Chemical Engineering Science 71 (2012) 5–17

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science

journal homepage: www.elsevier.com/locate/ces

Seeding strategies and residence time characteristics of continuous

preferential crystallization
Shamsul Qamar a,b,n, Martin Peter Elsner a,c, Iltaf Hussain b, Andreas Seidel-Morgenstern a
a
Max Planck Institute for Dynamics of Complex Technical Systems, Sandtorstr. 1, 39106 Magdeburg, Germany
b
Department of Mathematics, COMSATS Institute of Information Technology, Park Road Chak Shahzad, Islamabad, Pakistan
c
Georg Simon Ohm University of Applied Sciences, Prinzregentenufer 47, 90489 Nuremberg, Germany

a r t i c l e i n f o abstract

Article history: This contribution investigates the effects of different seeding strategies and residence time character-
Received 4 September 2011 istics on the dynamics of a Mixed Suspension Mixed Product Removal (MSMPR) crystallizer equipped
Received in revised form with a fines dissolution unit. For the first time continuous preferential enantioselective crystallization is
9 December 2011
investigated. The fines dissolution is included as recycle streams around the MSMPR crystallizer.
Accepted 20 December 2011
Moreover, primary heterogeneous and secondary nucleation mechanisms along with size-dependent
Available online 29 December 2011
growth rates are taken into account. A semi-discrete high resolution finite volume scheme (FVS) is
Keywords: employed for discretizing the derivatives with respect to the length coordinate. The resulting ordinary
Population balances differential equations (ODEs) are solved by a Runge–Kutta method of order four. Several numerical case
Continuous preferential crystallization
studies are carried out. The results support process design and optimization.
Fines dissolution
& 2011 Elsevier Ltd. All rights reserved.
Seeding strategies
High resolution schemes
Goal functions

1. Introduction is kinetically inhibited. During a starting-up period, which

An attractive process for gaining pure enantiomers from
racemic mixtures is the so-called preferential crystallization
(PC), which has been realized up to now in a discontinuous
operation mode, see for example Jacques et al. (1981), Collet
(1999), Alvarez-Rodrigo et al. (2004), Elsner et al. (2005),
Coquerel (2007), Czapla et al. (2009), and Elsner et al. (2011).
For elucidating the principle of a continuous enantioselective
process one might consider a suspension crystallizer revealing
MSMPR characteristics, i.e. a perfectly mixed tank (concerning
both phases), which is continuously fed with a solution posses-
sing a racemic composition of two enantiomers, see Fig. 1. Solid
particles and liquid phase are continuously withdrawn. By a
continuous supply of homochiral seed crystals of the preferred
target enantiomer the preferential crystallization (PC) of only this
enantiomer is initialized, i.e. growth of the seed crystals and
possibly secondary nucleation of crystals of the seeded enantio-
mer, provided the crystallization takes place within the meta-
stable zone where spontaneous, uncontrolled primary nucleation
n
Corresponding author at: Max Planck Institute for Dynamics of Complex
Technical Systems, Sandtorstr. 1, 39106 Magdeburg, Germany.
Tel.: þ92 51 8318471; fax: þ 92 51 8318499.
E-mail addresses: qamar@mpi-magdeburg.mpg.de,
shamsul.qamar@comsats.edu.pk (S. Qamar).
strongly depends on the properties of the system as well as on
the process parameters, the concentration of the target enantio-
mer is decreasing until a steady state is reached where the
composition is determined by the mean residence time. Due to
different kinetic mechanisms and their inherent different time
constants, a different depletion of the supersaturation for each
enantiomer can be realized by an appropriate choice of the
process conditions. As long as a critical mean residence time,
where primary nucleation may appear, is not exceeded, the
concentration of the undesired counter enantiomer remains
constant during the whole time. This fact reveals a benefit of this
continuous process in comparison to the batch one. An optimal
selection of the process conditions allows a constant production
of the goal enantiomer at a high purity level. The mathematical
modeling of PC as well as the optimization of the essential
operating conditions requires high numerical precision (Qamar
et al., 2008).
The population balance based modeling is regarded as a well
suited and commonly accepted modeling approach for describing
crystallization processes. The theory of population balances began
in 1960s when Hulburt and Katz (1964) introduced it in the field
of chemical engineering. A comprehensive overview on popula-
tion balance modeling, nucleation and growth kinetics terms and
methods of solution can be found in the book by LeVeque (2002),
Mersmann (2001), Nývlt et al. (1985), and Ramkrishna (2000).

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.12.030
6 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17
Fig. 1. Illustration of the principle of continuous enantioselective crystallization.
During the last decades many efficient methods were developed
for solving population balance models (PBMs) such as the method
of characteristics introduced in various fields by Rhee et al. (1986)
and adopted for PBMs by Lim et al. (2002), Qamar et al. (2009),
and Ramkrishna (2000), the method of weighted residual or
orthogonal collocation by Rawlings et al. (1992), the Monte Carlo
simulation by Smith and Matsoukas (1998) and Tandon and
Rosner (1999), the fixed and moving pivot techniques by Kumar
and Ramkrishna (1996a,b), and the high resolution finite volume
schemes by Gunawan et al. (2004) and Qamar et al. (2006).
In this work a dynamic mathematical model is derived for
simulating a continuously operated ideally mixed MSMPR crystal-
lizer applied for continuous preferential enantioselective crystal-
lization. The developed model includes the phenomena of
primary heterogeneous nucleation, secondary nucleation, and
size-dependent growth. A recycle pipe is attached to the crystal-
lizer for the fines dissolution and it is assumed that fines are
completely dissolved at the other end of the pipe. The effects of
different seeding and operating strategies are investigated in
simulations. The model and its parameters are based on experi-
mental data obtained from batchwise crystallization. The semi-
discrete flux-limiting finite volume scheme of Koren (1993) is
implemented to solve the given model equations. The numerical
results demonstrate the high order accuracy, efficiency and
potential of the proposed numerical method for solving models
describing continuous preferential crystallization of enantiomers
and the potential of the process. The same performance of finite
volume schemes was also observed by Gunawan et al. (2004),
Hermanto et al. (2008), and Qamar et al. (2006) for solving
polymorphic and batch crystallization models. Moreover, differ-
ent contributions are available in the literature on continuous
seeding and seed dissolution strategies in the batch crystalliza-
tion process, see for example the articles of Aamir et al. (2010),
Chung et al. (1999), and Woo et al. (2011) and reference therein.
The paper is organized as follows: in Section 2, a dynamic
mathematical model based on population balances will be devel-
oped for a continuous MSMPR preferential crystallizer. In Section
3, the numerical method for solving the governing equations will
be presented. In Section 4, some numerical case studies will be
carried out. Finally, conclusions will be drawn in Section 5.
2. Continuous crystallization model
The model of the continuous crystallization process considers
a two-phase dispersed system. The first phase is a continuous
liquid phase containing the dissolved enantiomers and the
solvent. The conditions in this phase determine the growth and
 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17 7

nucleation rates of crystals during the crystallization process and The term corresponding to the dissolution unit in Eq. (1) is
have significant influence on the crystal size distributions (CSDs). expressed by
The second phase is the dispersed solid phase of individual
ðkÞ hðxÞF ðkÞ ðt,xÞ
crystals. The growth of crystals strongly depends on the degree F_ pipe ðt,xÞ ¼ , ð6Þ
of supersaturation of the liquid phase.
tpipe
Description of the CSD for a dispersed system in a crystallizer where h(x) represents the classification (death) function.
needs an adequate differential equation accounting for the effects The i-th moment of the CSD is defined as
of nucleation, growth, and other phenomena involving the change Z 1
of crystals population, for instance fines dissolution in the current mðkÞ
i
ðtÞ ¼ xi F ðkÞ ðt,xÞ dx, i ¼ 0; 1,2, . . . , kA fp,cg: ð7Þ
0
study. The population balance modeling is a suitable approach for
predicting CSD in the solid phase of a disperse system. According to Fig. 1, the corresponding mass balances for the liquid
An one-dimensional mathematical model for an ideally mixed phase are given as
continuous crystallizer is given by the following population ðkÞ Z 1
dmL ðtÞ
balance equation (PBE) for the number density function F ðkÞ ðt,xÞ _ ðkÞ
¼m _ ðkÞ
L,in ðtÞm L,out ðtÞ3rS kv x2 GðkÞ ðt,xÞF ðkÞ ðt,xÞ dx, k A fp,cg
dt 0
(e.g. Abegg et al., 1968; Hulburt and Katz, 1964; Motz et al., 2003)
ð8Þ
@F ðkÞ @ðGðkÞ F ðkÞ Þ _ ðkÞ ðkÞ ðkÞ
¼ þ F seeds F_ out F_ pipe , kA fp,cg, ð1Þ with the initial conditions
@t @x
where p stands for the preferred enantiomer and c for the counter mðkÞ ðkÞ
L ðt ¼ 0Þ ¼ mL,0 ¼ rL,0 V L , ð9Þ
enantiomer. In Eq. (1) the term on the left hand side denotes the where rL,0 ¼ rL ðt ¼ 0Þ is the density of fresh (initially) super-
accumulation of crystals of size x. The first term on the right hand saturated solution. The incoming and outgoing mass flow rates
side represents the convective transport in the direction of the are defined as
property coordinate x due to size-dependent crystal growth rate
ðkÞ _ ðkÞ
m _ ðkÞ
ðtÞ ¼ m _ ðkÞ
þm ðtÞ, _ ðkÞ
m _ ðkÞ _ ðkÞ
L,out ðtÞ ¼ m L,out,c ðtÞ þ m L,out,pipe ðtÞ:
GðkÞ . The term F_
seeds denotes the incoming particle number flux L,in L,in,c L,in,pipe
ðkÞ ð10Þ
due to seeding, F_ out is the particle number flux due to product
ðkÞ
ðkÞ
removal, and F pipe denotes the particle number flux to the fines The inflow rate m _ L,in is the sum of two incoming streams to the
crystallizer, the first one m _ ðkÞ
L,in,c stands for the incoming flux of
dissolution unit (c.f. Fig. 1).
fresh solution to the crystallizer and the second one m _ ðkÞ
The initial conditions for the PBE in Eq. (1) are based on the L,in,pipe
denotes the incoming flux of particle-free solution from the
assumption that the crystallizer initially (at time t ¼0) contains
dissolution pipe. Similarly, the mass outflow rate m _ ðkÞ
L,out is the
only pure liquid phase. In other words
sum of two outgoing streams, i.e. mass flow rate of the utilized
F ðkÞ ðt ¼ 0,xÞ ¼ 0, kA fp,cg: ð2Þ solution from the crystallizer and the mass flow rate of the
solution containing the fraction of fines being taken out from
Assuming further, that primary nucleation leads to a crystal
the crystallizer to the dissolution unit. The mass flow rate of the
nuclei of minimum size xmin and that the number density function
incoming fresh supersaturated solution is defined as
F ðkÞ vanishes for a arbitrary large crystal size xmax , holds
_ ðkÞ ðkÞ _
BðkÞ m L,in,c ¼ w ðt ¼ 0ÞrL,0 V L , k A fp,cg, ð11Þ
ðkÞ 0 ðtÞ ðkÞ
F ðt,x ¼ xmin Þ ¼ , F ðt,x ¼ xmax Þ ¼ 0, ð3Þ
Gðt,xmin Þ where wðkÞ represents the mass fraction of k-th enantiomer which
where BðkÞ
denotes the nucleation rate. The crystallizer is only is defined as
0
seeded with preferred enantiomer, thus
( ðpÞ mðkÞ
L ðtÞ
wðkÞ ðtÞ ¼ , kA fp,cg: ð12Þ
_ mL ðtÞ þ mLðcÞ ðtÞ þ msolvent
ðpÞ
_F ðkÞ ðxÞ ¼ F seeds ðxÞ for k ¼ p, ð4Þ
seeds
0 for k ¼ c: The terms mðkÞ ðpÞ
and mðcÞ
L , mL L represent the masses of k-th,
For the operation of the configuration shown in Fig. 1, three types preferred, and counter enantiomers, respectively. Moreover,
of residence times are relevant: msolvent is mass of the solvent (water). The composition depen-
dence solution density is defined as
1. The liquid mean residence time in the perfectly mixed crystal- rL ðtÞ ¼ 1000ðro þK 3 wtot ðtÞÞ, ð13Þ
lizer, tL ¼ V L =V_ L , where V L denotes the volume of the liquid
phase in the crystallizer and V_ L is the volumetric flow rate of where ro is the density of water
the liquid phase. 1
2. A characteristic solid mean residence time in a perfectly mixed ro ¼ : ð14Þ
K1 þ K2T2
_ ðpÞ
crystallizer, tS ¼ V S =ðrS =m seeds
Þ, where V S represents the overall
volume of the solid phase, rS is the density of the solid phase and The parameters K1, K2, K3 are given in Table 1 and T denotes the
m_ ðpÞ is the mass flow rate of the seeds of preferred enantiomer. temperature which is kept constant at 33 1C. The total mass
seeds
3. The residence time in the pipe (plug flow), tpipe ¼ V pipe =V_ pipe . fraction wtot is given as
Here, V pipe denotes volume of the pipe and V_ pipe is the wtot ðtÞ ¼ wðpÞ ðtÞ þ wðcÞ ðtÞ: ð15Þ
volumetric flow rate to the pipe.
The outgoing mass fluxes of the solution from the crystallizer are
The number density function in the withdrawn crystal stream is defined as
assumed to be equal to the number density function inside the _ ðkÞ
m ððkÞ
ðtÞrL ðtÞV_ L ,
L,out,c ðtÞ ¼ w k A fp,cg: ð16Þ
crystallizer. Therefore, the right hand sided second and third
terms of Eq. (1) are given as The outgoing mass flux from the crystallizer to the dissolution
unit is given as
ðkÞ F ðkÞ ðxÞ ðkÞ F ðkÞ ðt,xÞ
F_ seeds ðxÞ ¼ seeds
, F_ out ðt,xÞ ¼ , k A fp,cg: ð5Þ _ ðkÞ
m ðtÞ ¼ wðkÞ ðtÞrL ðtÞV_ pipe , k A fp,cg: ð17Þ
tS tS L,out,pipe
8 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17

Respecting the fines dissolution and the residence in the tube, the Table 1
incoming mass flux from the dissolution unit to the crystallizer is Values of physicochemical parameters related to threonine–water system (Elsner
et al., 2005, 2011).
given as
Z
kv rS 1 3 Parameters Symbols Value Unit
m_ ðkÞ
L,in,pipe
ðtÞ ¼ _
m ðkÞ
L,out,pipe
ðtt pipe Þ þ x hðxÞF ðkÞ ðttpipe ,xÞ dx:
tpipe 0
Volume shape factor kv 0.122 –
ð18Þ Density of solid phase rS 1250 kg/m3
Crystal growth rate exponent kg,eff,0 2:98  109 m/min
Due to racemic solution, the initial masses of both enantiomers Nucleation rate exponent m  3/min
kb,sec,0 2:38  1026
are the same and the mass fluxes of both enantiomers are equal in Volume of the liquid phase VL m3
4:496  104
the feed stream, holds Volumetric flow rate V_ L 0:4  10 5 m3/min

mðpÞ ðcÞ
¼ mL,0 , _ ðpÞ
m ¼ ðcÞ
_ L,in,c
m : ð19Þ Volume of the pipe V pipe 4:0  105 m3
L,0 L,in,c
Volumetric flow rate V_ pipe 0:4  10 5 m3/min
A size-dependent growth rate approach is applied (e.g. Mass of the solvent msolvent 0.4843 kg
Constant for density K1 1.0023 cm3 g  1
Mersmann, 2001; Elsner et al., 2011)
Constant for density K2 4:68  106 cm3 g  1 K  2
GðkÞ ðt,xÞ ¼ kg,eff ðTÞðSðkÞ ðtÞ1Þg ð1 þaASL xÞdASL : ð20Þ Constant for density K3 0.3652 cm3 g  1
Density of water ro 0.9947 g cm  3
The exponent g denotes the growth order and the constants aASL Density of fresh solution rL,0 1:067  103 kg m  3
and dASL represent the size dependency. The temperature depen- Parameter for temperature KT 1874.4 K
dependence
dence of the growth rate constant kg,eff is given by an Arrhenius
Parameter for mass fraction KW 0.290 –
approach dependence
  Universal gas constant R 8.314 J mol/K
EA,g
kg,eff ¼ kg,eff,0 exp  : ð21Þ Growth rate exponent g 1.1919 –
RT Parameter for crystal growth aASL m1
2:0209  104
Parameter for crystal growth dASL 4:066  101 –
Here, the symbol kg,eff,0 is the pre-exponential factor of the
Activation energy for crystal EA,g 75:54  103 J mol  1
growth rate constant, EA,g is the activation energy, and R is the
growth
universal gas constant. Their values are given in Table 1. The J mol  1
Activation energy for EA,b 63:83  103
symbol SðkÞ ðtÞ denotes the supersaturation of the k-th enantiomer nucleation
which is defined as Secondary nucleation bsec 4.80 –
exponent
wðkÞ ðtÞ Exponent for third moment nsec 3.0258 –
SðkÞ ðtÞ ¼ , ð22Þ Nucleation constant kb,prim 3:847  102 ðkg m3 Þ7=3 =min K
wðkÞ
eq ðtÞ
Constant for exponential law aprim 4:304  10 3 –
where wðkÞ
eq is the saturated mass fraction of the k-th enantiomer
Constant of surface area A1  6.9584 m2
dependency
defined as
Solubility constant (Eq. (23)) aeq 9:83  102 –
wðkÞ
eq ðtÞ ¼ aeq þ beq w
ðkÞðtÞ
: ð23Þ Solubility constant (Eq. (23)) beq 7:45  102 –
Seeds distribution constant a1 0.014 –
The terms aeq and beq are the solubility constants given in Table 1. (Eq. (43))
The nucleation rate is defined as the sum of primary (hetero- Seeds distribution constant a2 0.0009 m
(Eq. (43))
geneous) nucleation rate and secondary nucleation rate (e.g.
Seeds distribution constant a3 0.288 –
Mersmann, 2001; Elsner et al., 2011) (Eq. (43))

BðkÞ ðkÞ ðkÞ

0 ðtÞ ¼ B0,prim ðtÞ þ B0,sec ðtÞ, k A fp,cg: ð24Þ

The primary nucleation rate is given by a semi-empirical equation constant is given as

derived from Mersmann model (see Mersmann, 2001; Elsner  
E
et al., 2011) kb,sec ¼ kb,sec,0 exp  A,b : ð29Þ
! RT
3
ðkÞ ðkÞ
aprim lnðrS =C ðkÞ
eq Þ
B0,prim ðtÞ ¼ Z ðtÞ exp , ð25Þ The symbol kb,sec,0 is the pre-exponential factor of the secondary
ðln SðkÞ ðtÞÞ2
nucleation rate constant and EA,b is the corresponding activation
where energy.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!ffi Although the steady state results are crucial for an assessment
P u
K T =T
ðkÞ
k w ðtÞ
u
tln rS  ðkÞ 7=3 of the process, a better experimental realization can be achieved
ðkÞ
Z ðtÞ ¼ kb,prim Te exp S ðtÞC ðkÞ
eq ðtÞ :
KW ðkÞ
C eq ðtÞ by investigating the process dynamically.

ð26Þ

Here, C ðkÞ
eq denotes the concentrations of the k-th enantiomer at
equilibrium as defined below 3. Numerical schemes

C ðkÞ ðkÞ
eq ðtÞ ¼ rL ðtÞweq ðtÞ: ð27Þ In this section, the high resolution finite volume scheme of
Koren (1993) is implemented for discretizing the derivative of
The secondary nucleation rate is given by an overall power law
length coordinate in the population balance model. For the
expression
derivation and explanation of the scheme, it is convenient to
bsec nsec
BðkÞ ðkÞ
0,sec ðtÞ ¼ kb,sec ðS ðtÞ1Þ ðmðkÞ
3 ðtÞÞ , ð28Þ re-write Eq. (1) in the following form:

where bsec is the secondary nucleation rate exponent and nsec is @F ðkÞ ðt,xÞ @½GðkÞ ðt,xÞF ðkÞ ðt,xÞ
¼ þ Q ðkÞ ðt,xÞ, ð30Þ
the third moment exponent. The secondary nucleation rate @t @x
 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17 9

where order four was used. In the case of fines dissolution with time-
ðkÞ ðkÞ ðkÞ delay the residence time in the pipe was taken as the integer
Q ðkÞ
ðt,xÞ ¼ F_ seeds ðxÞF_ out ðt,xÞF_ pipe ðt,xÞ: ð31Þ multiple of the time step. This facilitates to keep the old values in
In order to apply a numerical scheme, the first step is to discretize memory and to avoid the linear interpolation. The computer
the computational domain which is here the crystal size x. Let N program is written in the Matlab 7.9.1 (R2009b).
be a large integer, and denote by ðxið1=2Þ Þ, i A ð1; 2, . . . ,N þ 1Þ the
partitions of cells in the domain ½xmin ,xmax , where xmin is the
minimum and xmax is the maximum crystal length of interest. For 4. Numerical test problems
each i ¼ 1; 2, . . . ,N, Dx represents the cell width, the points xi refer
to the cell centers, and the points xi 7 ð1=2Þ denotes the cell In this section the simulation results of a MSMPR preferential
boundaries. The integration of Eq. (30) over the cell crystallizer are presented for different operating strategies.
Oi ¼ ½xjð1=2Þ ,xj þ ð1=2Þ  yields the following cell centered semi- Firstly, the crystallizer is operated without fines dissolution unit
discrete finite volume schemes for IðkÞ ¼ ðGðkÞ F ðkÞ Þi 7 ð1=2Þ and is either seeded continuously or periodically with seeds of the
i 7 ð1=2Þ
Z Z preferred enantiomer. In this case the last terms on the right hand
@F ðkÞ ðt,xÞ side of Eqs. (1) and (10) are neglected. Secondly, the crystallizer is
dx ¼ ðIðkÞ ðkÞ
i þ ð1=2Þ ðtÞIið1=2Þ ðtÞÞ þ Q ðkÞ ðt,xÞ dx: ð32Þ
Oi @t Oi equipped with a fines dissolution unit and is continuously seeded
Let F ðkÞ and Q ðkÞ denote the average values of the number density with seeds of the preferred enantiomer. In this process, it is
i i
and source term in each cell Oi , i.e. assumed that all fines are dissolved in the pipe before re-entering
Z Z the solution back to the crystallizer.
1 1 In order to judge the quality of the process some goal functions
F ðkÞ
i ðtÞ ¼ F ðkÞ
ðt,xÞ dx, Q ðkÞ
i ðtÞ ¼ Q ðkÞ ðt,xÞ dx: ð33Þ
Dx Oi Dx Oi can be used such as product purity, productivity, yield and mean
Then, Eq. (32) can be written as crystal size of the preferred enantiomer. These goal functions give
ðkÞ ðkÞ
detailed information about the success and potential of contin-
ðkÞ Ii þ ð1=2Þ ðtÞIið1=2Þ ðtÞ
dF i ðtÞ uous preferential crystallization.
¼ þ Q ðkÞ
i
ðtÞ, i ¼ 1; 2, . . . ,N, k A fp,cg:
dt Dx Purity: It is typically of key importance in enantioselective
ð34Þ crystallization
Here, N denotes the total number of mesh elements in the _ ðpÞ
m S,out
computational domain. The accuracy of finite volume discretiza- PuðpÞ ¼ , where
_ ðpÞ
m _ ðcÞ
S,out þ m S,out
tion is mainly determined by the way in which the cell face fluxes Z 1
are computed. Assuming that the flow is in positive x-direction, a _ ðkÞ x3 F ðkÞ ðt,xÞ dx, k A fp,cg:
m S,out ¼ kv rS ð39Þ
first order accurate upwind scheme is obtained by taking the 0
backward differences It is a ratio of the mass of preferred enantiomer to the sum of the
IiðkÞ ðtÞ ¼ ðGðkÞ ðtÞF ðkÞ ðtÞÞi , IðkÞ ðtÞ ¼ ðGðkÞ ðtÞF ðkÞ ðtÞÞi1 : ð35Þ masses of preferred and counter enantiomers.
þ ð1=2Þ ið1=2Þ
Productivity: The productivity of a continuous crystallization
To get high order accuracy of the scheme, one has to use better process can be defined as the mass flow of solid produced per unit
approximation of the cell interface fluxes. According to the high size. It can be described by the following equation:
resolution finite volume scheme of Koren the flux at the right
boundary xi þ ð1=2Þ is approximated as PrðpÞ ¼ mass flow of solid produced ðpreferredÞ per unit volume

1 _ ðpÞ
m _ ðpÞ
S,out m seeds
IiðkÞ ðkÞ
þ ð1=2Þ ðtÞ ¼ Ii ðtÞ þ fðrðkÞ ðkÞ ðkÞ
i þ ð1=2Þ ðtÞÞðIi ðtÞIi1 ðtÞÞ: ð36Þ ¼
VL
: ð40Þ
2
Similarly, one can approximate the flux at left cell boundary. The Yield: The yield can be defined as the ratio of the mass flow of
flux limiting function f according to Koren (1993) is defined as solid target particles (p) produced over mass flow of this enan-
0 0 0 111 tiomer in the feed solution introduced
  1 2r iðkÞ
þ ð1=2Þ
ðtÞ
f riðkÞ ðtÞ ¼ max@0,min@2r iðkÞ ðtÞ, min@ þ ,2AAA:
þ ð1=2Þ þ ð1=2Þ 3 3 _ ðpÞ
m _ ðpÞ
S,out m seeds
Y ðpÞ ¼ : ð41Þ
ð37Þ _ ðpÞ
m L,in:c

Here, r ðkÞ
i þ ð1=2Þ ðtÞ is the so-called upwind ratio of the consecutive Mean crystal size: Besides the purity, the crystal size also plays an
flux gradients important role in most industrial applications. For that reason, the
mean crystal size of the preferred enantiomer is instructive. It is
IðkÞ ðkÞ
i þ 1 ðtÞIi ðtÞ þ e
r ðkÞ ðtÞ ¼ , ð38Þ defined as
i þ ð1=2Þ
IðkÞ
i
ðtÞIðkÞ
i1
ðtÞ þ e
mðpÞ
1
where e is a small number to avoid division by zero. This scheme x ðpÞ ¼ , ð42Þ
mðpÞ
0
is not applicable up to the boundary cells because it needs values
of the cell nodes which are not present. To overcome this where mðpÞ
1 is the total length of crystals and mðpÞ
0 is the total
problem, the first order approximation of the fluxes was used at number of crystals (c.f. Eq. (7)).
the interfaces of the first two cells on the left-boundary and at the In the following three test problems, the minimum and
interfaces of the last cell on the right-boundary. At the remaining maximum crystal sizes of interest are taken as xmin ¼ 1:0 
interior cell interfaces, the high order flux approximation of Eq. 1010 m and xmax ¼ 1:0  102 m, respectively. The corresponding
(36) was used. It should be noted that, the first order approxima- computational domain is subdivided into 200 grid points. The
tion of the fluxes in the boundary cells does not effect the overall remaining parameters are given in Table 1. The physicochemical
accuracy of the proposed high resolution scheme. parameters correspond to the enantiomers of the amino acid
The resulting system of ordinary differential equations (ODEs) threonine and the solvent water, see Elsner et al. (2005, 2011).
in Eq. (34) together with Eqs. (36)– (38) can be solved by a The temperature of the crystallizer was assumed to be constant at
standard ODE-solver. In this study a Runge–Kutta method of 33 1C.
10 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17

The mean residence time of the liquid phase remained fixed at Figs. 2 and 3 also justify and support the above discussion.
112.41 min, while the residence time of the solid phase was They show a comparison of goal functions for different mass flows
considered as either tS ¼ tL or tS ¼ 2tL . In actual operation, it may of seeds and different residence times under isothermal condi-
be difficult to decouple the residence times of the liquid and solid tion. It can bee seen that supersaturation decreases with increas-
phases. Thus, it is most likely to set tS ¼ tL . However, larger ing the mass flow of seeds. As explained above, an increase in the
residence times of the solid phase compared to the liquid phase mass flow (amount) of seeds enhances the consumption rate of
could eventually be realized by installing a filter for the solid the solute mass (in the solution). Moreover, the time needed to
particles at the outlet, thus, allowing a longer residence tS of the achieve steady state becomes shorter by increasing the mass flow
solid phase. of seeds which is also clear from Table 2. The left and right hand
In the numerical calculations, the steady state was identified side figures show the same behavior but an increase in the
when the relative deviation error in the supersaturation was residence time of the solid phase reduces the steady state value
below 5  107 . of the supersaturation. As the residence time increases, the
crystals remain in the crystallizer for a longer time and the
surface area of crystals increases, resulting in a high depletion
4.1. Test problem 1: continuous seeding without fines dissolution of the supersaturation. It can be further observed that purity
increases by increasing the mass flow of seeds. The mean crystal
In this problem, the crystallizer is continuously seeded with size, productivity, and yield plots also agree with the data in the
the preferred enantiomer. The corresponding seeds size distribu- tables. Initially, overshoots can be observed in the results when
tion is given as the ratio of residence times is 0.3. It means that, the productivity
  2 ! increases for a certain time and then bounces back. The reason
a1 1 x
F ðpÞ
seeds
¼ exp 0:5 ln , ð43Þ can be the high initial supersaturation which is depleted with
xAa a3 a2
time before reaching to the steady state condition. As the yield
where the values of constants a1, a2 and a3 are given in Table 1. depends on the productivity, it shows a similar behavior. At the
Moreover, the normalization factor is given as startup, the values of productivity and yield remain negative for a
certain time and then become positive. During that period, the
kv rS crystallizer is seeded with the seeds of preferred enantiomer
Aa ¼ mðpÞ
3 : ð44Þ
_ ðpÞ
m seeds
 tS which stay for a certain time before being taken out as a product.
In other words, the productivity does not starts instantaneously
The initial masses of the preferred and counter enantiomers in the due to the residence time of the solid phase. Moreover, in Fig. 2
crystallizer are taken as mðpÞ ðcÞ
0 ¼ m0 ¼ 0:0478 kg. purity drops temporarily before reaching the steady state due to
The simulation results corresponding to steady state condi- the stronger relative impact (negative) of primary nucleation of
tions are displayed in Table 2 for different mass flow of seeds. It counter enantiomer. This effect reduces gradually when steady
can be observed that an increase in the ratio of solid to the liquid state condition is approaching.
residence times, tS =tL , diminishes the purity of the preferred Finally, Fig. 4 presents the percentage relative errors in the
enantiomer due to the production of increasing amount of mass balances at steady state condition and the crystal size
counter enantiomer as an impurity, while the productivity, yield, distributions for different mass flows of seeds. It can be observed
and mean crystal size of the preferred enantiomer are improved. that the error in the mass balances is moderately increasing until
As tS increases, the crystals of the preferred enantiomer have m_ ðpÞ  0:04 g=min and then bounces back. Moreover, bigger
seeds
more time to grow and, hence, the overall productivity, yield and crystals are produced for lower mass flow rate of seeds which
the mean crystal size are enhanced. The productivity and yield also justifies the results of Table 2.
along with the investment of seeds are increased by increasing In summary, the purity diminishes by increasing the ratio
the mass flow of seeds. It is clear that with large amount of between residence times of the solid to the liquid phases, while
seeding more crystals are produced, providing the potential for the productivity, yield, and mean crystal size are enhanced. An
high productivity. However, the mean crystal size reduces on increase in this ratio reduces the mass flow of seeds and increases
increasing the mass of seeds. The reason is obvious, by increasing the time needed to achieve steady state. Moreover, the super-
the mass flow (amount) of seeds the existing supersaturation is saturation and mean crystal size are reduced with increasing the
consumed by large number of crystals and, therefore, the mean mass flow of seeds, while productivity, yield and purity are
crystal size becomes smaller. However, by investing less seeds, improved. The errors in mass balances are below 0.15% for a
the existing supersaturation is utilized by only a fewer crystals, wide range of mass flow of seeds.
leading to larger mean size of the crystals. Finally, the steady-
state is reached after a longer time when the ratio tS =tL is larger. 4.2. Test problem 2: periodic seeding without fines dissolution
However, the time needed for steady state reduces by increasing
the mass flow of seeds. In this test problem, continuous crystallization along with
periodic seeding of preferred enantiomer is analyzed. The model
equations are exactly the same as for the continuous seeding case
Table 2 in test problem 1. In Table 3, ton represents the time at which
Problem 1: values of goal functions for at tL ¼ 112:41 min. seeding is switched on and toff denotes the period of time at which
seeding is switched off. The seed flow rate can be varied in
tS =tL tstd _ ðpÞ
m ðg=minÞ PuðpÞ Pr ðpÞ ðkg=min m3 Þ Y ðpÞ x ðpÞ contrast to the continuous seeding. In this table, a comparison of
seed
(min) (%) (%) (mm)
simulation results for continuous and periodic seeding strategies
1.0 646 0.0089 99.74 0.037 3.91 1.31 are presented. A schematic diagram of the periodic seeding is
2.0 1411 0.0044 98.95 0.043 4.54 1.54 shown in Fig. 5. In that figure, the shaded areas represent the
1.0 676 0.0356 99.92 0.060 6.29 1.13 seeding times (ton) and the non-shaded areas depict the periods
2.0 1315 0.0178 99.54 0.062 6.60 1.27 when seeding is switched off (toff).
1.0 721 0.0623 99.94 0.067 7.08 1.08
2.0 1318 0.0311 99.68 0.069 7.29 1.18
In Table 3, the simulation results are presented for different
ratios of the residence times. The increase in tS reduces the purity
 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17 11

Fig. 2. Problem 1: comparison of the results for different masses of seeds: left tS =tL ¼ 1:0, right: tS =tL ¼ 2:0.

of the preferred enantiomer in the periodic case. As residence
time increases, more nuclei of the counter enantiomer are
produced as an impurity of the product and, thus, the purity is
lower. It can also be observed that the purity has decreasing
behavior for increasing toff. On the other hand, as tS increases the
crystals of the preferred enantiomer get more time to grow and,
thus, the overall productivity and yield enhances. Additionally,
the productivity and yield are decreased by increasing toff. It is
obvious that when seeding is switched off for a longer time, the
mean crystal size will increase by increasing the residence time tS
due to longer stay of seeds in the crystallizer and because of
utilizing the existing supersaturation by fewer crystals. The mean
crystal size also increases by increasing toff for all considered
residence times due to the utilization of existing supersaturation
12 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17

Fig. 3. Problem 1: comparison of the results for different masses of seeds: left tS =tL ¼ 1:0, right: tS =tL ¼ 2:0.

Fig. 4. Problem 1: errors in mass balances and CSDs for different mass flows of seeds.

by fewer crystals. Table 3 also shows that the feed flow rate of case in Table 3 shows that continuous seeding gives better purity,
seeds decreases by increasing the residence time. A comparison of productivity and yield, while the mean crystal size is smaller and
the continuous seeding case in Table 2 with the current periodic investment (mass flow of seeds) remains the same in all
 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17 13

Table 3
Problem 2: a comparison of continuous (C) and periodic (P) seeding strategies for mðpÞ
seeds
¼ 4 g and tL ¼ 112:4 min.

tS =tL Seeding type t std ðminÞ ton (min) toff (min) _ ðpÞ
m ðg=minÞ PuðpÞ ð%Þ Pr ðpÞ ðkg=min m3 Þ Y ðpÞ ð%Þ x ðpÞ ðmmÞ
seed

1.0 P 670 10 5 0.0244 99.87 0.051 5.50 1.18

1.0 P 670 10 10 0.0189 99.84 0.045 4.67 1.23
1.0 P 670 10 20 0.0126 99.79 0.040 4.26 1.27
1.0 C 676 – – 0.0356 99.92 0.060 6.29 1.13
2.0 P 1320 10 5 0.0125 99.41 0.044 4.67 1.34
2.0 P 1321 10 10 0.0089 99.30 0.033 3.47 1.40
2.0 P 1321 10 20 0.0051 99.10 0.021 2.25 1.46
2.0 C 1315 – – 0.0178 99.54 0.062 6.60 1.27

reaching the steady state due to the stronger relative impact

(negative) of primary nucleation of counter enantiomer. This
effect reduces gradually when steady state condition is
mass flow of seeds [g/min]

approaching.
In summary, the periodic seeding reduces the productivity and
yield, while improves the mean crystal size. The purity is slightly
seeding on reduced with periodic seeding (approx. 0.1%). Moreover, the
investment of seed crystals is reduced.

4.3. Test problem 3: continuous seeding with fines dissolution

In this problem, a continuous seeding with fines dissolution is

seeding off
time [min]
Fig. 5. Schematic diagram for periodic seeding.
considered residence times. On the other hand, the periodic
seeding gives larger crystals at low investment (feed flow rate)
of seeds per residence time. Finally, the time needed to achieve
steady-state becomes longer on increasing tS , while keeping the
liquid residence time tL fixed. This behavior is exactly similar to
the continuous seeding case.
Figs. 6 and 7 show the goal functions at different periods of
seeding and for different ratios of residence times. It can be
observed that supersaturation decreases with reducing time
period of seeding, because a decrease in toff of seeds speeds up
the consumption of solute mass. Both figures show the same
trends. However, a comparison of the plots in Figs. 6 and 7 shows
that, an increase in the residence time of solid phase gives a
decrease in the steady state supersaturation. The reason is clear,
as the solid residence time increases the crystals stay in the
crystallizer for a longer time and, thus, the surface area of crystals
increases by consuming more supersaturation. Further, the
steady-state time delays as the residence time of the solid phase
increases. Oscillations are also visible which indicate the condi-
tions of periodic seeding. These oscillations are less prominent in
the right hand side plots of Figs. 6 and 7 due to an increase in the
residence time of the solid phase. The purity plots show that it
decreases by increasing toff. Initially, the decrease in purity is due
to increasing amount of counter enantiomer but after a certain
time the amount of preferred enantiomer starts increasing which
improves the purity. It can also be observed that an increase in toff
produces large size crystals which is also clear from Table 3. The
behavior of productivity is also similar to that in Table 3. It is
evident that larger toff gives lower productivity. Yield is directly
related to the productivity, thus, it has similar behavior to the
productivity. Moreover, in Fig. 6 purity drops temporarily before
investigated. Fines particles below the critical size xcrit ¼ 6:0 
104 m are taken out from the crystallizer along with the solution
into the dissolution unit (the pipe). After a certain time delay the
particle-free solution re-enters to the crystallizer. The fines
dissolution unit is usually equipped with a heat exchanger for
dissolving fines followed by a heat sink which brings back the
solution temperature to that of the crystallizer. A complete model
equation (1) is employed in this problem. The selection (death)
function is taken as
(
0:6 if z r xcrit ,
hðxÞ ¼ ð45Þ
0 otherwise:
In this case, the simulation results with continuous seeding
and fines dissolution are analyzed. Table 4 shows a comparison of
fines dissolution (yes) and without fines dissolution (no) by
considering different ratios of residence times and fixed mass
flow of seeds. It can be seen that the purity is higher in the case of
fines dissolution compared to that without fines dissolution, as
the fines dissolution diminishes the number of counter enantio-
mer crystals in the crystallizer. From Table 4, it can be realized
that the productivity and yield are slightly lower in the case of
fines dissolution for both ratios of residence times. The fines
dissolution enhances the supersaturation of the solution which in
turn increases the growth rate, thus, improves the mean
crystal size.
5. Conclusion
A dynamic mathematical model was derived for simulating
continuously operated ideally mixed MSMPR crystallizer applied
for continuous preferential enantioselective crystallization. The
model incorporates the removal and subsequent dissolution of
fines in the outer fines dissolution loop connected to the crystal-
lizer. The effects of different seeding strategies and residence time
characteristics were analyzed. The effect due to varying the ratio
of residence times of solid and liquid phases was also analyzed.
A semi-discrete high resolution finite volume schemes was used
to solve the given model. The resulting ODE system was solved by
14 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17

1.1 1.1
ton=10, toff=5 ton=10, toff=5
t =10, t =10 t =10, t =10
1.09 on off 1.09 on off
ton=10, toff=20 ton=10, toff=20

1.08 1.08
supersaturation [−]

supersaturation [−]
1.07 1.07

1.06 1.06

1.05 1.05

1.04 1.04

1.03 1.03
0 500 1000 1500 2000 0 500 1000 1500 2000 2500 3000 3500 4000
time [min] time [min]
100 100
t =10, t =5
on off
99.9 99.9 ton=10, toff=10
99.8 99.8 ton=10, toff=20

99.7 99.7
ton=10, toff=5
99.6 99.6
purity [%]

purity [%]

ton=10, toff=10
99.5 ton=10, toff=20 99.5

99.4 99.4

99.3 99.3

99.2 99.2

99.1 99.1

99 99
0 500 1000 1500 2000 0 500 1000 1500 2000 2500 3000 3500 4000
time [min] time [min]
0.06 0.05

0.045
0.05
0.04
productivity [kg/min m3]

productivity [kg/min m3]

0.035
0.04
0.03

0.03 0.025
ton=10, toff=5
ton=10, toff=10 0.02
0.02 t =10, t =20
on off 0.015 t =10, t =5
on off
0.01 ton=10, toff=10
0.01
t =10, t =20
0.005 on off

0 0
0 500 1000 1500 2000 0 500 1000 1500 2000 2500 3000 3500 4000
time [min] time [min]

Fig. 6. Problem 2: comparison of the results for different periods of seeding: left tS =tL ¼ 1:0, right: tS =tL ¼ 2:0.

a Runge–Kutta method of order four. Several case studies were
carried out. It was observed that supersaturation and mean
crystal were reduced on increasing the mass flow of seeds, while
productivity, yield and purity were improved. The errors in mass
balances were found below 0.15% for a wide range of mass flow of
seeds. The periodic seeding slightly diminished the purity, pro-
ductivity and yield but improved the mean crystal size and the
investment (mass flow) of seeds was also smaller. The fines
dissolution improved the mean crystal size and purity but slightly
reduced the productivity and yield. These results could be used
to find the optimum operating conditions for improving the
product quality and for reducing the operational cost of contin-
uous preferential crystallization. For the case study carried out
(resolution of an amino acid enantiomers) a production of
 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17 15

6 5

4.5
5
4

3.5
4
3

yield [%]
yield [%]

3 ton=10, toff=5 2.5

ton=10, toff=10 2
2 ton=10, toff=20
1.5 ton=10, toff=5

1 ton=10, toff=10
1
ton=10, toff=20
0.5

0 0
0 500 1000 1500 2000 0 500 1000 1500 2000 2500 3000 3500 4000
time [min] time [min]
1.35 1.6

1.3
1.5

1.25
mean crystal size [mm]

mean crystal size [mm] 1.4

1.2

1.3
1.15

1.1 1.2 ton=10, toff=5

ton=10, toff=5 ton=10, toff=10
1.05 ton=10, toff=10 ton=10, toff=20
1.1
ton=10, toff=20
1
1
0.95
0 500 1000 1500 2000 0 500 1000 1500 2000 2500 3000 3500 4000
time [min] time [min]

Fig. 7. Problem 2: comparison of the results for different periods of seeding: left tS =tL ¼ 1:0, right: tS =tL ¼ 2:0.

Table 4
Problem 3: results with fines dissolution (yes) and without fines dissolution (no).

tS =tL FD t std ðminÞ _ ðpÞ

m ðg=minÞ PuðpÞ ð%Þ Pr ðpÞ ðkg=minÞ Y ðpÞ ð%Þ x ðpÞ ðmmÞ
seed

1.0 No 676 0.0356 99.92 0.060 6.29 1.13

1.0 Yes 648 0.0356 99.99 0.060 6.25 1.15
2.0 No 1315 0.0178 99.54 0.062 6.60 1.27
2.0 Yes 1303 0.0178 99.97 0.062 6.56 1.29

100 kg=ðday m3 Þ appears to be feasible. Altogether, the process aeq temperature dependent parameter for solubility
appears to possess large potential and deserves practical realiza- (dimensionless)
tion which is currently in progress. aprim lumped constant of primary nucleation in Eq. (25)
(dimensionless)
beq temperature dependent parameter for solubility
Nomenclature (dimensionless)
bsec exponent (secondary nucleation) (dimensionless)
a1 fitted parameter for log-normal distribution B0 overall nucleation rate at negligible particle size
function (dimensionless) x ¼0 (s  1)
a2 fitted parameter for log-normal distribution B0,prim nucleation rate for primary heterogeneous
function (dimensionless) mechanism (s  1)
a3 fitted parameter for log-normal distribution B0,sec nucleation rate for secondary mechanism (s  1)
function (dimensionless)
C eq (mass) solubility concentration at temperature T eq
aASL parameter for size dependency of crystal growth
(equilibrium) (kg/m3)
(m  1)
16 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17

dASL parameter for size dependency of crystal growth tL mean liquid phase residence time (min)
(dimensionless) tS mean solid residence time (min)
EA,b activation energy for secondary nucleation (J/mol) tpipe residence time in the pipe (min)
EA,g activation energy for crystal growth (J/mol)
F number density (m  1) Subscripts
F seeds number density function for seed crystals (m  1)
g exponent in crystal growth term (dimensionless) in inflow stream to crystallizer
G crystal growth rate (m/s) c crystallizer (dimensionless)
h selection (death) function for fines dissolution eff effective (dimensionless)
(dimensionless) out outlet stream from the crystallizer
kb,prim lumped constant for nucleation std (subscript) short notation for steady state
(dimensionless)
ððkg m3 Þ7=3 =min KÞ
eq equilibrium (dimensionless)
kb,sec rate constant for nucleation rate (m  3/min)
L liquid (dimensionless)
kb,sec,0 pre-exponent factor for nucleation (m  3/min)
prim primary (dimensionless)
kg,eff effective crystal growth constant (m/min)
sec secondary (dimensionless)
kg,eff,0 pre-exponent factor for growth (m/min)
kv volume shape factor (dimensionless) Superscripts
K1 (fitted) constant for density (m3/kg)
k enantiomer k A fp,cg (p: preferred, c: counter)
K2 (fitted) constant for density (m3/(kg K2))
(dimensionless)
K3 (fitted) constant for density (m3/kg)
KT (fitted) parameter for temperature dependence (K) Abbreviations
KW (fitted) parameter for mass fraction dependence
(dimensionless) CSD crystal size distribution
m mass (kg) FD fines dissolution
mL solute mass in the liquid (kg) ODE ordinary differential equation
_ seeds
m mass flow of seeds (kg/min) PBE population balance equation
_ L,in
m incoming mass flow rate to the crystallizer (kg/min) PBM population balance model
_ L,out
m outgoing mass flow rate from the crystallizer (kg/min)
_ L,in,c
m incoming mass flow rate of fresh solution (kg/min)
_ L,out,c
m outgoing stream of exhausted solution (kg/min)
_ L,in,pipe
m incoming stream from the fines dissolution unit
(kg/min) Acknowledgments
m_ L,out,pipe outgoing stream to the fines dissolution unit (kg/min)
mm minmod limiter function This work was partially supported by Higher Education Com-
nsec exponent in secondary nucleation (dimensionless) mission (HEC) of Pakistan. The third author is thankful to the
N number of grid points (dimensionless) University of Engineering and Technology, Peshawar (Khyber
3 Pakhtunkhwa), Pakistan for study leave.
Pr ðpÞ productivity per unit volume ðkg=ðmin min ÞÞ
Pu ðpÞ purity (dimensionless)
R gas constant (J/(mol K)) References
S supersaturation degree (dimensionless)
t time (min) Aamir, E., Nagy, Z.K., Rielly, C.D., 2010. Optimal seed recipe design for crystal size
ton on time of periodic seeding (min) distribution control for batch cooling crystallisation processes. Chem. Eng. Sci.
toff off time of periodic seeding (min) 65, 3602–3614.
Abegg, C.F., Stevens, J.D., Larson, M.A., 1968. Crystal size distributions in contin-
t std is the time at which steady state is achieved (min) uous crystallizers when growth rate is size dependent. AIChE J. 14, 118–122.
T temperature(K) Alvarez-Rodrigo, A., Lorenz, H., Seidel-Morgenstern, A., 2004. On-line monitoring
T eq equilibrium temperature (min) of preferential crystallization of enantiomers. Chirality 16, 499–508.
Chung, S.H., Ma, D.L., Braatz, R.D., 1999. Optimal seeding in batch crystallisation.
VL volume of the liquid in the crystallizer (m3) Can. J. Chem. Eng. 77, 590–595.
V_ L volumetric flow rate (m3/min) Collet, A., 1999. Separation and purification of enantiomers by crystallization.
Enantiomer 4, 157–172.
w mass fraction (dimensionless)
Coquerel, G., 2007. Preferential crystallization. In: Sakai, K., Hirayama, N., Tamura,
weq mass fraction at equilibrium temperature R. (Eds.), Novel Optical Resolution Technologies. Topics in Current Chemistry,
(dimensionless) vol. 269; 2007, pp. 1–51.
x particle size (m) Czapla, F., Haida, H., Elsner, M.P., Lorenz, H., Seidel-Morgenstern, A., 2009.
Parameterization of population balance models for polythermal auto seeded
x ðpÞ mean crystal size (m) preferential crystallization of enantiomers. Chem. Eng. Sci. 64, 753–763.
x0 ,xmax minimum and maximum crystal sizes (m) Elsner, M.P., Fernández Menéndez, D., Alonso Muslera, E., Seidel-Morgenstern, A.,
2005. Experimental study and simplified mathematical description of prefer-
Y ðpÞ yield (dimensionless) ential crystallization. Chirality 17, 183–195.
Elsner, M.P., Ziomek, G., Seidel-Morgenstern, A., 2011. Simultaneous preferential
Greek symbols crystallization in a coupled batch operation mode. Part II. Experimental study
and model refinement. Chem. Eng. Sci. 66, 1269–1284.
IðkÞ numerical flux for the k-th enantiomer Gunawan, R., Fusman, I., Braatz, R.D., 2004. High resolution algorithms for multi-
(dimensionless) dimensional population balance equations. AIChE J. 50, 2738–2749.
Hermanto, M.W., Braatz, R.D., Chiu, M.-S., 2008. High-order simulation of poly-
mi i-th moment of the CSD
morphic crystallization using weighted essentially nonoscillatory methods.
rL density of liquid phase (kg/m3) AIChE J. 55, 122–131.
rL,0 density of fresh liquid at t¼ 0 (kg/m3) Hulburt, H.M., Katz, S., 1964. Some problems in particle technology. Chem. Eng.
Sci. 19, 555–574.
rS density of solid phase (kg/m3) Jacques, J., Collet, A., Wilen, S.H., 1981. Enantiomers, Racemates and Resolutions.
ro density of water (kg/m3) Wiley, New York.
 S. Qamar et al. / Chemical Engineering Science 71 (2012) 5–17
Koren, B., 1993. A robust upwind discretization method for advection, diffusion
and source terms. In: Vreugdenhil, C.B., Koren, B. (Eds.), Numerical Methods
for Advection–Diffusion Problems. Notes on Numerical Fluid Mechanics, vol.
45. , Vieweg Verlag, Braunschweig, pp. 117–138.
Kumar, S., Ramkrishna, D., 1996a. On the solution of population balance equations
by discretization. I. A fixed pivot technique. Chem. Eng. Sci. 51, 1311–1332.
Kumar, S., Ramkrishna, D., 1996b. On the solution of population balance equations
by discretization. II. A moving pivot technique. Chem. Eng. Sci. 51, 1333–1342.
LeVeque, R.J., 2002. Finite Volume Methods for Hyperbolic Problems. Cambridge
University Press, New York.
Lim, Y.I., Lann, J-M.L., Meyer, X.M., Joulia, X., Lee, G., Yoon, E.S., 2002. On the
solution of population balance equation (PBE) with accurate front
tracking method in practical crystallization processes. Chem. Eng. Sci. 57,
3715–3732.
Mersmann, A., 2001. Crystallization Technology Handbook, second ed. Marcel
Dekker, New York.
Motz, S., Mitrović, A., Gilles, E.-D., Vollmer, U., Raish, J., 2003. Modeling, simulation
and stabilizing-control of an oscillating continuous crystallizer with fines
dissolution. Chem. Eng. Sci. 58, 3473–3488.
Nývlt, J., Söhnel, O., Matuchová, M., Broul, M., 1985. The Kinetics of Industrial
Crystallization. Chemical Engineering Monographs, vol. 19. Elsevier,
Amsterdam.
17
Qamar, S., Elsner, M.P., Angelov, I., Warnecke, G., Seidel-Morgenstern, A., 2006. A
comparative study of high resolution schemes for solving population balances
in crystallization. Comput. Chem. Eng. 30, 1119–1131.
Qamar, S., Ashfaq, A., Angelov, I., Elsner, M.P., Warnecke, G., Seidel-Morgenstern,
A., 2008. Numerical solutions of population balance models in preferential
crystallization. Chem. Eng. Sci. 63, 1342–1352.
Qamar, S., Mukhtar, S., Seidel-Morgenstern, A., 2009. An efficient numerical
technique for solving one-dimensional batch crystallization models with
size-dependent growth rates. Chem. Eng. Sci. 64, 3659–3667.
Ramkrishna, D., 2000. Population Balances: Theory and Applications to Particulate
Systems in Engineering. Academic Press, San Diego, USA.
Rawlings, J.B., Witkowski, W.R., Eaton, J.W., 1992. Modelling and control of
crystallizers. Powder Technology 69, 3–9.
Rhee, H.-K., Aris, R., Amundson, N.R., 1986. First Order Partial Differential
Equations, vol. 1. Prentice-Hall, Englewood Cliffs, NJ, USA.
Smith, M., Matsoukas, T., 1998. Constant-number Monte Carlo simulation of
population balances. Chem. Eng. Sci. 53, 1777–1786.
Tandon, P., Rosner, D.E., 1999. Monte Carlo simulation of particle aggregation and
simultaneous restructuring. Colloid Interface Sci. 213, 273–286.
Woo, X.Y., Tan, R.B.H., Braatz, R.D., 2011. Precise tailoring of the crystal size
distribution by controlled growth and continuous seeding from impinging jet
crystallizers. Cryst. Eng. Commun. 13, 2006–2014.