Vous êtes sur la page 1sur 11

d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

available at www.sciencedirect.com

journal homepage: www.intl.elsevierhealth.com/journals/dema

Formation of gaps at the llerresin interface induced by

polymerization contraction stress
Gaps at the interface
Li Feng a, , Byoung I. Suh a , Adrian C. Shortall b

Research and Development, Bisco Inc., Schaumburg, IL 60193, USA

School of Dentistry, University of Birmingham, Birmingham, UK

a r t i c l e

i n f o

a b s t r a c t

Article history:

Objectives. To investigate the movement of resin matrix with respect to the ller particles of

Received 10 June 2009

lled composites during their photo cure without or with polymerization contraction stress

Received in revised form


7 January 2010

Methods. Two types of composites were prepared. Glass beads as macroscopic llers were

Accepted 5 March 2010

placed into the center of a bis-GMA/TEGDMA resin to make single bead-embedding composites and a variety of llers of different compositions, sizes, and shapes were mixed with
another bis-GMA/TEGDMA resin to make lightly lled composites. They were photo cured in


a cavity constructed with an acrylic or aluminum ring sitting on a polyester strip. Bonding to


the ring constrained the polymerization shrinkage and thus produced a PCS. The formation

Fillerresin interface

of gaps between the ller and the resin was detected by optical microscopy for the glass


beadresin systems, and by light attenuation and scanning electron microscopy (SEM) for


the llerresin composites.


Results. In general, for composites with untreated llers, the optical microscopy and
SEM revealed gaps at the llerresin interface only when they were cured under constrained shrinkage conditions. These composites attenuated more light when cured under
constrained shrinkage conditions than when under non-constrained conditions. For the
composites with silane-treated llers, no gaps were observed. Some did not show any signicant difference in light attenuation when cured under either constrained or non-constrained
Conclusions. The resin tends to move away from the ller particles under the inuence of
PCS. Strengthening the llerresin interaction, such as by the use of silane-treated ller,
may help prevent the resin departure and thus the formation of gaps.
2010 Published by Elsevier Ltd on behalf of Academy of Dental Materials. All rights



Polymerization contraction stress (PCS) has been studied

extensively in the eld of contemporary restorative dentistry,

having spawned a plethora of research papers [19]. This is

largely due to two related reasons: the unique curing conguration that produces stress and the grim consequence the
stress leads to. PCS is the result of the interplay between
polymerization induced contraction, conned volume for the

Corresponding author at: Bisco Inc., 1100 W Irving Park Road, Schaumburg, IL 60193, USA. Tel.: +1 847 534 6096; fax: +1 847 891 6705.
E-mail address: lfeng@bisco.com (L. Feng).
0109-5641/$ see front matter 2010 Published by Elsevier Ltd on behalf of Academy of Dental Materials. All rights reserved.


d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

Table 1 Materials used.

Glass bead
Bonding agent
Separating agent



Bis-GMAa /TEGDMAb (50:50 for beads and 60:40 for llers) plus 0.50%
CQc and 0.50% EDMABd
All Bond 2 One Step
316 Silicone Release Spray

Bisco, Inc., Schaumburg, IL

Bisco Bisco
Dow Corning, Midland, MI

Bis-phenol A glycidyl dimethacrylate (Esstech).

Triethylene glycol dimethacrylat (Esstech).
Camphorquinone (Hampford Research, Stratford, CT).
Ethyl-4-dimethylamino benzoate (Lancaster Synthesis, Windham, NH).

restorative, and constrained shrinkage by dental bonding.

Since it always tends to pull the restorative material away from
the tooth cavity wall, PCS presents a constant threat to the
integrity of the restoration.
PCS is essentially a tensile stress [1014]. As a restorative material (usually a resin composite, which will be used
thereafter) cures with constrained shrinkage, it is like being
stretched in order to accommodate the conned volume. This
strain produces a tension and thus the tensile stress. Under
a free shrinking condition, such as that with no bonding to
the cavity wall, the composite will not develop PCS, at least
not the type we usually refer to in dentistry. Since a composite has to bond to the cavity wall to prevent microleakage and
to withstand the load, PCS is perpetually present as long as
the composite shrinks upon curing. In addition to imposing a
pulling stress to the cavity wall, PCS simultaneously acts as
an internal tensile stress on the composite as well. In contrast
to that on dental bonding, however, the effect of PCS on the
curing material itself has been largely ignored. It is probably
because, compared with an interfacial bonding, the composite is a much stronger material. So presumably no signicant
damage can be expected. From the surface, that may be true.
But this does not mean the composite is not affected. Based
on the results of a handful of publications [5,12,1521], the
mechanical properties, especially the strength, appeared to
be lower for the composites cured with PCS than those without [12,1721]. For example, 30 min after light cure and with
its shrinkage constrained, hybrid composite Z100 showed a
tensile modulus of 3.5 GPa [21]. In comparison, its Youngs
modulus was at least 6 GPa after the similar length of cure
without constraint [22]. It is likely that the PCS may have had
altered the structure of the cured composite and therefore
undermined its mechanical properties in the former case.
One of the common characteristics among the aforementioned researches is that they all used ller reinforced resin
composites as the restorative materials [1522]. An important
parameter with such materials is the llerresin interactions
[23,24]. An intimate contact between the two materials is
a prerequisite for optimizing the mechanical properties. We
hypothesized that, due to its volume change, the curing resin
will move with respect to the ller particle and this rearrangement can alter their contacting state. Moving toward the
particle will generate hydrostatic pressure onto it, improving the contact. Conversely, moving away from the particle
will create gaps (voids) at the interface, which weakens the
llerresin interaction and subsequently compromises the
properties. Therefore, studying the alteration of their inter-

face induced by PCS can shed some light on its inuence on

the mechanical properties.
This work was intended to investigate the movement of
resin matrix with respect to the ller particle during photo
cure of dimethacrylate based resins and their slightly lled
composites. Its occurrence was inferred by the formation, or
the lack, of gaps between the ller and resin. Three techniques
were utilized to detect the change of interface before and after
cure: (1) Gap formation between a macroscopic bead and the
resin by an optical microscope, (2) Light attenuation by gapenhanced scattering by a radiometer, and (3) Gap formation
between ller particles and the resin by a scanning electron
microscope (SEM).


Materials and methods

Table 1 lists the materials used in this study. The glass bead,
served as a macroscopic ller particle for direct visualization, was made from a glass ber (0.3 mm in diameter) by
aming one of its ends with a propane torch. The resin composition comprised of a typical dimethacrylate blend used for
commercial dental composites. The bonding agents created
a constrained shrinkage condition by bonding the composite
to the ring while the separating agent freed the shrinkage by
preventing the composite from adhering to the ring.
Table 2 shows the seven tested llers with varying compositions, shapes, and sizes. They were either used untreated
(as received) or silane treated to strengthen their interactions
with the resin. A composite slurry was made by thoroughly
blending each of the llers into the 60:40 bis-GMA/TEGDMA
resin by grind-mixing with a ceramic mortar and pestle. The
resin was only lightly lled (20 wt% for Spherical 1, 40 wt% for
Irregular 2, and 10 wt% for the rests) to allow enough light to
pass through so sufciently accurate irradiance readings could
be accomplished.
The constraint of polymerization shrinkage was provided
by a circular cavity, constructed by sitting either an acrylic
or aluminum ring (6.0 mm in inner diameter, 9.5 mm in outer
diameter, and 2.0 mm in height) on a clear polyester strip and
covered with another strip. Because of its high compliance,
the polyester strip deformed easily under PCS, contributing little constraint to the shrinkage of a curing resin or composite.
Therefore, the top and bottom of the ring can be approximately
treated as unbonded surfaces and the C-factor was 1.5 for
both types of rings. Acrylic rings were either sandblasted and
primed with bonding agent One Step to constrain shrinkage,


d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

Table 2 Fillers tested.


Trade name

Spherical 1
Spherical 2
Spherical 3

Zeeospheres W-210

Irregular 1
Irregular 2
Irregular 3

Raysorb T-4000
Aerosil OX-50

Aluminum silicate
Barium aluminum
Barium glass
Strontium glass
Strontium glass

3M St. Paul, MN
Suzuki Yusu Industrial Osaka, Japan

Schott Glass Landshut, Germany

Esstech Essington, PA
Ferro Cleveland, OH
Degussa Theodore, AL



All the llers were used either as received (untreated), or silanized with -methacrxypropyltrimethoxy silane (silane treated).

or coated with the separating agent to allow free shrinkage

(non-constraint). The aluminum ring was in addition primed
with bonding agent All Bond 2 before One Step coating. All
Bond 2 produced stronger bonding to the aluminum surface
than One Step to better withstand the higher PCS using this
more rigid material. Note that the non-constrained condition
is relative to the constrained one in the context. Even with the
use of the separating agent, there was still some constraint
to shrinkage from interactions between the composite and
the ring and between the composite and the polyester strips.
But their contributions were much smaller than PCS and thus
considered negligible.

For optical microscope observations, a specimen was prepared by pipetting the 50:50 bis-GMA/TEGDMA resin into a
cavity formed by an acrylic or aluminum ring and then carefully dropping one glass bead (0.51.0 mm in diameter) into
the resin near the center. No cover strip was applied for this
set of experiments. Three types of glass beads were fabricated: smooth, smooth but with a nick (cut with a scalpel
when the bead was still hot and pliable), and roughened
(by sandblasting with 50 m Al2 O3 ). They were used without
treatments, coated with the separating agent, or treated with
silane. Treatments for the last one included 9.5% hydrouoric acid etching (Porcelain Etchant, Bisco), water rinsing,

Fig. 1 Optical microscope images of smooth glass beads in the 50:50 bis-GMA/TEGDMA resin, cured with non-constrained
polymerization shrinkage (Scale bar = 0.5 mm): (A) untreated and uncured as the control, (B) untreated, (C) separating agent
treated, and (D) silane treated. Like the uncured (A), none of the cured (BD) showed the gap formation at the beadresin


d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

air drying, and silane coating (Bis-Silane, Bisco). The 50:50

GMA/TEGDMA ratio was thought to give better wetting to
roughened glass beads than the 60:40 one. The specimen
on a white mixing pad was placed on the platform of TechnoLook TW-TL10M microscope (Caltex Scientic, Irvine, CA).
Optical images were recorded both before and after the resin
was light cured for 20 s at 500 mW/cm2 by an LED curing
unit (Skylight by DMETE, Korea). Each set of experiment was
conducted at least three times to ensure the reproducibility.
For the measurement of light attenuation, each of the
lightly lled composites was transferred into a cavity formed
by an aluminum ring, slightly over the edge of the ring, and
then covered with another clear polyester strip, which reduced

the thickness of the material to that of the ring (2.0 mm). Aluminum rings were exclusively used here because they would
not interfere with the light measurements because of their
opacity. The composite was irradiated with a halogen curing
unit (VIP, Bisco) under 500 mW/cm2 for 2 20 s from both top
and bottom. After the strips were removed, the cured specimen was placed between the light guide of the VIP and a digital
radiometer (Cure Rite 8000, EFOS, Mississauga, ON). Acting as a
lter window, the cured composite attenuated the penetrating
light by scattering as well as absorbing. The way of measuring
the attenuated light was very similar to that used by Musanje
and Darvell [25]. The formation of gaps at the llerresin interface increased the mismatch of refractive indices, making the
material less transparent [26]. A chemical cured composite

Fig. 2 Optical microscope images of smooth glass beads in the 50:50 bis-GMA/TEGDMA resin, cured with constrained
polymerization shrinkage (Scale bar = 0.5 mm): (A) untreated and uncured as the control, (B) untreated (acrylic ring), (C)
separating agent treated (acrylic ring), (D) untreated (aluminum ring), and (E) silane treated (aluminum ring). After cure,
(BD) developed gaps while (E) did not.

d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729


Fig. 3 Optical microscope images of smooth glass beads with a nick in the 50:50 bis-GMA/TEGDMA resin, cured with
non-constrained polymerization shrinkage (Scale bar = 0.5 mm): (A) untreated and uncured, (B) the same specimen as in (A)
but cured, (C) separating agent treated and uncured, and (D) the same specimen as in (C) but cured. The nicks not seen in
the uncured (A) and (C) were now visible after cure in (B) and (C), indicated by the arrows.

with 10 wt% TiO2 powder (P25, Degussa) was prepared for estimating light transmission around the peripheral of the cured
composite: there could be gaps along the margin between the
composite and the separating agent treated ring after the polymerization shrinkage. Four to ve specimens were measured
for each condition. The t-test was used to compare the means
of measured light irradiance of those without and with polymerization shrinkage constraint.
Specimens for SEM observations were similarly cured as
those above. But this time acrylic rings were used instead since
the cured composite inside the aluminum ring had a strong
tendency to debond from the ring during polishing. After cure
and removal of the strips, the top of the specimen was ground
into a at surface with a 320 grit SiC paper (3M). The surface
was then consecutively polished with a 1500 grit SiC paper
(3M), 1 m, and 0.3 m Al2 O3 papers (both from Buehler, Lake
Bluff, IL) in the absence of water. The polished specimens (two
for each experimental condition) were air cleaned, sputtercoated with gold, and observed with SEM SM-510 (Topcon,
Tokyo, Japan) under 15 kV.



Fig. 1 shows optical microscope images of smooth glass beads

embedded in the resin (one bead in each specimen), cured
with non-constrained polymerization shrinkage. Just like the

uncured one (Fig. 1(A)), none of the specimens showed gaps

at the beadresin interface, regardless of the treatments the
beads had received (Fig. 1(B)(D)). All the triplicates reproduced
the same results. When the shrinkage was constrained by
bonding the resin to the ring, however, only that with silanetreated ller consistently remained gap free (Fig. 2(E)); the
other two developed distinguishable gaps (Fig. 2(B) and (C)).
The gap formation was more prominent when the resin was
constrained by the aluminum ring (Fig. 2(D)) than the acrylic
ring (Fig. 2(B)). In fact, for the untreated beads, gaps appeared
in all three specimens with aluminum rings while only two in
six with acrylic rings. For the separating agent treated beads,
all the specimens showed gaps.
Both smooth glass beads in Fig. 3 had a nick carved
with a scalpel when they were still hot. While they were
invisible when the resin was uncured (Fig. 3(A) and (C)) the
nicks became recognizable after cure on the beads untreated
(Fig. 3(B)) and treated with the separating agent (Fig. 3(D)).
Note that the formation of the gaps was under the condition that the shrinkage was non-constrained, which did not
produce any gaps on the smooth glass beads in Fig. 1. Fig. 4
compares some subtle changes in transparency of one smooth
glass bead and two roughened glass beads. All the specimens
were cured with non-constrained shrinkage so no PCS was
expected. There was no noticeable change for the smooth
bead before and after cure (Fig. 4(B) vs. (A)), whereas the transparency was somewhat reduced after cure for the roughened


d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

Fig. 4 Optical microscope images of glass beads in the 50:50 bis-GMA/TEGDMA resin, cured with non-constrained
polymerization shrinkage (Scale bar = 0.5 mm): (A) smooth and uncured, (B) the same specimen as in (A) but cured, (C)
roughened and uncured, (D) the same specimen as in (C) but cured, (E) roughened, separating agent treated, and uncured,
and (F) the same specimen as in (E) but cured. After cure, while the transparency of (B) remained unchanged in comparison
with that of (A), the transparencies of (D) and (F) became lower than those in (C) and (E), respectively.

one (Fig. 4(D) vs. (C)). The change was more discernable for the
roughened and separating agent treated bead (Fig. 4(F) vs. (E)).
It is worth noting that we took a great effort to sure that these
reductions of transparencies were not an optical illusion, such
as defocusing, by ne tuning the focus as well as by using three
identically fabricated beads.
Fig. 5 presents the light irradiance attenuated by composites with various llers untreated and silane-treated, and
cured non-constrained and constrained. Also given were four
different controls in the caption: the irradiances for (I) a neat
resin only (with no ller), (II) a completely clear specimen (with

an empty ring), (III) a cured opaque composite (with TiO2 ) with

potential gaps at the margin, and (IV) a totally opaque ring
(with a black disk covering the opening). As the data show,
the transmission of light was negligible either through the
margin (Control III) or outside the ring (Control IV). While the
neat resin yielded statistically the same attenuated irradiance
(Control I), all the composites with untreated llers produced
a slightly but signicantly lower attenuated irradiance when
their shrinkage was constrained than those non-constrained
(Fig. 5(A)). Solely based on the results, one cannot tell the
absolute status of gaps under the either curing conditions.

d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729


Fig. 5 Comparison of light irradiance through the

composites bonded aluminum rings cured with and
without constrained polymerization shrinkage. (A)
Untreated llers and (B) silane-treated llers. The pairs
with two different horizontal bars were statistically
different (p < 0.05). For every ller-containing pair in (A), the
composite cured without constraint showed a higher
irradiance than that with constraint. For three of the seven
ller-containing pairs in (B), however, an identical
irradiance was produced in both curing conditions. There
were four controls: (I) that with neat resin only for
assessing the effect of ller (the rst two columns in (A)
and (B)); (II) that with no materials in the ring for the
non-attenuated irradiance 450(10) mW/cm2 ; (III) that with
the chemical cured resin and 10% TiO2 for the irradiance
leaked through the margin 3(1) mW/cm2 ; and (IV) that
with a black disk covering the ring for the irradiance
coming from outside of the ring 1(0) mW/cm2 .

Nevertheless, they did suggest that more gaps were produced

under the constrained condition because of the lower translucency. Conversely, the attenuated irradiance became equal
for three of the seven composites consisting of silane-treated
llers (Fig. 5(A)), suggesting that strengthening the llerresin
interaction could resist the pulling away stress and thus prevent the formation of gaps.
The use of a curing light unit together with a radiometer
to assess the attenuated irradiance may not be as sophisticated as some other techniques. But nevertheless, the results
were reproducible and therefore reliable, thanks to the precise
controls of the thickness of specimens, the consistency of the
irradiance from the curing unit, and the position of specimens
on the radiometer.
Finally, Fig. 6 shows SEM images of composites with 20%
Spherical 1. No gaps were observed with the one with the
untreated ller and cured with its shrinkage non-constrained
(Fig. 6(A)). Only the specimen with the untreated ller and
cured with its shrinkage constrained showed the presence of
gaps around the llers (Fig. 6(B)). Again, no gaps were present
with the silane-treated ller and cured with its shrinkage constrained (Fig. 6(C)). These results are consistent with those in
Figs. 1, 2 and 5.



By detecting the formation of gaps (Figs. 2, 5 and 6), all

the three techniques were consistent in showing that, under
the constrained shrinkage conditions, gaps could develop
between the ller particle and the resin matrix, suggesting
that PCS pulls the resin away from the particle surface. Two
of the techniques, optical microscopy and SEM, in addition,
clearly revealed that without the PCS the llerresin interface remained gapless (Figs. 1 and 6), even when a separating
agent was applied (Fig. 1). Despite giving indirect evidence, the
light attenuation measurements also followed the same trend
(Fig. 5). The three methods also indicated that the enhanced
interaction at the llerresin interface by the silane treatment may be able to prevent the formation of gaps under
constrained shrinkage.
Under certain conditions, two sources for generating artifacts had been noticed in our SEM observations. The rst
was the disappearance of gaps if the cured composite was
removed from the ring. It was presumably due to the relief of
the PCS, allowing the initially formed gaps to close. Therefore,
the cured composite was always with the ring throughout the
experiment. Even so, there was no guarantee for those using
aluminum rings. Spontaneous debonding often occurred during polishing, a consequence of its high PCS. For this reason,
only acrylic rings were utilized. Aluminum rings would have
produced much more gaps, as will be discussed later. The second artifact was the appearance of gaps around the untreated
llers once the cured specimen was in contact with water,
even for the composites cured without constraint. It was likely
due to the high hydrophilicity of the ller that preferred to
interact with water. Consequently, the cured specimens were
not allowed to contact water, even during polishing. After
identied these artifacts, we believed that the images in Fig. 6
were genuine.
We can thus picture the following processes. When the
composite undergoes non-constrained shrinkage, its internal
voids shrink, too. In the absence of solid cores, the volume of
voids (now pores) will become smaller. In the presence of solid
cores (such as ller particles), their volume cannot change but
their interaction with the cores becomes stronger under the
compressive stress. This agrees with previous observations
that polymerization contraction generates a hoop stress [27],
forcing the matrix to clamp particles more tightly [28]. The
process is analogous to the shrink tting technique [29,30],
which uses either cooling shrinkage of the outer ring or heating expansion of the inner shaft to t the two parts together.
Just as such a pre-stress makes the joint tighter, the hoop
stress can strengthen the llerresin interaction as well.
On the other hand, when the composite undergoes constrained shrinkage, the generated PCS expands the voids,
which can partially compensate the shrinkage induced volume loss and thus help reduce the stress. This is one of
the mechanisms on why the PCS was reduced by the use of
porous resins [31]. The presence of solid cores in the voids
complicates the situation. Unlike the empty void case, the
expansion force can be partially balanced by the attraction
forces. Too weak attraction cannot prevent resin from leaving
the interface. In fact, an approach to reducing PCS has been


d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

Fig. 6 SEM micrographs of composites with: (A) untreated Spherical 1 and cured with non-constrained shrinkage, (B)
untreated Spherical 1 and cured with constrained shrinkage, and (C) silane-treated Spherical 1 and cured with constrained
shrinkage. The arrows in (B) indicate gaps, which were about 10% of all the visible interfaces on the image. Also more gaps
appeared to be around bigger llers than smaller ones.

demonstrated by the use of non-bonded nano-ller to deliberately weaken the llerresin interactions [14]. Conversely, too
low a PCS is unable to overcome the attraction, no separation
would occur. Whether or not the PCS can actually separate the
resin from the ller and then expand the void will depend on
its magnitude with respect to other opposing forces. An analogous process of generating a porous structure in polymers
has been utilized in manufacturing breathable lms [32,33]. A
particulate (CaCO3 ) lled composite is stretched and its deformation causes debonding between the ller and the polymer
matrix, producing pores. The debonding stress is a function
of the thermal stress, the reversible work of adhesion, the
Youngs modulus of the matrix, and the radius of particles
The reversible work of adhesion is related to the magnitude of llerresin interactions in our systems. We observed
that the use of acrylic rings, which due to its low elastic modulus (3.3 GPa [34]) produced a PCS of 4 MPa, could always create
gaps around a smooth glass bead if the bead was coated with
the separating agent (Fig. 2(C)). With untreated beads, gaps
could still form but only at about 33% of the time (Fig. 2(B)).
Gaps appeared every time again when the shrinkage was constrained by aluminum rings, which produced a PCS of 20 MPa
owning to their much higher elastic modulus (69 GPa [35])

(Fig. 2(D)). Finally, no gaps were ever found around silanetreated beads even when aluminum rings were used (Fig. 2(E)).
The last observation, together with that from Fig. 5(B), suggests
that with a proper silane treatment, the llerresin interfacial
bonding can be strong enough to resist the high PCS created
by rigid aluminum rings. By the same token, however, the fact
that the composites using a silane-treated ller still resulted
in a lower attenuated irradiance, Fig. 5(B) implies that the efciency of silane treatment as well as mixing status may play
important roles, too.
The aforementioned processes are schematically summarized in Fig. 7 by using 2D illustrations. When the resin shrinks
during cure, the void also shrinks (Fig. 7(A)), which reduces its
surface area and is thus energetically favorable by reducing the
total surface free energy. The shrinkage of the void appears to
increase the resin volume as the total volume equals those of
the resin and the void. However, since the total volume also
shrinks, this compensates both reduced volumes of void and
resin. The system is stable. Therefore, we suggest that free
shrinkage and the reduction of surface area together are the
driving force for the resin to move toward the particle during
cure (Fig. 7(B)).
There are several scenarios that do not fulll the two
requirements and thus may see the resin being separated from

d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

the particle. The rst scenario is the constrained total volume

shrinkage, where the rst requirement is not met. A sufcient constraint of shrinkage produces enough stress (PCS)
that forces the void to expand even if it means to increase
the total surface free energy (Fig. 7(C)). The system is less
stable, however, due to the presence of created gaps and the
associated higher free energy even if the volume constraint is
later removed. Such a higher structural instability and formation of gaps at the llerresin interface may explain why the
specimens fabricated with their shrinkage constrained had
lower mechanical properties than those without [12,1721].
As the strength of a composite depends much more on the
resinller adhesion than the modulus [23,24], it is probably
why the tensile strengths showed signicant differences while
the hardness did not [19,20]. After all, hardness can be nicely
correlated to the modulus [36].
The second scenario is the curvature of the llerresin
interface. When it is convex with respect to the ller, the
resin moving away from the particle means an increase of its


surface area. Such an energetically unfavorable process will

not happen spontaneously under the free shrinking condition.
However, when there is a concave surface, moving away from
the particle in fact reduces the surface area even if the total
shrinkage is not constrained (Fig. 7(D)). That is why gaps were
produced in Fig. 3(B) and (D) on the smooth glass beads with
nicks, which possessed a local concave curvature. By the same
token, a rough surface provides many microscopically concave
surfaces, which could initiate gap formation under polymerization shrinkage. That is probably the reason why the cured
specimen was slightly less transparent than the uncured
because its microscopic voids scattered more light (Fig. 4(B),
(D), and (F)). (A potential mismatch between the refractive
index of the separating agent and that of the resin should not
mater because the comparison is made between the uncured
and the cured specimens and curing would change the resins
refractive index very little [26].) In addition, our optical microscope observations had showed that debonding was mostly
likely initiated from the joint area between bead and its han-

Fig. 7 2D schematics showing effects of different scenarios on the shrinkage direction of the resin matrix with respect to
the ller particles. (A) Non-constrained shrinkage with an empty void, (B) non-constrained shrinkage with a global convex
surface (ller particle), (C) constrained shrinkage with a global convex surface, (D) non-constrained shrinkage with a global
convex surface but a local concave one, and (E) non-constrained shrinkage with resin ssures and a resin chamber.


d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

dle, where the surface was slightly concave (Figs. 2(BD) and
The third scenario is not proven in our experiments
but worthy of some speculation based on what has been
discussed. In composites with highly agglomerated ller particles, their touching one another may create local chambers,
where the shrinkage of resin is constrained (Fig. 7(E)) and the
total resin shrinkage will have no effects. Also, local concavity
is not limited to individual particles. Resin dead-ends, such
as pockets, valleys, bits, and ssures may form from closely
packed ller particles as well (Fig. 7(E)). All of these can serve as
concave surfaces and the trapped resin will likely move away
from the interface under polymerization shrinkage regardless how the global composite shrinksdoing so will always
reduce the surface area as well as the PCS.
We only added a small fraction of ller in the composites
for light attenuation measurements in order to maintain a
high curing homogeneity as well as measuring accuracy. With
a ller loading around 6080 wt% in a regular composite, the
population of gaps may be different. From our limited experiments, we noticed that our model composites with higher
ller loading seemed to produce greater irradiance differences
between those with constrained shrinkage and those without,
even with silane-treated llers. It could have been the results
of the enhanced disparities due to the presence of a larger
interfacial area and thus more associated gaps. It is possible
that commercial composites could show a greater structural
and property disparity because they contain much more ller
than those used in this research.
Finally, as suggested by this work as well as the other
studies [12,1721], it is very plausible that the structure and
properties of a composite, when it is cured with the polymerization shrinkage constrained, may differ from those
non-constrained. In reality, almost all composites used in vivo
are cured under some kind of PCS, while the specimens prepared for the evaluation of mechanical properties are not. Yet,
the data of the latter are always unequivocally used to represent those of the former. It is an unfortunate fact. Hopefully
this paper will encourage research workers and manufacturers to address this issue.



By the use of three different measuring techniques and model

composites, we proved that, during photo cure of a composite, there is a tendency to form gaps between the resin matrix
and the ller particles if its polymerization shrinkage was constrained. This is the consequence of the resin being separated
from the ller particles driven by polymerization contraction stress. Effective silane treatments of llers may reduce
the occurrence of gap formation by enhancing the llerresin


[1] Davidson CL, Feilzer AJ. Polymerization shrinkage and

polymerization shrinkage stress in polymer-based
restoratives. J Dent 1997;25:43540.

[2] Giachetti L, Scaminaci Russo D, Bambi C, Grandini R. A

review of polymerization shrinkage stress: current
techniques for posterior direct resin restorations. J Contemp
Dent Pract 2006;7:7988.
[3] Braga RR, Ferracane JL. Alternatives in polymerization
contraction stress management. Crit Rev Oral Biol Med
[4] Braga RR, Ballester RY, Ferracane JL. Factors involved in the
development of polymerization shrinkage stress in
resin-composites: a systematic review. Dent Mater
[5] Sakaguchi RL, Wiltbank BD, Murchison CF. Cure induced
stresses and damage in particulate reinforced polymer
matrix composites: a review of the scientic literature. Dent
Mater 2005;1:436.
[6] Lingois P, Berglund L, Greco A, Maffezoil. Chemically
induced residual stresses in dental composites. J Mater Sci
[7] Peutzfeldt A, Asmussen E. Determinants of in vitro gap
formation of resin composites. J Dent 2004;32:10915.
[8] Loguercio AD, Reis A, Schoeder M, Balducci I, Versluis A,
Ballester RY. Polymerization shrinkage: effects of boundary
conditions and lling technique in composite restorations. J
Dent 2004;32:45970.
[9] Li J, Li H, Fok SL. A mathematical analysis of shrinkage
stress development in dental composite restorations during
resin polymerization. Dent Mater 2008;24:92331.
[10] Feilzer AJ, De Gee AJ, Davidson CL. Quantitative
determination of stress reduction by ow in composite
restoration. Dent Mater 1990;6:16771.
[11] Kemp-Scholte CM, Davidson CL. Marginal sealing of curing
contraction gaps in class V composite resin restorations. J
Dent Res 1988;67:8415.
[12] Davidson CL, Zeghbroeck LV, Feilzer AJ. Destructive stresses
in adhesive luting cements. J Dent Res 1991;70:8802.
[13] Kinomoto Y, Torii M. Photoelastic analysis of polymerization
contraction stress in resin composite restorations. J Dent
[14] Condon JR, Ferracane JL. Reduced polymerization stress
through non-bonded nanoller particles. Biomaterials
[15] Davidson CL, De Gee AJ. Relaxation of polymerization
contraction stresses by ow in dental composites. J Dent Res
[16] Loguercio AD, Reis A, Ballester RY. Polymerization shrinkage:
effects of constraint and lling technique in composite
restorations. Dent Mater 2004;20:23643.
[17] Choi KK, Ferracane JL, Ryu GJ, Choi SM, Lee MJ, Park SJ.
Effects of cavity conguration on composite restoration.
Oper Dent 2004;29:4629.
[18] Nayif MM, Nakajima M, Foxton RM, Tagami J. Effect of light
irradiation time on the mechanical properties of two
owable composites with different initiation systems in
bonded and unbonded cavities. Dent Mater J 2007;26:
[19] Nayif MM, Nakajima M, Aksornmuang J, Ikeda M, Tagami J.
Effect of adhesion to cavity walls on the mechanical
properties of resin composites. Dent Mater 2008;24:
[20] Nayif MM, Nakajima M, Foxton RM, Tagami J. Bond strength
and ultimate tensile strength of resin composite lled into
dentine cavity: effect of bulk and incremental lling
technique. J Dent 2008;36:22834.
[21] Kleverlaan CJ, Feilzer AJ. Polymerization shrinkage and
construction stress of dental resin composites. Dent Mater
[22] Dauvillier BS, Paarnts MP, Feilzer AJ. Modeling of the
viscoelastic behavior of dental light-activated resin
composites during curing. Dent Mater 2003;19:27785.

d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 719729

[23] Ahmed S, Jones FR. A review of particulate reinforcement

theories for polymer composites. J Mater Sci
[24] Wypych G. Handbook of llers. 2nd ed. Toronto: PDL; 1999.
[25] Musanje L, Darvell BW. Curing-light attenuation in
lled-resin restorative materials. Dent Mater 2006;22:80417.
[26] Shortall AC, Palin WM, Buetscher P. Refractive index
mismatch and monomer reactivity inuence composite
curing depth. J Dent Res 2008;87:848.
[27] Soderholm KJM. Inuence of silane treatment and ller
fraction on thermal expansion of composite resins. J Dent
Res 1984;63:13216.
[28] Debnath S, Ranade RSL, McCool J, Boderick K, Baran G.
Interface effects on mechanical properties of
particle-reinforced composites. Dent Mater 2004;20:67786.
[29] Xue Q, Nesterenko VF, Meyer MA. Evaluation of the
collapsing thick-walled cylinder technique for shear-band
spacing. Int J Impact Eng 2008;28:25780.
[30] Yeo HT, Choi Y, Hur KD. Analysis and design of the
prestressed cold extrusion die. Int J Adv Manuf Tech


[31] Alster D, Feilzer AJ, Dee Gee AJ, Mol A, Davidson CL. The
dependence of shrinkage stress reduction on porosity
concentration in thin resin layers. J Dent Res
[32] Pukanszky B, Voros G. Mechanism of interfacial interactions
in particulate ller composites. Compos Interfaces
[33] Renner K, Yang MS, Moczo J, Choi HJ, Pukanszky B. Analysis
of the debonding process in polypropylene model
composites. Eur Polym J 2005;41:25209.
[34] Domininghaus H. Plastics for engineers: materials,
properties, applications. Munich: Hanser; 1988.
p. 268.
[35] Smith WF. Principles of materials science and engineering.
New York: McGraw-Hill; 1986. p. 251.
[36] Xu HHK, Smith DT, Schumache GE, Eichmiller FC, Antonucci
JM. Indentation modulus and hardness of
whisker-reinforced heat-cured dental resin composites.
Dent Mater 2000;16:24854.