d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134

available at www.sciencedirect.com

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

Strengthening of a model composite restoration using shape

optimization: A numerical and experimental study

H. Li a,∗ , X. Yun b , J. Li a , L. Shi c , A.S. Fok a,d , M.J. Madden d , J.F. Labuz e
a Minnesota Dental Research Center for Biomaterials and Biomechanics, School of Dentistry, University of Minnesota, 16-212 Moos
Tower, 515 Delaware Street SE, Minneapolis, MN 55455, USA
b State Key Laboratory of Oral Diseases, Sichuan University, China
c Institute of Nuclear and New Energy Technology, Tsinghua University, China
d Department of Restorative Sciences, School of Dentistry, University of Minnesota, USA
e Department of Civil Engineering, Institute of Technology, University of Minnesota, USA

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

Article history: Objective. This study aims to validate a cavity shape optimization approach for improving
Received 8 July 2009 the debonding resistance of dental restorations by carrying out fracture tests on restored
Received in revised form model teeth with standard and optimized cavity designs.
1 September 2009 Method. The bio-mimetic stress-induced material transformation (SMT) optimization
Accepted 9 September 2009 method was incorporated into the finite element (FE) program ABAQUS as a user mate-
rial (UMAT) subroutine. The method uses stress minimization to optimize the cavity shape
of a MOD restoration in an artificial premolar with special reference to the tooth–restoration
Keywords: interface under occlusal loads. The mechanical performance of the optimized design was
Shape optimization first verified through FE analysis and then compared with that of the conventional design
Dental restoration using fracture tests on model teeth.
Cavity shape Results. The SMT optimization process indicated a T-shape cavity as a more favorable design
Debonding resistance for the MOD restoration in the artificial premolar. Compared with the conventional parallel
Finite element method wall, or undercut design, the T-shape cavity was shown numerically to reduce the interfacial
stresses by up to 69%, and experimentally to increase the mean debonding resistance of the
model teeth by 23% (p < 0.05).
Significance. Cavity shape optimization can help increase the debonding resistance of
restored teeth by reducing the interfacial stresses between tooth and restoration under
occlusal loads.
© 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.

by fracture of the tooth. Therefore, maintaining the integrity

1. Introduction
Secondary caries is caries that occur at the degraded mar-
gins of an existing dental restoration. Recent clinical surveys
indicated that secondary caries was the most common reason
for the failure and replacement of restorations [1,2], followed
of the tooth–restoration interface is essential in ensuring the
longevity of dental restorations [3,4].
Much work has been done to explore the possible fac-
tors that lead to marginal deterioration of dental restorations.
Among these, tooth–restoration interfacial stresses, produced
by either shrinkage of the dental composite during light-

∗
Corresponding author. Tel.: +1 612 626 2556.
E-mail address: lihaiyan2002@hotmail.com (H. Li).
0109-5641/$ – see front matter © 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.dental.2009.09.005

Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134
curing or the occlusal loads during mastication, have been
shown to compromise the marginal integrity of compos-
ite restorations by causing interfacial debonding [5,6]. Using
appropriate techniques and material properties, for example,
an appropriate adhesive layer thickness or a more flexible
restorative material, can help reduce shrinkage stress [7,8].
Geometrical configuration of the cavity will also affect the
magnitude of the interfacial stresses [9–14]. Using both numer-
ical and analytical methods, Hubsch et al. [11,12] studied the
interfacial shrinkage stresses of restorations with different
cavity shapes. They showed that a drawn-out bevel could
help reduce the stress singularity at the surface margin of
the restoration. De Vree et al. [13] compared the interfacial
stresses in amalgam restorations with conventional and mod-
ified cavities under occlusal loads. Using the finite element
(FE) method, they showed that the modified design resulted
in fewer stress concentrations and lower stress values. In Ref.
[14], fracture tests on human teeth with different amalgam
restoration designs were conducted. The experimental results
showed a relationship between the fracture strength and cav-
ity dimensions.
Recently, modern shape optimization techniques have
been applied to the design of dental restorations, including
cavity preparations, bridges and implants [15–17]. Combined
with the powerful stress analysis capability of the FE method,
shape optimization techniques can improve significantly the
mechanical performance of structures with complex geome-
tries and non-linear material properties. Based on the adaptive
growth process of biological structures [18], Shi et al. [15–17]
have developed shape optimization methods in the form
of stress-induced material transformation (SMT) and stress-
induced volume transformation (SVT) and incorporated them
into the commercial FE package ABAQUS [19] via its user mate-
rial (UMAT) subroutine. In [16], the shape of a MOD restoration
in a premolar was optimized using the above techniques and
the resulting diamond-shape cavity was shown to have greatly
reduced interfacial stresses under external load.
The work presented in this paper extends the theoretical
work of Shi et al. [16] on cavity shape optimization, the aim
being to validate the approach using destructive mechanical
tests. As a first attempt, preclinical artificial teeth were used in
the experiments to minimize the variations in tooth anatomy
and material properties. The SMT method was thus applied to
the optimization of a MOD cavity in a model maxillary premo-
lar to minimize the stresses at the tooth–restoration interface.
FE simulations and mechanical tests were then conducted to
verify the effectiveness of the new cavity shape in increasing
the resistance of the restored tooth against interfacial debond-
ing.
Table 1 – Dimensions (mm) of restorations.
Standard
W D
FE model 1.88 4.47
Test samples (mean, SD) 1.91, 0.11 4.51, 0.23
Note: refer to Fig. 4.
W, width of standard cavity; D, depth of standard cavity (from tip of higher cusp); W1, width of upper part of T-shape cavity; W2, width of lower
part of T-shape cavity; D1, depth of upper part of T-shape cavity; D2, depth of lower part of T-shape cavity; SD, standard deviation.
Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
127
2. Materials and methods
All the FE analyses presented in this study, including the shape
optimization, were performed using the commercial FE pack-
age ABAQUS [19].
2.1. Shape optimization using SMT [16,17]
To reduce the interfacial stresses between the tooth material
and the restoration, the stress-induced material transforma-
tion (SMT) method was utilized to optimize the cavity shape
[16,17]. SMT is one of the simplest structural topology opti-
mization methods that mimic the biological mineralization
process in living bone: material is deposited in loaded areas
while non-load-bearing material is gradually removed. This
process can be simulated in a FE analysis by making Young’s
modulus (E) of the material a function of the mechanical
stress/strain, which will be non-uniform over the volume of
structure being optimized. The following equation shows the
changes in E between two time steps or increments:
Ei+1 = Ei + k(i − ref ) (1)
where Ei and  i are Young’s modulus and stress at the ith
increment, respectively;  ref is a reference or threshold stress
above which changes to Young’s modulus will occur; and k is a
parameter that controls the rate of changes of Young’s modu-
lus. A more detailed description of the SMT method has been
presented elsewhere [16,17].
The above optimization method has been incorporated
into the finite element software ABAQUS via a user material
(UMAT) subroutine [19]. During the optimization process, the
UMAT subroutine is called upon repeatedly until the spatial
distribution of Young’s modulus reaches a steady state.
2.2. Shape optimization of MOD cavity in a maxillary
premolar
2.2.1. FE model
The Lava Scan ST Scanner (3M ESPE, USA) was used to obtain
the geometric model of the artificial maxillary premolar. A 2-
D FE model was then constructed in ABAQUS from the 3D
geometric model; see Fig. 1(a). By St. Venant’s principle [20],
assuming that the tooth is well supported by the surround-
ing bone, the exclusion of its root from the model should have
little effect on the stress distribution around the restoration.
Therefore, only the crown of the tooth was modeled.
Optimized
W1 W2 D1 D2
2.88 1.88 3.10 4.47
2.86, 0.20 1.89, 0.10 3.35, 0.20 4.51, 0.19
128 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134

Fig. 1 – Cavity shape optimization: (a) 2-D FE model, (b) results—Young’s modulus distribution indicating regions (red) that
require reinforcement, and (c) optimized cavity shape with key positions (A–H) marked along the interface. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

The constraints on the tooth from the bones were simpli-
fied by fixing all degrees of freedom at the base of the tooth
model. A standard MOD restoration was then constructed, as
shown in Fig. 1(a). The dimensions of the restoration are given
in Table 1. In total, the model contained 8051 plane-strain
elements (CPE3 and CPE4 [19]) and 8148 nodes.
A compressive load was applied to the tooth through a
semi-cylindrical steel loading head with an outer radius of
3 mm (Fig. 1(a)). A uniform pressure of 15.15 MPa was applied
on the top surface of the loading head, giving a total force of
400 N, which represented the maximum occlusal force typi-
cally reported in the literature. Polymerization shrinkage of
the composite was not considered.
The mechanical properties of the three materials are shown
in Table 2. Young’s moduli of the artificial tooth and composite
restoration were determined using 3-point bend tests on small
beam specimens of the materials. That of the steel loading
head was based on typical values from the literature. Poisson’s
ratios of all the materials were assumed to be 0.3.
2.2.3. Shape optimization process
The shape optimization process began with the initial model
as shown in Fig. 1(a). The first step of the process involved
loading the tooth with the compressive force to create
stresses. Areas of stress concentration would appear around
the tooth–restoration interface and these were the areas
that required reinforcement through material transformation.

2.2.2. Material properties

There were three components with different materials in the
Table 2 – Material properties for the FE tooth model.
FE model (Fig. 1(a)): the artificial tooth, composite restora-
tion and steel loading head. All materials were assumed to Tooth Restoration Loading head
be linearly elastic. The composite restoration was assumed Young’s modulus (GPa) 6 12 210
to be fully cured without the presence of residual stresses. Poisson’s ratio 0.3 0.3 0.3

Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134
Thus, in the second step, the SMT process based on Eq. (1) was
executed via the user material (UMAT) subroutine to trans-
form the artificial tooth material around the restoration with
overtly high stresses to the stronger restorative material which
had a higher Young’s modulus. The optimized shape of the
restoration was therefore expected to be bigger than the orig-
inal standard shape. The SMT process took place in small
increments, with the stresses,  i , updated at the end of each
increment until convergence or a steady state was achieved.
The parameters in Eq. (1) were set as:  ref = 11 MPa and k = 50.
The maximum principal stress was chosen to be the stress to
cause material changes since debonding was considered to be
caused predominantly by a tensile stress.
2.3. Verification of optimized design by means of FEA
To verify the potential effectiveness of the optimized design in
improving debonding resistance, a 2-D FE stress analysis based
on the model in Fig. 1(a) was carried out to evaluate the inter-
facial stresses of the restored tooth for both optimized and
non-optimized cavities. The loading and boundary conditions
were the same as those shown in Fig. 1(a), and the material
properties were also the same as those shown in Table 2. A tied
contact interaction [19] was defined between the tooth and
restoration surfaces to facilitate extraction of the interfacial
stresses that were normal and parallel to the interface.
2.4. Verification of optimized design by means of
experiment
Even though FE analysis could provide very useful numeri-
cal results for evaluating the effectiveness of the optimized
design, experimental proof was still necessary. Therefore,
mechanical tests were performed to compare the resistance
against debonding of restored artificial model teeth with the
optimized cavity against those with the standard design.
2.4.1. Sample preparation
Artificial maxillary premolars were used in this seminal study
to minimize the variations in geometry and material prop-
erties, as would be found with natural human teeth. 50
specimens, separated into 3 groups, were tested: 12 intact
Fig. 2 – Compressive test: (a) mounting of a test specimen (A—resin, B—nylon mounting ring, C—model tooth specimen,
D—positioning index, E—cutout for pouring resin and F—positioners) and (b) clamping jig and loading head.
Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
129
teeth (Group I), 20 teeth with a standard restoration (Group S)
and 18 teeth with an optimized restoration (Group O). Group I
was used as a control group.
The standard and optimized cavities were prepared by the
same operator following standard clinical procedures with a
high-speed handpiece and dental cutting burs. The dimen-
sions of the cavities for both the numerical model and actual
specimens are shown in Table 1. Ethyl-2-cyanoacrylate-based
superglue, used as a bonding agent between the restoration
and artificial tooth, was applied thinly on the cavity walls. The
cavities were then filled with a dental composite (Z100, 3M
ESPE, USA) incrementally to minimize shrinkage stress and
to ensure complete polymerization of the composite [21]. A
blue curing light (3M ESPE, USA) was used to cure the dental
composite.
The restored samples were potted into Teflon mounting
rings using an orthodontic resin (DENTSPLY International Inc.,
USA), as shown in Fig. 2(a). To ensure all the specimens have
the same position and orientation relative to the load axis, an
index was made using a vinyl polysiloxane impression mate-
rial (3M ESPE, USA) to accurately locate the specimen within
the mounting ring (Fig. 2(a)). The positioning index was fixed
to the ring via two diametrically opposite positioners that fit
into the two corresponding holes of the ring. The resin was
then poured through a cutout of the index to fill the ring up to
the level of the cementum–enamel junction, so as to be con-
sistent with the boundary condition used in the FE analysis
(Fig. 1(a)).
2.4.2. Compressive test
The compressive tests of the artificial teeth were conducted
on a universal test machine (858 Mini Bionix II, MTS, USA). A
clamping jig, which was placed on to the load cell of the test
machine, was used to secure the mounting ring containing
the specimen, as shown in Fig. 2(b). The loading head, with a
diameter of 6 mm, was connected to the moving piston of the
test machine. The compressive load was applied to the speci-
men in a stroke-control mode at a rate of 0.1 mm/min until the
specimen fractured. Test data, including time, displacement,
and load were recorded during the whole loading history for
each specimen.
130 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134

Fig. 3 – Interfacial stresses along normalized interfacial distance (starting from the palatal cusp and moving towards the
buccal cusp): (a) normal stress and (b) shear stress, and distributions of maximum principle stress within tooth: (c) standard
design and (d) optimized design.

regions with a high E value close to that of the composite

3. Results material. The results show that increases in Young’s modu-
lus occurred on both sides of the restoration. However, a larger
3.1. Shape optimization area with increased Young’s modulus was found on the palatal
side of the restoration than on the buccal side. The former
Fig. 1(b) shows the converged Young’s modulus E distribution also extended to the occlusal surface, whereas the latter was
from the shape optimization analysis. The red color indicates subsurface.

Fig. 4 – Typical load–displacement curves: (a) with clear debonding before final fracture and (b) with no clear debonding
before final fracture.

Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134
Fig. 5 – Failure modes of (a) intact tooth, (b) restored tooth with T-shape restoration: debonding at palatal interface (mode P),
and (c) restored tooth with standard restoration: debonding at buccal interface (mode B).
Based on the results in Fig. 1(b), a new T-shape cavity was
proposed as shown in Fig. 1(c). Points a–d and A–H are the key
points along the interfaces of the standard and new restora-
tion designs where stresses were evaluated. Even though the
change in Young’s modulus on the buccal side of the restora-
tion was smaller and subsurface, compared to that on the
palatal side, similar changes were made to both sides of the
cavity for ease of preparation.
3.2. Verification by means of FEA
Fig. 3(a) and (b) shows the normal and shear stresses along
the interface between the tooth and restoration for both the
standard and optimized designs. For ease of comparison, the
normalized distance along the interface starting from the
palatal side was used to construct the curves. Refer to Fig. 1
for the positions of points a–d and A–H along the interfaces.
It can be seen that, with the optimized cavity design,
there was significant reduction in the interfacial normal stress
between points A and C and between points F and H; see
Fig. 3(a). At point A, for example, the reduction was 64%
whereas at point H it was 69%. Between points B and C, and
between points F and G, i.e. at the steps of the T-shape cavity,
the normal stress changed from tensile to compressive. The
magnitude of the interfacial shear stresses (Fig. 3(b)) was also
greatly reduced at points A and H, i.e. the palatal and buccal
surface margins, but increased markedly at the steps of the
new cavity. Between points C and F, the stresses of the two
models were similar because of the similar geometric profiles
and dimensions.
Fig. 3(c) and (d) shows the stress distributions within
the tooth for standard and optimized designs, respectively.
Besides the areas near the cavity, the distributions of stresses
within the main body of the tooth are very similar.
3.3. Compressive tests
3.3.1. Load–displacement curve
Fig. 4 shows two typical load–displacement curves obtained
from the compressive tests of the artificial model teeth. In
Fig. 4(a), a small but noticeable drop in load can be seen when
the load was between 100 and 200 N. Another penetration test
(5 samples were tested) using a dye penetrant (Basic Fuchsin,
Sigma–Aldrich Inc., USA) revealed that debonding between the
Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
131
tooth and restoration had taken place at this point of the load-
ing history. Therefore, this intermediate load, at which the
small drop in load occurred, is called the debonding load, P1,
and is the parameter used to evaluate the resistance of the
restored tooth against debonding. The ultimate failure load,
P2, is defined as the maximum load sustained by the restored
tooth before final fracture. However, about 15% of the restored
samples did not show a small drop in load at the intermediate
load level, as shown in Fig. 4(b), indicating that debonding of
the restoration might not have occurred in these specimens
prior to final fracture, or that the intermediate drop in load
might have been too small to be detected [22]. Therefore, two
methods were used to treat the data of specimens without P1
in their load–displacement curves: (A) exclude them from the
analysis and (B) use the final failure load as its debonding load,
i.e. P1 = P2. The intact teeth also followed a load–displacement
curve similar to that in Fig. 4(b), albeit with a higher final fail-
ure load.
3.3.2. Failure mode
Fig. 5(a) shows the consistent failure mode of the intact speci-
mens with fracture taking place in the middle of the tooth. For
the restored specimens, fracture along the tooth–restoration
interface could start from either the palatal side (mode P), as
shown in Fig. 5(b), or the buccal side (mode B), as shown in
Fig. 5(c). The ratios of mode-P to mode-B fractures for Group
S (standard cavity) and Group O (optimized cavity) are 4:1 and
9:1, respectively. The palatal side of the restoration therefore
seemed more susceptible to debonding under the compressive
load used in the experiment.
3.3.3. Statistic analysis
The mean value and standard deviation (SD) of P1 and P2
are given in Table 3 for all the tooth specimens. The results
show that the restored teeth were 32% weaker than the intact
teeth in terms of the final fracture load. However, compared
to the standard restoration, the optimized cavity increased the
debonding strength, P1, by 23.9% when method A was used for
the analysis and 24.5% when method B was used. Using the
optimized cavity did not change the final fracture strength, P2,
of the restored teeth significantly, with only a 3.1% decrease
compared to the group with the standard cavity. The variations
in strength between specimens were large, especially among
the restored teeth (Groups S and O), with the coefficient of vari-
132 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134

Table 3 – Mean and standard deviation (SD) of debonding strength, P1, and final fracture strength, P2, of model teeth.
Group Sample size Mean (SD) of P1 (N) Mean (SD) of P2 (N)

Method Aa Method Bb
I 12 – – 496.4 (64.1)
S 20 175.1 (49.0) 189.4 (64.7) 350.8 (81.4)
O 18 217.0 (52.4) 235.8 (67.2) 339.9 (66.4)
Improvement 23.9% 24.5% −3.1%
a
Samples without noticeable debonding in load–displacement curve are excluded.
b
Samples without noticeable debonding in load–displacement curve are assumed to have P1 = P2.

ation in P1 being larger than that in P2. The largest coefficient
of variation was 34% for P1 of Group S.
Fig. 6 shows the cumulative distribution function (CDF)
plots of P1 and P2 for the restored teeth. The CDF of P2 for
the intact teeth is also plotted for comparison. In Fig. 6(a)
and (b), the difference in P1 between the teeth with standard
and optimized restorations can be seen to be very significant,
irrespective of the method used for analyzing the debonding
data. For a certain debonding probability, the tooth with the
optimized design could endure a higher load than that with
a non-optimized design. Alternatively, for a particular load
level, the debonding probability of the tooth with the opti-
mized design was much lower than that with the standard
design. Fig. 6(c) confirms that the two sets of restored teeth
had very similar P2 values, but these were much lower than
those of the intact teeth.
A t-test analysis was carried out to evaluate the statistical
significance of the difference in strength between the teeth
with standard and optimized cavities. The p-values that indi-
cate the significance are shown in Table 4. It can be seen that
there was a significant difference (p < 0.05) in P1 between the
two cavity shapes, again irrespective of the method used for
debonding analysis. The difference in P2 between the two cav-
ity shapes, however, was not significant (p = 0.65).
4. Discussion
There are a number of reasons why shape optimization has
not been widely used in the practice of dentistry. Modern mini-
mal intervention techniques along with caries control, tend to
limit the application of shape optimization, since the preven-
tion and control of small carious lesions take precedence over
the engineering strength of the tooth. However where carious
lesions are large and/or final cavity size is large relative to the
crown size, optimization techniques can significantly inform
the dentist what the final cavity form should be. Following on
the success in the design of engineering structures, modern
Table 4 – Statistical significance (p-value) of the
difference in strength between teeth with standard and
optimized cavities.
Debonding Final fracture
strength, P1 strength, P2
Method A Method B
p-Value 0.02 0.04 0.65
shape optimization methods coupled with the use of validat-
ing experimental data should encourage greater use of shape
optimization for larger cavities in the chairside practice of den-
tistry. The work presented in this paper aims to contribute to
this process.
In this study, the stress-induced material transformation
(SMT) method [16,18] was used to optimize a MOD restoration
in an artificial model tooth in order to minimize the inter-
facial stresses during occlusal loading. A T-shape cavity was
indicated by the optimization analysis. The reduction in mag-
nitude of the interfacial stresses was verified using FE analysis,
and significant improvement in resistance against debonding
with the new cavity design was demonstrated using mechan-
ical tests that simulated occlusal loading. The results have
shown clearly that shape optimization is a very useful tool for
the design of stronger dental restorations, and have provided
a good basis for further investigation using natural human
teeth.
The main reason for using artificial model teeth in this
seminal work was to minimize the effects of the significant
variations in material properties and geometries of natural
human teeth, which were also more difficult to acquire in
large quantities. Note that a large number of specimens were
required for the mechanical tests to produce statistically use-
ful results because of the inherent scatter associated with
the fracture strength of such brittle materials. This was evi-
dent in the test data of the model teeth, even for the intact
ones with nominally identical shapes and dimensions. The
restored teeth had more scattered data (larger coefficients of
variation) probably because of the additional variations in the
dimensions, material properties, preparation procedures and
bonding conditions of the composite restorations.
The optimization process indicated material strengthen-
ing on both sides of the restoration (Fig. 1(b)), with more
strengthening being required on the palatal side than on the
buccal side. The former also extended to the occlusal sur-
face, whereas the latter was subsurface. This showed that the
palatal cusp borne more of the load than the buccal cusp, most
probably because of the less favorable angle of its occlusal
surface relative to the vertical loading axis. The numerical
results were also consistent with the experimental results
which showed a higher number of debonding cases at the
palatal side of the restoration than at the buccal side prior to
final fracture; the ratio being 4:1 for the standard design and
9:1 for the optimized design.
Referring to Fig. 3, the interfacial normal stresses at points
A and G of the optimized restoration were predicted to be
lower by 64% and 69%, respectively, than those in the standard

Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134
Fig. 6 – CDF plots of (a) P1 with method A, (b) P1 with
method B, and (c) P2.
design. These reductions in the interfacial normal stresses at
points A and H should increase the resistance to debonding
between the tooth and dental restoration by roughly the same
percentage. The fracture test, however, only showed a 23%
increase in the debonding resistance (Table 3). The possible
reasons are as follows.
Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
133
Note that the FE analysis used a plane-strain 2D model,
where the restoration was assumed to have a uniform cross-
section. In the actual specimens, the cross-section of the
restoration varied through the thickness, with the cavity depth
at the mesial and distal surfaces being much smaller than that
at the midsection. The actual stresses in the restored teeth
could therefore be quite different from those given by the 2D
approximation.
Another possible reason for the discrepancy between the
expected and actual increase in debonding resistance was that
shear stresses could also cause debonding. It can be seen from
Fig. 3 that there were high stress concentrations along the
tooth–restoration interface between points B and C, as well
as between points F and G, in the optimized model. These
were caused by the step changes in geometry (from vertical
to horizontal) of the new cavity. Among these stress concen-
trations, the normal stresses were compressive and, therefore,
did not contribute to the debonding. In actual fact, they might
even have helped to suppress it. The greatly increased shear
stresses at these positions could, however, promote debond-
ing. Since the shear stresses in the T-shape restoration were
higher than those in the standard design, the overall improve-
ment in debonding resistance was not as high as one would
expect from considering the normal stress results alone. This
implies that, for even better mechanical performance, the
optimization should be based on an effective stress that com-
bines both the normal and shear components of the stresses.
A further possible reason for the discrepancy between the-
ory and experiment was the presence of shrinkage stress,
which was not considered in the FE calculation. Even though
the incremental filling method was adopted for constructing
the restoration, shrinkage stress could still be produced, which
acted as a preload on the interface.
In the fracture tests, a small number (15%) of the restored
specimens did not show any debonding behavior in the
load–displacement curves prior to final fracture; see Fig. 4(b).
It is possible that there was no debonding until final frac-
ture, or that debonding led directly to final fracture in these
specimens. However, it is more likely that debonding did
occur before final fracture but the process was so gradual
and the change in load so small that it was not detected by
the load-measuring instrument. Using an acoustic emission
(AE) system, Ereifej et al. [22] measured the onset of frac-
ture in ceramic crowns. They were able to detect the initial
subcritical cracking events, including those not shown in the
load–displacement curve. Using the AE method may there-
fore allow the more latent debonding of the restoration to be
detected.
The optimization results are greatly influenced by the
parameters in Eq. (1). The effect of these parameters, notably
k, on the numerical solution has been studied by Shi [17] and
recommendations have been made on the range of values to
use to ensure convergence. The size of the region that would
undergo material transformation depends on the reference
stress,  ref , as well as the external load. If  ref , which is a
measure of the strength of the material or structure, is large
compared to the load, the areas affected and the amount of
changes in E will both be small. On the other hand, if  ref
is small compared to the load, the above changes will be
large.
134 d e n t a l m a t e r i a l s 2 6 ( 2 0 1 0 ) 126–134
The optimization results are also dependent on the initial
material properties and their spatial distribution of the object
to be optimized and the form of the external load to which it
is being subjected. For a human premolar containing tissues
(enamel and dentin) with different properties, a diamond-
shape, instead of a T-shape, cavity was obtained for a MOD
restoration using the same SMT optimization technique [16].
This will be the basic cavity shape adopted in further in vitro
study using real human teeth. However, the analysis may have
to be refined by using a more realistic load simulation than the
cylindrical load head used in this and previous studies.
Despite the simplifications involved in the current study,
the model system was considered to have exhibited suffi-
cient similarities in terms of mechanical behavior (debonding
and final fracture) with an actual restored tooth to war-
rant further studies using real human teeth to validate the
shape optimization approach. However, because of the much
greater variations in loadings, geometries and material distri-
butions in real human teeth, much bigger sample sizes may
be required for the corresponding mechanical tests in order
to differentiate the differences in mechanical performance of
the standard and optimized restoration designs.
5. Conclusions
• Using the SMT shape optimization method and a maximum
normal stress criterion, the cavity shape of a MOD restora-
tion in an artificial model premolar was modified into a
T-shape design.
• FE analysis showed the modified design to have lower (>60%)
normal stresses but higher (180%) shear stresses along the
tooth–restoration interface.
• Compared with the traditional design, the optimized design
showed significant (24%) improvement in the resistance
against debonding under compressive load.
• Shape optimization methods can be used to improve the
service life of dental restorations.
Acknowledgements
This study was supported by the Grant-in-Aid program of the
University of Minnesota. The authors would like to acknowl-
edge the Minnesota Dental Research Center for Biomaterials
and Biomechanics (MDRCBB) for providing the testing devices
and the Minnesota Supercomputing Institute (MSI) for provid-
ing the computing services for this study.
references
[1] Forss H, Widstrom E. Reasons for restorative therapy and the
longevity of restorations in adults. Acta Odontologica
Scandinavica 2004;62(2):82.
[2] Opdam NJM, Bronkhorst EM, Roeters JM, Loomans BAC. A
retrospective clinical study on longevity of posterior
composite and amalgam restorations. Dental Materials
2007;23(1):2.
Downloaded for Anonymous User (n/a) at Egyptian Knowledge Bank from ClinicalKey.com by Elsevier on January 11, 2021.
For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.
[3] Goldberg AJ. Deterioration of restorative materials and the
risk for secondary caries. Advances in Dental Research
1990;4:14.
[4] Hayashi M, Wilson NHF. Marginal deterioration as a
predictor of failure of a posterior composite. European
Journal of Oral Sciences 2003;111(2):155.
[5] Ferracane JL, Mitchem JC. Relationship between composite
contraction stress and leakage in Class V cavities. American
Journal of Dentistry 2003;16(4):239.
[6] Davidson CL, Abdalla AI. Effect of occlusal load cycling on
the marginal integrity of adhesive Class V restorations.
American Journal of Dentistry 1994;7(2):111.
[7] Ausiello P, Apicella A, Davidson CL. Effect of adhesive layer
properties on stress distribution in composite
restorations—a 3D finite element analysis. Dental Materials
2002;18(4):295.
[8] Kahler B, Kotousov A, Borkowski K. Effect of material
properties on stresses at the restoration–dentin interface of
composite restorations during polymerization. Dental
Materials 2006;22(10):942.
[9] Feilzer AJ, De Gee AJ, Davidson CL. Setting stress in
composite resin in relation to configuration of the
restoration. Journal of Dental Research 1987;66(11):1636.
[10] Braga RR, Boaro LCC, Kuroe T, Azevedo CLN, Singer JM.
Influence of cavity dimensions and their derivatives (volume
and ‘C’ factor) on shrinkage stress development and
microleakage of composite restorations. Dental Materials
2006;22(9):818.
[11] Hubsch P, Middleton J. Asymptotic analysis of the stress field
in adhering dental restorations. Journal of Biomechanical
Engineering 2000;122:408.
[12] Hubsch P, Middleton J, Knox J. The influence of cavity shape
on the stresses in composite dental restorations: a finite
element study. Computer Methods in Biomechanics and
Biomedical Engineering 2002;5(5):343.
[13] De Vree JHP, Peters MCRB, Plasschaert AJM. The influence of
modification of cavity design on distribution of stresses in a
restored molar. Journal of Dental Research 1984;63(10):1217.
[14] Mondelli J, Steagall L, Ishikiriama A, de Lima Navarro MF,
Soares FB. Fracture strength of human teeth with cavity
preparations. The Journal of Prosthetic Dentistry
1980;43(4):419.
[15] Couegnat G, Fok ASL, Cooper JE, Qualtrough AJE. Structural
optimization of dental restorations using the principle of
adaptive growth. Dental Materials 2006;22:3.
[16] Shi L, Fok ASL, Qualtrough A. A two-stage shape
optimization process for cavity preparation. Dental
Materials 2008;24:1444.
[17] Shi L. Shape optimization of dental restorations using the
principle of biological adaptive growth, Ph.D. Thesis,
University of Manchester; 2008.
[18] Mattheck C. Design in nature: learning from trees. Springer;
1998. p. 290.
[19] ABAQUS version 6.5 user manuals, Pawtucket, RI, USA:
Hibbett, Karlsson and Sorensen Inc.; 2004.
[20] Cook RD, Young WC. Advanced mechanics of materials. 2nd
ed. Prentice Hall; 1999.
[21] Deliperi S, Bardwell DN. An alternative method to reduce
polymerization shrinkage in direct posterior composite
restorations. The Journal of the American Dental
Association 2002;133:1387.
[22] Ereifej N, Silikas N, Watts D. Initial versus final fracture of
metal-free crowns, analyzed via acoustic emission. Dental
Materials 2008;24:1289.