Antos, D. Chem. Eng. Sci. 2011, 66, 5638−5647.
(23) Sloan, G. J.; McGhie, A. R. Techniques of melt crystallization; J.
Wiley & Sons: New York, 1988.
(24) Lin, S. W.; Ng, K. M.; Wibowo, C. Comput. Chem. Eng. 2008,
32, 956−970.
AUTHOR INFORMATION
Corresponding Author
*E-mail: dorota.antos@prz.edu.pl.
■
Notes
The authors declare no competing financial interest.
2566
dx.doi.org/10.1021/cg300166g | Cryst. Growth Des. 2012, 12, 2557−2566