d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453

available at www.sciencedirect.com

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

A two-stage shape optimization process

for cavity preparation

Li Shi a , Alex S.L. Fok a,∗ , Alison Qualtrough b

a School of Mechanical, Aerospace and Civil Engineering, The University of Manchester,
P.O. Box 88, Sackville Street, Manchester M60 1QD, UK
b School of Dentistry, The University of Manchester, Higher Cambridge Street Manchester M15 6FH, UK

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

Article history: Objectives. Clinical data indicate that previously restored teeth are more likely to fracture
Received 13 November 2007 under occlusal loads. The reason for this is attributed to the high stresses at the tooth-
Accepted 3 March 2008 restoration interface, especially following debonding of the restoration from the tooth. This
work aims to minimise these interfacial stresses by optimizing the cavity shape using mod-
ern shape optimisation techniques.
Keywords: Methods. Shape optimisation methods based on the principle of biological adaptive growth
Structural optimization were incorporated into a finite element program and used to optimize the design of cavity
Cavity preparations preparations as previous work had successfully used one such method to minimise stresses
Dental restorations at the internal line angles of conventional restorations with defective bonds. The overall
Interfacial stresses shapes of the cavity preparations were maintained while the profiles of the internal line
Finite element method angles were modified. In the present study, the overall shape of the cavity preparation
was also subject to modification in the optimization process. A topological optimization
method which placed the restorative material according to the stress distribution was first
used to obtain a draft design for the cavity shape, assuming perfect bonding at the tooth-
restoration interface. The draft shape was then refined using the method employed in the
previous study, to allow for deterioration in the interfacial bond strength. These optimiza-
tion methods were incorporated into the commercial finite element package ABAQUS as a
User Material Subroutine (UMAT) to automate the optimization process.
Results. Compared with the conventional design, the stress level at the tooth-restoration
interface in the optimized design was reduced significantly, irrespective of the bonding
condition.
Conclusions. Finite-element based shape optimization methods provide a useful tool for min-
imizing the interfacial stresses in dental restorations. The longevity of restored teeth using
the optimized designs is therefore expected to be prolonged.
© 2008 Published by Elsevier Ltd on behalf of Academy of Dental Materials.
All rights reserved.

1. Introduction was found to be the most common reason for replacement

(36%, 52% and 41% for composite, glass ionomer and amal-
There are many factors that cause failure of dental restora- gam, respectively), while fracture of tooth or restoration and
tions. In a survey of 3455 restorations [1], secondary caries lost composite restorations were the other two main causes.

∗
Corresponding author at: School of Dentistry, University of Minnesota, 515 Delaware Street SE, Minneapolis, Minnesota 55455, USA.
Tel.: +1 612 625 5406; fax: +1 612 626 1484.
E-mail address: [email protected] (A.S.L. Fok).
0109-5641/$ – see front matter © 2008 Published by Elsevier Ltd on behalf of Academy of Dental Materials. All rights reserved.
doi:10.1016/j.dental.2008.03.016
 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453 1445

In another survey involving amalgam and composite resin ite restorations [14–16]. Results from these studies have shown
restorations [2], the main reasons for failure were found to be that restorations with larger diameters and depths, i.e. higher
caries (34%), endodontic treatment (12%) and fracture of the C-factors, generated higher levels of shrinkage stress and
tooth (13%). Caries was found to be more common in compos- microleakage. For occlusal loading, finite element analysis has
ite restorations, while fracture of the tooth was more common also shown that the stress distribution in teeth with deep cav-
in amalgam restorations. This was attributed to the fact that ities is unfavourable [17].
amalgam restorations are not adhesively bonded to the cusps, Mondelli et al. [18] reported the fracture strength of max-
causing them to be more susceptible to cracking from the illary amalgam-restored premolars with three types of cavity
highly stressed internal line angles [2]. preparation – Class I, Class II and Mesiocclusodistal (MOD) cav-
Secondary caries occurs at the margin of an existing ities. Different buccolingual isthmus widths (one fourth, one
restoration. In amalgam restorations, marginal fracture is third, and one half of the intercuspal distance) were consid-
considered as a precursor to secondary caries [3]. Research ered. A compressive axial load was applied on the test teeth via
suggests that reducing marginal fracture could reduce the risk a steel sphere of 4 mm in diameter. The results showed that all
of secondary caries in teeth restored with amalgam [4]. Even occlusal cavity preparations decrease the strength of teeth in
in the absence of secondary caries, marginal fracture may proportion to the width of the preparation. The restored teeth
lead to premature replacement of the restoration [3]. Marginal with the narrower isthmus width resulted in better fracture
fracture, therefore, plays an important role in the failure of resistance for the same preparation.
amalgam restorations. For composite restorations, an anal- Some new cavity shapes were proposed by Porte et al. [19]
ysis of the findings of a multi-centre clinical trial revealed who investigated shrinkage stresses. Cavities with large CSA,
that restorations with marginal deterioration at year 3 were which were considered to lead to low stress concentrations
more likely to have failed by year 5 than restorations without near the free surface, resulted in better marginal integrity. A
marginal deterioration at year 3 [5]. There are many rea- follow-up finite element study using axi-symmetric models
sons that can cause marginal deterioration. Experiments have supported these findings [20]. Similar cavity shapes were also
shown that both occlusal loading and shrinkage of composite investigated by Hembree [21] who measured microleakage in
undermine marginal integrity [6,7], with the stresses produced composite restorations with different carvo-surface designs.
by these mechanisms causing failure of the tooth-restoration His results, however, indicated that etching the CSA and apply-
interface. ing enamel bond were more important than the type of cavity
Much work has been carried out to explore methods of shape used in minimizing microleakage.
reducing marginal deterioration. By means of finite element Douvitsas [22] examined the width of the gap at the cervical
analysis, Ausiello et al. [8] found that an appropriate adhe- wall following cyclic thermal loading in Class II compos-
sive layer thickness in a composite-restored tooth could lead ite resin restorations with rectangular and spherical cavity
to maximum shrinkage stress relief and improved interface designs. Human molars were prepared for each cavity design
integration. More flexible restorations were also shown to by two different procedures, with or without etching and
reduce the shrinkage stress magnitude and produce a more bonding. After restoration, the specimens were stored for
uniform stress distribution in the adhesive layer [9,10]. How- 24 h, and then thermally loaded for 1500 cycles. The results
ever, the materials and their interfaces also need to withstand indicated that the gap widths in the spherical cavities were
the stresses due to occlusal loading. smaller than those in the rectangular cavities, irrespective of
Other studies have addressed the influence of cavity shape the restorative material and procedure used. Moreover, the
on the magnitude of the interfacial stresses. An ideal cavity restorations placed with the use of etching and bonding agents
shape should minimise stress concentrations along the tooth- recorded the smallest gap width compared with the other
restoration interface due to sharp angles or differences in groups.
material properties. If the interfacial stresses can be kept lower A saucer-shaped cavity design was proposed and investi-
than the mechanical strength of the bond, marginal deterio- gated for composite restorations [23–25]. In a clinical trial, 51
ration in a composite restoration can be expected to reduce, preparations with such a cavity shape were completed and
thus prolonging the lifetime of the restoration. Moreover, a evaluated annually. Observations up to 10 years indicated that
lower stress level along the interface, with or without bonding, 70% of the restorations were acceptable for continued use
could also reduce the likelihood of fracture of the restoration [24]. The authors concluded that the saucer-shaped prepa-
and tooth [2]. The cavo-surface angle (CSA) and amalgam mar- ration was preferable to the box preparation in composite
gin angle (AMA) were considered to be important factors that restorations. A more recent study compared the survival
might affect the marginal integrity of amalgam restorations of restorations placed in saucer-shaped cavities to that of
[11]. Research has shown that cavity walls at 90◦ to the tooth restorations placed in tunnel preparations [25]. After a mean
surface would provide a maximum bulk of amalgam in the service life of 28.8 and 30.3 months, the proportion of the tun-
form of a butt joint [12]. In such restorations the enamel rods nel and saucer-shaped restorations survived was 46% and 76%,
will be parallel to the cavity wall. A CSA of 105◦ –115◦ and an respectively. Moreover, saucer-shaped restorations showed
AMA of at least 70◦ were clinically considered to be practical lower caries development than the tunnel restorations after
and sufficient to minimize the risk of enamel fracture [13]. an observation period of 24 months.
The effects of the cavity configuration factor (C-factor), i.e. the The alternative cavity shapes proposed in the above works
ratio of the bonded surface area in a cavity to the unbonded were mostly based on experience and intuition of the inves-
surface area, and the remaining dentine thickness (RDT) on tigators. More recently, attempts have been made by Proos
shrinkage stress development have been studied in compos- et al. [26] and Couegnat et al. [27] at applying modern shape
1446 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453

optimization techniques to the design of dental restorations.

Table 1 – Materials properties of tooth and restoration
They modified the designs of ceramic dental bridges and [33–36]
cavity shapes, respectively, using nature-inspired structural
Young’s modulus Poisson’s ratio
optimization methods to minimize stresses in the restora-
tions under occlusal loading. The work by Couegnat et al. Composite 19 GPa 0.24
[27] considered standard cavity shapes with weakened tooth- Enamel 80 GPa 0.30
restoration bonding, and the modifications introduced were Dentin 20 GPa 0.31
Pulp 2.1 MPa 0.45
mainly to the internal line angles only. The aim of the present
paper was to use shape optimization techniques to derive
alternative cavity shapes to minimize the interfacial stresses 2.1. Preliminary finite element (FE) calculations
for perfectly bonded restorations under occlusal loads. The
new cavity shape was then refined further to account for Fig. 1(a) shows a 2D finite element model for the crown of a
the possibility of deteriorated bonding between the tooth and premolar with a MOD composite restoration. The model was
restoration. The two optimization methods, one for each stage created from CT images of a real tooth (Fig. 1(b)) using the
of the optimization process, are called Stress-induced Material finite element software MSC PATRAN [32]. It contained 9981
Transformation (SMT) and Stress-induced Volume Transfor- eight-node elements and 30,172 nodes. A thickness of 5 mm
mation (SVT), respectively. They are based on methods which in the medio-distal direction was adopted. Material proper-
were originally proposed by Mattheck [28] and have gained ties of enamel, dentine and composite were assumed to be
a favourable reputation in German industry. Both methods isotropic and linearly elastic; their values are shown in Table 1
mimic the adaptive growth of biological structures such as [33–36]. Since the pulp is much softer than the other tissues
bones and trees stimulated by mechanical loads. SMT is used or materials (see Table 1), it was excluded from the finite
to determine a rough but efficient layout of materials in a element model to simplify the analysis. Perfect bonding was
structure and thus provides a draft design from which the opti- assumed between the tooth and the composite restoration.
mal shape could be obtained. SVT is then performed to refine The model was fully constrained at the lower boundary close
the design by smoothing the geometries to remove high stress to the enamel-dentine junction. A rigid ball with a diameter of
concentrations in the structure. Note that only SVT was used 4.7 mm was simulated to apply occlusal loading to the tooth.
in the previous study [27]. The applied load was 400 N. The actual analysis was carried
out using another finite element package, ABAQUS [31].
The aim of this preliminary finite element analysis was
2. Materials and method to identify the most highly stressed regions within the
restored tooth, especially those along the tooth-restoration
To better understand the mechanical behaviour of restored interface. These would be the regions to which shape opti-
teeth, the finite element (FE) method has been widely used in mization would be applied. Thus, the mesh around the
the field of dental research [29]. The advantages of finite ele- tooth-restoration interface was made much finer than those
ment simulations have been summarized as higher efficiency, in the other areas, with an average element edge length of
lower costs and lower risks, compared with experiments on 0.04 mm.
real teeth [30]. The method also facilitates the shape opti-
mization process. For the present work, SMT and SVT were 2.2. Stress-induced material transformation in
implemented into the finite element package ABAQUS [31] to structural topology optimisation
design cavity shapes with more favourable interfacial stresses
for both bonded and unbonded restorations under occlusal There has been considerable interest in the application of
loading. structural topology and shape optimization techniques to

Fig. 1 – (a) Finite element model of a tooth with restoration under occlusal loading and (b) CT image of the tooth used to
create the finite element model.
 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453 1447

determining the most efficient material distribution in struc- The reference stress  ref is the failure stress of the interfacial
tural design [37], and the finite element method is the main bond.
tool used to implement these techniques. SMT is one of the A User Material Subroutine (UMAT) was created to imple-
simplest structural topology optimization methods. It mimics ment the SMT method within the ABAQUS finite element suite
the biological mineralization process in living bone: material is [31]. The UMAT contains a constitutive material model which
deposited in loaded areas while non-load-bearing material is is able to change its properties according to the stresses as
carefully weeded out to provide a pre-optimized light-weight described above. The program was used iteratively during the
design. This can be achieved computationally using the finite finite element analysis until the spatial distribution of Young’s
element method. modulus converged.
Firstly, the design space is filled with a material of uni-
form Young’s modulus E and the corresponding finite element 2.3. Application of SMT in the optimization of cavity
model under the anticipated loading condition is analysed to shape
obtain the stress distribution in the structure. Next, the local
E value is modified according to the stresses calculated in the SMT was applied to both the enamel and dentine around the
previous step using a predefined function (E =f()). For exam- tooth-restoration interface of the restored tooth. Bonding at
ple, E can be set to be proportional to the von Mises stress. the interface was assumed to be perfect at this first stage of the
In this way, more heavily loaded zones are strengthened and optimization process, the objective of which was to improve
the lightly loaded zones are softened, which is analogous to the conventional cavity shape by finding an alternative design
the deposition and resorption processes in living bone. After with a more favourable interfacial stress distribution under
that, the new and inhomogeneous structure is subjected to occlusal loading.
the same loading condition and the process is repeated again The finite element model which was subjected to SMT,
and again until a converged state is achieved, i.e. the mate- together with its boundary and loading conditions, were the
rial is iteratively re-distributed within the design space. The same as those used in the preliminary analysis. Two user
resulting material layout provides the optimal starting point materials (UMAT) were defined, one for the enamel region
for the design. and the other for the dentine region. Young’s modulus of an
The key in this first stage of the optimization process is element at the tooth-restoration interface was changed from
the function E = f(), which could affect both the results and that of dentine or enamel to that of the restorative material if
the efficiency of the process. Several functions have been the local interfacial stress exceeds the reference stress, which
proposed. The simplest one used by Mattheck [28] is the one- was taken to be the micro tensile bond strength. The tooth-
to-one function (E = ), but it has low efficiency. Moreover, restoration interface gradually shifted in space until all the
structures derived using this function have a wide range of interfacial stresses fell below the reference stress. In order
Young’s modulus values, making them difficult to manufac- to limit the size of the problem, the maximum width of the
ture. A better function, also proposed by Mattheck [28], uses optimal cavity was limited to 3 mm.
the stress-increment-controlled method. With this method, The reference stress,  ref , was chosen to be the strength
Young’s modulus of an element is changed according to: of the bond between the tooth tissue and restoration, on the
assumption that debonding of the restoration is the more fre-
Ei+1 = Ei + k(i − ref ) (1) quent failure mode. Depending on the type of adhesive, the
method of surface preparation and test procedure, a wide
where Ei and  i are Young’s modulus and stress at the ith range of bond strengths have been reported [38–41]. In this
increment, respectively;  ref is a reference or threshold stress paper, the tensile strengths of bonds created using UNICEM
above which changes to Young’s modulus will occur; and k [38], 19.6 MPa for enamel-restoration and 15.9 MPa for dentine-
is a parameter that controls the rate of change of Young’s restoration, respectively, were chosen to be the basic reference
modulus. The stress-increment-controlled method has high stresses for the two user materials. Further, assuming that the
efficiency and can be used to adjust volume of the struc- bond strength would be reduced over time, two other sets of
ture being optimized by choosing different  ref . However, the reference stresses, which were 25% and 40% lower than the
structure thus created still has a wide distribution of Young’s original values, were also considered.
modulus values. To avoid this problem, Eqs. (2) and (3) below
were added to constrain the change of Young’s modulus: 2.4. Application of SVT in the optimization of cavity
shape
Ei+1 = E1 if Ei+1 < E1 (2)
SMT is used on the premise that there is no debonding along
Ei+1 = E2 if Ei+1 > E2 (3) the tooth-restoration interface. Significant deterioration of the
interfacial bond or complete debonding of the restoration from
E1 and E2 are, respectively, the maximum and minimum the tooth tissues will lead to high stress concentrations under
values of Young’s modulus allowed in the structure. The func- occlusal loading, notably at the internal line angles of the
tions (1), (2) and (3) dictate that there can only be two materials cavity, which could result in tooth fracture. The other shape
in the structure throughout the optimization process. In the optimization technique, SVT, can be used to refine the cav-
present work, E1 is Young’s modulus of the tooth tissue, either ity shape obtained from SMT to account for the possibility
enamel or dentine, depending on the region of the structure of interfacial debonding. With SVT, instead of changing the
concerned, while E2 is Young’s modulus of the composite. stiffness, the volume of the overloaded regions, usually on
1448 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453

the surface of a structure, is increased to reduce the stress enamel-restoration junctions. The maximum value of about
concentration. When the finite element method is used to 24 MPa, which is higher than the tensile bond strengths used
implement SVT, the surface layer of interest can be conve- in the present work, could lead to marginal fracture along the
niently given a stress-dependent, pseudo temperature field interface. Therefore, it is essential to reduce such stress con-
which will result in volume expansion in areas with high centrations. Moreover, the interfacial stresses should be kept
stresses. The mechanical-pseudo thermal process is repeated as low as possible considering the fact that bond strengths
until a uniform stress distribution in the surface layer is could decrease through time in the oral environment and
achieved under occlusal loading. SVT and its implementation interfacial debonding could occur through fatigue under cyclic
within ABAQUS is described in detail in ref. [27] and there- loading with stresses lower than the static bond strength.
fore will not be repeated here. Since the exact condition of The maximum interfacial shear stress (dotted line in
a deteriorated bond is difficult to establish, in this paper we Fig. 2(b)) is about 8 MPa, much lower than the peak ten-
assume, as a worst case scenario, that the bond between the sile stress. More importantly, the interfacial shear strengths
tooth and the restoration has totally failed when performing reported in the open literature are mostly higher than 20 MPa
shape optimization using SVT. [42,43]. Compared to the tensile stresses, the interfacial shear
stresses are therefore less important and are not considered
in the following optimization work.
3. Results
3.2. Optimizing cavity design using SMT
3.1. Preliminary finite element analyses
The results from using SMT to optimize the cavity shape
Fig. 2(a) shows the maximum principal stress contours in the
are shown in Fig. 3 in the form of Young’s modulus dis-
2D tooth model with the conventional cavity design and per-
tribution. Regions where material transformation has taken
fect bonding conditions. Focusing on the tooth-restoration
place are indicated in dark blue. Using the first set of bond
interface, the stresses normal and parallel to the interface are
strengths as the reference stresses and an occlusal load of
plotted in Fig. 2(b) which shows peaks of tensile stress in the
400 N resulted only in a very small change to the enamel
region (Fig. 3(a)). When the reference stresses for enamel and
dentine were reduced to 14.7 and 12 MPa, respectively, more
enamel and dentine were transformed to composite, giving a
more extended cavity space (Fig. 3(b)). Further reduction of
the reference stresses, to 11.8 and 9.5 MPa, led to an even
larger cavity. In this case, however, the transformed region
reached the boundary set for SMT, which means that the
boundary will need to be extended if the full cavity is to be
obtained.
The above results indicate that, under perfect bonding con-
dition, the optimal cavity should have a wider cavo-surface
opening in order to minimise the interfacial stresses under
occlusal loads. The results also show that there are no mate-
rial changes at the floor of the cavity, which suggests that a
deeper cavity is not required.
An improved cavity shape was then drawn up based on
the SMT results for the second set of reference stresses/bond
strengths; see Fig. 3(b). The width of the cavity was increased
where the stresses were highest, while the depth was kept the
same as in the original design. The new cavity thus has the
approximate shape of a hexagon, with the CSA being ∼100◦ .
Further finite element analysis was then performed to verify
the interfacial stresses in the new design.

3.3. Verification of the optimised design by means of

FEA

Fig. 4 shows the maximum principal stress contours and the

corresponding normal stress distribution along the tooth-
restoration interface. Compared with the stresses in the
Fig. 2 – (a) Maximum principal stress contours and (b) stress original design (Fig. 2), the two stress peaks at the enamel-
distribution along the tooth-restoration interface for a tooth composite junctions were reduced from 24 to 12 MPa, which is
with a conventional MOD cavity design.(For interpretation lower than the bond strength of 19.6 MPa for enamel and com-
of the references to color in this figure legend, the reader is posite. Although the maximum stress at the dentin-composite
referred to the web version of the article.) interface was slightly increased from 11.2 to 12.8 MPa, it is
 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453 1449

Fig. 3 – Cavity shape optimization using SMT with an occlusal load of 400 N and different reference stresses. Regions where
material transformation has taken place are shown in cyan. The indicated new cavity shape is shown in black in (b).

still lower than the bond strength for dentine and composite 3.4. Further shape optimization using SVT
(15.9 MPa). The FEA results therefore confirmed that the new
cavity design would reduce the interfacial stresses between When debonding of the restoration occurs, the mechanical
the composite restoration and tooth tissues, and thus could behavior of the restored tooth will be very different from that
potentially prevent the marginal fracture and prolong the life- with a perfectly bonded restoration. As shown in Fig. 5(a), the
time of the restoration under occlusal loading. maximum principal stresses due to occlusal loading in the
tooth with a debonded restoration are much higher than those
given in Fig. 4 for the bonded case. Peak stresses, as high as
125 MPa, could now be found at the internal line angles of
the cavity floor. These stresses could cause fracture of the
tooth, resulting in substantial or even irreparable structural
damage.
To reduce the stress concentrations resulting from debond-
ing of the restoration, as shown in Fig. 5 (a), a second stage of
shape optimization using SVT was conducted. Fig. 5(b) shows
the finite element model with the new cavity design which was
subjected to further shape optimization using SVT. The layer
of cavity surface elements which underwent SVT can be seen.
The results after 15 iterations are shown in Fig. 5(c) and (d). It
can be seen that the internal line angles of the cavity became
more rounded and, compared with the model in Fig. 5(a), the
stress concentrations there were reduced significantly.

3.5. Further verification using 3D finite element model

Strictly speaking, the 2D models considered in this paper are

only relevant to MOD restorations that have uniform dimen-
sions mesio-distally, which may not be representative. It
would be useful to see whether the cavity shape obtained from
the 2D optimization could also reduce interfacial stresses in a
more realistically shaped MOD restoration. 3D finite element
models with the conventional as well as the optimized cav-
ity design were therefore created for comparison. The model
with the conventional cavity shape and a cutout showing its
internal mesh density are shown in Fig. 6(a–c). The cavity was
extended to cover the central fissure and the pits. The depth
Fig. 4 – Maximum principal stress contours and normal of the cavity was similar to that in the 2D model. In prac-
stresses along the tooth-restoration interface for the SMT tice, all the corners on the cavity floor should be rounded to
optimized cavity design with perfect bonding. (Reproduced facilitate packing of the restorative material. However, to ease
with permission from Dental biomaterials: Imaging, testing their construction, sharp corners were used in the 3D models.
and modeling, Edited by RV Curtis and TF Watson, The model contains 558,789 elements and 115,612 nodes. A
Woodhead Publishing Ltd., 2008.) rigid ball was used to apply an occlusal load of 400 N axially
1450 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453

Fig. 5 – Maximum principal stress contours for the cavities before and after SVT optimization (a and c), finite element model
for SVT (b) and stresses along the interface for the final optimized cavity design after 15 increments (d).

(Fig. 6(a)). A 3D finite element model based on the 2D optimiza- Fig. 7(a) shows contours of interfacial stresses normal to
tion results was also created to verify the new cavity shape. the tooth-composite interface for a 3D model with the conven-
The model with a hexagonal shape for the midsection of its tional cavity design. The tooth, with its restoration removed
cavity is shown in Fig. 6(d–e). for clarity, is split mesio-distally into two approximately equal

Fig. 6 – 3D finite element model of the tooth with an MOD restoration: (a) whole model, (b – e) cavity shape and mesh
densities at the interfaces for the conventional and optimized cavity.
 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453 1451

Fig. 7 – Normal stresses at the tooth-composite interface of 3D model with conventional and optimized cavity design.

halves to reveal stress distributions on the cavity surfaces. compared with the original design. It is interesting to note that
As shown in this figure, very high tensile stresses (red) exist the new cavity shape is rather similar to the spherical cav-
in the enamel region on the buccal and lingual surfaces of ity design proposed by Douvitsas [22] in his study on the gap
the cavity. Under an occlusal load of 400 N, the maximum width at the cervical wall in Class II composite restorations
tensile interfacial stress is about 31.9 MPa. Such high stress following cyclic thermal loading and the saucer-shaped cavity
concentrations could cause marginal deterioration or even used by Nordbo et al. [23,24] in their clinical study. Both cavity
fracture of the tooth. These 3D results confirm those of the 2D designs have been shown to be superior to the conventional
FEA shown in Fig. 2, which indicate that the most vulnerable box cavity design. Although only MOD restorations have been
areas in terms of debonding are within the enamel-composite considered in this paper, it is expected that the optimization
junctions. Although the 2D interfacial stresses are lower in methods used can also be applied to other classes of restora-
magnitude, the fact that they have a very similar distribu- tion. Other restorative materials and bond strengths can also
tion to the 3D results justifies the use of 2D models in the be considered by changing the Young’s modulus and reference
optimization exercise. stresses in the optimization process.
Contours of the normal stresses along the tooth- A 400 N axial load was applied to the occlusal surface via
restoration interface in optimal design are shown in Fig. 7(b), a rigid ball. The load represented approximately the average
which has the same colour spectrum for stresses as those in of the maximum bite forces (MBF) in the premolar region
Fig. 7(a) to allow direct comparison. It can be seen that, com- which have been recorded [44–46]. It should be noted that a
pared with the conventional cavity design, the proportion of wide range of MBF have been reported (131–608 N), and the
areas with high tensile stresses was reduced significantly. measured biting forces appear to be related to the measur-
ing method, sex and age of the object [47]. The choice of the
applied load will affect the size and shape of the optimized
4. Discussion cavity design.
During SMT, there was transformation of dentine to com-
The combination of SMT and SVT has proved to be a poten- posite which took place away from the tooth-restoration
tially useful tool for cavity shape optimization in dental interface. In fact, it took place at part of the enamo-dentinal
restoration. The optimized cavity shape for an MOD restora- junction where the stresses exceeded the reference stress for
tion has been shown to have reduced interfacial stresses when dentine; see Fig. 6(b). The strength of the enamo-dentinal junc-
1452 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453

tion is not known precisely. If it is higher than the reference [6] Ferracane JL, Mitchem JC. Relationship between composite
stress, the change of material would be unnecessary, which contraction stress and leakage in Class V cavities. Am J Dent
means a more conservative cavity design than the one pro- 2003;16(4):239.
[7] Davidson CL, Abdalla AI. Effect of occlusal load cycling on
posed would be indicated.
the marginal integrity of adhesive Class V restorations. Am J
Changes to the cavity shape due to SVT are not significant Dent 1994;7(2):111.
(see Fig. 5), indicating that the shape given by the preced- [8] Ausiello P, Apicella A, Davidson CL. Effect of adhesive layer
ing SMT step is already close to the final optimized one for properties on stress distribution in composite restorations –
an unbonded restoration. The resulting reduction in stresses, a 3D finite element analysis. Dent Mater 2002;18(4):295.
however, is considerable, which shows the importance of SVT. [9] Braga RR, Ballester RY, Ferracane JL. Factors involved in the
When the structure to be optimized is more irregular in shape development of polymerization shrinkage stress in
resin-composites: a systematic review. Dent Mater
or contains more sharp angles, the changes produced by SVT
2005;21(10):962.
will be more apparent [27]. [10] Kahler B, Kotousov A, Borkowski K. Effect of material
The 3D finite element models confirm to a large extent properties on stresses at the restoration-dentin interface of
the validity of the 2D optimization results. However, since composite restorations during polymerization. Dent Mater
the 3D ‘optimized’ cavity shape was extrapolated from the 2006;22(10):942.
2D model without actually performing a 3D optimization, [11] Stratis S, Bryant RW. The influence of modified cavity design
and finishing techniques on the clinical performance of
some stress concentrations still exist in parts of the tooth-
amalgam restorations: A 2-year, clinical study. J Oral Rehabil
restoration interface (Fig. 7). To eliminate these remaining
1998;25(4):269.
high stresses, optimization with a 3D finite element model is [12] Mahler DB, Terkla LG. Relationship of cavity design to
necessary. However, the much increased computational effort restorative materials. Dent Clin North Am Mar
required is beyond the scope of the present work. 1965;23:149–57.
Possible limitations of this optimization exercise include [13] Elderton RJ. Cavo-surface angles, amalgam margin angles
the inaccuracies in the assumed linear elastic material prop- and occlusal cavity preparations. Br Dent J 1984;156(9):319.
[14] Yoshikawa T, Sano H, Burrow ME, Tagami J, Pashley DH.
erties and bonding conditions due to inadequate experimental
Effects of dentin depth and cavity configuration on bond
data. The step changes in material/tissue properties across the strength. J Dent Res 1999;78(4):898.
various interfaces might also be too simplistic. Moreover, only [15] Braga RR, Boaro LCC, Kuroe T, Azevedo CLN, Singer JM.
occlusal loading was considered in the present work; the other Influence of cavity dimensions and their derivatives (volume
important mechanical factor that can affect the interfacial and ‘C’ factor) on shrinkage stress development and
stress distributions, i.e. shrinkage stresses due to polymer- microleakage of composite restorations. Dent Mater
ization of the composite, was not considered. The optimized 2006;22(9):818.
[16] Feilzer AJ, De Gee AJ, Davidson CL. Setting stress in
cavity design in this paper has a larger C-factor, which could
composite resin in relation to configuration of the
lead to greater shrinkage stresses. Future work will attempt to restoration. J Dent Res 1987;66(11):1636.
address these limitations. [17] Goel VK, Khera SC, Gurusami S, Chen RC. Effect of cavity
depth on stresses in a restored tooth. J Prosthet Dent
1992;67(2):174.
5. Concluding remarks [18] Mondelli J, Steagall L, Ishikiriama A, de Lima Navarro MF,
Soares FB. Fracture strength of human teeth with cavity
preparations. J Prosthet Dent 1980;43(4):419–22.
Shape optimization techniques (SMT, SVT) coupled with the [19] Porte A, Lutz F, Lund MR, Swartz ML, Cochran MA. Cavity
finite element method can be used to derive cavity shapes designs for composite resins. Operative Dent 1984;9(2):50.
with lower interfacial stresses under occlusal loading. The [20] Hubsch P, Middleton J, Knox J. The influence of cavity shape
optimized cavity shapes can potentially prolong the life of on the stresses in composite dental restorations: a finite
restorations and reduce the incidence of fracture in restored element study. Comput Methods Biomech Biomed Eng
teeth. 2002;5(5):343.
[21] Hembree Jr JH. Microleakage of composite resin restorations
with different cavosurface designs. J Prosthet Dent
references 1980;44(2):171.
[22] Douvitsas G. Effect of cavity design on gap formation in
class II composite resin restorations. J Prosthet Dent
1991;65(4):475–9.
[1] Forss H, Widstrom E. Reasons for restorative therapy and the [23] Nordbo H, Leirskar J, von der Fehr FR. Saucer-shaped cavity
longevity of restorations in adults. Acta Odontol Scand preparation for composite resin restorations in class II
2004;62(2):82. carious lesions: three-year results. J Prosthet Dent
[2] Opdam NJM, Bronkhorst EM, Roeters JM, Loomans BAC. A 1993;69(2):155–9.
retrospective clinical study on longevity of posterior [24] Nordbo H, Leirskar J, Von Der Fehr FR. Saucer-shaped cavity
composite and amalgam restorations. Dent Mater preparations for posterior approximal resin composite
2007;23(1):2. restorations: observations up to 10 years. Quintessence Int
[3] Mahler DB, Engle JH. Clinical evaluation of amalgam bonding 1998;29(1):5–11.
in class I and II restorations. J Am Dent Assoc 2000;131(1):43. [25] Hörsted-Bindslev P, Heyde-Petersen B, Simonsen P, Baelum
[4] Goldberg AJ. Deterioration of restorative materials and the V. Tunnel or saucer-shaped restorations: a survival analysis.
risk for secondary caries. Adv Dent Res 1990;4:14. Clin Oral Invest 2005;9(4):233–8.
[5] Hayashi M, Wilson NHF. Marginal deterioration as a [26] Proos K, Steven G, Swain M, Ironside J. Preliminary studies
predictor of failure of a posterior composite. Eur J Oral Sci on the optimum shape of dental bridges. Comput Methods
2003;111(2):155–62. Biomech Biomed Eng 2000;4(1):77.
 d e n t a l m a t e r i a l s 2 4 ( 2 0 0 8 ) 1444–1453 1453

[27] Couegnat G, Fok SL, Cooper JE, Qualtrough AJE. Structural [38] De Munck J, Vargas M, Van Landuyt K, Hikita K, Lambrechts
optimization of dental restorations using the principle of P, Van Meerbeek B. Bonding of an auto-adhesive luting
adaptive growth. Dent Mater 2006;22(1):3. material to enamel and dentin. Dent Mater 2004;20(10):
[28] Mattheck C. Design in nature: learning from trees. Springer; 963.
1998, 290. [39] Lafuente JD, Chaves A, Carmiol R. Bond strength of dual-
[29] Geng JP, Tan KBC, Liu GR. Application of finite element cured resin cements to human teeth. J Esthet Dent (Canada)
analysis in implant dentistry: a review of the literature. J 2000;12(2):105.
Prosthet Dent 2001;85(6):585. [40] Hikita K, Van Meerbeek B, De Munck J, Ikeda T, Van Landuyt
[30] Magne P. Efficient 3D finite element analysis of dental K, Maida T, Lambrechts P, Peumans M. Bonding effectiveness
restorative procedures using micro-CT data. Dent Mater of adhesive luting agents to enamel and dentin. Dent Mater
2007;23(5):539. 2007;23(1):71.
[31] ABAQUS version 6.5 user manuals, Pawtucket, RI, USA.: [41] Van Noort R, Noroozi S, Howard IC, Cardew G. A critique of
Hibbett, Karlsson and Sorensen Inc. 2004. bond strength measurements. J Dent 1989;17(2):61.
[32] MSC.Patran, user’s manual version 2001, MSC.Software [42] Barkmeier WW, Erickson RL. Shear bond strength of
Corporation, 2001. composite to enamel and dentin using Scotchbond
[33] Kinney JH, Marshall SJ, Marshall GW. The mechanical Multi-Purpose. Am J Dent 1994;7(3):175.
properties of human dentin: a critical review and [43] Sfondrini MF, Cacciafesta V, Pistorio A, Sfondrini G. Effects of
re-evaluation of the dental literature. Crit Rev Oral Biol Med: conventional and high-intensity light-curing on enamel
Off Publ Am Assoc Oral Biol 2003;14(1):13. shear bond strength of composite resin and resin-modified
[34] Habelitz S, Marshall SJ, Marshall Jr GW, Balooch M. glass-ionomer. Am J Orthod Dentofacial Orthop
Mechanical properties of human dental enamel on the 2001;119(1):30.
nanometre scale. Arch Oral Biol 2001;46(2):p173. [44] Kleinfelder JW, Ludwig K. Maximal bite force in patients
[35] Abe Y, Lambrechts P, Inoue S, Braem MJA, Takeuchi M, with reduced periodontal tissue support with and without
Vanherle G, Van Meerbeek B. Dynamic elastic modulus of splinting. J Periodontol 2002;73(10):1184–7.
‘packable’ composites. Dent Mater 2001;17(6):520. [45] Widmalm SE, Ericsson SG. Maximal bite force with centric
[36] Braem M, Van Doren VE, Lambrechts P, Vanherle G. and eccentric load. J Oral Rehabil 1982;9(5):445–50.
Determination of Young’s modulus of dental composites: a [46] Ferrario VF, Sforza C, Serrao G, Dellavia C, Tartaglia GM.
phenomenological model. J Mater Sci 1987;22(6):2037. Single tooth bite forces in healthy young adults. J Oral
[37] Hassani B, Hinton E. A review of homogenization and Rehabil 2004;31(1):18–22.
topology optimisation. I. Homogenization theory for media [47] Bakke M. Bite Force and occlusion. Semin Orthod
with periodic structure. Comput Struct 1998;69(6):707. 2006;12(2):120–6.