Available online at www.sciencedirect.

com

Acta Biomaterialia 4 (2008) 1715–1724

www.elsevier.com/locate/actabiomat

Finite element modeling as a tool for predicting the fracture

behavior of robocast scaffolds
Pedro Miranda *, Antonia Pajares, Fernando Guiberteau
Departamento de Ingenierı́a Mecánica, Energética y de los Materiales, Universidad de Extremadura, Avda de Elvas s/n, 06071 Badajoz, Spain

Received 10 January 2008; received in revised form 13 May 2008; accepted 22 May 2008
Available online 5 June 2008

Abstract

The use of finite element modeling to calculate the stress fields in complex scaffold structures and thus predict their mechanical behav-
ior during service (e.g., as load-bearing bone implants) is evaluated. The method is applied to identifying the fracture modes and esti-
mating the strength of robocast hydroxyapatite and b-tricalcium phosphate scaffolds, consisting of a three-dimensional lattice of
interpenetrating rods. The calculations are performed for three testing configurations: compression, tension and shear. Different testing
orientations relative to the calcium phosphate rods are considered for each configuration. The predictions for the compressive configu-
rations are compared to experimental data from uniaxial compression tests.
Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Finite element analysis; Robocasting; Scaffolds; Strength; Fracture modes

1. Introduction properties of the materials comprising the structure are

Tissue engineering for bone regeneration requires the
use of a porous scaffold to serve as a template for cell inter-
actions and the formation of the extracellular matrix, as
well as to provide structural support for the newly formed
tissue [1]. These scaffolds should be able to withstand some
degree of loading during their use in vivo, besides provid-
ing the required biological response. However, since por-
ous scaffolds are inherently weak (especially those based
on calcium phosphates) [2], it is necessary to optimize their
mechanical response so that they can provide the appropri-
ate support while bone regenerates and they are slowly
resorbed [3]. Characterizing and predicting the mechanical
behavior of these complex three-dimensional (3D) struc-
tures is therefore an essential task.
Finite element modeling (FEM) has been shown to be
capable of predicting the behavior of complex structures,
such as multilayer systems [4–10], provided the mechanical
*
Corresponding author. Tel.: +34 924 28 9600; fax: +34 924 28 9601.
E-mail address: pmiranda@unex.es (P. Miranda).
known. Consequently, the objective of the present work
was to analyze the capabilities of finite element simulation
as a technique for calculating the stress fields in complex
scaffolds and thereby estimating their mechanical behavior
under different loading configurations. In particular, FEM
is applied to identify the fracture modes and estimate the
strength of hydroxyapatite (HA) and b-tricalcium phos-
phate (b-TCP) scaffolds fabricated by robocasting.
Robocasting, or direct-write assembly, is a solid free-
form fabrication (SFF) technique that consists of the
robotic deposition of highly concentrated colloidal suspen-
sions (inks) capable of fully supporting their own weight
during assembly [11–13]. Robocasting allows ceramic scaf-
folds to be built using water-based inks with minimal
organic content (<1 wt.%) and without the need for a sac-
rificial support material or mold. Using this technique a
network of semisolid ink rods is printed directly, hence
the term direct-write assembly, by extrusion through a
deposition nozzle to get the desired 3D structure. Fig. 1
shows a schematic diagram of the process.
The prototype structure considered for this study
consisted of a 3D tetragonal lattice of interpenetrating rods

1742-7061/$ - see front matter Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actbio.2008.05.020
1716 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724
Fig. 1. Scheme of the robocasting fabrication process. The ceramic
scaffold is built layer by layer from a computer design. A three-axis robotic
arm moves the injection syringe while pressing the ceramic ink through the
conical deposition nozzle, immersed in an oil bath, to create a self-
supporting 3D network of ceramic rods.
– fabricated from either HA or b-TCP inks. The mechani-
cal behavior of these scaffolds was simulated by FEM
under three testing configurations: uniaxial compression,
tension and shear. Different testing orientations relative
to the rods comprising the structure were considered for
each configuration. The most probable fracture mode
was determined from the location of the maximum tensile
stresses in the structure. This approach had been found
to give excellent results in an earlier study of the compres-
sive fracture modes of HA scaffolds [14]. The predicted
compressive, tensile and shear strengths of the scaffolds
for each testing orientation were estimated from the
FEM results assuming a critical stress criterion. Finally,
the predictions for the compressive configuration were
compared to experimental data from uniaxial compression
tests.
2. Experimental procedure
Details of the fabrication of the HA and b-TCP inks and
the corresponding robocast scaffolds have been presented
elsewhere [2,15] and, as they are used here only for the exper-
imental verification of the FEM results obtained in this
work, only the relevant data will be included in this section.
The 3D scaffolds, consisting of a lattice of ceramic rods,
were constructed layer by layer via extrusion of the corre-
sponding inks through conical deposition nozzles, using a
computer-controlled robotic deposition device (3D Inks,
Stillwater, OK), as shown schematically in Fig. 1. The
deposition was done in a non-wetting oil bath to prevent
non-uniform drying during assembly. The samples were
removed from the bath and dried in air at room tempera-
ture for 24 h, followed by a burn-out at 400 °C (ramp at
1 °C min 1 and a 1 h hold) to evaporate organics. They
were finally sintered at 1300 °C (ramp at 3 °C min 1 and
a 2 h hold). The resulting scaffolds were imaged (Fig. 2)
using scanning electron microscopy (SEM) (S-4300SE/N,
Hitachi, USA). The average rod diameter was measured
to be d = 220 ± 10 lm and the average center-to-center
spacing between rods in the printing plane s = 300 ±
10 lm, with a rod overlap between adjacent layers of about
50 ± 10 lm, yielding a vertical spacing between like layers
of around h = 340 ± 10 lm (Fig. 3).
The intrinsic mechanical properties of the individual cal-
cium phosphate rods were evaluated to provide essential
input parameters for the numerical simulations. Instru-
mented indentation (Nanotest, Micro Materials Ltd.,
Wrexham, UK) was used to determine the elastic moduli
of the HA and b-TCP rods. Berkovich indentation tests
were performed on polished sections (to 1 lm finish) per-
pendicular to the rod axis. Single indentations of about
5.5 lm depth and 40 lm side were placed at the center of
the rods. This indent size is large enough compared to
the grain sizes in both materials (3.2 ± 0.5 and
7.4 ± 0.7 lm for HA and b-TCP, respectively [2]) to pro-
vide meaningful information about the mechanical proper-
ties of the rods (and not of individual grains) but small
enough to avoid a significant influence of the free surface
of the rods – ASTM E 384 recommends a center-to-side
spacing between Vickers indents of 2.5 times the diagonal.
The inert fracture strength (i.e., the strength in the
absence of chemical fatigue or slow crack growth) of the
rods was estimated from high-load-resolution three-point
bending tests (Electroforce 3200, Bose Corp., Eden Prairie,
MN) performed on individual rods printed and sintered for
the purpose. The tests were performed in air at a constant
crosshead speed of 30 mm min–1. Since the fracture load
for rods of 220 lm diameter was close to the sensitivity
limit of the load cell, thicker rods of 360 lm diameter were
used instead. Although the geometry and size of these iso-
lated rods deviate from the real situation, these tests are the
best possible means available to measure this material
property (much more accurate than testing bulk samples
of like materials, for example). Besides, in the opinion of
the authors, the actual values of the rod strength should
lay within the errors of these imperfect measurements.
Finite element simulations were carried out using ABA-
QUS/StandardÒ software (Simulia, Providence, RI). The
algorithm models an approximately cubic scaffold
(2  2  2.09 mm) consisting of 12 alternating orthogonal
layers of parallel HA or b-TCP rods of 220 lm diameter
spaced 300 lm from center to center, with each layer over-
lapping the adjacent ones by 50 lm (Fig. 3). The FEM grid
 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724
Fig. 2. SEM micrographs showing the morphology of HA (a, c) and b-TCP (b, d) scaffolds after sintering at 1300 °C for 2 h: (a, b) printing plane view and
(c, d) cross-section view.
Fig. 3. Finite element grid used to simulate the scaffolds, consisting of 12
layers of rods of 220 lm diameter, d, spaced s = 300 lm from center-to-
center, and with a layer overlap of 50 lm that yields a like-layer spacing of
h = 340 lm along the printing direction (direction 3). The dimensions of
the elements at the rod surfaces are around 25 lm, but become larger
towards the interior.
consists of more than 1.25  106 linear tetrahedral elements
of about 25 lm in the vicinity of the external surfaces of the
rods, but progressively increasing in size towards their inte-
rior. Isotropic elastic behavior is assumed for the whole
1717
system, using the elastic moduli obtained from the indenta-
tion tests and a Poisson’s ratio of 0.28 [16] for both
materials.
The scaffold is placed between two parallel rigid plates
(Fig. 4), already in contact with the structure from the
beginning. The contact is considered frictionless for the
compressive test simulations and infinitely rough and
sticky for the tensile and shear test simulations. One of
the rigid plates is fixed while the other moves under the
action of a linearly increasing applied force (up to
500 N). For both the uniaxial tensile and compressive test
simulations, two different orientations were considered,
one with the load applied perpendicular to the printing
plane (Fig. 4a and c) and the other with the load applied
parallel to the rods (Fig. 4b and d), i.e., along the equiva-
lent axis 1 or 2 in Fig. 3. For the shear test simulations
the three different load orientations depicted in Fig. 4e–g
were considered.
The evolution of stresses with the applied stress – esti-
mated by normalizing the applied load by the correspond-
ing initial cross-sectional area of the simulated sample –
was recorded during the simulation. The predicted
strengths (compressive, tensile, or shear) of the scaffolds,
i.e., the applied stress acting on the structure at failure,
for each testing configuration were calculated assuming a
critical stress criterion. First the greatest tensile stress act-
ing at each point in the structure, rt, is determined from
the FEM calculated stress tensor. Failure is then presumed
to occur when the maximum of rt equals the inert fracture
strength, rF, of the rods (as determined from the bending
1718 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724

Fig. 4. Scheme of the different testing configurations simulated in this study: compression along (a) the printing direction or (b) a rod axis; tension along
(c) the printing direction or (d) a rod axis; shear (e) on the printing plane along a rod axis, or on a plane orthogonal to a rod axis along (f) the printing
direction or (g) a rod axis. The rigid plates used in the FEM simulations to transfer the load to the structure are indicated. The bottom plate is held fixed
while a load is applied to the top plate in the direction indicated by the arrows.

tests) – this implicitly assumes an opening fracture mode the structure was estimated as the maximum applied load
(mode I), which is that most commonly activated in brittle divided by the square external cross-section of the sample,
materials [17]. For the calculation of the maximum value of and the results were compared with the FEM predictions.
rt, the near-contact regions were ignored as they are not More than 10 samples were tested in each case in order
representative of the macroscopic behavior of the structure to get statistically reliable values.
under each load configuration (i.e., uniaxial tension, com-
pression or shear), but rather of the boundary conditions 3. Results and discussion
selected for the simulations. Furthermore, contact stresses
will only produce localized damage [14] and would not be Table 1 summarizes the results of the characterization of
responsible for the ultimate failure of the structure except the individual rods comprising the HA and b-TCP scaf-
under high-loading-rate, e.g., impact, conditions. folds. The elastic modulus, E, of each material was calcu-
Experimental uniaxial compression tests were per- lated from the effective modulus, E* = E/(1 m2),
formed along the aforementioned two directions (Fig. 4a obtained from the indentation tests, assuming a value of
and b) on cubic blocks of about 2 mm side cut from the sin-
tered HA and b-TCP scaffolds. The tests were carried out
on a universal testing machine (AG-IS10kN, Shimadzu Table 1
Intrinsic mechanical properties of HA and b-TCP rods
Corp., Kyoto, Japan) in air at a constant crosshead speed
of 0.6 mm min 1 (the rate of 30 mm min 1 used in bending E* (GPa) ma E (GPa) rF (MPa)
tests was impractical for the compression tests because it HA 83 ± 4 0.28 82 ± 4 68 ± 12
caused premature failure from contact damage and thus TCP 38 ± 8 0.28 36 ± 7 27 ± 9
a
a slower rate was selected). The compressive strength of From Ref. [16].
 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724
0.28 [16] for the Poisson’s ratio. As can be clearly appreci-
ated, the HA rods exhibited more than twice the stiffness
and strength of the b-TCP rods.
These elastic properties were used as input for the FEM
simulations of the different testing configurations depicted
in Fig. 4. The FEM results confirmed that the stresses
developed in the scaffolds for each configuration are obvi-
ously independent of the elastic modulus of the material
considered, as shown for the case of uniaxial compression
in Fig. 5, where the maximum of rt is represented vs. the
applied stress for both HA and b-TCP. Consequently, the
analysis that follows of the stress fields generated and the
associated damage modes that are to be expected in each
case is applicable to any scaffold with a similar geometry.
Evidently, the strains do vary from one material to
another, and the elastic modulus of the material compris-
ing the individual rods will determine the effective modulus
of the scaffold.
Fig. 6 shows the contours of the greatest tensile stress,
rt, developed in the scaffold during uniaxial compression
at 250 N load (which is around the critical load for fracture
in the experimental tests) for the two orientations consid-
ered (Fig. 4a and b). When compression is applied perpen-
dicularly to the printing plane (Fig. 6a), i.e., along direction
3, rt presents a maximum at the centers of the unsupported
segments – specifically at the top and bottom surfaces of
those segments – of all the rods (although not so clearly vis-
ible in Fig. 6a for those aligned along direction 2). At those
points, rt is directed along the corresponding rod axis (i.e.,
rt = r11 or rt = r22) and therefore, assuming an opening
fracture mode (mode I), which, as mentioned above, is
the most commonly activated in brittle materials [17], will
Fig. 5. Plot of maximum tensile stress in the structure (maximum of rt) vs.
the applied stress (applied load normalized by initial cross-sectional area)
for both HA and b-TCP scaffolds in the two compression directions
considered: perpendicular to the printing plane (direction 3, Fig. 4a) and
along the rod axes (direction 1 or 2, Fig. 4b). Note that the results are
independent of the material.
1719
induce cracking perpendicular to the rod axes over the
entire structure at those points. This type of cracking has
indeed been observed in experimental compressive tests
with this same orientation [14]. Of course, crack initiation
does not necessarily occur at the exact location of the max-
imum tensile stresses because it also depends on the loca-
tion of the largest flaws [18].
When the compressive load is applied along one of the
rod axes (directions 1 or 2), the maximum tensile stresses
are again located on the top and bottom surfaces of the
horizontal rods (Fig. 6b), but now close to the joints with
the vertical rods (see the inset). The orientation of these
maximum tensile stresses is such that it will induce cracking
(again, assuming mode I fracture) of the horizontal rods
along those joints, effectively detaching them from the ver-
tical rods that remain largely in compression. This predic-
tion was also consistent with the curvature of the crack
surfaces observed in the experimental tests [14]. The rela-
tive magnitude of the maximum stresses in the two orienta-
tions will be discussed in detail below.
For the case of uniaxial tension tests, Fig. 7 shows the rt
contours developed in the scaffold at 250 N load (for com-
parison with Fig. 6) for the two orientations considered
(Fig. 4c and d). Note that, since stresses increase linearly
with load (Fig. 5), contours at any other load can be calcu-
lated just by multiplying the stresses in the legend by the
appropriate factor. When the loading is along direction 3
(Fig. 7a), and ignoring the values near the contacts, rt pre-
sents its maximum values at the joints between adjacent
rod layers, as clearly shown in Fig. 7a for the rods aligned
along direction 1 but also present in those aligned along
direction 2. At a certain load, these maximum tensile stres-
ses will induce fracture along those joints, separating each
layer from its neighbors. When the tension is applied along
one of the rod axes (directions 1 or 2), the tensile stresses
are largely located on the vertical rods (Fig. 7b), with their
maxima in the free segments of those rods close to (though
not exactly at) the joints with the horizontal rods (see
inset). The orientation of these maximum tensile stresses
is largely along the rod axes, so that cracks originated by
them will tend to break the vertical rods transversally.
Fig. 8 shows analogous contour plots of rt at 250 N load
for the three different shear configurations selected (Fig.
4e–g). For the sake of clarity, the direction of the applied
load is included in the figures (recall that the bottom sur-
face is kept fixed). Although the three configurations are
significantly different, the stress contours generated exhibit
many similarities. For all three orientations, the tensile
stresses organize into bands of alternating intensity ori-
ented at 45° with respect to the load axis. The maxima of
these stresses, disregarding the near-contact field, are
located at the joints between the different rod layers, the
edges at the intersections acting as stress concentrators.
Therefore, for the three configurations, cracking will occur
at these locations and the different rod layers will tend to
detach from each other. However, the actual cracking pro-
cess will be slightly different in the case of the configuration
1720 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724

Fig. 6. FEM-generated stress contours corresponding to the highest tensile stress, rt, generated when a compressive load of 250 N is applied (a)
orthogonal to the rods (direction 3) or (b) along a rod axis (direction 1, equivalent to direction 2).

in Fig. 8c, where adjacent layers tend to rotate relative to
each other instead of simply one displacing the others, as
in the cases of Fig. 8a and b.
Concerning the relative intensities of the maximum
stresses developed in each of the configurations analyzed,
Fig. 9 shows the absolute maximum value of rt anywhere
in the scaffold vs. the applied stress (i.e., applied load nor-
malized by the corresponding scaffold’s external cross-sec-
tion) for each testing configuration. This representation
and the conclusions that derive from it are independent
of the material considered and therefore have a universal
validity for any scaffold with the same morphology. As
can be clearly appreciated, there are very large differences
in the maximum tensile stress levels that develop in the
scaffolds depending on the type of load applied. As
expected, compressive loads are the mildest, with stresses
about three times lower than those that develop in the ten-
sion tests and more than an order of magnitude lower than
those of the shear configurations. Of the two compression
configurations considered, the stresses are higher when
the load is applied perpendicular to the printing plane
(direction 3), but this situation is reversed under tensile
loads, which yield higher stresses when applied along the
rod axes (direction 1 or 2). The most deleterious configura-
tion analyzed is that corresponding to shear load applied in
planes perpendicular to the printing plane, along the direc-
tion of the rod axes (Fig. 4g). The torsion-like relative dis-
placement of the rod layers generates a stress field around
the joint that is significantly greater than in the two other
shear configurations, which nonetheless both give stress
levels more than twice as large as those of the tension con-
figurations. An immediate, though anticipated, conclusion
from these results is that brittle porous scaffolds fabricated
by robocasting, regardless of the material they are made of,
will exhibit far superior strength properties under compres-
sive loads; tensile and shear stresses acting on them should
be minimized, if not altogether suppressed, to improve
their mechanical performance.
 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724 1721

Fig. 7. FEM-generated stress contours corresponding to the greatest tensile stress, rt, generated when a tensile load of 250 N is applied (a) orthogonal to
the rods (direction 3) or (b) along a rod axis (direction 1, equivalent to direction 2).

From the plots in Fig. 9 it is simple to predict the
mechanical strength of a scaffold for each testing configura-
tion by using a critical stress criterion, provided the intrin-
sic strength of the constituent material is known. Indeed,
these plots allow one to determine the applied stress at
which the maximum tensile stress in the scaffold will equal
the strength of the material, i.e., the structure strength
under that type of load. Fig. 10 shows the strength values
calculated following this procedure for the two materials
considered in this study, HA and b-TCP, using the values
of the inert strength, rF, from Table 1. The error bars rep-
resent standard deviations of the predictions due to uncer-
tainties in the measurement of rF from the corresponding
three-point bending tests. As expected, the compressive
strength values exhibited by both materials are significantly
higher than their corresponding tensile and, especially,
shear strengths. That the HA values are more than twice
those corresponding to b-TCP is a simple reflection of their
respective inert strengths.
Finally, Fig. 11 shows the comparison between the pre-
dicted compressive strength and the experimental results
obtained in the uniaxial compressive tests performed on
the actual scaffolds for the two testing orientations consid-
ered (Fig. 4a and b). Again, the error bars represent the
standard deviation of the data. The FEM predictions are
seen to agree with the experimental compressive strengths
1722 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724

Fig. 8. FEM-generated stress contours corresponding to the greatest tensile stress, rt, generated when a shear load of 250 N is applied on the printing
plane along a rod axis (direction 1, equivalent to direction 2) (a), or on a plane orthogonal to a rod axis (direction 1, equivalent to direction 2) along the
printing direction (direction 3) (b) or along the perpendicular rod axis (direction 2, equivalent to direction 1) (c).

within uncertainties for both the HA and the b-TCP scaf- able for b-TCP, could be attributed to the use of an inert
folds. The apparently systematic slight (within the errors) strength value, while the actual experiments were per-
overestimate of the compressive strength, especially notice- formed at a low speed, where slow crack growth could be
 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724
Fig. 9. Plot of maximum tensile stress in the scaffold (maximum of rt) vs.
the applied stress (applied load normalized by initial cross-sectional area)
for all the testing configurations analyzed: compression (two solid lines),
tension (two dashed lines) and shear (three lines with dots). Legend
notation is the same as used in Fig. 4.
Fig. 10. Bar graph showing HA and b-TCP scaffold strength values
predicted by FEM for the different testing configurations, using the inert
strength, rF, values of Table 1. Error bars represent standard deviations of
predictions due to uncertainties in the measurement of rF. Label notation
on the x-axis is the same as used in Fig. 4.
reducing the fracture strength of the scaffold. The TCP
strength values used to calculate FEM estimates are less
reliable than the corresponding HA values because, even
after increasing the rod diameters, the critical loads for
cracking lay close to the sensitivity threshold of the load
cell used in the three-point bending test, which could
explain the more significant deviation of FEM prediction
from experimental values for this material. In sum, there-
1723
Fig. 11. Bar graph showing comparison between experimental measure-
ments and FEM predictions of the compressive strength of HA and b-TCP
scaffolds for the two compression orientations considered. Error bars
represent standard deviations of the data.
fore, these results confirm the ability of FEM to predict
the fracture behavior of robocast scaffolds both qualita-
tively (i.e., to predict their fracture modes [14]) and quanti-
tatively (i.e., to calculate their strength), which was the
main objective of the present work.
4. Conclusions and implications
Computer-aided fabrication methods, such as robocast-
ing, are ideal for fabricating scaffolds with a designed pore
structure that enables their biological and mechanical
responses to be optimized to levels unattainable with other
techniques. In particular, optimizing the mechanical per-
formance of porous brittle scaffolds is critical since that is
precisely what prevents their widespread application for
regeneration of load-bearing bone defects. To perform
such optimization, it is necessary to have a predictive tool
as an aid in the design of the most favorable scaffold archi-
tecture. As the present results show, FEM could be that
tool, as it enables the analysis and prediction of failure
modes and the quantitative evaluation of important
mechanical parameters, such as strength.
In particular, the present study has successfully analyzed
the stress fields and the expected fracture modes under
seven different loading conditions (Fig. 4) of a robocast
scaffold consisting of a tetragonal lattice of ceramic rods.
From these results (Fig. 9), a key guideline for the design
of brittle robocast scaffolds was derived: implant shape
and location should be designed to minimize the tensile
and, above all, shear stresses acting on the structure. This
last type of load is extremely prejudicial for the mechanical
performance of the scaffold as it generates intense tensile
stresses that can lead to premature failure of the system.
1724 P. Miranda et al. / Acta Biomaterialia 4 (2008) 1715–1724
Fortunately, in most bone implant applications the loads
will be largely compressive and therefore scaffold perfor-
mance will not be so seriously jeopardized. For this partic-
ular type of load, another recommendation would be to
preferentially orient the scaffolds so that the compression
is parallel to one of the rod axes in order to minimize the
tensile stresses generated in the structure.
The FEM results were also used to predict the fracture
strength of HA and b-TCP scaffolds fabricated by this
technique under each type of load by applying a critical
tensile stress criterion using their respective inert strengths,
as measured in three-point bending tests, as the critical
stress values. The compressive strengths predicted by the
FEM simulations (for both materials in the two orienta-
tions considered) were successfully validated by compari-
son with experimental uniaxial compression test data,
justifying the suitability of the present methodological
approach for the optimization task at hand.
Indeed, predictions obtained by FEM could allow one
to determine how the different geometrical variables (rod
thickness, spacing, interpenetration, relative angles, etc.)
affect the mechanical behavior of the structure. Work in
this direction is currently in progress. The results of those
studies will make it possible to create an intelligent design
of the scaffold’s initial computer-aided design model and
thereby optimize its mechanical performance. Although
we have focused here on the estimation of fracture modes
and strength, nothing prevents this method from being
applied to the optimization of other mechanical parame-
ters, such as the elastic modulus of the structure. In sum,
FEM is a powerful tool for predicting the mechanical per-
formance of scaffolds fabricated by robocasting or other
SFF techniques, and thus optimizing their mechanical per-
formance through the intelligent design of their geometry,
which is an unavoidable prerequisite for their use in load-
bearing bone tissue engineering applications.
Acknowledgements
This work was supported by the Ministerio de Educa-
ción y Ciencia (Spanish Government) and the Fondo So-
cial Europeo (MAT2006-08720).
References
[1] Karageorgiou V, Kaplan D. Porosity of 3D biomaterial scaffolds and
osteogenesis. Biomaterials 2005;26:5474–91.
[2] Miranda P, Pajares A, Saiz E, Tomsia AP, Guiberteau F. Mechanical
properties of calcium phosphate scaffolds fabricated by robocasting. J
Biomed Mater Res A 2008;85A:218–27.
[3] Adachi T, Osako Y, Tanaka M, Hojo M, Hollister SJ. Framework
for optimal design of porous scaffold microstructure by computa-
tional simulation of bone regeneration. Biomaterials 2006;27:
3964–72.
[4] Zhao H, Miranda P, Lawn BR, Hu XZ. Cracking in ceramic/metal/
polymer trilayer systems. J Mater Res 2002;17:1102–11.
[5] Chai H, Lawn BR. Cracking in brittle laminates from concentrated
loads. Acta Mater 2002;50:2613–25.
[6] Miranda P, Pajares A, Guiberteau F, Deng Y, Lawn BR. Designing
damage-resistant brittle-coating structures: I. Bilayers. Acta Mater
2003;51:4347–56.
[7] Miranda P, Pajares A, Guiberteau F, Deng Y, Zhao H, Lawn BR.
Designing damage-resistant brittle-coating structures: II. Trilayers.
Acta Mater 2003;51:4357–65.
[8] Hsueh CH, Luttrell CR, Becher PF. Analyses of multilayered dental
ceramics subjected to biaxial flexure tests. Dent Mater 2006;22:460–9.
[9] Kim JH, Miranda P, Kim DK, Lawn BR. Effect of an adhesive
interlayer on the fracture of a brittle coating on a supporting
substrate. J Mater Res 2003;18:222–7.
[10] Deng Y, Miranda P, Pajares A, Guiberteau F, Lawn BR. Fracture of
ceramic/ceramic/polymer trilayers for biomechanical applications. J
Biomed Mater Res A 2003;67A:828–33.
[11] Cesarano III J, Segalman JR, Calvert P. Robocasting provides
moldless fabrication from slurry deposition. Ceram Ind
1998;148:94–102.
[12] Cesarano J, Calvert P. Freeforming objects with low-binder slurry.
US Patent 6027326 (2000).
[13] Smay JE, Cesarano J, Lewis JA. Colloidal inks for directed assembly
of 3D periodic structures. Langmuir 2002;18:5429–37.
[14] Miranda P, Pajares A, Saiz E, Tomsia AP, Guiberteau F. Fracture
modes under uniaxial compression in hydroxyapatite scaffolds
fabricated by robocasting. J Biomed Mater Res A 2007;83A:646–55.
[15] Miranda P, Saiz E, Gryn K, Tomsia AP. Sintering and robocasting of
beta-tricalcium phosphate scaffolds for orthopaedic applications.
Acta Biomater 2006;2:457–66.
[16] Grenoble DE, Dunn KL, Katz JL, Gilmore RS, Murty KL. Elastic
properties of hard tissues and apatites. J Biomed Mater Res
1972;6:221–33.
[17] Lawn BR. Fracture of brittle solids. Cambridge: Cambridge Univer-
sity Press; 1993.
[18] Miranda P, Pajares A, Guiberteau F, Cumbrera FL, Lawn BR. Role
of flaw statistics in contact fracture of brittle coatings. Acta Mater
2001;49:3719–26.