Materials Science and Engineering C 76 (2017) 406–416

Contents lists available at ScienceDirect

Materials Science and Engineering C

journal homepage: www.elsevier.com/locate/msec

Effects of plasma electrolytic oxidation process on the mechanical

properties of additively manufactured porous biomaterials

Zahra Gorgin Karaji a,b,⁎, Reza Hedayati a, Behdad Pouran a,c, Iulian Apachitei a, Amir A Zadpoor a
a
Department of Biomechanical Engineering, Faculty of Mechanical, Maritime, and Materials Engineering, Delft University of Technology (TU Delft), Mekelweg 2, 2628, CD, Delft, The Netherlands
b
Department of Mechanical Engineering, Kermanshah University of Technology (KUT), 67156-85420, Kermanshah, Iran
c
Department of Orthopedics, UMC Utrecht, Heidelberglaan 100, 3584CX Utrecht, The Netherlands

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

Article history: Metallic porous biomaterials are recently attracting more attention thanks to the additive manufacturing tech-
Received 22 November 2016 niques which help produce more complex structures as compared to conventional techniques. On the other
Received in revised form 24 December 2016 hand, bio-functional surfaces on metallic biomaterials such as titanium and its alloys are necessary to enhance
Accepted 10 March 2017
the biological interactions with the host tissue. This study discusses the effect of plasma electrolytic oxidation
Available online 12 March 2017
(PEO), as a surface modification technique to produce bio-functional layers, on the mechanical properties of ad-
Keywords:
ditively manufactured Ti6Al4V scaffolds based on the cubic unit cell. For this purpose, the PEO process with two
3D printed scaffolds different oxidation times was applied on scaffolds with four different values of relative density. The effects of the
Additive manufacturing PEO process were studied by scanning electron microscopy (SEM), energy dispersive X-ray spectroscopy (EDS),
Bio-functional surfaces optical microscopy as well as static and dynamic (fatigue) mechanical testing under compression. SEM results in-
Titanium alloys dicated pore formation on the surface of the scaffolds after oxidation with a thickness of 4.85 ± 0.36 μm of the
And mechanical properties oxide layer after 2 min and 9.04 ± 2.27 μm after 5 min oxidation (based on optical images). The static test results
showed the high effect of relative density of porous structure on its mechanical properties. However, oxidation
did not influence most of the mechanical properties such as maximum stress, yield stress, plateau stress, and en-
ergy absorption, although its effect on the elastic modulus was considerable. Under fatigue loading, none of the
scaffolds failed even after 106 loading cycles at 70% of their yield stress.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction
In bone tissue engineering, scaffolds play essential roles in mimick-
ing bone characteristics and supporting the formation of new tissue.
Not only should bone-mimicking scaffolds be biocompatible and capa-
ble of inducing bone ingrowth and osseointegration, they should also
have proper mechanical properties relative to the surrounding tissue
[1–3]. Hence, fabrication of scaffolds with biomechanical properties
similar to those of the surrounding bone is one of the main concerns
in bone tissue engineering [4].
Due to the higher strength of metallic scaffolds as compared to other
materials, they have become the key material in load-bearing applica-
tions such as orthopedic and dental implants [5]. One of the most
well-known metallic biomaterials applicable in dental and orthopedic
⁎ Corresponding author at: Department of Biomechanical Engineering, Faculty of
Mechanical, Maritime, and Materials Engineering, Delft University of Technology (TU
Delft), Mekelweg 2, 2628, CD, Delft, The Netherlands.
E-mail address: z.gorgin@kut.ac.ir (Z. Gorgin Karaji).
implants are titanium and its alloys due to their excellent properties
such as biocompatibility, high strength-to-weight ratio, and closer elas-
tic modulus to bone as compared to some other metallic biomaterials
[6]. However, even for implants made of titanium, aseptic loosening
may happens due to stress shielding which is the result of mismatch be-
tween the elastic moduli of bone and metallic implant as well as the
micro-motions of the implant in the implanted zone. Porous implants
have been introduced as one of the most suitable solutions to overcome
this deficiency. Porosity within porous implants decreases the elastic
modulus of the metallic structure to values around those of bone
which balances the load transfer through the implant leading to de-
creased effect of stress shielding. The interconnected hollow pores
with adequate pore size inside porous structures also provide space
for nutrient delivery, vascularization, and finally bone ingrowth and im-
plant fixation [7–9]. Additive manufacturing (3D printing) is one of ad-
vanced techniques for fabrication of porous metallic scaffolds. This
method allows for manufacturing of interconnected porous structures
with desired micro-architectures. By applying this method and control-
ling the porosity and designing pore morphology and size, predictable
mechanical properties are achievable [10,11].

http://dx.doi.org/10.1016/j.msec.2017.03.079
0928-4931/© 2017 Elsevier B.V. All rights reserved.
 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416
Another way to overcome aseptic loosening is improving bone in-
growth which is feasible through surface treatment. Improvement of
the biological properties of scaffolds is possible through surface modifi-
cation and inducing new surface properties [5]. Plasma electrolytic oxi-
dation (PEO) is an effective surface modification technique for synthesis
of bio-functional oxide layers with interconnected micro/nano porosity
on titanium materials. This surface layer is beneficial for increasing cell
affinity and growth, which in turn results in better bone-scaffold inte-
gration [8]. Overall, surface modification might influence the mechani-
cal properties of porous structures more than those of bulk materials
due to the much higher surface area of the porous structures [12]. Pre-
vious studies [13,14] on bulk titanium alloys have shown that the pres-
ence of the PEO layer may affect the fatigue behavior of these materials.
Currently, there are no reports addressing the effect of PEO layers on the
mechanical behavior of additively manufactured titanium scaffolds.
In the present study, we aim to investigate the effects of bio-
functionalization of additively manufactured Ti6Al4V scaffolds through
the PEO process on their static and dynamic mechanical properties. The
PEO process with the same current density but two different oxidation
times (i.e. 2 min and 5 min) in calcium acetate and calcium glycero-
phosphate electrolytes were applied to scaffolds with the cubic unit
cell and with four different porosities (relative densities) produced
using selective laser melting (SLM). Both the as-manufactured and oxi-
dized scaffolds were characterized by SEM, energy dispersive X-ray
spectroscopy (EDS), optical microscopy, and static and dynamic me-
chanical compression tests. Furthermore, we used the finite element
(FE) method to predict the mechanical properties of the porous metallic
biomaterials.
2. Materials and methods
2.1. Additive manufacturing of scaffolds
Porous scaffolds were additively manufactured using the selective
laser melting (SLM) method (Layerwise NV, Leuven, Belgium). File
preparation was performed using the Magics (Materialise, Belgium)
and DMP Control (3D Systems, Belgium) software. A customized 3D
Systems ProX DMP 320 machine was used to manufacture the speci-
mens. The low oxygen level in this machine (b50 ppm) is ideally suited
for production of Ti6Al4V-ELI scaffolds (according to ASTM B348, grade
23) on top of a rigid plate. All scaffolds were cylinders (approximately
7 mm length and 10 mm diameter) with a porous structure based on
the cubic unit cell, but with four different strut sizes and porosities
Table 1
Physical properties of the additively manufacture scaffold.
Scaffold Unit cell shape Strut size (μm) Pore size (μm) Relative density (%)
A Cubic 823 ± 230 1020 ± 311 37
B 693 ± 200 1155 ± 354 28
C 654 ± 190 1139 ± 359 24
D 451 ± 147 1413 ± 366 13
407
(designated as types A, B, C, and D in Table 1). The samples were
scanned using a micro-computed tomography machine (micro-CT)
(20 μm × 20 μm × 20 μm voxel size, 90 kV tube voltage, 180 μA tube cur-
rent, 3600 projections, Quantum FX, Perkinelmer, USA). The 3D recon-
structed images were then sliced into 2D stacks using Analyze 11.0
(in-built micro-CT software). The 2D stacks were locally segmented in
imageJ v.1.47. The pore size and distribution, strut size, porosity, total
surface area per unit volume, and volume of the as-manufactured scaf-
folds were determined using boneJ (plugin of imageJ) (Table 1).
2.2. Plasma electrolytic oxidation of scaffolds
In the first step, the additively manufactured specimens were
degreased in acetone, ethanol, and deionized water (5 min each one).
The PEO process was applied to the electrolyte prepared by dissolving
24 g/L calcium acetate (P99%, Sigma–Aldrich) and 4.2 g/L calcium glyc-
erophosphate (Dr. Paul Lohmann, Germany) in demineralized water. A
cylindrical stainless steel (Goodfellow, AISI 316 Hard) electrode was
used as cathode whereas the scaffold was the anode. A current density
of 20 A/dm2 was supplied through an AC power supply (50 Hz) type
ACS 1500 (ET Power Systems Ltd., UK) in galvanostatic mode. The tem-
perature during PEO was controlled by using a cooling bath and a dou-
ble-walled electrolytic cell to maintain the temperature at 9 ± 7 °C.
More experimental details on the applied PEO process can be found in
a previous study of our group [15]. Two different oxidation times (i.e.
2 min and 5 min) were considered for each group. The numbers 0, 2,
and 5 were used to specify samples with respectively 0 min, 2 min,
and 5 min of oxidation time (Table 2).
2.3. Surface morphology and chemical characterization
Scanning electron microscopy (SEM) (Dual Beam strata 235, EI,
USA) was employed to image the topography of the scaffolds, and
their elemental composition was determined by energy dispersive X-
ray spectroscopy (EDS) (EDAX, UK).
2.4. Thickness of the TiO2 layer
To measure the thickness of TiO2 coating formed on the scaffold's
struts, a specimen was embedded into a cold mounting epoxy resin
(ClaroCit, Struers, Denmark). The mounted specimen was successively
ground with SiC grit paper and polished with Chem cloth and colloidal
silica paste (0.04 μm, OP-S suspension) after which the cross-section
Porosity (%) Total surface area/volume (mm−1) Macro structure (scale bar: 250 μm)
63 46
72 51
76 58
87 77
408 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416

Table 2 2.5.2. Dynamic mechanical testing

Scaffolds labeling. The compression-compression fatigue behavior of the scaffolds was
Scaffold As-manufactured 2 min oxidation time 5 min oxidation time characterized using a hydraulic test frame (MTS, Minneapolis, US) with
A A0 A2 A5
a 10 kN load cell. A sinusoidal waveform with a frequency of 15 Hz and
B B0 B2 B5 load ratio of R = 0.1 was applied to each scaffold for a load level equaling
C C0 C2 C5 0.7 Fy where Fy is the yield force determined for each type of specimen.
D D0 D2 D5 For each sample type, two specimens were tested. 106 cycles was cho-
sen as the ultimate cycle number because 105–106 loading cycles are as-
sumed to be the appropriate fatigue limit for biomedical applications
of the struts was revealed. An optical microscope (Olympus BX60M, [18].
Japan) was used for imaging the cross-sections of the samples. The anal-
ySIS auto5 software (Olympus, Japan) was used to measure the layer 2.6. Statistical analysis
thickness on the captured optical images.
The data presented are expressed as mean ± standard deviation
2.5. Mechanical tests (SD) and were analyzed in SPSS using one-way ANOVA. The differences
between groups were assumed to be statistically significant different at
Cylindrical specimens with lengths of approximately 7 mm and di- p-value b 0.05.
ameters of approximately 10 mm were used to study the mechanical
behavior of as-manufactured specimens as well as the specimens sub- 2.7. Finite element (FE) modeling
jected to both types of PEO treatments (i.e. 2 and 5 min) (Table 2).
The geometry of specimens remained the same in all other experiments FE models with dimensions and geometries similar to those of the
performed here. test samples were created (Fig. 1). ANSYS - a commercial FE package -
was used for creating the geometry, discretizing the struts, and solving
the FE models. For discretizing the struts, the beam element type 189
2.5.1. Static mechanical testing
(Timoshenko beam theory) was implemented. The material properties
To study the static mechanical behavior of the scaffolds, quasi-static
used for the FE modeling were those of the bulk titanium alloy
compressive tests were carried out using a Zwick Z100 machine (Zwick
Ti6Al4V. The struts were considered rigid at the vertices.
GmbH & Co. KG, Germany). The testing condition was according to the
The bottom nodes of the structure were not allowed to move in the Y
ISO 13314:2011 standard [16]. The cross head velocity during the com-
direction to represent the boundary conditions imposed by the lower
pression tests was set to 0.8 mm/min and each test was continued up to
grip. The upper nodes were displaced downward in such a way that
80% compressive strain. The obtained stress-strain curves were ana-
1% strain was generated in the lattice structure. Both the top and bottom
lyzed for evaluating the mechanical properties of the specimens. The
nodes were allowed to move in the directions parallel to the grip plates
following parameters were obtained for each test: maximum compres-
(i.e. in the X and Z directions) in order to better capture the lateral be-
sive strength (σmax) which is the first observed peak in the stress-strain
havior of the structure. All the degrees of freedom of one of the bottom
curve, strain at σmax (εmax), plateau stress (σpl) that is the arithmetical
nodes of the structure were fixed in the space in order to avoid rigid
mean of the stresses at 20% and 40% strains, yield stress (σy) as the com-
body motion. The elastic modulus of the structure was calculated
pressive stress at 0.2% offset of quasi elastic line, elastic gradient (or
using the well-known formula [19] for simple compression δ = PL/AE,
Young's modulus) as the gradient of the elastic straight line between
where δ is the applied displacement on the structure, P is the resultant
points having the stresses of 20% and 70% of yield stress, quasi-elastic
force on the bottom nodes in the Y direction, L and A are the length
gradient as the slope of the linear deformation region before σmax, ener-
and cross-section area of the cylindrical sample, and E is the elastic mod-
gy absorption that is determined by calculating the area under the
ulus of the structure. The yield stress of the structure was calculated
stress-strain curve up to 50% strain, ductility that is determined by cal-
using a simple cross multiply: if the applied stress on the structure σ
culating the area under the stress-strain curve up to fracture point, den-
causes the maximum local stress to be σmax, the yield stress value σy
sification strain εd, and densification stress σd that is referred to the
that causes the maximum local stress in the structure σys would be
stress at the densification strain [16,17]. In this study, three different
σy =σysσ/σmax.
methods were used for defining and determining the densification
Each FE model was discretized in two ways: in the first model, all
strain εd: method I) the strain at the last local maximum in energy ab-
struts were discretized with elements with the same cross-section di-
sorption efficiency-strain curve, method II) the strain at the starting
mension as of the mean cross-section of the test samples. In the second
point of densification which was assumed to be equal to the strain of
model, the elements constructing each strut could have different cross-
the intersection of the plateau and post-densification elastic straight
section areas. The edge length of the cross-section of each element,
lines, and method III) the strain at the last local minimum before the
hereafter referred to as strut size, was obtained from a Gaussian distri-
stress rises steeply. The obtained static mechanical data are reported
bution function available in ANSYS with mean and standard deviation
as a function of relative density. To calculate the relative density, the
values set to those measured from the test specimens (Table 1).
bulk density of the titanium alloy was assumed to be 4420 kg/m3 [2].
In order to normalize the elastic modulus of the structures, the elastic
modulus of the bulk material was assumed to be 122.3 ± 2.5 GPa [2]. 3. Results
The energy absorption efficiency ƞ(ɛ) was obtained using the follow-
ing eq. [17]: 3.1. Surface bio-functionalization and characterization

The oxidation of scaffolds led to quite similar voltage-time transients

1 ɛ (Fig. 2). The data derived from the recorded voltage-time diagrams of
ƞðɛ Þ ¼ ∫ σ ðɛ Þdɛ ð1Þ
σ ðɛ Þ 0 the scaffolds during the PEO process, such as the growth rate of the
TiO2 layer in the anodizing step (dV/dt)1, sparking voltage (Vsp), voltage
increase rate following the spark initiation (dV/dt)2, and final voltage
The equation gives the area under the stress-strain curve up to a cer- (Vf) are summarized in Table 3. All the scaffolds had quite similar values
tain value of strain. of dV/dt and Vsp.
 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416 409

Fig. 1. The (a) side, (b) top, and (c) 3D views of the cylindrical porous structure FE model with regular cross-section. (d) The side view of the cylindrical porous structure FE model with
irregular cross-section.

The morphology of the titanium scaffolds before and after applica-
tion of the PEO process revealed by SEM is presented in Fig. 3. A clear
difference in the strut size and pore size of the scaffolds can be observed
from the SEM micrographs (Fig. 3, I). The melted sphere-like lumps
present on the struts of the porous structure have smooth surfaces
(Fig. 3, II). Different oxidation times did not change the macro-structure
(i.e. the geometry of the sphere-like lumps) (data not shown), however,
they created micro-pores on the surface of the specimens. The forma-
tion of typical PEO micro-pores on the scaffolds (which are very differ-
ent in size) is depicted in Fig. 3, III. Apparently, as the oxidation time
increased from 2 to 5 min, the number of pores decreased while their di-
ameter increased (Fig. 3, IV).
The elemental composition of the titanium scaffolds following the
EDS analysis is indicated in Fig. 4. The EDS spectra revealed the presence
of oxygen and the constitutive elements of Ti6Al4V alloy along with the
incorporation of P and Ca species from the electrolyte. Furthermore,
with the increase of the oxidation time from 2 to 5 min, the concentra-
tion of the alloying elements (Ti, Al and V) had decreased, whereas the
weight percentage of Ca and P had increased. Subsequently, the atomic
ratio of Ca to P increased from 1 to 1.5 (Table 4).
The oxide layer thickness of A2, B2, C2 and D2 scaffolds measured by
optical microscopy was 4.7 ± 0.7 μm, 4.7 ± 1.5 μm, 5.1 ± 1.1 μm, and
4.9 ± 1.3 μm, respectively. The difference between the oxide layer thick-
nesses was not significant (p-value N 0.05) and the average of the
410 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416

reported values was 4.9 ± 0.36 μm. By increasing the oxidation time to
5 min, the average layer thickness increased to 9 ± 2.2 μm.

3.2. Mechanical properties

3.2.1. Static mechanical properties

The static mechanical properties of as-manufactured and PEO-treat-
ed scaffolds are presented in Fig. 5a-g. According to these data, 2-min
oxidation did not cause any significant difference in all the mechanical
properties including maximum stress σmax,elastic modulus E, relative
elastic modulus E/Es, yield stress σy, plateau stress σpl, and energy ab-
sorption U (Fig. 5a–e,g). 5-min oxidation decreased elastic modulus, rel-
ative elastic modulus, yield stress, and plateau stress but again did not
have any effect on the maximum stress and energy absorption (Fig. 5,
Fig. 2. PEO voltage-time (V-t) transients of Ti6Al4V scaffolds (Vsp-sparking voltage, Vf-
p-value b 0.05). The plateau to yield stress ratio decreased by increasing
final voltage, (dV/dt) – voltage increase rate).
oxidation time from 0 to 2 min and then to 5 min (Fig. 5f). The scaffolds
with higher values of relative density showed higher energy absorption
capacities (Fig. 5g).
Table 3
Derived data of V-t curve (Fig. 2): (dV/dt)-voltage increase rate, Vsp-sparking voltage, Vf-
The stress-strain responses (Fig. 6, I) suggested similar deformation
final voltage. behavior (Fig. 6, II). For all types of specimens, the initial part of the
stress-strain curves suggested a linear trend. As the relative density de-
Scaffold (dV/dt)1 (V/s) Vsp (V) (dV/dt)2 (V/s) Vf (V)
creased, the maximum strength was reached in lower strains. Type D
A2 6.95 139 0.53 192 scaffolds (with the lowest relative density) therefore experienced their
B2 6.57 138 0.57 194
maximum strength at the lowest strain as compared to the other type
C2 6.85 137 0.61 198
D2 6.48 136 0.62 197 of scaffolds. Oxidized scaffold types A, B, and C showed higher ductility
A5 5.91 136 0.26 209 as compared to the as-manufactured specimens with the same relative
B5 5.58 134 0.29 215 density. All type D scaffolds showed similar ductility. The densification
C5 5.78 133 0.3 215
values for both strain and stress calculated using the three different
D5 6.14 135 0.31 221
methods of I (based on Fig. 6, II), II, and III were largely similar (Fig. 7).

Fig. 3. SEM micrographs of as-manufactured (I, II), 2 min (III) and 5 min (IV) PEO treated scaffolds.
 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416 411

elastic modulus values were higher than their corresponding experi-

mental values. The elastic moduli estimated with FE for regular and ir-
regular cross-sections were not close to the experimental data.
Weakening of the FE structure due to applying irregularity in the
cross-section size improved the agreement between the numerical
and experimental values of the elastic modulus.

4. Discussion

Although PEO improves the bio-functionality of titanium scaffolds

through local generation of specific topographical cues (micro/nano-po-
rosity and micro/nano-roughness) as well as Ca and P incorporation, it
may also be detrimental to their mechanical properties. Therefore, un-
derstanding the relationship between the mechanical properties and
oxide layer structure of additively manufactured scaffolds undergone
the PEO process represents an important research aspect towards
their bio-functionalization.

4.1. Oxide layer formation and characterization

Voltage-time curves showed that Vsp, Vf, and (dV/dt) parameters de-
pend mainly on the current density and titanium alloy composition.
Furthermore, the oxidation time did not affect the values of sparking
voltage among all the scaffolds. However, the oxidation time influences
the thickness of the oxide layer formed. It was observed that by increas-
ing the oxidation time from 2 min to 5 min, the average layer thickness
increased approximately by a factor of 2. These results are in line with
previous PEO studies on bulk titanium alloys [5,8,14]. At similar oxida-
tion time, the thickness and the pore morphology of the oxide layers
were similar for all the specimens. In addition, after 5 min of PEO treat-
ment, the Ca/P ratio increased to about 1.5 which is very close to that of
the hydroxyapatite phase (1.67).

4.2. Mechanical properties

Investigation of the mechanical properties of porous biomaterials is

important from several perspectives. First, one needs to ensure a satis-
factory match between the stiffness of a porous biomaterial and the
bone it replaces. The similarity of stiffness of bone substitutes and the
bone itself helps avoid stress shielding. Second, porous biomaterials
should be capable of providing sufficient mechanical support without
failure under physiological mechanical loading. The plateau stress as
well as the yield and first maximum stress values reported here could
Fig. 4. EDS analysis of as-manufactured, 2 min, and 5 min oxidation scaffolds. be of great importance in this regard. Since porous biomaterials are usu-
ally used as substitutes in areas that experience repetitive (cyclic) load-
3.2.2. Fatigue strength ing, certainty about the fatigue strength of these materials is critical as
In this study, the fatigue strength measurements were performed by well.
applying 0.7 σy load to each group of the specimens. Neither non-oxi- The scaffold types A, B, C, and D represented different static mechan-
dized scaffolds nor oxidized ones failed after 106 cycles of loading. ical properties. It must be noted that A0, B0, C0, and D0 are similar in
their micro-scale features (Fig. 3). All the scaffolds were also made of
the same bulk material with similar manufacturing parameters. There-
3.3. FE modeling fore, the difference in the mechanical properties can only be attributed
to their different pore sizes. This fact also holds for the oxidized samples
The elastic modulus versus relative density plots obtained numeri- which had similar chemical composition. All oxidation times resulted in
cally, analytically, and experimentally for different morphologies is pre- similar microstructures of A2, B2, C2 and D2 (Fig. 3, III) scaffolds or A5,
sented in Fig. 8. For analytical part, the formulas presented in [20] were B5, C5 and D5 specimens (Fig. 3, IV).
σ
used (EEs ¼ ðκaÞ2 and σ ysy ¼ ðκaÞ2). The numerically obtained yield stress and Consistent with a previous study [21], with cubic cell shape, the
values of plateau and yield stress were significantly different. In addition
to porosity, our results indicate that the plateau to yield stress ratio may
Table 4
also be affected by the thickness of the oxide layer (Fig. 5f).
analysis of scaffolds before and after PEO treatment.
Reaching significantly different elastic moduli in scaffolds with dif-
Element Ti Al V O P Ca Ca/P ferent oxidation times illustrates that the elastic modulus is more sensi-
(wt%) (At)
tive to the formation of the ceramic oxide layer than the other
As-manufactured 88.78 7.38 3.84 – – – – parameters.
2 min 37.99 2.69 1.78 44.88 5.53 7.14 1 With regards to stress-strain responses, the PEO treatment of titani-
5-min 31 2.45 1.4 44.57 6.99 13.59 1.5
um did not change the trend of stress-strain curves of the scaffolds. For
412 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416

Fig. 5. Static mechanical behavior parameters; a) maximum stress, b) elastic modulus, c) E/Es, d) yield stress, e) plateau stress, f) plateau stress/yield stress and g) energy absorption of the
scaffolds.

instance, all scaffold types showed linear trends in the initial parts of
their stress-strain curves. Lower relative density caused earlier failure,
which is expected. In the second part of the stress-strain curves, a dis-
tinct plateau with multiple fluctuations in stress appeared. This region
is related to the collapse of different layers. After the compression
tests started, the weaker struts broke and caused the upper layers col-
lapse on the lower ones. The subsequent local peaks in the plateau
part of the stress-strain curve represent the failure of the subsequent
layers of the porous structure. The value of local peaks corresponding
to the collapse of the subsequent layers of the porous structure is
lower than the peak value corresponding to the collapse of the first
layer due to the fact that after the collapse of the first layer, a fraction
of struts in the next layers are also damaged. These fluctuations of the
stress continue up to the collapse of all the layers, which marks the
end point of the plateau. Subsequently, the densification happened
and final linear part with high rate slope could be observed in the
stress-strain curves. This behavior occurs frequently in the stress-strain
curves of porous structures and is relevant from the energy absorption
 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416
Fig. 6. I) Stress vs. strain curves, II) efficiency vs. strain curve for all scaffolds.
viewpoint. Any stress increase implies that the structure still has the
ability of absorbing energy [22].
The higher ductility of oxidized scaffolds types A, B, and C compared
to the as-manufactured ones and also similar ductility in type D scaf-
folds with different oxidation times demonstrates that the heat treat-
ment through the PEO process can slightly affect the ductility of
additively manufactured scaffolds. For the type D scaffolds, which
have larger pores, heat can be more easily transferred. Therefore, this
type of scaffolds has been subjected to lower temperatures, meaning
that the PEO process has not substantially affected the ductility of type
D scaffolds. This behavior was also observed for the elastic modulus
for the same reason.
In the PEO process, the natural TiO2 layer grows with two different
mechanisms. Ti oxidation contributes to 70% of the overall TiO2 layer
growth, while elemental deposition during the PEO process contributes
to 30% of it [23]. The micro-discharge which causes high local tempera-
tures during the PEO process can anneal [24] and sinter separated par-
ticles which eventually reduce the potentially failing parts such as
notches and other irregularities. Hence, it might provide a balance be-
tween the reduction of the mechanical strength because of the
413
formation of pores and the microcracks on the surface on one hand
and the improvement of mechanical properties through sintering
on the other hand. Moreover, the high temperature associated with
the manufacturing process and subsequent rapid cooling [25] - re-
peated several times - result in accumulation of residual stresses in
the scaffold structure. The heat produced during the PEO process
may release some of the residual stresses generated during the
manufacturing process primarily due to possible stress relaxation ef-
fects [12,24]. Residual stresses play important roles in most of the
properties of the coated scaffolds, namely hardness, adhesion,
wear- and corrosion-resistance as well as the static and dynamic
mechanical characteristics [24].
Densification is an important parameter for characterization of the
mechanical behavior of scaffolds. Our results confirm the accuracy of
defining strain and stress densification based on energy absorption
curve which was reported before by Li et al. [17].
Fatigue test is an important mechanical test to simulate the condi-
tions experienced by load-bearing orthopedic implants. None of the
as-manufactured and oxidized scaffolds failed even after 106 cycles of
loading. This finding is in line with what was reported by Amin Yavari
414 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416

Fig. 7. (a) Densification strain, and (b) densification stress of the scaffolds.

et al. [18] for as-manufactured scaffolds with the same pore morpholo-
gy, porosity, and materials but with different strut size to strut length ra-
tios. The very long fatigue life of additively manufactured porous Fig. 8. Comparison of (a) relative elastic modulus, and (b) relative yield stress vs. relative
density, between experimental results and FE results obtained from structures with
biomaterials with the cubic unit cell may be due to the fact that gener-
regular and irregular cross-sections.
ally only compressive stresses develop in such type of porous structures.
Since compressive stresses could hamper crack growth, possibly
have less pores present” [27]. The local pores inside the struts of the ad-
through crack closure, the fatigue life of structures subjected to purely
ditively manufactured porous structures are expected to severely affect
compressive stresses tend to be very long, if not infinite. Application
the experimental yield stress as well. However, the results of the current
of the PEO process does not fundamentally change this phenomenon
study as well as the results of [7,9,26] demonstrated that the discrepan-
which is why the fatigue life of the PEO-treated specimens remained
cy present between the yield stresses obtained from experimental tests
very long.
and analytical predictions is much smaller as compared to the discrep-
ancy in the experimental and analytical elastic moduli (Fig. 8). Visual
4.3. FE modeling
observations of the deformation of the porous structure based on the
cubic unit cell showed that before any local yielding, the buckling of
Comparison of the mechanical properties of the experimental results
the struts (and therefore the buckling of the whole structure) is the
on one hand and the numerical and analytical results on the other hand
main cause of decrease in the slope of the stress-strain curve after the
demonstrated that the elastic moduli obtained from the experimental
initial linear part (see deformation of the porous structure in Fig. 9).
tests are severely smaller than those obtained using the analytical/nu-
Therefore, the criterion used in this study to obtain the yield stress of
merical approaches. Such big discrepancies between the experimental
the porous structure (0.2% right offset from the linear part of stress-
and analytical elastic moduli of porous biomaterials based on cubic
strain curve) may not be representative of the actual yielding of the
unit cell have already been observed in other studies, see e.g [7,9,26].
struts of the porous structure, but may be representative of the start of
In all the noted studies, there has been a large difference between the
the buckling regime in the structure.
elastic moduli obtained from numerical and analytical studies on one
In the next step, it was decided to see whether or not there is a big
hand and the elastic moduli obtained from experimental tests on the
discrepancy between analytical and experimental buckling stresses
other hand. This large difference between the experimental and numer-
(similar to the descripancy observed in elastic modukus). The analytical
ical/analytical results can be attributed to the large percentage of hori-
buckling stress in the porous structure based on cubic unit cells is
zontal struts in the cubic structure. In the cubic structure, 2/3 of the
σbuckling = π2EsI/Kl4 [28]. The experimental observations showed that
struts are horizontal. In porous biomaterials based on other unit cell
the struts buckle with a condition similar to columns with one end
types, the percentage of horizontal struts is much lower. A study [27]
fixed and the other end pinned (Fig. 9). For such a structure, K = 2.
showed that the amount of enclosed pores inside additively
Substituting K = 2, I = a4/12, and mechanical properties of Ti6Al4V
manufactured material decreases with “the inclination of the strut,
σ buckling 2Es a4
meaning horizontal struts have lots of porosities while diagonal struts into the relationship π2EsI/Kl4 gives σy ¼ 4 . The analytical
12 l σ y
 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416
Fig. 9. The deformation of the porous structure based on cubic unit cells under compressive loading after a) 0 s, b) 27 s, c) 67 s and d) 88 s compression.
buckling stress was one order of magnitude larger than the stress at the
first point of non-linearity in the stress-strain curve of the porous struc-
ture. After considering the effects of internal pores on the value of the
analytical buckling stress and therefore decreasing its value to 1/12 of
its actual value (similar to the decrease in the elastic modulus of porous
structure due to presence of micro-pores, Fig. 8), the analytical and ex-
perimental buckling stresses approached each other.
The big difference between the failure stress predicted by the analyt-
ical and numerical models and the experimental yield stress in the po-
rous structure based on cubic unit cell has not been observed in the
porous structures based on other types of unit cells [28]. One of the
main reasons behind this difference may be that in the porous struc-
tures based on the cubic unit cell, the loaded struts are all parallel to
the loading direction. The irregularities, e.g. caused by the manufactur-
ing process, could therefore easily cause buckling and bending. The an-
alytical and numerical models in the cubic porous structure only
consider compression in the struts for predicting the yield stress. This
is in contrast with what happens in reality when other deformation
modes such as bending and buckling occur a very short time after the
specimen compression has started. In the analytical and numerical
models considered for porous structures based on other unit cell
types, bending deformation and its effect on the deformation of the
unit cell is already included. Therefore, it seems that the analytical and
numerical models created for the porous structures based on cubic
unit cell need to be modified to consider the complexities caused by
the irregularities created during the additive manufacturing process.
5. Conclusions
In this study, additively manufactured Ti6Al4V scaffolds based on
the cubic unit cell with four different relative densities in the range of
13–37% were oxidized through the PEO process with a current density
of 20 A/dm2 and for two different oxidation times (i.e. 2 min and
5 min). SEM data revealed the formation of specific PEO micro-pores
on the entire surface of the oxidized scaffolds. By extending the oxida-
tion time from 2 to 5 min, the number of pores decreased while the
415
pores became more homogenously distributed and their size increased.
EDS analysis indicated the incorporation of Ca and P species from the
electrolyte in the oxide structure of titanium scaffolds. The Ca/P atomic
ratio increased from 1 to 1.5 after 5 min of oxidation time. The thickness
of the oxide layer also increased from 4.9 ± 0.4 μm to 9 ± 2.3 μm as a
consequence of extending the oxidation time. The oxidation process
with 2 min duration did not change the mechanical properties (elastic
modulus, yield stress, and plateau stress) of the scaffolds considerably.
The mechanical properties, however, decreased by 0 to 30% after
5 min oxidation. The decrease in the mechanical properties became
more significant in the scaffolds with higher relative densities. The fa-
tigue tests revealed that none of the scaffolds for any of the oxidation
conditions failed even after 106 cycles of loading at 70% of their yield
strength. The analytical and numerical studies carried out in this study
suggested that (a) the elastic modulus of additively manufactured po-
rous structures based on cubic unit cell is much smaller than the values
predicted by the analytical and numerical studies due to the presence of
several micro-pores, and (b) the mechanism governing the failure of
porous structures based on cubic unit cell is not yielding but it is buck-
ling. By considering the effects of micro-pores inside the micro-struc-
ture of the porous specimens and the buckling of the struts (instead of
yielding), the analytical, numerical, and experimental values of the elas-
tic modulus and failure stress become very close to each other.
Acknowledgement
The authors would like to acknowledge the kind support of 3D Sys-
tems Layerwise for production of the specimens used in the current
study. Karel Lietaert from the same company kindly read the paper
and provided feedback.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.msec.2017.03.079.
416 Z. Gorgin Karaji et al. / Materials Science and Engineering C 76 (2017) 406–416
References
[1] V. Karageorgiou, D. Kaplan, Porosity of 3D biomaterial scaffolds and osteogenesis,
Biomaterials 26 (2005) 5474–5491.
[2] S.M. Ahmadi, G. Campoli, S. Amin Yavari, B. Sajadi, R. Wauthle, J. Schrooten, H.
Weinans, A.A. Zadpoor, Mechanical behavior of regular open-cell porous biomate-
rials made of diamond lattice unit cells, J. Mech. Behav. Biomed. Mater. 34 (2014)
106–115.
[3] D.W. Hutmacher, Scaffolds in tissue engineering bone and cartilage, Biomaterials 21
(2000) 2529–2543.
[4] L. Moroni, J.R. de Wijn, C.A. van Blitterswijk, 3D fiber-deposited scaffolds for tissue
engineering: influence of pores geometry and architecture on dynamic mechanical
properties, Biomaterials 27 (2006) 974–985.
[5] I. Apachitei, A. Leoni, A.C. Riemslag, L.E. Fratila-Apachitei, J. Duszczyk, Enhanced fa-
tigue performance of porous coated Ti6Al4V biomedical alloy, Appl. Surf. Sci. 257
(2011) 6941–6944.
[6] R. Zhou, D. Wei, S. Cheng, Y. Zhou, D. Jia, Y. Wang, B. Li, The effect of titanium bead
diameter of porous titanium on the formation of micro-arc oxidized TiO2-based
coatings containing Si and Ca, Ceram. Int. 39 (2013) 5725–5732.
[7] R. Hedayati, M. Sadighi, M. Mohammadi-Aghdam, A.A. Zadpoor, Effect of mass mul-
tiple counting on the elastic properties of open-cell regular porous biomaterials,
Mater. Des. 89 (2016) 9–20.
[8] A. Leoni, I. Apachitei, A.C. Riemslag, L.E. Fratila-Apachitei, J. Duszczyk, In vitro fatigue
behavior of surface oxidized Ti35Zr10Nb biomedical alloy, Mater. Sci. Eng. C 31
(2011) 1779–1783.
[9] G. Campoli, M.S. Borleffs, S. Amin Yavari, R. Wauthle, H. Weinans, A.A. Zadpoor, Me-
chanical properties of open-cell metallic biomaterials manufactured using additive
manufacturing, Mater. Des. 49 (2013) 957–965.
[10] S. Amirkhani, R. Bagheri, A. Zehtab Yazdi, Effect of pore geometry and loading direc-
tion on deformation mechanism of rapid prototyped scaffolds, Acta Mater. 60
(2012) 2778–2789.
[11] J. Kadkhodapour, H. Montazerian, A.C. Darabi, A.P. Anaraki, S.M. Ahmadi, A.A.
Zadpoor, S. Schmauder, Failure mechanisms of additively manufactured porous bio-
materials: effects of porosity and type of unit cell, J. Mech. Behav. Biomed. Mater. 50
(2015) 180–191.
[12] S. Amin Yavari, S.M. Ahmadi, J. van der Stok, R. Wauthle, A.C. Riemslag, M. Janssen, J.
Schrooten, H. Weinans, A.A. Zadpoor, Effects of bio-functionalizing surface treat-
ments on the mechanical behavior of open porous titanium biomaterials, J. Mech.
Behav. Biomed. Mater. 36 (2014) 109–119.
[13] B. Lonyuk, I. Apachitei, J. Duszczyk, The effect of oxide coatings on fatigue properties
of 7475-T6 aluminium alloy, Surf. Coat. Technol. 201 (2007) 8688–8694.
[14] I. Apachitei, B. Lonyuk, L.E. Fratila-Apachitei, J. Zhou, J. Duszczyk, Fatigue response of
porous coated titanium biomedical alloys, Scr. Mater. 61 (2009) 113–116.
[15] B.S. Necula, L.E. Fratila-Apachitei, S.A.J. Zaat, I. Apachitei, J. Duszczyk, In vitro antibac-
terial activity of porous TiO2–Ag composite layers against methicillin-resistant
Staphylococcus aureus, Acta Biomater. 5 (2009) 3573–3580.
[16] I. Standard, ISO 13314, (E)(2011) Mechanical testing of metals—ductility
testing—compression test for porous and cellular metalsRef Number ISO. 13314
2011 1–7.
[17] Q.M. Li, Compressive strain at the onset of densification of cellular solids, J. Cell.
Plast. 42 (2006) 371–392.
[18] S. Amin Yavari, S.M. Ahmadi, R. Wauthle, B. Pouran, J. Schrooten, H. Weinans, A.A.
Zadpoor, Relationship between unit cell type and porosity and the fatigue behavior
of selective laser melted meta-biomaterials, J. Mech. Behav. Biomed. Mater. 43
(2015) 91–100.
[19] A.C. Ugural, Stresses in Beams, Plates, and Shells, CRC Press, Third, 2009.
[20] A.A. Zadpoor, R. Hedayati, Analytical relationships for prediction of the mechanical
properties of additively manufactured porous biomaterials, J. Biomed. Mater. Res.
A 104A (2016) 3164–3174.
[21] S. Ahmadi, S. Yavari, R. Wauthle, B. Pouran, J. Schrooten, H. Weinans, A. Zadpoor, Ad-
ditively manufactured open-cell porous biomaterials made from six different space-
filling unit cells: the mechanical and morphological properties, Materials. 8 (2015)
1871–1896.
[22] B. Gorny, T. Niendorf, J. Lackmann, M. Thoene, T. Troester, H.J. Maier, In situ charac-
terization of the deformation and failure behavior of non-stochastic porous struc-
tures processed by selective laser melting, Mater. Sci. Eng. A 528 (2011) 7962–7967.
[23] Y.M. Wang, Plasma electrolytic oxidation treatment of aluminium and titanium al-
loys, Surf. Eng. Light Alloys Alum. Magnes. Titan. Alloys 110 (2010).
[24] R.H.U. Khan, A. Yerokhin, X. Li, H. Dong, A. Matthews, Surface characterisation of DC
plasma electrolytic oxidation treated 6082 aluminium alloy: effect of current densi-
ty and electrolyte concentration, Surf. Coat. Technol. 205 (2010) 1679–1688.
[25] F. Brenne, T. Niendorf, H.J. Maier, Additively manufactured cellular structures: im-
pact of microstructure and local strains on the monotonic and cyclic behavior
under uniaxial and bending load, J. Mater. Process. Technol. 213 (2013) 1558–1564.
[26] J. Parthasarathy, B. Starly, S. Raman, A. Christensen, Mechanical evaluation of porous
titanium (Ti6Al4V) structures with electron beam melting (EBM), J. Mech. Behav.
Biomed. Mater. 3 (2010) 249–259.
[27] R. Wauthle, B. Vrancken, B. Beynaerts, K. Jorissen, J. Schrooten, J.-P. Kruth, J. Van
Humbeeck, Effects of build orientation and heat treatment on the microstructure
and mechanical properties of selective laser melted Ti6Al4V lattice structures,
Addit. Manuf. 5 (2015) 77–84.
[28] R. Hedayati, M. Sadighi, M. Mohammadi-Aghdam, A.A. Zadpoor, Mechanical proper-
ties of regular porous biomaterials made from truncated cube repeating unit cells:
analytical solutions and computational models, Mater. Sci. Eng. C 60 (2016)
163–183.