Chemical Engineering Science 66 (2011) 848–858

Contents lists available at ScienceDirect

Chemical Engineering Science

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

Study of sol–gel transition in calcium alginate system by

population balance model
Yinyan Zhao a,n, Fengping Hu b, Jeffrey J. Evans b, Michael T. Harris a
a
School of Chemical Engineering, Purdue University, 480 Stadium Mall Drive, West Lafayette, IN 47907-2100, USA
b
Department of Electrical and Computer Engineering Technology, Purdue University, West Lafayette, IN 47907-2021, USA

a r t i c l e i n f o a b s t r a c t

Article history: The growth of calcium alginate (Ca-alginate) clusters and its sol–gel transition phenomena were
Received 15 September 2010 simulated with a population balance model. The classical statistical concept proposed by Flory–
Received in revised form Stockmayer has been adopted and incorporated by the factor of cyclic structures formation into the
4 November 2010
model. Ca-alginate gelation was described as a random cross-linking of multifunctional monochains. The
Accepted 11 November 2010
resulting discretized kinetic equations were solved numerically by parallel computation. After demon-
Available online 23 November 2010
strating the accuracy of this model, the effect of the cyclic structure, the reaction rate constants and
Keywords: the physical properties of the reactants (i.e., alginate and gelling agent of Ca2 + ions) were examined. The
Calcium alginate present work clarifies the mechanisms for forming ionotropic Ca-alginate gels and presents the
Polymers
concentration ranges where the transition between the sol state and the gel state occurs. The model
Cluster
allows for the estimation of the relative magnitude of stepwise reaction rate constants and the prediction
Gels
Population balance of the sol–gel transition boundary. Finally, a reasonable agreement was obtained between the
Phase change experimental results and model predictions.
& 2010 Elsevier Ltd. All rights reserved.

1. Introduction Ca-alginate gels are physical gels, which are different from the
majority of synthetic chemical gels produced by multifunctional
Alginate is a polysaccharide, widely present in the cell walls of precursors. Chemical gels are considered as networks or as being
brown algae as well as some bacteria (Haug et al., 1974). Chemi- true macromolecules, in which covalent bonds are formed. Con-
cally, alginates are linear unbranched copolymers composed of sequently, they have an infinite molecular weight and possess an
(1-4) linked b-D-mannuronate residues (M-block) and its C-5 infinite lifetime (Osada and Khokhlov, 2002), whereas the physical
epimer a-L-guluronate residues (G-block). Within the alginate, the gel systems have non-covalent interactions and their junction
M and G monomers are sequentially assembled in either repeating zones hold the chains together through physical forces, ionic
(MM or GG) or alternating (MG) blocks with varying proportions of interactions, and hydrogen or van der Walls bindings. Physical
monomers and sequence of units (see Fig. 1). In the presence of gels share much in common with their covalent analogues if the
divalent cations, such as commonly used Ca2 + ions, the functional lifetime of the cross-links is sufficiently long (Fuchs et al., 1997).
carboxyl groups of alginate residues cooperatively interact with The system is a liquid before cross-linking starts and remains
Ca2 + ions to form cross-linked hydrogels. It has been known that liquid-like until the viscosity becomes infinite at the gel point. But,
the alginate monomer composition, the M/G ratio, has a major because the cross-links of physical gels are not permanent, they
impact on the gelation with the metal ions and the gel structure may show thermo-reversibility and tend to have rheological creep
(Hassan et al., 1988a, 1988b; Gu et al., 2004). In general, alginates behavior at long times, which makes it difficult to define the gel
with a low M/G ratio form dense and brittle gels, whereas alginates point rigorously by experiments. Traditional rheological measure-
with a high M/G ratio give elastic gels. This is presumably due ments, such as Static-light-scattering, require dilution of the
to the influence of the different chain stiffnesses of the residues. sample, which is not feasible for reversibly associating systems
The backbone conformation of G-blocks was suggested to be of the since dissolution of the clusters will occur on dilution. Ca-alginate,
‘‘buckled ribbon’’ form and is stiffer than that of M-blocks which in particular, is a physical biopolymer having ion-mediated junc-
has a ‘‘flat-ribbon’’ form (Matsumoto et al., 1992; Draget et al., tion zones. Although many experimental studies have been
1994; Martinsen et al., 1989). devoted to observe the gelation process of this system, most
findings regarding the gelation time and the conditions to initiate
gel do not agree with each other (Wang et al., 1993; Matsumoto and
n
Corresponding author. Mashiko, 1990; LeRoux et al., 1999). The discrepancies are attrib-
E-mail address: yinyan.purdue@gmail.com (Y. Zhao). uted to the nature of this type of non-permanent gel networks and

0009-2509/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2010.11.025
 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858 849

2 2
(a)

2 2

2 2

2 2

(b)

Fig. 1. Schematic drawings of (a) guluronate chains and (b) mannuronate chains.

the precursor alginate samples used in the experiments, which boundary under the homogeneous reaction conditions. Detailed
might have different M/G ratios, unit sequences and molecular understanding of the mechanisms of gelation kinetics of the Ca-
weight of subchains. Moreover, the degree of spatial inhomogene- alginate system would not only elucidate the characteristics of
ities depends strongly on the conditions at the preparation of the ionotropic physical gels, and resolve the dispute coming from
network and the internal distribution of the cross-linker, subse- different experimental studies, but would also benefit in the
quently, affecting the sol–gel transition boundary (Skjakbraek precise design of the Ca-alginate gels with a homogeneous
et al., 1989). Homogeneous Ca-alginate gelation is difficult to structure. A population balance model was introduced where the
achieve because Ca-alginate gels prepared by commonly used cross-linking process is modeled as a random binding of multi-
methods using an external Ca2 + source can be visualized as having functional monochains. This model recasts the polymerization
a strongly cross-linked surface and a less densely cross-linked kinetics simply as a statistical scheme and allows us to simulate
center due to the retardation of the diffusion of free Ca2 + ions in the network growth. By using this model, the influence of the
regions near the interface where the gel has formed (Quong et al., physical properties of reactants on the gel time and the sol–gel
1998). Several methods have been used to form homogeneous gel transition boundary was investigated, and the step-reaction rate
structures, such as the in-situ release of internal calcium salt (i.e., was predicted. The present work clarifies the mechanisms of
CaCO3, Ca-EDTA) by decreasing the pH to avoid the concentration forming ionotropic Ca-alginate gels and defines the transition
gradient of Ca2 + in the gel (Poncelet et al., 1992; Liu et al., 2003; boundary between the sol state and the gel state as a function of
Poncelet, 2001). the initial concentration of calcium source and alginate. The
Although the Ca-alginate gel system has wide applications, transient sol–gel diagram predicted by this model was compared
there are only a few studies that focus on the kinetics of the sol–gel with the previous experimental observations to validate the
transition and the critical behavior in the aqueous solutions theoretical model at a macroscopic level.
induced by adding divalent cations (Wang and Spencer, 1998; The remainder of the paper is organized as follows. The
Khairou et al., 2002; Hassan et al., 1988a, 1988b, 1989). The experimental section presents the recap of the experimental
detailed pathway of the Ca2 + ions binding to the alginate chains preparation of Ca-alginate microspheres using an internal calcium
has been a point of controversy, presumably due to the recognized source as reported in our previous paper (Zhao et al., 2007). The
complexity of the system. Different interpretations for the forma- theoretical formulation section presents the assumptions, the
tion of junctions between the Ca2 + ions and the alginate functional initial conditions, and the equations that govern the reactions
sites have been reported (Hassan et al., 1989; Fang et al., 2007; between calcium binding and the intermolecular chain interac-
Braccini and Perez, 2001; Grant et al., 1973). These junctions may tions. Next, the results and discussion are presented. Finally, the
be intermolecular, where the carboxyl and hydroxyl groups belong last section summarizes the key findings of the paper and provides
to different chains and/or with the combination of intramolecular concluding remarks.
bindings, where the carboxyl and hydroxyl groups are in the same
chain (Hassan et al., 1989). It is generally believed that the Ca2 + ions
bind preferentially to the guluronate units of alginate rather than 2. Experimental section
the mannuronate units (Martinsen et al., 1989). Moreover, some
researchers (Braccini and Perez, 2001; Smidsrod and Skjåk-Bræk, 2.1. Alginate sample
1990) reported that Ca2 + ions can replace the hydrogen on the
carboxyl functional group of the guluronate blocks, which can Low-viscosity sodium alginate was obtained from Sigma–
naturally zip into a cavity shape. This is not the case for the carboxyl Aldrich (St. Louis, MO, USA) with the following characteristics:
functional group of the mannuronate blocks. The overall ratio of Ca/ 3.8–1000 cP for 0.2–4.8 wt%/vol water solution (0.2–4.8 g sodium
G is stoichiometric. Alginate chains adopt a helical conformation alginate per 100 mL of solution), MW! 6–28 kDa, FG ¼0.31 and
which is described by the ‘‘egg-box model’’ (Grant et al., 1973). FM ¼1 # FG, where FG and FM represent the fractions of guluronic and
Our group has conducted an extensive experimental study to mannuronic groups, respectively. The fraction of residues was
prepare Ca-alginate gel microspheres by utilizing calcium carbo- determined from 1H NMR (Grasdalen et al., 1979; Grasdalen, 1983).
nate nanoparticles as the internal calcium source (Zhao et al., The molecular weight of the alginate sample was determined from
2007). As part of this research, we observed the sol–gel transition size exclusion chromatography calibrated by protein standards.
850 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858

In this way, the molecular weight was overestimated because the gel point, which is equivalent to a divergence of the weight
protein coils are more flexible than alginate chains. In the present average molecular weight. The Flory–Stockmayer model provides
study, only guluronic residues were considered to be functional for the simplest kinetic scheme of polyfunctional polymerization.
the gelation reaction. Therefore, the functionality, f, of the alginate Although the model has been criticized often for its oversimplifica-
sample can be calculated by MW/mmonomer $ FG, which is in the tion of the network structure, the model captures the ideal state of
approximate range between 10 and 45 and where mmonomer is the network structure. The Flory–Stockmayer model assumes that
the mass of a monomer unit and MW is the molecular weight of the the probability for bond formation is the same between any pair of
alginate. ‘‘Zero-shear’’ or ‘‘Newtonian’’ viscosity of dilute solutions clusters neglecting size effect and no intramolecular bonds are
of alginate in water was measured by a Brookfield DV-II+ Pro allowed. The second assumption excludes the formation of rings
rotational viscometer equipped with an SC4-18 spindle. All mea- and cycles within the branched macromolecule. As already men-
surements were performed at room temperature. tioned in the introduction section, Ca2 + ions induced chain–chain
association, which constitute the junction zones responsible for gel
2.2. Preparation of Ca-alginate gels and the sol–gel diagram formation has been derived experimentally. The egg-box model has
led to the conclusion that the mechanism of cross-linking in
Ca-alginate gels were prepared following a procedure described alginate gels involves co-operative binding of Ca2 + ions and
in our previous paper (Zhao et al., 2007). Briefly, an internal calcium extensive formation of cycles between aligned polyguluronate
source, CaCO3 nanoparticles, was uniformly suspended in a sodium ribbons. Therefore, the classical noncycliaztion assumption cannot
alginate solution with different concentrations. Then, the aqueous meet the requirement to describe the Ca-alginate gelation precisely
mixture of the sodium alginate and the CaCO3 nanoparticles was in reality. In the present study, the F–S theory of gelation was
atomized by a high electric field into a secondary continuous applied in a more realistic way by incorporating a cluster growing
organic phase containing acetic acid to form water-in-oil emul- factor, p, where p represents the probability of a functional group
sions. As mass transfer of the acetic acid from the continuous phase on an alginate cluster forming intermolecular binding. 1 # p is then
to the aqueous droplet occurred, the free Ca2 + ions were released the cyclization probability of intramolecular binding that leads to a
by the dissolution of CaCO3 nanoparticles to initiate gelation of the cyclic structure without increasing the cluster size. The detailed
alginate. The time for diffusion of the acetic acid followed by the incorporation of this growing factor, p, to the model was described
dissolution of CaCO3 nanoparticles was estimated from literature in the next sections.
values (Davies and Wiggill, 1960; Blinkova and Eliseev, 2005),
which is on the order of 0.1 s. In comparison to the gelation 3.1. Problem definition
reaction,25 the dissolution of CaCO3 nanoparticles is essentially
instantaneous. The gelation of the alginate is the rate-controlling The model system is considered to consist of a drop of uniformly
step during formation of the gel-microsphere. Therefore, this is distributed alginate monochains and inner source of Ca2 + ions
assumably a uniformly triggered gelation process for the potential mixture. Fig. 2 represents the proposed schematics of the multiple-
formation of homogeneous gel structures. step binding of Ca2 + ions to alginate. The alginate molecule
The densely cross-linked gel network gave rise to a sufficient contains hydroxyl and hydrocarbonate groups that can coordinate
difference in the refractive index relative to that of liquids; hence, with free Ca2 + ions. Once two alginate chains align with each other,
the morphology of the gel beads can be examined under a regular multiple Ca2 + ions can react with available functional groups on the
light microscope. A drop of emulsion was placed on a glass slide opposite position from the other alginate chain, accordingly. The
and the nascent Ca-alginate particles were observed by a Nikon reaction steps for the ionic Cal-alginate junction formation are
Labophot-2 Binocular Brightfield Microscopy with a Javelin video given by
camera, E plan 2 $ , 10 $ , 20 $ objectives and equipped with a k1
Micro-Lite FL3000 fiber optic light source. The transition from the Na#AlgðsolÞ þ Ca2 þ ##!-Alg#Ca þ þNa þ ð1Þ
sol state to gel state was observed and a sol–gel transition diagram
k2
was constructed using the data obtained from a large number of Alg#Ca þ þ Na#AlgðsolÞ ##!Alg#Ca#AlgðgelÞ þNa þ ð2Þ
experiments. As is evident from Fig. 5a in Ref. 25, a critical alginate
concentration of approximately 0.6 wt%/vol divides the final This is a two-step reaction with the replacement of the counter
products into two different morphologies. Below this concentra- ions, Na + ions, which leave the macromolecular chains of the
tion, instead of generating a continuous gel, only polymer clusters alginate, by a stoichiometrically equivalent amount of Ca2 + ions.
were formed and viscous liquid droplets were observed with a Step I describes the process of free Ca2 + ions (empty dots) attaching
background of the continuous oil phase. Above this concentration, to functional groups on the alginate monochains, and subse-
densely cross-linked gel beads appeared. The sol–gel diagram will quently, these functional groups loss their ‘‘reactivity’’. In Step II,
be compared to the sol–gel diagram that was computed using the the bound Ca2 + ions (gray dots) are still active and further react
theoretical model. with available functional groups to complete the junction forma-
tion (black dots for inactive Ca2 + ions). In this paper, we use the
term of semi-occupied Ca2 + ions for those that have already reacted
3. Theoretical formulation with one alginate chain, but not yet cross-linked with another
alginate chain. Adopted from the work of Flory, at any stage during
A simple view of gelation by the Flory theory (Flory, 1953) is the reaction, all unreacted functional groups on clusters were
defined as clusters grown by chain cross-linking either with itself to considered to be equally reactive, regardless of the size of the
form intramolecular cycles, or with another chain intermolecularly molecule to which they are attached or the history of other
to create branched chains. The size of the cluster increases as more functional groups on the same molecule. Therefore, only two
and more intermolecular cross-links are formed, and eventually at reaction rate constants, k1 and k2, were introduced describing
the gel point, the largest cluster span the whole system and the serial reactions involving free Ca2 + ion binding and junction
presents an infinite network. In statistical mechanics, gelation is formation for semi-occupied Ca2 + ions. We also assume that the
modeled by Flory and Stockmayer (Flory, 1941; Stockmayer, 1943, alginate precursors are mono-dispersed with fixed chain length.
1944) as the random cross-linking of multifunctional monochains This assumption prevents exact agreement with the reality that
and the second moment of the molar mass distribution diverges at naturally occurred polysaccharides are invariably highly polydisperse
 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858 851

Step I Step II
2+
Ca

Monochains Monocomplex Clusters

2+
Free Ca ions Semi-occupied Ca2+ ions Bounded Ca2+ ions

Fig. 2. Schematic representation of the multiple-step binding of Ca2 + ions to alginate.

(Mw/Mn Z10) (Morris et al., 1981). It should be possible, however, to

f1=10, P2, j=2
provide for this circumstance within the general framework of the
present method. In addition to that, the two-step reactions were
considered to be irreversible due to the high stability of the formed
metal alginate gel complexes (Wang and Spencer, 1998; Hassan and
Abdalla, 1991). For modeling purposes, the model was casted in
spherical coordinates, whose origin lies at the center of the drop. In
this paper, the system was taken to be spherically symmetric with
homogeneous reaction throughout the droplet, so the problem was
independent of the radial and angular position.

3.2. Cyclization model

As have been mentioned above, intramolecular reactions, lead-

f2=pf1=8, P2, j=2
ing to a cyclic structure occur frequently inside the Ca2 + ions
induced alginate gel system. The complexity of the cyclization
reaction often prohibits the generic statistical approaches; there-
fore, we applied a simplified model by incorporating the effect of
cyclization into the traditional statistical model. A connection
between the resultant reactivity change from a cyclic by-product
and the impact to the reaction equation is constructed. An example
of this circumstance is shown in Fig. 3a. A pair of sample alginate
chains with the number of functional groups per chain, f1 ¼10, form Fig. 3. Illustration of cyclization model.
a dimer structure with one nascent binding and additional two
functional groups cyclized (black dots). The molar concentration of
a cluster structure is afterwards denoted by Pi,j; here, i is the size of
the cluster (i.e., the total number of polymer chains that the cluster As the final check of the mass balance, the amount of free Ca2 + ions
is composed of) and j represents the number of semi-occupied Ca2 + available in the system is not affected by the formation of the cyclic
ions on the cluster. For the dimer structure shown in Fig. 3a, where structures.
two sites have been occupied by Ca2 + ions (gray dots) but not cross-
linked yet, i¼2 and j¼2. Evidently, the cyclic parts of the dimer are
not able to participate in additional reactions, indicating the 3.3. Population balance equations
cyclization process is identical to lose of functionalities on the
pair polymer chains. In this particular case, there are two functional The F–S theory of polymerization may be described directly
groups forming cyclic structure out of ten for each individual chain, by a series of kinetic equations in a form first discussed by
the resultant probability of continuous growth per chain is eight Smoluchowski (1917) and Rubinstein and Colby (2003). At the
out of a total of ten, which is equal to 0.8. By introducing the discrete regime, the governing population balance equations for
prefactor, p, the circumstances can be equally described by alginate clusters and free Ca2 + ions were derived based on the
decreasing the number of functional groups on a single chain by classical concept proposed by Smoluchowski as follows (Galina and
a factor of p. Therefore, the new functionality f2 ¼pf1 ¼8 shown in Lechowicz, 1998):
Fig. 3b and the available active sites for further intermolecular @Pi,j ! "
reactions can be determined by f2i–2(i–1)–j ¼12 as employed by ¼ #k1 Pi,j CCa2 þ pfi#2ði#1Þ#j
@t
the F–S theory for a non-cyclic form. Assuming the intermolecular X1 X a
chain interaction is a random process, the introduction of a single #k2 Pi,j Pm,n fj½pfm#2ðm#1Þ#n) þ n½pfi#2ði#1Þ#j)g
value p for various chains on the same cluster as well as different m¼1n¼0
! "
clusters excludes the position and molecule size dependent of p. þ k1 Pi,j#1 CCa2 þ pfi#2ði#1Þ#ðj#1Þ
852 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858

1
g
i#1 X
X semi-occupied Ca2 + ions. Pi,j (i¼1, N) designates the cluster size
þ k2 Pm,n Pi#m,j#n þ 1 fn½pf ði#mÞ#2ði#m#1Þ#ðj#n þ 1Þ) distribution. This distribution can be described by the following
2 m¼1n¼b
three moments:
þ ðj#n þ 1Þ½pfm#2ðm#1Þ#n)g ð3Þ
p fi#2ði#1Þ
X X
m0,0 ¼ Pi,j ð8Þ
where i¼1 j¼0

p fi#2ði#1Þ
a ¼ minfpfm#2ðm#1Þ,ðpf #2Þðiþ mÞ#j þ 3g ð4Þ X X
m1,0 ¼ iPi,j ð9Þ
i¼1 j¼0

b ¼ maxf0,j#1#ðpf #2Þði#mÞg ð5Þ

p fi#2ði#1Þ
X X
m2,0 ¼ i2 Pi,j ð10Þ
g ¼ minfpfm#2ðm#1Þ,j þ 1g ð6Þ i¼1 j¼0

These three moments contain all the information of the cluster

X1 pfi#2ði#1Þ
X
@CCa2 þ size distribution. The zeroth moment, m0,0, represents the total
¼ #k1 CCa2 þ Pi,j ½pfi#2ði#1Þ#j) ð7Þ
@t number of molecules in the system. In retrospect, the zeroth
i¼1 j¼0
moment decreases as the polymerization proceeds. The first
In the above Eq. (3), if i¼1, p¼1 and otherwise, p was determined moment, m1,0, is the total number concentration of monochains,
from experimental deduction, which is discussed in the next section. which cannot disappear or be generated in the system. Therefore, it
The first term expresses the disappearance of i-mer clusters with j should always be equal to the initial value. This value is used to
semi-occupied Ca2 + ions by reaction with any free Ca2 + ions. Since i- check the accuracy of the numerical results for the conservation of
mer clusters with j semi-Ca2 + ions bear pfi–2(i–1)–j active groups mass. The second moment, m2,0, carries the molecular weight
and this is the number of distinguishable ways in which i-mer information. In this paper, the number-average degrees of poly-
clusters and free Ca2 + ions can react. A factor of pfi–2(i–1)–j was merization (DPn) and the weight-average degree of polymerization
introduced in the first term. The second term expresses the (DPw) by the ratios of successive moments are defined as
consumption of i-mer clusters with j semi-Ca2 + ions by reacting
DPn ¼ m1,0 =m0,0 ð11Þ
with any other m-mer clusters with n semi-occupied Ca2 + ions.
Statistically, the possible reaction is the addition of the ways in DPw ¼ m2,0 =m1,0 ð12Þ
which j semi-occupied Ca2 + ions on the i-mer cluster can react with
the available sites on the m-mer cluster and the ways in which n The gel point is where a cascading growth in size leads to the
semi-occupied Ca2 + ions on the m-mer cluster can react with the formation of an incipient gel that occupies the whole system.
active sites on the i-mer. The sum over m is given by all size range According the classical F–S theory, this phenomena is represented
clusters and the upper limit of n is no longer simply restricted by the numerically by the divergence of DPw and m2,0. Eqs. (3) and (7) were
available functional groups of m-mer, but also from the subse- solved simultaneously with the initial conditions, Pi,j ¼ Pi,0 and
quently formulated complex. The quality of the upper limit of n CCa2 þ ¼ C0 at t ¼0 by the explicit Euler’s method. Parallel computing
denoted as a has to meet the relation such as 0rnrpfm–2(m–1) was used to solve the series of equations using a Message Passing
and 0rj+n–1rpf(i+m)–2(i+m–1), where one semi-occupied Ca2 + Interface library on a computer cluster with 128 nodes.
ion in the system is consumed to form an i+m-mer complex. These
constraints are reflected in Eq. (4). The additional restriction for a
from the resultant complex is, in fact, to exclude the impossible 4. Results and discussion
reaction between two clusters that both are saturated by semi-
occupied Ca2 + ions. The third and fourth terms of Eq. (3), respec- 4.1. Experimental determination of the prefactor, p
tively, represent the formation of i-mer cluster with j semi-occupied
Ca2 + ions due to the additional free Ca2 + ion binding to i-mer clusters Double logarithmic plots of Zsp (specific viscosity, defined in the
with j# 1 semi-occupied Ca2 + ions and the combination between footnote to Fig. 4) against the alginate concentration, P, show a
two smaller size clusters (m-mer and i–m-mer). The possible pronounced increase in gradient above a specific critical concen-
reaction factors for the third term and the fourth term were derived tration. Similar behavior is well known for synthetic polymer
in the same manner as those of the first term and the second term. solutions, the turning point of which is defined as the overlap
The factor of 1/2 appears in the fourth term because of the two concentration (Dea, 1989). This point appears because of the
indistinguishable reactions between m-mers and i–m-mers. The transition from dilute solution conditions, where individual poly-
values of the down and upper limits of n in last term are denoted as b mer molecules are present as isolated chain coils, to concentrated
and g, respectively, which were determined from the restriction of solutions where the volume of solution can no longer accommo-
the active sites on both individual and resultant complex. The sum date the swollen molecular coils without chain entanglement. In
over n is confined by 0rnrpfm–2(m–1) and 0rj–n+1rpf(i–m) – general, the onset of coil overlap is under extremes of concentration
2(i–m–1), by which Eqs. (5) and (6) were derived. Eq. (7) expresses for stiff polysaccharide systems, which is always experimentally
the free Ca2 + ions consumption during the condensation reaction by inaccessible. The break in Fig. 4a is different from the overlap
reacting with all size range clusters. The factor of pfi–2(i–1)–j follows concentration commonly noted in synthetic polymer systems. The
the same explanation as that of the first term in Eq. (3). The diffusion turning point in this case denotes the transition from dilute
# $ # @C
$ solution behavior (low degree of coil overlap) to concentrated
@P
terms r12 @r@ D @ri,j r 2 and r12 @r@ DCa2 þ Ca@r2 þ r 2 were neglected from
solution behavior (total interpenetration of random coil mole-
Eqs. (3) and (7) to simply the numerical solution to the problem and cules), which indicates that specific interchain associations, ana-
spatially invariant concentrations were assumed inside the drop. logous to those in polysaccharide gels but of shorter timescales,
As implied in the preceding discussion on alginate, not all may occur in concentrated solutions of otherwise disordered
clusters are generally expected to have the same size. A distribution polysaccharide (Morris et al., 1981; Dea, 1989). The terms ‘‘con-
of cluster sizes is expected during and after the reaction, which we centrated’’ and ‘‘dilute’’ will be used in this paper to represent the
denote by Pi,j, the molar concentration of species with size i and j concentration above and below the turning point.
 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858 853

The complexity of alginate system prevents the exact determi-

2.5
nation of the function of p with alginate concentration. Unless
otherwise specified, the data set (a ¼0.3, b¼1.0) was used for the
Low degree of coil overlap
2 rest of calculation in this paper, which yielded reasonable predic-
Slope ≈ 2.6 tions. The present assumption helps clarify the significance of the
cyclic structure formation within the Ca-alginate system in differ-
1.5 ent concentration regimes.
∗
log η sp

4.2. Effect of reaction rate constants (k1, k2)

1
Slope ≈ 1.2
Multiple binding steps of Ca2 + ions to alginate often occurs
simultaneously after the initiation of reaction, therefore, the
0.5
Total interpenetration of experimental determination of the distinguished reaction rate
random coil constants becomes very difficult. Previous studies have shown
0 that the free divalent metal ions binding rate, k1, is measurable, but
the values of the exchange rate constant, k2, vary with the type of
-1 -0.5 0 0.5 1 1.5 metal ions, ionic strength of the reaction mixture and pH values
log P in the system (Khairou et al., 2002; Hassan et al., 1988a, 1988b;
Hassan, 1991). The value of k1 for Ca-alginate system has not been
strictly defined before, we preferentially use an approximate value,
k1 ¼0.72 L/mol h, that is lower than what was reported for a
copper-alginate binding. k2 was treated as a variable in the initial
calculation. The model was then used to investigate the polymer-
ization of alginate and to map the sol–gel diagram.
Fig. 5a, b is the typical numerical results of the sol and gel
formation, respectively, plotted using numerical solutions at
0.5 k1 ¼0.72 L/mol h and k2 ¼1.44 L/mol h. It can be seen that in the
p

sol regime, the time evolution of m2,0 indicated by a solid line, that
carries the information about the growth of the polymer cluster
size, stops after a certain time. That represents the formation of
polymer clusters with limited size. However, in the gel regime, the
size of the branched chains becomes larger as more and more
intermolecular cross-links are formed. The weight average mole-
cular weight increases faster and faster and eventually becomes
0 infinite as one of the branched chains span the whole system.
-1.5 -1 -0.5 0 0.5 1 1.5 Mathematically, the infinite network is represented by the diver-
gence of m2,0 along the polycondensation. m0,0 denotes the total
log P number concentration of molecules as shown by a dotted curve. In
Fig. 4. (a) Generalized concentration, P, dependence of specific viscosity, Zsp, for sodium both cases, m0,0 decreases as the polycondensation proceeds. The
alginate system; (b) proposed prefactor, P, as a function of sodium alginate concentra- total number concentration of monochains, m1,0 (represented by
tion, P. nSpecific viscosity, Zsp is the fractional enhancement of viscosity due to the the dash–dot lines) remains constant since no polymer chains are
solvent, i.e., Zsp ¼ (Z # Zs)/Z, where Z is the solution viscosity and Zs is that of solvent.
created or destroyed in the system, which verifies the accuracy of
the population balance model.

As the degree of coil overlap increases, the freedom of movement

of individual polymer chains is progressively restricted, with con-
µ 2,0 µ 2,0
sequent increase in time required to align the adjacent chains. In the
µ 1,0 µ 1,0
dilute polymer concentration region, once the linkage between a pair
µ 0,0 µ 0,0
of adjacent chains occurs, these two chains have more time to interact
with each other before the third chain diffusively approaches them. 5×10-2
-5
The probability of forming cyclic structure with intramolecular 2×10
multiple linkages is greatly increased. Therefore, the prefactor, p,
for a single alginate chain to have different chain interactions for a
continuous growth, should be less than 0.5. In contrast, at the Sol Gel
concentrated region, the local interactions severely limit the possible
orientations of adjacent residues and the system is dominated by
intermolecular interactions. In this case, the prefactor, p, is assumed to
be close to a unit. The cyclic structure formation may ultimately occur
mainly within the postgel regime in this type of system. Thus, as
illustrated in Fig. 4b, a simple assumption was made to describe this
circumstance, at which the prefactor, p, is a step function of alginate 0 1 2 0 0.2 0.4
concentration: t ×10-1 h t ×10-1 h
(
a 0 o a o0:5 when P1,0 ðt ¼ 0Þ o 0:6 wt%=vol Fig. 5. Evolution in time of m0,0, m1,0, m2,0 with k1 ¼0.72 L/mol h, k2 ¼1.44 L/mol h,
p¼ ð13Þ
b 0:5 ob o 1:0 when P1,0 ðt ¼ 0Þ 41:2 wt%=vol f¼ 20, and initial conditions: (a) P1,0(t ¼0) ¼ 0.6 wt%/vol, CCa2 þ ðt ¼ 0Þ ¼ 0:2 mol=L;
(b) P1,0(t ¼0) ¼ 1.2 wt%/vol, CCa2 þ ðt ¼ 0Þ ¼ 0:2 mol=L.
854 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858

The effect of the two reaction rate constants, k1 and k2, was Table 1
obtained by analyzing the change of the trends of m2,0 with time Effect of three different sets of reaction rate constants on the Ca-alginate sol–gel
formation with P1,0 (t ¼0) ¼ 1.2 wt%/vol and k1 ¼ 0.72 L/mol h.
evolution. As indicated by solving the population balance model
based on Eqs. (3) and (7), the absolute values of k1 and k2 affect the CCa2 þ ðt ¼ 0Þ ðmol=LÞ k2 ¼ 0.36 L/mol h k2 ¼0.72 L/mol h k2 ¼ 1.44 L/mol h
duration of reaction. However, the change of the reaction rate
constants with a fixed ratio does not affect the trends of m2,0, which 0.05 O O O
has been verified by multiple data sets. The plots of m2,0 using 0.10 O O O
0.20 $ O O
different reaction rate constants at the same ratio are not shown in
0.40 $ $ O
this paper since the curves overlap very well after normalization by
time. By increasing the value of k2 at a constant k1 with other Note: O gel, $ sol.
conditions specified to be same, a change from the sol state to the
gel state occurs as Fig. 6 shows, e.g., when k1 ¼0.72 L/mol h,
situations in which the values of k2 are smaller (k2 ¼0.36 L/mol h), hence, the subsequent semi-occupied Ca2 + binding becomes easier,
equal (k2 ¼0.72 L/mol h), or higher (k2 ¼1.44 L/mol h) than k1 show which explains the fact that k2 is greater than k1 during the
a systematical increase of m2,0 until it divergences at the higher stepwise Ca-alginate reaction. The fact that k2 4k1 is consistent
value of k2. This result indicates that the second step reaction for with the results from a recent publication about the experimental
interchains binding influences the growth of cluster size signifi- investigation of the multistep binding rate of calcium to alginate in
cantly. Gelation occurs when the value of k2 is larger relative to k1. the extremely dilute solutions.21 The data set, k1 ¼ 0.72 L/mol h and
The effect of the three different sets of reaction rate constants on k2 ¼1.44
the Ca-alginate sol–gel formation was examined and the results are L/mol h, were used in this paper unless indicated otherwise.
demonstrated in Table 1 with the initial conditions P1,0 (t ¼0) ¼
1.2 wt%/vol and the values of CCa2 þ ðt ¼ 0Þ varied systematically.
Comparing these data to those obtained from various reported 4.3. Effect of alginate/Ca2 + ion concentrations and functionality, f
experiments (Wang et al., 1993; Khairou et al., 2002; Morch et al.,
2006), the estimated values of k2 were examined. These findings The computed effect of the alginate property and the gelling
suggest that k2 4k1 is required to guarantee a gel formation for a agent, free Ca2 + ions, are reported in Figs. 7–9. Fig. 7 shows the
alginate system with the concentration of 1.2 wt%/vol in the evolution in time of DPw with various initial concentrations of
examined concentration range of free Ca2 + ions as well as for alginate, P1,0(t¼ 0), in the range of 0.2 to 2.0 wt%/vol at f ¼20 and
the prediction of a reasonable gelation time. During the early stage CCa2 þ ðt ¼ 0Þ ¼ 0:2 mol=L. DPw represents the average number of
of cross-linking, the free Ca2 + attachment is suspected to happen chains on a cluster chosen by randomly selecting a chain. The larger
when the polymer chains are pre-coordinated with each other, the DPw value, the bigger the size of alginate clusters formed. Fig. 7
shows that a profound change from the gel state to the sol state
begins to take place as the initial alginate concentration drops from
1.2 to 0.6 wt%/vol. In the concentrated alginate region, the size of
the cluster formed in the polymer system with a higher concentra-
tion is larger and the gelation is quicker than that in systems with a
k1/k2 = 0.5 lower concentration, as shown in the inset (1) to Fig. 7. The
k1/k2 = 1.0 cascading growth in size leads to an incipient gel where DPw
-2 k1/k2 = 2.0 rapidly approaches a large value. The gelation time decreases
5×10
substantially from 5.6 h for 1.2 wt%/vol alginate system to 4.0 h for
2.0 wt%/vol system, indicating a strong dependence of polymer
concentration on the gelation reaction. From the statistical view,
the higher concentration of polymer generates a higher probability
of forming junctions. However, in a dilute polymer solution, where
the initial concentration of alginate is less than 0.6 wt%/vol, DPw is
suppressed to less than 2. Inset (2) to Fig. 7 shows that the size of
the cluster decreases with the decrease of alginate concentration,
µ 2,0 where the growth of the cluster is restricted by the amount of
polymer chains.
Moreover, the effect of functionality, f, on the gelation time is
significant in the same manner as the effect of alginate concentra-
tion on the change of the total functional groups in the system.
Fig. 8a shows that a small increase of functionality, f, can
dramatically decrease the gelation time. The decrease is more
sensitive to the change of functionality, f, in comparison to the
polymer concentration. This finding is depicted more clearly in
Fig. 8b by comparing the increases of the same proportion of
functional groups (e.g., 25%) by increasing f from 20 to 25 or by
adding more initial alginate chains from 1.6 to 2.0 wt%/vol, where
the functionality effect is more evident.
0 1 2 The initial concentration of the gelling agent, free Ca2 + ions,
-1 plays an important role for gelation as well. The gelation times for
t ×10 h the system with different CCa2 þ ðt ¼ 0Þ (0.05, 0.1, 0.2 and 0.4 mol/L)
are shown in Fig. 9a, where it takes the gelation to be completed in
Fig. 6. Evolution in time of m2,0 with three different sets of reactions rate constants
by fixing k1 ¼ 0.72 L/mol h and varying k2 with f¼ 20 and initial conditions 9.5, 7.2, 5.6 and 5.2 h, respectively. The higher the Ca2 + ions
P1,0(t¼ 0) ¼1.2 wt%/vol, CCa2 þ ðt ¼ 0Þ ¼ 0:4 mol=L. concentration, the less the time to reach the gel point. The
 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858 855

3 20 1.2
(1) (1) (2)

10 1.1

DPw

2 0 0.2 0.4 0.6

1
0 0.2 0.4 0.6
-1
t ×10 h t ×10-1 h
P1, 0 (t = 0 ) = 2.0 wt % / vol
P1, 0 (t = 0 ) = 1.6 wt % / vol
P1, 0 (t = 0 ) = 1.2 wt % / vol
P1, 0 (t = 0 ) = 0.6 wt % / vol
(2)
P1, 0 (t = 0 ) = 0.4 wt % / vol
1
0 0.2 0.4 0.6 P1, 0 (t = 0 ) = 0.2 wt % / vol
-1
t ×10 h
Fig. 7. Evolution in time of DPw with various alginate concentrations at k1 ¼0.72 L/mol h, k2 ¼ 1.44 L/mol h, f ¼20 and initial conditions CCa2 þ ðt ¼ 0Þ ¼ 0:2 mol=L.

f = 20 P1,0 (t = 0) = 1.6 wt%/vol

f = 20
f = 20 P1,0 (t = 0) = 2.0 wt%/vol
f = 25
20 f = 30 20 f = 25 P1,0 (t = 0) = 1.6 wt%/vol
DPw

DPw

10 10

0
0 0.2 0.4 0.6 0 0.2 0.4 0.6
-1 -1
t ×10 h t ×10 h
Fig. 8. Evolution in time of DPw: (a) with various functionality, f at k1 ¼ 0.72 L/mol h, k2 ¼ 1.44 L/mol h, and initial conditions P1,0(t¼ 0)¼1.2 wt%/vol, CCa2 þ ðt ¼ 0Þ ¼ 0:2 mol=L;
(b) with comparison between increasing 25% total functional groups by increasing the functionality, f, from 20 to 25 and by increasing P1,0(t¼0) from 1.6 wt%/vol to 2.0 wt%/vol.

computations reveal that the acceleration of the gelation reaction is 4.4. Mechanism of the sol–gel transition
restricted once the concentration of free Ca2 + ions is sufficiently
high, where the gelation time reaches a limited low value and To elucidate the physics of phase transition that were depicted
cannot further decrease. For this case, the polymer concentration is in the previous section, the evolution in time of semi-occupied Ca2 +
the controlling parameter. On the other hand, if the polymer has ions on a large cluster was examined. Fig. 10a, b shows, respec-
sufficient sites for the reaction but not the free Ca2 + ions, the tively, situations for a 10-mer in which the amount of gelling agent
gelation time decreases as the increase of polymer concentration is excessive or deficient (see, also, Fig. 9b). The binding of Ca2 + ions
(or f) as seen in Figs. 7 and 8a. Moreover, extreme high or low can be characterized by the distribution of semi-occupied Ca2 + ions
CCa2 þ ðt ¼ 0Þ can cause the sol–gel transform to occur as Fig. 9b along the polymer cluster. As Fig. 10a shows, all functional groups
shows. The mechanisms of the phase change are discussed in the (pfi–2(i–1)¼ 42, where p¼0.3) on the 10-mers are rapidly occupied
following section. by free Ca2 + ions, indicated by the rapid appearance of the peak of
856 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858

0 0.2 0.4 0.6

20 4

1.08
15
3
DPw

DPw
10
1.04
2

1
0 0.2 0.4 0.6 0.8 0 10 20
t ×10 -1 h t ×10 -1 h

CCa2+ ( t = 0 ) = 0.4 mol/L CCa 2+ ( t = 0 ) = 1.0 mol/L

CCa2+ ( t = 0 ) = 0.2 mol/L CCa 2+ ( t = 0 ) = 10-5 mol/L
CCa2+ ( t = 0 ) = 0.1 mol/L
CCa2+ ( t = 0 ) =0.05 mol/L

Fig. 9. Evolution in time of DPw with various CCa2 þ ðt ¼ 0Þ at k1 ¼ 0.72 L/mol h, k2 ¼1.44 L/mol h, f ¼20 and initial conditions P1,0(t ¼0) ¼ 1.2 wt%/vol: (a) at the gel state and (b) at
the sol state.

t = 8.0 h t = 30 h
t = 6.2 h P t = 23 h
t = 4.4 h t = 17 h
t = 2.7 h t = 10 h
5E-17

t ×10 h
t ×10 h
5E-17

t = 4.5 h
t = 4.3 h
t = 3.7 h
t = 2.7 h
t = 1.6 h

t ×10 h 5E-8

Fig. 10. Semi-occupied Ca2 + ions distribution on 10-mers at k1 ¼0.72 L/mol h, k2 ¼1.44 L/mol h, f¼20 and initial conditions: (a) with P1,0(t¼ 0)¼0.2 wt%/vol, CCa2 þ ðt ¼ 0Þ ¼
0:2 mol=L forming the sol state, (b) with P1,0(t¼ 0)¼1.2 wt%/vol, CCa2 þ ðt ¼ 0Þ ¼ 10#5 mol=L forming the sol state, (c) with P1,0(t¼0)¼1.6 wt%/vol, CCa2 þ ðt ¼ 0Þ ¼ 0:2 M forming the
gel state.

P10,j at the maximum j. In this case, all clusters, not only 10-mers, interactions. Hence, the growth of the cluster is terminated. On
are saturated by semi-occupied Ca2 + ions, therefore, the clusters no the other hand, at an extremely low Ca2 + ions regime as shown in
longer have available active sites for further intermolecular Fig. 10b (only low value of j plotted), the occupation of Ca2 + ions is
 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858 857

only visible in the vicinity of j¼1. The three-dimensional network 0.4

cannot extend continuously throughout the whole system because
of the lack of gelling agent for the cross-linking reaction. In these
two circumstances, the sol formation is due to either the Ca2 + ions
saturation along the cluster or the depletion of Ca2 + ions. However, 0.3
in the gel regime, Fig. 10c makes plain that the migration of the
peak of the semi-occupied Ca2 + ions is around the center of the

Ca2+ mol/L
I II
whole range of j (maximum pfi–2(i–1) ¼182, where p¼ 1.0) and
thereby both functional groups and semi-occupied Ca2 + ions are 0.2
available for further reactions. The rapid increase of the peak value
until the reaction for 4.3 h and a decrease afterwards explains the
growth of these 10-mers and further consumption for forming
bigger size clusters. 0.1
Fig. 11 further demonstrates the weight distribution of the
clusters with different sizes in the proceeding system. The weight
P
fraction is calculated from fi#2ði#1Þ
j¼0
iPi,j =m1,0 . The inset is a blow-up III
at the low size regime of Fig. 11. It is easily seen by noting that the 0
majority of the mass of the system is from the small size clusters in 0.6 1.2 1.8 2.4 3
both the sol state and the gel state. Gelation does not mean that the Alginate (wt%/vol)
entire mass of the system is in the gel, but rather that the incipient Computation data points forming sol Experimental data points forming sol
gel has a cascading growth in size considered to be infinitely large Computation data points forming gel Experimental data points forming gel
but with nearly zero mass fraction.
Approximate sol-gel transition boundary based on computation data
Approximate sol-gel transition boundary based on experimental data
4.5. Sol–gel diagram and comparison to experimental results
Fig. 12. Sol–gel transition diagram from both the experimental and the computational
studies (at k1 ¼ 0.72 L/mol h, k2 ¼ 1.44 L/mol h, f¼ 20). Region I: sol state; region II: gel
Using the data obtained from a large number of computational
state; and region III: additional gel state from the computational results.
results, the sol–gel transition associated with the reactant concentra-
tions was examined. Fig. 12 summarizes the structural information of
Ca-alginate at different reactant conditions from both the experi-
mental and the computational studies. The results are represented region II), the densely cross-linked gel appears. The competition
by the filled symbols25 and the empty symbols, respectively. between sufficient free Ca2 + ions and alginate chains in the system
As is evident from Fig. 12, a generic critical boundary at alginate enhances the gelation process in the forward direction. This means
concentrations approximately between 0.6 and 1.2 wt%/vol was that the cluster size of Ca-alginate network grows rapidly and DPw
observed from both the experimental (solid curve proposed) and diverges before the depletion of either reactant. Meanwhile, the sol
the computational results (dash line proposed), which divided the state also exits at the extremely low free Ca2 + ions region, which is
final products into two different morphologies. Below this boundary theoretically lower than experimental observation. The sol state
(represented by region I), only polymer clusters were formed instead formation at the extremely low concentration of Ca2 + ions is due
of generating a continuous gel with DPw numerically approaching to the metal ions depletion. In reality, the effect of inhomogeneous
a plateau. It is suggested from the previous section that a junction formation locally could result in a higher consumption of
mechanism of Ca2 + ions saturation results in a sol state at a low gelling agent than that of the theoretical prediction. Moreover, the
polymer concentration region. Above this boundary (represented by occurrence of an additional gel window between 0.2 and 1.2 wt%/vol
alginate concentration (shown in region III) has been reported
numerically. Not too surprisingly, the phenomenon can be explained
1 because there is always a theoretical gap between Ca2 + ions satura-
1
tion and Ca2 + ions depletion at a certain polymer concentration for a
0.8 gel to occur. It is worth to note that in contrast to the gel regime II, the
0.8 gelation time in this gel window is extremely long (42 days), where
0.6 low concentrations of both polymer chains and gelling agent prevent
the interchain reactions from happening too frequently. Therefore,
Weight fraction

0.4 the sluggishness of gelation in this region and the small size of this
0.6 region may explain why gelation was not observed experimentally in
0.2
this region. In general, the computational sol–gel diagram and the
0
experimental observation are in a good agreement.
0.4 2 4 6 8 10
i

5. Conclusions
0.2
Sol at P1,0 (t = 0) = 0.2 wt%/vol This work provides a model of the physicochemical behavior of
Gel at P1,0 (t = 0) = 1.6 wt%/vol an ionic physical gel under different experimental conditions
corresponding to either structural or environmental parameters.
0 This model can be very useful to examine the effect of relative
200 400
values of the reaction rate constants on the sol–gel boundary and
i the gelation time. By using the population balance model, the
growth of the Ca-alginate cluster was simulated. The sol–gel
Fig. 11. Disrtibution of the weight fraction of cluster with size i in both the sol state
and the get state at k1 ¼0.72 L/mol h, k2 ¼1.44 L/mol h, f¼ 20 and initial condition transition occurs at the boundary where the competition between
CCa2 þ ðt ¼ 0Þ ¼ 0:2 mol=L. gelling agent and alginate chains in the system either enhances or
858 Y. Zhao et al. / Chemical Engineering Science 66 (2011) 848–858

reduces the gelation process in the forward direction. The model Flory, P.J., 1953. Principles of Polymer Chemistry. Cornell University Press, Ithaca, NY
allows for evaluation of the mechanisms by which ionotropic Ca- (Chapter IX).
Fuchs, T., Richtering, W., Burchard, W., Kajiwara, K., Kitamura, S., 1997. Polymer Gels
alginate gels are formed, and the effect of chemical properties on and Networks 5, 541–559.
the sol–gel transition boundary. The comparisons of the model Galina, H., Lechowicz, J.B., 1998. In Grafting/Characterization Techniques/Kinetic
prediction and the experimental data of the sol–gel boundary were Modeling. New York Inc., Berlin, Germany.
Grant, G.T., Morris, E.R., Rees, D.A., Smith, P.J.C., Thom, D., 1973. FEBS Letters 32,
assessed to be within reasonable agreement. 195–198.
Grasdalen, H., 1983. Carbohydrate Research 118, 255–260.
Grasdalen, H., Larsen, B., Smidsrod, O., 1979. Carbohydrate Research 68, 23–31.
Nomenclature Gu, F., Amsden, B., Neufeld, R., 2004. Journal of Controlled Release 96, 463–472.
Hassan, R.M., 1991. Journal of Materials Science 26, 5806–5810.
Hassan, R.M., Abdalla, M.A., 1991. Acta Polymerica 42, 447–450.
Pi,j the concentration of alginate cluster with size of i and j Hassan, R.M., Elshatoury, S.A., Mousa, M.A., Hassan, A., 1988a. European Polymer
sites of semi-Ca2 + ions (M or wt%/vol) Journal 24, 1173–1175.
Hassan, R.M., Summan, A.M., Hassan, M.K., Elshatoury, S.A., 1989. European Polymer
CCa2 þ free Ca2 + ion concentration in the system (mol/L) Journal 25, 1209–1212.
k1 reaction rate constant between alginate cluster and free Hassan, R.M., Wahdan, M.H., Hassan, A., 1988b. European Polymer Journal 24, 281–283.
Ca2 + (L/mol h) Haug, A., Larsen, B., Smidsrod, O., 1974. Carbohydrate Research 32, 217–225.
Khairou, K.S., Al-Gethami, W.M., Hassan, R.M., 2002. Journal of Membrane Science
k2 reaction rate constant between intermolecular alginate 209, 445–456.
clusters (L/mol h) Khokhlov, Y.O.A.R., 2002. Polymer Gels and Networks, Part A. Marcel Dekker,
f number of functional groups per chain (dimensionless) New York.
LeRoux, M.A., Guilak, F., Setton, L.A., 1999. Journal of Biomedical Materials Research
p the probability for a single alginate chain to have inter-
47, 46–53.
molecular interactions (dimensionless) Liu, X.D., Bao, D.C., Xue, W.M., Xiong, Y., Yu, W.T., Yu, X.J., Ma, X.J., Yuan, Q., 2003.
Journal of Applied Polymer Science 87, 848–852.
Martinsen, A., Skjakbraek, G., Smidsrod, O., 1989. Biotechnology and Bioengineering
33, 79–89.
Matsumoto, T., Kawai, M., Masuda, T., 1992. Journal of the Chemical
Acknowledgement Society—Faraday Transactions 88, 2673–2676.
Matsumoto, T., Mashiko, K., 1990. Biopolymers 29, 1707–1713.
Morch, Y.A., Donati, I., Strand, B.L., Skjak-Braek, G., 2006. Biomacromolecules 7,
This research was supported by Purdue Research Foundation
1471–1480.
Shreve Fund. We thank Adaptive Computing Systems Lab (ACSL) of Morris, E.R., Cutler, A.N., Ross-Murphy, S.B., Rees, D.A., Price, J., 1981. Carbohydrate
Purdue University for the constructive input for parallel computing Polymers 1, 5–21.
Poncelet, D., 2001. Bioartificial Organs III: Tissue Sourcing, Immunoisolation, and
using Message Passing Interface library (MPI).
Clinical Trials 944, 74–82.
Poncelet, D., Lencki, R., Beaulieu, C., Halle, J.P., Neufeld, R.J., Fournier, A., 1992.
References Applied Microbiology and Biotechnology 38, 39–45.
Quong, D., Neufeld, R.J., Skjak-Braek, G., Poncelet, D., 1998. Biotechnology and
Bioengineering 57, 438–446.
Blinkova, E.V., Eliseev, E.I., 2005. Russian Journal of Applied Chemistry 78, Rubinstein, M., Colby, R.H., 2003. Polymer Physics. Oxford University Press, USA
1064–1066. (Chapters 1 and 5).
Braccini, I., Perez, S., 2001. Biomacromolecules 2, 1089–1096. Skjakbraek, G., Grasdalen, H., Smidsrod, O., 1989. Carbohydrate Polymers 10, 31–54.
Davies, J.T., Wiggill, J.B., 1960. Proceedings of the Royal Society of London Series A Smidsrod, O., Skjåk-Bræk, G., 1990. Trends in Biotechnology 8, 71–78.
255, 277–291. Smoluchowski, M.V., 1917. Zeitschrift fur Physikalische Chemie 92, 129–168.
Dea, I.C.M., 1989. Pure and Applied Chemistry 61, 1315–1322. Stockmayer, W.H., 1943. Journal of Chemical Physics 11, 45–55.
Draget, K.I., Braek, G.S., Smidsrod, O., 1994. Carbohydrate Polymers 25, 31–38. Stockmayer, W.H., 1944. Journal of Chemical Physics 12, 125–131.
Fang, Y.P., Al-Assaf, S., Phillips, G.O., Nishinari, K., Funami, T., Williams, P.A., Li, L.B., Wang, X.W., Spencer, H.G., 1998. Polymer 39, 2759–2764.
2007. Journal of Physical Chemistry B 111, 2456–2462. Wang, Z.Y., Zhang, Q.Z., Konno, M., Saito, S., 1993. Journal de Physique II 3, 1–7.
Flory, P.J., 1941. Journal of the American Chemical Society 63, 3083–3090. Zhao, Y.Y., Carvajal, M.T., Won, Y.Y., Harris, M.T., 2007. Langmuir 23, 12489–12496.