Sei sulla pagina 1di 12

Chemical Engineering Science 56 (2001) 3173}3184

Kinetic modeling of the e!ect of solvent concentration on primary


cyclization during polymerization of multifunctional monomers
Jeannine E. Elliott , Jay W. Anseth , Christopher N. Bowman  *
Department of Chemical Engineering, University of Colorado, Boulder, Engineering Centre, ECCH 111, Campus Box 424,
Boulder, CO 80309-0424, USA
Dental School, University of Colorado Health Science Center, Denver, CO 80045-0508, USA
Received 19 May 2000; received in revised form 28 November 2000; accepted 5 December 2000

Abstract
Controlling the swelling ratio, di!usion rate, and mechanical properties of a crosslinked polymer is important in hydrogel design for
biomedical applications. Each of these factors depends strongly on the degree of crosslinking. Primary cyclization, where a propagating radical reacts intramolecularly with a pendant double bond on the same chain, decreases the crosslinking density and increases the
molecular weight between crosslinks. Processing conditions, speci"cally the solvent concentration, strongly a!ect the extent of
primary cyclization. In this work the e!ects of solvent concentration and comonomer composition on primary cyclization are
investigated using a novel kinetic model and experimental measurement of mechanical properties. Two divinyl crosslinking agents
were investigated, diethyleneglycol dimethacrylate (DEGDMA) and polyethyleneglycol 600 dimethacrylate (PEG(600)DMA), and
each was copolymerized with hydroxyethyl methacrylate (HEMA) and octyl methacrylate (OcMA). The model is further used to
predict the gel point conversion and swelling ratio of PAA hydrogels polymerized in the presence of varying amounts of water. Model
results show how increasing the solvent concentration during the polymerization increases the molecular weight between crosslinks
by nearly a factor of three and more than doubles the swelling ratio. Where possible, experimental results provide quantitative
agreement with model predictions.  2001 Elsevier Science Ltd. All rights reserved.
Keywords: Polymer; Gels; Crosslinking; Cyclization; Solvent e!ects; Simulation

1. Introduction
The use of crosslinked polymer hydrogels as biomaterials is a growing area of biomedical technology:
consequently, research on the network formation process
occurring in the presence of solvents is important. Free
radical copolymerization of multivinyl monomers with
hydrophilic monovinyl monomers leads to the formation
of hydrogels that swell signi"cantly but do not dissolve in
the presence of water. Because of their biocompatibility
and hydrophilic nature, hydrogels have biomedical
applications in contact lenses, wound bandages and
dressings, bioadhesives, cell immobilization, tissue engineering, and drug delivery systems (Wichterle & Lim,

* Correspondence address: Department of Chemical Engineering,


University of Colorado, Boulder, Engineering Centre, ECCH 111,
Campus Box 424, Boulder, CO 80309-0424, USA. Tel.: #1-303-4927471; fax: #1-303-492-4341.
E-mail address: bowmanc@colorado.edu (C. N. Bowman).

1960; Peppas, 1987; Bae & Kim, 1993; Ende & Peppas,
1996; Jen, Wake, & Mikos, 1996; Wheeler, Woods, Cox,
Cantrell, Watkins, & Edlich, 1996; Kao, Manivannan
& Sawan, 1997). Understanding of the polymer network
formation in the presence of a solvent and the resulting
network structure and properties is essential to develop
hydrogels with controlled swelling and properties for
speci"c biomedical applications.
During polymerization or copolymerization involving
multivinyl monomers, primary cyclization can occur
when a pendant double bond reacts with the radical on
the same propagating chain that created the pendant.
The degree of primary cyclization strongly e!ects the
network structure created and its resulting properties.
The mechanical integrity of a hydrogel is obtained from
the crosslinks in the network, particularly in solution
where `physicala crosslinks are hardly present. The mesh
size of a polymer, i.e., the distance between crosslinks,
controls the degree of swelling and di!usion in the
polymer, which are important to many applications,
especially in the development of drug delivery materials.

0009-2509/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 0 ) 0 0 5 4 7 - 9

3174

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

As drug release rates are a function of the degree of


crosslinking of the polymer, larger mesh sizes correlate
with greater di!usion of drug molecules through the
polymer (Peppas & Khare, 1993; Lehr, Bouwstra,
Vanhal, Verhoef, & Junginger, 1992). By reducing
the crosslinking density, primary cyclization changes the
mechanical properties, swelling, and di!usion through
the hydrogel from what would be predicted in an ideal
homogeneous polymer. Because the crosslinking density
controls so many important hydrogel properties, understanding what a!ects the extent of crosslinking is key in
developing biomaterials for speci"c applications.
This work investigates how the comonomer composition and the amount of solvent (generally water in hydrogel formation) used during polymerization in#uence the
degree of primary cyclization using a numerical modeling
approach and experimental measurements of mechanical
properties. Two sets of experiments were performed
for this study using diethyleneglycol dimethacrylate
(DEGDMA) or polyethyleneglycol 600 dimethacrylate
(PEG(600)DMA) as the crosslinking agents. DEGDMA
was chosen because it is a commonly used crosslinking
agent in soft contact lenses. To compare the e!ect of the
crosslinking molecule size PEG(600)DMA was also
evaluated. In the "rst set of experiments hydroxyethyl
methacrylate (HEMA) copolymers were photopolymerized using methanol as a solvent. HEMA was
studied because, due to its biocompatibility and
hydrophilic properties, it is used in many biomedical
applications including contact lenses, wound bandages,
and cell immoblization (Wichterle & Lim, 1960; Montheard, Chatzopoulos, & Chappard, 1992; Jen et al.,
1996; Ng & Tighe, 1976). Methanol was selected as the
solvent because the monomers, initiator, and polymer all
were highly miscible with it. In the second set of experiments octyl methacrylate (OcMA) copolymers were polymerized in varying amounts of hexanol. Although OcMA
is not a typical biopolymer, it was chosen for the second
set of experiments because it will thermally polymerize to
complete conversion, and thus allowed for more controlled experiments and better comparisons of molecular
weight between crosslinks. Hexanol was used in these
experiments because the polymer, monomer, and initator
were miscible with it. In conjunction with the experiments, this research utilizes a kinetic model to investigate
the e!ect of solvent concentration during polymerization
on the structure and properties of polymer hydrogels.
The model is compared with the experimental results and
further used to predict properties of polyacrylic acid
(PAA) hydrogels.
The model solves the di!erential kinetic balances on
the reacting species to determine the relative formation of
crosslinks and cycles during the polymerization. When
a multivinyl monomer is incorporated into the polymer
chains, a pendant double bond is formed. This pendant
double bond can react with a radical in the bulk solution

Fig. 1. Polymer networks with low and high degrees of cyclization.

to form a crosslink or react with the radical on its own


propagating chain to form a primary cycle. In the di!erential kinetic balances, the rates of crosslinking and primary cyclization are controlled by the concentration of
radicals in the bulk solution (bulk radical concentration)
and the e!ective concentration of radicals on the same
propagating chain as the pendant (local radical concentration), respectively. These radical concentrations
change with time as radicals are created through initiation reactions and terminate with each other. Additionally, the local radical concentration varies from the time
the pendant was formed, as the proximity of radicals on
the propagating chain is a function of how long the
pendant has existed and how far the radical which formed it has propagated. Using the idea of pendant birth
time developed by Tobita (Tobita & Hamielec, 1989;
Tobita, 1992), each pendant double bond is tracked separately to incorporate the varying reactivity with birth
time. By inclusion of the two radical concentrations (i.e.
the local and bulk concentration), the model predicts
how the rate of crosslinking and cyclization change with
conversion.
Development of a model that includes primary cyclization and varying pendant reactivity is important for
modeling hydrogels and predicting network structure.
Cyclization causes a much more loosely crosslinked material to be formed than would be predicted by the
conversion if only crosslinking was occurring. Fig. 1 gives
a visual representation of how crosslinking and cyclization a!ect the subsequent swelling of the polymer
network. When a polymer system is more highly crosslinked, the overall structure is more tightly held together,
adding rigidity and enhanced mechanical strength, while
reducing the subsequent swelling as shown in Fig. 1a.

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

When a polymer is more cyclized, as shown in Fig. 1b, the


overall network structure is "lled with ring structures,
and the backbone polymer chains are able to swell
further apart. The mesh size or molecular weight between
crosslinks (Mc) is signi"cantly increased by primary cyclization, leading to increased swelling and reduced
modulus as well as numerous changes in other network
properties. Evidence of cyclization is observed indirectly,
as it cannot be explicitly measured experimentally. The
presence of cyclization in hydrogels is indicated by the
heterogeneity which is seen with small-angle neutron
scattering in a poly(acrylic acid) and methylene bisacrylamide copolymers (Moussaid, Candau, & Joosten,
1994). Further, the delayed gel-point conversion from
what is predicted by the classical Flory}Stockmayer theory is evidence of the existence of signi"cant cyclization
reactions (Walling, 1945; Galina, Dusek, Tuzar, & Stokr,
1980; Dusek & Spevacek, 1980; Dusek, 1982; Boots,
Kloosterboer, & Hei, 1985). Similarly, primary cyclization has been used to explain the increased swelling
ratio, beyond what is predicted for an ideal network, that
is measured in polyacrylamide gels (Okay, Balimtas,
& Naghash, 1997).
In studying cyclization it is important to use both
experimental and modeling techniques. Numerical
approaches for studying cyclization abound in part
because of the di$culty determining cyclization rates
experimentally. Experimental techniques to measure
primary cyclization are limited to measuring the amount
of extractable, unreacted monomer or measuring the
mechanical properties, and deducing the degree of primary cyclization (Kloosterboer, 1988; Anseth, Bowman,
& Brannon-Peppas, 1996). Numerical models previously
developed generally fall into three categories: statistical,
space-based simulations, and kinetic approaches. Statistical models include work by Dusek and Ilavsky and
others (Flory, 1953; Stockmayer, 1943; Gordon, 1962;
Dusek & Ilavsky, 1975; Macosko & Miller, 1976; Miller
& Macosko, 1976; Dusek & Spevacek, 1980; Miller
& Macosko, 1988; Gordon & Malcolm, 1966) where
reaction probabilities control the growth of radical
chains. More recently, combined statistical and kinetic
models have been developed by Dusek and Somvarsky as
well as others to model the network formation of crosslinked polymers (Dusek & Somvarsky, 1996; Luo, Weng,
Huang, & Pan, 1997).
Space-based simulations (Monte Carlo, percolation,
kinetic gelation) which use a lattice structure to simulate
network formation have also been widely used over the
last two decades to model chain polymerizations
(Manneville & Seze, 1981; Boots & Pandey, 1984; Bansil,
Herrmann, & Stau!er, 1984; Kloosterboer, 1988; Simon,
Allen, Bennett, Williams, & Williams, 1989; Bowman
& Peppas, 1992; Anseth & Bowman, 1994; Chiu & Lee,
1995; Schroder & Oppermann, 1997). These simulations
are useful for modeling and describing the structural

3175

evolution of highly crosslinked polymers in which the


e!ects of heterogeneity are prominent.
Additionally, work on kinetic models has been done by
Tobita and Hamielec and others (Tobita & Hamielec,
1988; Tobita & Hamielec, 1989; Okay, 1993; Okay, Kurz,
Lutz, & Funke, 1995; Naghash, Okay, & Yildririm, 1995;
Naghash, Yagci, & Okay, 1997) using a pseudokinetic
approach where the non-ideal spatial e!ects are generally
averaged into the rate constants. Early models speci"cally on gel formation developed by Okay and Naghash
neglected cyclization (Okay, 1993; Naghash et al., 1995).
In later research, the rate constant for primary cyclization was assumed constant throughout the polymerization (Okay et al., 1995).
Landin and Macosko developed a mathematically
simpler kinetic model that included cyclization (Landin
& Macosko, 1988) by using a proportionality factor for
the fraction of pendants that are consumed in primary
cyclization reactions. Cyclization is again assumed to
occur at a constant rate and model parameters for cyclization and pendant reactivity are determined from experimental data such that cyclization rates are approximated
rather than predicted (Dusek, 1998). To predict primary
cyclization rates more accurately, the variation of pendant reactivity should be included. The model developed
for this work assumes pendant reactivity for cyclization is
controlled by the local radical concentration. Its "rst
principles approach also gives it the #exibility to predict
behavior in a large variety of experimental systems.

2. Computational methods
The numerical kinetic model has been described in
detail in a previous paper (Elliott & Bowman, 1999a,b).
In general, the model is unique because it develops and
solves the di!erential kinetic equations accounting for
the di!erence in reactivity of the pendant double bonds
spatially and during the polymerization. Monomeric and
pendant double bonds are tracked separately to capture
the local dynamics and reactivity of the pendant double
bonds. Calculation of the rate of consumption of monomeric double bonds is based on the kinetic expression
for a bimolecular collision, using the kinetic parameter
k times the concentrations of monomeric double bonds
N
and radical species in bulk solution [R ]. The con@
centration of bulk radicals [R ] is calculated using the
@
pseudo-steady-state assumption. Once a multifunctional
monomer is consumed, a pendant double bond is created, which can react either by crosslinking or cyclization. As shown in Fig. 2, both of these two mechanisms
of propagation of pendant double bonds (R ) are con
sidered: the reaction of pendant double bonds with the
radical on the same propagating chain (local radicals) to
form cycles and the reaction of pendant double bonds
with bulk radicals to form crosslinks. Secondary cycles

3176

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

Fig. 3. Radius containing the local radical for a particular pendant.


Fig. 2. Mechanism of monomeric- and pendant-double-bond reactions.

can also be produced, but for this work they are considered equivalent to crosslinks. The di!erence in reactivity of the two competing mechanisms is incorporated
into the apparent radical concentrations relevant to the
crosslinking and cyclization reactions.
The local radical concentration is the apparent concentration of the radical on the same propagating chain
for a speci"c pendant double bond. Pendant double
bonds that have existed for di!erent lengths of time each
have their own local radical concentration, which is
a function of when the pendant was created (birth time
[t ] and the current time. When the pendant is "rst
@
created, it is very close to the propagating radical and the
local radical concentration (and therefore the rate of
primary cycle formation) is high. Conversely, after long
times the local radical concentration diminishes dramatically. To calculate the local radical concentration, a volume is de"ned which includes both the pendant vinyl and
the propagating radical on the same kinetic chain that
initially formed the pendant as shown in Fig. 3. This
volume has a radius, which represents the distance between the pendant double bond and the radical and can
be calculated using statistics. The expression for the local
radical concentration [R (t, t )], as function of time and
J @
pendant birth time, including termination of local radicals with bulk radicals is
[R (t, t )]"exp[!k [R ](t!t )]
J @
R @
@
1
.
(1)
;
N [4/3(r #Cn(t, t )l)]


L
@
Here, k is the termination kinetic constant, N is
R

Avogadro's number, r is the monomer size, C is the

L
characteristic ratio (Flory, 1969), n is the number of
carbon}carbon bonds between the pendant double bond
and radical, and l is the length of a carbon}carbon bond.
The assumptions included in this equation are the following: (1) the distance the chain propagates can be calculated using statistics assuming an unperturbed chain;
(2) the molecular size of the crosslinking agent (with the

pendant double bond on one end) and the length the


radical propagates can be combined additively to approximate the distance between the pendant and the
radical; and (3) local radicals terminate with bulk radicals
by a second-order bimolecular kinetic reaction.
As shown below in Eq. (2), the rate of pendant (Pen)
consumption, R , is the sum of the rate of reaction with

bulk monomer and local radicals. The consumption of
pendants by cyclization is evaluated for pendants of all
birth times and summed:
R (t)"k [ Pen(t)][R ]

VJ
@
k R@ R exp[!k [R ](t!t )][ Pen(t, t )]
R @
@
@ .
#  
[4/3(r #Cn(t, t )l)]
N

L
@
 R@ 
(2)
In Eq. (2) the kinetic constant for propagation that leads
to crosslinking and cyclization are k and k , respecVJ

tively, which are both generally assumed to be equal to
k , the propagation kinetic constant. Using these kinetic
N
expressions, the numerical model tracks the formation of
pendant double bonds and their subsequent reaction
with bulk or local radicals. In this manner we are able to
predict the extent of cyclization and crosslinking.
The useful quantity to calculate for comparison with
experimental data is the molecular weight between crosslinks. The molecular weight between crosslinks is de"ned
as the polymer density, , (total weight of polymer/vol)
divided by the moles of crosslinked chains per unit volume, v, as shown in Eq. (1):

Mc" .
v

(3)

To "rst calculate the theoretical Mc for an ideal crosslinked network with complete conversion and no cyclization, the maximum number of crosslinked chains must
be known. The relationship between the concentration of
crosslinking agent and the moles of crosslinked chains
per unit volume, v can be generalized as the number of
double bonds (ndb) of the crosslinking agent times the

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

concentration of crosslinking monomer, [M ] (Hasa


VJ
& Janacek, 1967):
v"ndb[M ].
(4)
VJ
For a loosely crosslinked system where the concentration
of crosslinking agent is much less then the concentration
of monovinyl monomer, the density will be approximately the initial monovinyl concentration times the molecular weight of the monovinyl monomer. When the
weight fraction of the crosslinking agent is not negligible,
as in a more highly crosslinked system, the density of the
polymer network will be the initial double bond concentration, [DB ] times the average molecular weight of

a repeat unit on a double-bond basis, Mr. Thus, the
theoretical Mc for the divinyl/monovinyl copolymerization system like DEGDMA/OcMA will be the following:
Mr[DB ]
 .
Mc "

2[M ]
VJ

(5)

Using the simulation results, Mc is determined as


a function of conversion, X, and the degree of crosslinking for the non-ideal cases which include cyclization. The
polymer density in Eq. (3) is calculated based on the
concentration of double bonds that have been incorporated into the network, [DB ]X. This result assumes that

all monomers with at least one double bond reacted are
part of the network and contribute to the density, a reasonable assumption at high conversions. The concentration of crosslinks, v, in the network will be a function of
both conversion and the extent of cyclization. Every fully
reacted divinyl molecule like DEGDMA will contribute
two crosslinked chains. Divinyl molecules that cycle or
have an unreacted pendant will form a linear structure
and will not contribute to the number of crosslinked
chains. The total concentration of crosslinked chains
formed for the divinyl copolymerization will then be
twice the number of pendants that react in crosslinking
reactions, [ Pen ], and Mc is calculated as follows
VJ
Mr[DB ]X
 .
(6)
Mc"
2[Pen ]
VJ
Several parameters are needed for the model to specify
the polymer system being simulated when determining
Mc. The molecular size of the crosslinking molecule, r ,

are input as 4.5 and 6.7 As to represent DEGDMA and
PEG(600)DMA, respectively. These values are calculated
from the molecular weights of monomers, assuming
a spherical molecule with the double bonds on the radius.
The length of the carbon}carbon bond, l, is 1.54 As . The
characteristic ratio, C , relates the mean-squared endL
to-end distance calculated for a freely jointed chain to the
mean-squared end-to-end distance for the actual unperturbed chain. The characteristic ratio is related to how

3177

extended the chains will be in solution and varies with


solvent quality and the degree of solvation of the polymer
chains. The value of C is speci"c to the polymer comL
position and solvent used and extremely di$cult to determine experimentally in crosslinked polymers. For
these reasons the value of C is "t to experimental data.
L
In the "rst experiments methanol was used with HEMA
copolymers. A characteristic ratio of 3.2 was used to "t
the model data to the experimental results. In the second
set of experiments with OcMA copolymers and hexanol,
5.9 was used for the characteristic ratio. For the simulations with PAA, the characteristic ratio was set to 4.3.
The kinetic parameters, k , k , k , k , all remain conN R  VJ
stant throughout the simulation. The rate constant for
cyclization (k ) and crosslinking (k ) are assumed to be

VJ
equivalent to the kinetic constant for propagation (k ), as
N
the varying pendant reactivity is captured in the radical
concentrations. The chemical reactivity of the monovinyl
and divinyl crosslinking agent is assumed to be equivalent. The values of the kinetic parameters for k and
N
k were taken from experimental data for HEMA for all
R
simulations (Goodner, Lee, & Bowman, 1997).

3. Experimental methods
The monomers used in the experimental work were
diethyleneglycol dimethacrylate (DEGDMA), polyethyleneglycol 600 dimethacrylate (PEG600DMA), hydroxyethylmethacrylate (HEMA) and octyl methacrylate
(OcMA). DEGDMA and OcMA were purchased from
Polysciences (Warington, PA). PEG(600)DMA was
obtained from Sartomer (West Chester, PA) and HEMA
was purchased from Aldrich (Milwaukee, WI). The
monomers were used as received without additional
purifying or inhibiting. The solvent used with the
HEMA copolymers was methanol from Fisher (Fair
Lawn, NJ) and the photoinitiator was ,-dimethoxy-phenylacetophenone (DMPA) from Ciby-Geigy (Hawthorne, NY). Solutions of HEMA containing 2 or 10 mol%
crosslinking agent (DEGDMA or PEG(600)DMA) as
well as 0, 50, or 80 vol% solvent were photopolymerized
with 0.1 wt% (relative to the monomers) DMPA using an
ultraviolet light source which operated at approximately
18 mW/cm for 30 min. In the second set of experiments
with OcMA the solvent used was hexanol from Aldrich
(Milwaukee, WI). Analogous to the "rst set of
experiments, samples were created with 2 and 10 mol%
crosslinking agent (DEGDMA or PEG(600)DMA)
copolymerized with OcMA. Samples were thermally
polymerized with 1.0 wt% (relative to the monomers)
2,2-azobisisobutyronitrite (AIBN) at 703C for 90 min in
Te#on molds sealed with vacuum grease. Solutions were
made with 0, 20 and 50% solvent by volume.
Time}temperature scans were performed using a dynamic mechanical analyzer (DMA). A sinusoidal tensile

3178

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

force was applied to the sample while raising the temperature 53C/min to obtain the storage modulus of the
polymer system while in the rubbery region. By knowing
the storage modulus of the polymer, the average molecular weight between crosslinks, Mc, is calculated using the
following equation:
3
.
Mc"
E

(3)

In the above equation,  is the density of the polymer


system, is the temperature in Kelvin where the
modulus was obtained, and E is the storage modulus of
the polymer in the rubbery region. This equation is valid
assuming the material behaves as an ideal rubber, chain
ends can be neglected, i.e. the kinetic chain length is much
greater than the distance between crosslinks, and the
storage modulus is much greater than the loss modulus.
All of the assumptions should be valid for the systems
studied.
To determine the "nal conversion of the HEMA
copolymer samples, infrared (IR) spectra were obtained
between 4000 and 400/cm. For each polymer sample, the
peak area of the carbonyl group in the IR spectra was
compared to the peak area of the C"C (stretch) bond at
1637/cm. The ratio of these two areas was then normalized by the ratio of the peak areas of these two bonds
in the monomer solution. The conversion of the system
can be directly obtained by comparing these ratios since
the C"O bond (1720/cm) is una!ected by the polymerization reaction whereas the radical involved in the polymerization reaction propagates through the C"C bond.
IR experiments were performed on a portion of each
HEMA polymer sample both before and after the DMA
experiments to measure the amount of additional curing
that occurred during heating. Knowing the double bond
conversion is essential to compare the experimental
values for Mc with the values for Mc predicted by the
numerical model. The average molecular weight between
crosslinks strongly depends on the double bond conversion. IR experiments veri"ed that OcMA samples achieve
nearly 100% conversion as polymerized.

4. Results and discussion


Using the kinetic simulations, DMA, and IR experimental techniques, the e!ects of crosslinking agent
concentration, crosslinking agent size, and solvent
concentration on primary cyclization were investigated.
Numerical modeling of copolymerization in the presence
of varying amounts of solvent was performed. As the
numerical model tracks the consumption and creation of
each species as a function of time, the change in the
pendant reactivity with varying amount of solvent is
easily observed. Model results of the total pendant

Fig. 4. Model prediction of normalized pendant concentrations for 2%


DEGDMA/98% DEGDMA with 0% solvent, (**); 50% solvent,
(- - - - - -); and 80% solvent, (} } -) (monomer size"4.5 As , light
intensity"18 mW/cm).

double bond concentration as a function of polymerization time are shown in Fig. 4 for 2% DEGDMA/98%
HEMA. Results were normalized by the initial crosslinking agent concentration. The addition of solvent
decreases the amount of pendants that are building up by
increasing the rate at which they react away by cyclization. These results show how the model captures the
varying pendant reactivity that results from increasing
dilution of monomeric double bonds. With no solvent
present during the polymerization, the normalized
pendant concentration at 200 s is almost eleven times as
high as the normalized pendant concentration in the
polymerization performed in 80% solvent.
The increased reactivity is caused by the dilution of
monomeric double bonds with the addition of solvent.
The rate of monomeric double-bond consumption
decreases with solvent as it is directly proportional to the
monomeric double bond and bulk radical concentration.
Similarly, the rate of pendant crosslinking also decreases
with the dilution of the radical concentration in the bulk
solution. Consequently, the propagation rate will
dramatically decrease with solvent addition because
monomer units are added more slowly. The local rate
radicals propagate away from the pendant double bonds
also decreases as the propagation rate is reduced. The
distance between a radical and pendants on its chain will
increase less rapidly and the apparent local radical concentration will drop o! more slowly when solvent is
present. This phenomena is also illustrated in the two
pictures in Fig. 5. In Fig. 5a, where the growing radical is
surrounded by more monomer units when little or no
solvent is present, the radical is able to add repeat units
rapidly and does not have a signi"cant amount of time to
react with the pendant double bond to cyclize. The pendant double bond will then have an increased chance of
crosslinking. In Fig. 5b, when solvent is present, the

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

3179

Fig. 5. Solvent concentration e!ect on cyclization.

Fig. 7. The e!ect of solvent on Mc for DEGDMA/HEMA copolymers.


Simulation data for 2% (), and 10% DEGDMA, (;). Experimental
data for 2% (), and 10% DEGDMA () (monomer size"4.5 As , light
intensity"18 mW/cm). Theoretical Mc assuming no cyclization and
100% conversion for 2%, (**) and 10%, (- - - - -) crosslinking agent.

Fig. 6. Integral fraction of reacting pendants which form primary cycles


at each time for DEGDMA with 0% solvent, (**); 50% solvent,
(- - - - - -); and 80% solvent (} } -); (monomer size"4.5 As , light
intensity"18 mW/cm).

concentration of unreacted monomeric double bonds


will be diluted and the slowly growing radical chain will
have an increased chance of encountering the pendant
double bonds, causing more cyclization. Therefore,
primary cyclization, unlike crosslinking reactions, is
facilitated by increasing with solvent concentration. The
increased pendant reactivity with increasing solvent concentration during polymerization seen in Fig. 6 is attributed to greater degrees of primary cyclization.
The simulation is also used to investigate the extent of
primary cyclization for di!erent solvent amounts. The
numerical model can predict the fraction of pendants
that react by primary cyclization at each time during the
polymerization. Fig. 6 shows the model results for the
integral fraction of reacting pendants forming cycles as
a function of conversion for 2% DEGDMA/98%
HEMA with varying amounts solvent. The highest fraction of pendant cyclization occurs at the beginning of the
reaction because of the limited propagation of radicals
away from newly created pendants. These model results
correlate well with recent experimental and modeling

data of Okay and coworkers (Okay et al., 1995; Naghash


et al., 1997). Their work found high cyclization rates at
low conversions for methyl methacrylate and ethylene
glycol dimethacrylate copolymers (Naghash et al., 1997).
As reaction time goes on, the fraction forming cycles
decreases. At higher solvent concentrations (lower monomer concentrations) the pendant double bonds form
more cycles because the e!ective local radical concentration drops o! more slowly while the monomeric double
bond concentration is dramatically decreased by the solvent. The fraction of pendant double bonds reacting
away by cyclization drops o! less steeply the more solvent that is added, showing higher cyclization rates
throughout the reaction and further con"rming why the
pendant double bond concentration does not increase
with higher solvent amounts in Fig. 4.
Primary cyclization cannot be measured directly to be
compared with the simulation data. As discussed
previously, one method for measuring the degree of
cyclization is to determine the mechanical properties of
the polymer, speci"cally the average molecular weight
between crosslinks (Mc). Experimental DMA and simulation results for the Mc of 2 and 10% DEGDMA
copolymerizations with HEMA are shown in Fig. 7. The
theoretical Mc, assuming no cyclization and 100%
double bond conversion, is also plotted. The experimental results are higher than the theoretical, showing
that incomplete conversion and cyclization are preventing the polymer from reaching its crosslinking potential.
Cyclization causes the distance between crosslinks to be
greater (higher Mc) as pendants are consumed by cyclization reactions. For systems with both the 2 and 10%
crosslinking agent increasing the solvent amount increases the Mc. As expected, the Mc is lower for the

3180

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184


Table 1
Final conversion measured by IR for HEMA copolymers
Solvent (%)

0
50
80

DEGDMA (%)

PEG(600)DMA (%)

10

10

0.9
0.86
0.75

0.88
0.87
0.7

0.8
0.82
0.65

0.8
0.82
0.62

Fig. 8. The e!ect of solvent concentration on Mc for


PEG(600)DMA/HEMA copolymers. Simulation data for 2% (), and
10% PEG(600)DMA, (;). Experimental data for 2% (), and 10%
PEG(600)DMA () (monomer size"6.7 As , light intensity"
18 mW/cm). Theoretical Mc assuming no cyclization and 100% conversion for 2%, (**) and 10%, (- - - - -) crosslinking agent.

copolymer with 10% crosslinking agent because of the


increased crosslinking potential. Fig. 7 also shows that
model predictions simulated for the same "nal conversion are consistent with the experimental results. Experimental error associated with the measurement of the
rubbery modulus can contribute error of up to 20%.
Approximately 10% error is also associated with determining the "nal conversion of polymer with IR. Within
the error, the kinetic model is thus accurately predicting
the cyclization and crosslinking reaction in this
copolymer system. Results for PEG(600)DMA/HEMA
copolymers are similar (Fig. 8) to those for the DEGDMA crosslinked with HEMA. The experimental
values of Mc were greater than the theoretical for all but
one case. The deviation from the theoretical value again
demonstrates the importance of cyclization as the average Mc increases with increasing solvent amount. The
presented model data for Mc is at the same conversion
that was measured by the IR for each run. IR results for
DEGDMA and PEG(600)DMA are presented in
Table 1.
Similar trends are seen with the OcMA copolymer crosslinked with DEGDMA in Fig. 9 and
PEG(600)DMA in Fig. 10. As before, the theoretical Mc
at 100% conversion assuming no cyclization is plotted
on the "gure along with the experimental and modeling
results. All Mc results are above the theoretical line,
indicating that cyclization is occurring. As OcMA reached near 100% conversion at these polymerization conditions, the elevated Mc observed cannot be attributed to
incomplete conversion. Model and experimental results
again match very well within experimental error. Here,
error from di!erences in conversion is minimized.

Fig. 9. The e!ect of solvent on Mc for DEGDMA/OcMA copolymers.


Simulation data for 2% (), and 10% DEGDMA, (;). Experimental
data for 2% (), and 10% DEGDMA () (monomer size"4.5 As ,
thermally polymerized). Theoretical Mc assuming no cyclization and
100% conversion for 2%, (**) and 10%, (- - - - -) crosslinking agent.

When the DMA results of DEGDMA and


PEG(600)DMA are compared at the same crosslinking
agent concentration, the e!ect of crosslinking molecule
size can be ascertained. Figs. 11 and 12 shows the results
for 2 and 10% crosslinking agent with OcMA. In both
"gures the higher molecular weight crosslinking agent
leads to a more crosslinked polymer. These results are
consistent with other researchers' where increasing crosslinking agent size decreased swelling (Gonzales, Fan,
& Sevoian, 1996). Primary cyclization decreases with
larger crosslinking molecule size because the end-to-end
distance is larger; the pendant is further from the
propagating radical. As cyclization is most likely to occur
when the pendant double bond has just been formed,
a change in the initial distance between the pendant and
the radical strongly a!ects cyclization.
In the two sets of experiments performed, model results have been shown to be consistent with experimental
measurements of Mc in polymer gels. The model is #exible enough to simulate a variety of monomers and their
properties. Besides predicting the degree of primary cyclization and the Mc, the model can predict other important properties of hydrogels that are related to primary
cyclization. For instance, the e!ect of cyclization on the

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

Fig. 10. The e!ect of solvent concentration on Mc for


PEG(600)DMA/OcMA copolymers. Simulation data for 2% (), and
10% PEG(600)DMA (;). Experimental data for 2% (), and 10%
PEG(600)DMA () (monomer size"6.7 As , thermally polymerized).
Theoretical Mc assuming no cyclization and 100% conversion for 2%,
(**) and 10%, (- - - - -) crosslinking agent.

Fig. 11. The e!ect of monomer size on Mc for a copolymer with 2%


crosslinking agent. Simulation data for 2% DEGDMA (), and 2%
PEG(600)DMA (;), copolymerized with OcMA. Experimental data
for 2% DEGDMA (), and 2% PEG(600)DMA ().

gel-point conversion and subsequent equilibrium swelling can also be predicted. As stated in the introduction,
the presence of cyclization was noted when the gel-point
conversion during polymerization was higher then
predicted by classical Flory}Stockmayer theory. In
Fig. 13 the model prediction for the gel-point conversion
as a function of solvent amount and crosslinking agent
concentration is shown in a three-dimensional plot for
polyacrylic acid (PAA) and DEGDMA with a rate of
initiation equal to 1;10\ mol/l s. The gel point increases with solvent concentration due to the increase in
cyclization. The most dramatic increase in gel-point con-

3181

Fig. 12. The e!ect of solvent on Mc for a copolymer with 10% crosslinking agent. Simulation data for 10% DEGDMA (), and 10%
PEG(600)DMA (;), copolymerized with OcMA. Experimental data
for 10% DEGDMA (), and 10% PEG(600)DMA ().

Fig. 13. Gel-point conversion prediction as a function of crosslinking


agent concentration and solvent concentration for PAA copolymerized
with DEGDMA. (rate of initiation"1;10\ mol/l s).

version occurs at about 85% solvent, and at very high


solvent amounts (above 93%) no gelation occurs. The gel
point decreases with increasing crosslinking agent
concentration as expected because of the increased crosslinking.
The experimental and numerical results demonstrate
that monomer size, monomer concentration, and solvent
concentration during polymerization, all e!ect the crosslinking reaction and the hydrogel properties. This result
has signi"cant implications for biomedical applications
where the mechanical, di!usion, and swelling properties
of the polymer are critical. The relationship between Mc
and solvent concentration is especially important when
speci"c swelling properties are critical to an application.
Using the model predictions for Mc and M for the 2%
L

3182

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

Fig. 14. Predicted swelling ratio of 2% DEGDMA/98% PAA


copolymer polymerized in varying amounts of solvent (rate of initiation"1;10\ mol/l s).

DEGDMA/98% PAA system, the equilibrium swelling


ratio (volume swollen/volume dry) of the polymers was
predicted by the Flory}Rehner equation (Flory, 1953).
The chi factor was assumed to be 0.25 for a generic
thermodynamically good solvent. Results shown in
Fig. 14 demonstrate that increasing the amount of solvent during polymerization increased the equilibrium
swelling ratio of the resulting polymer, especially at solvent amounts over 60%. Increasing the solvent amount
from 60 to 80% changes the equilibrium swelling ratio
from 16 to 36. This di!erence in swelling ratio is caused
by the decreased crosslinking when the solvent concentration during polymerization is increased. Mc increased
from 6000 to 16,000 g/mol when the solvent concentration present during polymerization was increased from
50 to 80 vol% solvent. The fact that the amount of
solvent present during polymerization changes the equilibrium swelling has important implications on hydrogel
design.

5. Conclusions
The kinetic model and experimental results presented
provide insight into the e!ect of solvent concentration,
crosslinking agent size, and crosslinking agent concentration on the extent of crosslinking and primary cyclization
during the photopolymerization of multifunctional
monomers. Simulation results using the kinetic model
demonstrate how pendant reactivity varies during the
polymerization. Adding solvent to the reaction increases
the probability of cycling, due to the diluted concentration of monomer, and slowed rate of polymerization
causing the local radical on its own chain to remain
longer in close proximity to pendant double bonds. In

both the modeling and experimental results, cyclization


rates change with monomer size and monomer concentration (solvent concentration). Simulation results also
show that increasing the amount of solvent will increase
the cumulative amount of reacting pendants that are
consumed by cyclization. Further, because the fraction of
pendant double bonds that form cycles is highest initially
after it is formed, the size of the monomer strongly e!ects
cyclization rates. With a smaller monomer, a pendant is
closer to the propagating radical that created it, and can
more readily react intramolecularly to form a cycle. The
model is able to predict experimental data quantitatively
for the e!ect of solvent concentration during polymerization on the average molecular weight between crosslinks for DEGDMA and PEG(600)DMA copolymerized
with HEMA or OcMA. The agreement of the model
with experimental data validates the model as a
useful tool in studying polymerization of highly crosslinked monomers, and the model can be used to predict
the gel point and swelling ratio of more typical PAA
hydrogels. The increase in cyclization and the average
Mc is enough to change the predicted swelling properties
of the polymer for solvent amounts above 60% of the
volume. The solvent concentration during polymerization and crosslinking agent concentration will also
change the gel-point conversion of polymer because of
varying degrees of cyclization. These changes are important when designing hydrogels for speci"c biomedical
applications.
In conclusion, accounting for primary cyclization is
key to understanding the e!ects of solvent concentration
and comonomer composition on hydrogel crosslinking
density. Experiments and modeling show that greater
primary cyclization results with more solvent and smaller
crosslinking molecules. Good agreement between the
model and experiments validate it as a valuable
predictive tool in determining the crosslinking density,
swelling, and gel-point conversion of polymer gels and
hydrogel systems.

Acknowledgements
The authors would like to acknowledge Jason Brown
for help with the DMA experiments, the Camille Dreyfus
Teacher-Scholar Program; National Institutes of Health
for its support through a research grant (DE10959-01A2);
and the Presidential Faculty Fellow Program at the
National Science Foundation for "nancial support.

References
Anseth, K. S., & Bowman, C. N. (1994). Kinetic gelation model predictions of crosslinked polymer network microstructure. Chemical Engineering Science, 49, 2207}2217.

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184


Anseth, K. S., Bowman, C. N., & Brannon-Peppas, L. (1996). Mechanical properties of hydrogels and their experimental determination.
Biomaterials, 17, 1647}1657.
Bae, Y. H., & Kim, S. W. (1993). Hydrogel delivery systems based on
polymer blends, block copolymers or interpenetrating networks.
Advanced Drug Delivery Reviews, 11, 109}135.
Bansil, R., Herrmann, H. J., & Stau!er, D. (1984). Computer simulation
of kinetics of gelation by addition polymerization in a solvent.
Macromolecules, 17, 998}1004.
Boots, H. M. F., & Pandey, R. B. (1984). Qualitative percolation study
of free-radical cross-linking polymerization. Polymer Bulletin, 11,
415}420.
Boots, H. M. J., Kloosterboer, J. G., & Hei, B. M. M. V. D. (1985).
Inhomogeneity during the bulk polmerisation of divinyl compounds: di!erential scanning calorimetry experiments and percolation theory. The British Polymer Journal, 17, 219}223.
Bowman, C. N., & Peppas, N. A. (1992). A kinetic gelation method for
the simulation of free-radical polymerizations. Chemical Engineering Science, 47, 1411}1419.
Chiu, Y. Y., & Lee, L. J. (1995). Microgel formation in the free radical
crosslinking polymerization of ethylene glycol dimethacrylate
(EGDMA). II Simulation.. Journal of Polymer Science: Part A:
Polymer Chemistry, 33, 269}283.
Dusek, K. (1982). Developments in polymerization, Vol. 3. Network formation and cyclization in polymer reactions. Englewood Cli!s, NJ:
Applied Science Publishers.
Dusek, K. (1998). Network formation involving polyfunctional polymer
chains.. In R. F. T. Stepto (Ed.), Polymer networks: principles of their
formation structure and properties. London: Blackie Academic
& Professional.
Dusek, K., & Ilavsky, M. (1975). Cyclization in crosslinking polymerization I. Chain polymerization of a Bis unsaturated monomer (monodisperse case). Journal of Polymer Science, 53,
57}73.
Dusek, K., & Somvarsky, J. (1996). Topological nanohomogeneities in
polymer networks. Macromolecular Symposia, 106, 119}136.
Dusek, K., & Spevacek, K. (1980). Cyclization in vinl}divinyl
copolymerization. Polymer, 21, 750}756.
Elliott, J. E., & Bowman, C. N. (1999a). Kinetics of primary cyclization
reactions in crosslinked polymers: an analytical and numerical
approach to heterogeneity in network formation. Macromolecules,
32, 8621}8628.
Elliott, J. E., & Bowman, C. N. (1999b). Primary cyclization reaction in
crosslinked polymers. In B. T. Stokke, & A. Elgsaeter (Eds.), Wiley
polymer networks group review series (Vol. 2). New York: Wiley
(pp. 27}38).
Ende, M. T. A., & Peppas, N. A. (1996). Transport of ionizable drugs
and proteins in crosslinked poly(acrylic acid) and poly(acrylic acidco-2-hydroxyethyl methacrylate) hydrogels. I. Polymer characterization. Journal of Applied Polymer Science, 59, 673}685.
Flory, P. J. (1953). Principles of polymer chemistry. Ithica, London:
Cornel University Press.
Flory, P. J. (1969). Statistical Mechanics of Chain Molecules. New York:
Interscience Publishers, (p. 11).
Galina, H., Dusek, K., Tuzar, Z., & Stokr, J. (1980). The structure of low
conversion polymers of ethylene dimethacrylate. European Polymer
Journal, 16, 1043.
Gonzales, D., Fan, K., & Sevoian, M. (1996). Synthesis and swelling
characterizations of a poly(gamma-glutamic acid) hydrogel. Journal
of Polymer Science Part A: Polymer Chemistry, 34, 2019}2027.
Goodner, M., Lee, H. R., & Bowman, C. N. (1997). Method for determining the kinetic parameters in di!usion controlled free radical
homopolymerizations. Industrial & Engineering Chemistry Research,
36, 1247}1252.
Gordon, M. (1962). Good's theory of cascade processes applied to the
statistics of polymer distributions. Proceedings Royal Society London, Series A, A268, 240}259.

3183

Gordon, M., & Malcolm, G. N. (1966). Con"gurational statistics of


copolymer systems. Proceedings Royal Society London, Series A,
A295, 29}54.
Hasa, J., & Janacek, J. (1967). Journal of Polymer Science Part C:
Polymer Symposium, 16 (Part 1), 317.
Jen, A. C., Wake, M. C., & Mikos, A. G. (1996). Review: Hydrogels for
cell immoblization. Biotechnology and Bioengineering, 50, 357}364.
Kao, F.-J., Manivannan, G., & Sawan, S. P. (1997). UV curable bioadhesives: Copolymers of N-vinyl pyrrolidone. Journal of Biomedical
Materials Research, 38, 191}196.
Kloosterboer, J. G. (1988). Network formation by chain crosslinking
photopolymerization and its applications in electronics. Advances in
Polymer Science, 84, 1}57.
Landin, D. T., & Macosko, C. W. (1988). Cyclization and reduced reactivity of pendant vinyls during the copolymerization of methyl methacrylate and ethylene glycol dimethacrylate. Macromolecules, 21, 846}851.
Lehr, C. M., Bouwstra, J. A., Vanhal, D. A., Verhoef, J. C., & Junginger,
H. E. (1992). Release of a peptide drug from microspheres of Poly(2Hydroxyethyl-methacrylate) relevant for the development of an oral
bioadhesive drug delivery system. European Journal of Pharmaceutics and Biopharmaceutics, 38, 55}60.
Luo, Y. W., Weng, Z. X., Huang, Z. M., & Pan, Z. R. (1997). Modeling
of e!ective crosslinking density of the gel in free-radical copolymerization of vinyl/divinyl monomers. Journal of Applied Polymer
Science, 64, 1691}1699.
Macosko, C. W., & Miller, D. R. (1976). A new derivation of average
molecular weights of nonlinear polymers. Macromolecules, 9, 199}206.
Manneville & de Seze, L. (1981). Numerical methods in the study of
critical phenomena. Springer: Berlin.
Miller, D. R., & Macosko, C. W. (1976). A new derivation of post gel
properties of network polymer. Macromolecules, 9, 206}211.
Miller, D. R., & Macosko, C. W. (1988). Network parameters for
crosslinking of chains with length and site distribution. Journal of
Polymer Science Physical Education, 26, 11.
Montheard, J. P., Chatzopoulos, M., & Chappard, D. (1992).
2-hydroxyethyl methacrylate (HEMA): chemical properties and
applications in biomedical "elds. Review of Macromolecules Chemical Physics, C32, 1}34.
Moussaid, A., Candau, S. J., & Joosten, J. G. H. (1994). Structural and
dynamic properties of partially charged poly(acrylic acid) gels:
Nonergodicity and in homogeneities. Macromolecules, 27, 2102}2110.
Naghash, H. J., Okay, O., & Yildririm, H. (1995). Gel formation in
free-radical crosslinking copolymerization. Journal of Applied Polymer Science, 56, 477}483.
Naghash, H. J., Yagci, Y., & Okay, O. (1997). Gel formation by
chain-crosslinking photopolymerizaiton of methyl methacrylate
and ethylene glycol dimethacrylate. Polymer, 38, 1187}1196.
Ng, C. O., & Tighe, B. J. (1976). Polymers in contact lens applications
VI. The &dissolved' oxygen permeability of hydrogels and the design
of material for use in continuous wear lenses. The British Polymer
Journal, 8, 118}123.
Okay, O. (1993). Kinetic modeling of network formation and properties
in free-radical crosslinking copolymerization. Polymer, 35, 786}807.
Okay, O., Balimtas, N. K., & Naghash, H. J. (1997). E!ects of cyclization and pendant vinyl group reactivity on the swelling behavior of
polyacrylamide gels. Polymer Bulletin, 39, 233}239.
Okay, O., Kurz, M., Lutz, K., & Funke, W. (1995). Cyclization and reduced
pendant vinyl group reactivity during the free-radical cross-linking
polymerization of 1,4-divinylbenzene. Macromolecules, 28, 2728}2737.
Peppas, N. A. (Ed.) (1987). Hydrogels in medicine and pharmacy. (Vol. II).
Boca Raton, FL: CRC Press.
Peppas, N. A., & Khare, A. R. (1993). Preparation, structure and
di!usional behavior of hydrogels in controlled release. Advanced
Drug Delivery Reviews, 11, 1}35.
Schroder, U. P., & Oppermann, W. (1997). Computer simulation of
network formation via crosslinking copolymerization. Macromolecular Theory and Simulations, 6, 151}160.

3184

J. E. Elliott et al. / Chemical Engineering Science 56 (2001) 3173}3184

Simon, G. P., Allen, P. E. M., Bennett, D. J., Williams, D. R. B.,


& Williams, E. H. (1989). Nature of residual unsaturation during
cure of dimethacrylates examined by CPPEMAS 13C NMR and
simulation using a kinetic gelation model. Macromolecules, 22,
3555}3561.
Stockmayer, W. H. (1943). Theory of molecular size distribution and gel
formation in branched polymers. Journal of Chemical Physics, 11, 45.
Tobita, H. (1992). Control of network structure in free-radical crosslinking copolymerization. Polymer, 33, 3647}3657.
Tobita, H., & Hamielec, A. E. (1988). A kinetic model for network
formation in free radical polymerization. Makromolekulare Chemie,
Macromolecular Symposia, 20/21, 501.

Tobita, H., & Hamielec, A. E. (1989). Modeling of network formation in free radical polymerization. Macromolecules, 22,
3098}3105.
Walling, C. J. (1945). Gel formation in addition polymerization. Journal
of American Chemical Society, 67, 441.
Wheeler, J. C., Woods, J. A., Cox, M. J., Cantrell, R. W., Watkins, F. H.,
& Edlich, R. F. (1996). Evolution of hydrogel polymers as contact
lenses, surface coatings, dressings, and drug delivery systems. Journal of Long-Term Ewects of Medical Implants, 6, 207}217.
Wichterle, O., & Lim, D. (1960). Nature, 185, 117.

Potrebbero piacerti anche