Sei sulla pagina 1di 16

J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

available at www.sciencedirect.com

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

Research paper

Microstructure and mechanical properties of open-cellular


biomaterials prototypes for total knee replacement implants
fabricated by electron beam melting

L.E. Murr a,b,∗ , K.N. Amato a,b , S.J. Li c , Y.X. Tian c , X.Y. Cheng c , S.M. Gaytan a,b ,
E. Martinez a,b , P.W. Shindo a,b , F. Medina b , R.B. Wicker b,d
a Department of Metallurgical and Materials Engineering, The University of Texas at El Paso, El Paso, TX 79968, USA
b W. M. Keck Center for 3D Innovation, The University of Texas at El Paso, El Paso, TX 79968, USA
c Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences, Shenyang, 110016, China
d Department of Mechanical Engineering, The University of Texas at El Paso, El Paso, TX 79968, USA

A R T I C L E I N F O A B S T R A C T

Article history: Total knee replacement implants consisting of a Co–29Cr–6Mo alloy femoral component
Received 11 February 2011 and a Ti–6Al–4V tibial component are the basis for the additive manufacturing of novel
Received in revised form solid, mesh, and foam monoliths using electron beam melting (EBM). Ti–6Al–4V solid
14 April 2011 prototype microstructures were primarily α-phase acicular platelets while the mesh
Accepted 6 May 2011 and foam structures were characterized by α′ -martensite with some residual α. The
Published online 12 May 2011 Co–29Cr–6Mo containing 0.22% C formed columnar (directional) Cr23 C6 carbides spaced
∼2 µm in the build direction, while HIP-annealed Co–Cr alloy exhibited an intrinsic stacking
fault microstructure. A log–log plot of relative stiffness versus relative density for Ti–6Al–4V
and Co–29Cr–6Mo open-cellular mesh and foams resulted in a fitted line with a nearly ideal
slope, n = 2.1. A stress shielding design graph constructed from these data permitted mesh
and foam implant prototypes to be fabricated for compatible bone stiffness.
c 2011 Elsevier Ltd. All rights reserved.

1. Introduction including austenitic stainless steels (especially 316L) and Co-


base (Co–Cr) alloys (Park, 2000). Between about 1950 through
Bone implant materials, especially metals and alloys, to 1990, α and (α + β) Ti alloys with solid elastic (Young’s)
have evolved over several millennia beginning with gold moduli of ∼110 GPa were developed, while more recently
cranioplasty, Ni plated steel and related alloy evolution for β-Ti alloys with lower elastic moduli (ranging from ∼50
fracture plates and a host of Ti alloy implants for a variety to 90 GPa) have been developed (Niinomi, 2007). As bone
of skeletal repair joint replacements and dental restorations implant materials, Ti and its alloys exhibit very suitable
(Sanan and Haines, 1997). More recently, other metallic characteristics for biomedical applications because they
alloys have been employed especially as bone implant alloys, exhibit high biocompatibility, strength, and often superior

∗ Corresponding author at: Department of Metallurgical and Materials Engineering, The University of Texas at El Paso, El Paso, TX 79968,
USA. Tel.: +1 915 747 6929; fax: +1 915 747 8036.
E-mail address: lemurr@utep.edu (L.E. Murr).
c 2011 Elsevier Ltd. All rights reserved.
1751-6161/$ - see front matter ⃝
doi:10.1016/j.jmbbm.2011.05.010
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1397

corrosion resistance (Niinomi, 2001, 2008). However, the vast structures from any available pre-alloyed powder by additive
majority of these implant applications utilize monolithic (layer-based) manufacturing directly from computer (CAD)-
(solid) Ti alloy appliances which have exhibited limited generated models using directed laser or electron beams have
lifetimes as a consequence of incomplete or lacking fusion been demonstrated. Harryson and Cormier (2005) described
or bone cell ingrowth, and elastic modulus mismatch. Other the fabrication of custom (or patient-specific) orthopedic
issues, relating to implant efficiency include the absence implants using electron beam melting (EBM). Hollander et al.
of vascularization, and in situations such as knee implant (2006), Multen et al. (2009) and Stamp et al. (2009) utilized
appliances in particular, the production of wear debris at selective laser melting (SLM) to produce regular or periodic
bearing surfaces. (reticulated) mesh structures and stochastic foam structures
Elastic modulus mismatch, or stress shielding, is a major while Heinl et al. (2007, 2008) produced Ti-alloy mesh arrays
feature of metallic implant—bone incompatibility. Since bone corresponding to porosities or densities having stiffnesses
tissue remodels itself in response to applied stress, stress (elastic constants) approaching bone stiffnesses (in excess of
reduction leads to bone loss or resorption. Consequently, 20 GPa). Cellular Ti–6Al–4V fabricated by EBM for orthopaedic
high modulus (or stiff) implant material dominates the load implants has also been described by Marin et al. (2010).
bearing regime in contrast to low modulus bone. Cortical or More recently, Murr et al. (2010) illustrated a wide range of
hard bone can have elastic moduli ranging from 3 to 30 GPa, Ti–6Al–4V reticulated mesh and stochastic foam prototypes
while the softer (intramedulary) trabecular or cancellous fabricated by EBM, and demonstrated that these open-cellular
bone has significantly lower elastic moduli: 0.02–2 GPa (Yang structures behaved similarly to a wide range of metallic
et al., 2001). This is in contrast to ∼110 GPa for variously foams, especially aluminum and aluminum alloys which
processed Ti–6Al–4V which has dominated the manufacture follow the Gibson and Ashby (1997) relationship:
of commercial implants over the past decade (Long and
E = α1 Es (ρ/ρs )n (1)
Rack, 1998; Chuna et al., 2003; Niinomi, 2007). Many knee
replacements also utilize Co–Cr–Mo alloys especially for where E is the elastic (Young’s) modulus (or stiffness at
superior wear resistance, but these alloys have even higher a corresponding density, ρ), and Es and ρs are the solid
moduli: ∼210 GPa. (full density) stiffness and full density, respectively. While
An effective strategy for reducing or eliminating stress generally n is approximated by 2, the experimental value
shielding and simultaneously enhancing osteoconductivity has been shown to vary from ∼1.8 to 2.2. Murr et al. (2010)
and bone cell ingrowth, has been the development of porous demonstrated that for Ti–6Al–4V mesh and foam prototypes,
metallic structures such as foams, which can be tailored to n = 2.3. In addition, it was demonstrated that complex, 3D
more closely match the modulus or stiffness of bone, thereby implant monoliths incorporating functionally-graded mesh
minimizing or eliminating stress shielding. Open cellular or foam or mesh/foam density variations and solid sections
foams and other porous structures are required for bone could be fabricated by EBM.
cell ingrowth to be effective in optimizing biocompatibility. In the present study, we focus on the fabrication of open
While there has been considerable effort in developing Ti cellular mesh and foam components of a Co–Cr–Mo alloy
and Ti-alloy foams (Dunand, 2004; Ryan et al., 2006; Singh (ASTM F-75) and Ti–6Al–4V using EBM for a range of densities
et al., 2010) the development of practical, biomedical implants and measured stiffnesses. In addition, these structures are in-
incorporating porosity in surface coatings has been minimally corporated into Co–Cr femoral and Ti–6Al–4V tibial prototypes
successful, and many commercial implants utilize some representing total knee replacement implants. The varying
form of surface porosity created by sintering sponge or density open cellular structures and their microstructures are
bead configurations or attaching other porous coatings to characterized by optical and scanning electron microscopy
the monolithic implant surface (Hacking et al., 2002; Bobyn and X-ray diffraction, while the corresponding microinden-
et al., 2005). The mechanical behavior of porous structures dation hardness for mesh strut structures and foam ligament
depends on pore volume fraction and size distribution. structures are measured along with dynamic Young’s modu-
In turn, these determine the strut size or ligament wall lus (dynamic stiffness number).
thickness and dimension partitioning the pores, and bearing
the load. In addition, pore size is also important for effective
bone ingrowth. The general consensus has been that a 2. Experimental procedures
minimum pore architecture or interconnectivity is in the
range of 100 µm (Holy et al., 2000; Freyman et al., 2001). 2.1. Fabrication of biomedical implant prototypes: The
Correspondingly, Yang et al. (2001) and Ryan et al. (2006) have EBM system
suggested larger pore sizes in the range of 500 µm–1 mm
promote the formation of fibrous tissues to optimize the Fig. 1 shows a schematic view of the Arcam A2 electron beam
biocompatibility. However, in contrast to solid, monolithic melting (EBM) system along with examples of the Ti–6Al–4V
implants fixed with acrylic cement, porous or porous-coated and Co-base (29% Cr, 6% Mo, 0.22% C, 0.25% Ni, 0.7% Si, 0.5%
implants must be initially stabilized with a precise press-fit to Mn, balance Co; in weight%) pre-alloyed powders used to
assure optimal bone tissue ingrowth. fabricate solid (fully dense) and more complex open cellular
While it is very difficult to custom fabricate monolithic structures as well as multifunctional monolithic biomedical
and complex Ti-alloy implants having designed porosity components. As illustrated in Fig. 1, an electron beam is
or stiffness features, new approaches utilizing solid free- generated in the electron gun at (1), and focused (2) and
form fabrication or rapid prototyping to build complex electromagnetically scanned (3) onto powder layers gravity
1398 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

fed from cassettes (4) and mechanically raked (5) into layers
roughly 75–100 µm thick. The powder layer thickness is
smaller for the smaller average powder size, which in the case
of the Ti–6Al–4V powder is ∼30 µm in contrast to the Co-base
alloy powder with an average diameter of ∼40 µm (Fig. 1). The
beam is selectively scanned by computer aided design (CAD)
software to melt specific layer portions which are added to
the next layer, ultimately forming a 3D (multilayer) structure
as the building component is lowered with each added layer
(6 in Fig. 1). Prior to the melt scan, the beam is rastered over
the newly raked layer in multiple (∼11) passes to preheat
the layer. The preheat scan rate is ∼104 mm/s with a beam
current of ∼30 mA, in contrast to the melt scan rate of
∼400 mm/s at a reduced beam current of 6–10 mA. Preheat
temperatures averaged ∼ 640 ◦ C–840 ◦ C for the Ti-alloy and
Co-base alloy, respectively. This is in contrast to melting
temperatures of 1625 ◦ C and 1430 ◦ C for the Ti-alloy and
the Co-base alloy, respectively. The building of 3D structures
by EBM from limited melt pools in the additive layers is a
form of directional solidification, with the solidification rate
determined in part by the beam scan and focus parameters
as well as the building structure geometries and dimensions,
and associated heat conduction and cooling rate. Thijs
et al. (2010) have recently discussed microstructural-build
related features, even epitaxial or columnar-grain orientation
variations in selective laser melting (SLM) fabrication of
Ti–6Al–4V in the context of scanning strategies, which include
beam scan direction or direction sequence, beam current,
scan speed, beam focus, etc. as discussed above.

2.2. Mesh and foam building software and modeling Fig. 1 – EBM system schematic (Arcam A2). Components
strategies are discussed in the text. Upper left SEM view shows
Ti–6Al–4V powder while lower left SEM view shows
Reticulated mesh structures and stochastic foams for Co–29Cr–6Mo powder.
Ti–6Al–4V and Co–29Cr–6Mo alloys were fabricated from
CAD software models developed from Materialise/Magics
(http://www.materialise.com) software and CT-scan models, foam samples (utilizing the CAD models demonstrated in
respectively as discussed in detail elsewhere (Murr et al., Fig. 2), using a resonant frequency analyser (IMCE-HTVP-
2010). The mesh array structure generator utilized a single 1750-C) for rectangular test specimens. The elastic modulus,
unit cell or lattice-structure unit referred to as a dode thin E, referenced to as the dynamic stiffness or stiffness-
element which varies in geometrical structure from 2-fold related number is proportional to the resonant frequency
to 3-fold symmetry axes with a rotation of 45◦ around the squared. These rectangular samples were supported by
longitudinal, reference axis (Fig. 2(a)). This build element small beams impacted lightly by a small hammer creating
feature and corresponding 3D model are illustrated in Fig. 2(a) a measurable resonant frequency. The test samples were
and (b), respectively. By varying the element dimensions nominally 2.3 cm × 2.3 cm × 3.6 cm in dimension or carefully
(Fig. 2(a)) the model size can be changed, including the machined to this dimension, which complied with the
strut diameters, and this produces density or corresponding general requirements discussed for metal foams by Ashby
porosity variations in the EBM fabricated components.
et al. (2000).
Utilizing CT scans for standard aluminum stochastic
open cellular foams, bitmap files representing image or
model slices with specific pixel dimensions were created 2.4. Mechanical testing for solid, full-density components
as illustrated in Fig. 2(c) and (d). These file representations and mesh and foam samples
(Fig. 2(c) and (d)) can be altered in linear dimension to create
variations in cell sizes and cell ligament dimensions resulting Mechanical behavior for solid and open-cellular structures for
in density and porosity variations (Murr et al., 2010). the Ti–6Al–4V and Co–29Cr–6Mo alloys fabricated by EBM was
investigated using Vickers microindentation hardness (HV)
2.3. Dynamic stiffness measurements for mesh and foam (Shimadzu HMV-2000 tester using a 100 gf load) and Rockwell
samples C-scale hardness (HRC) (using a 150 kgf load). HV and HRC
measurements were made for the solid EBM components
The elastic moduli (E) or Young’s moduli were measured for while HV measurements only using the 25 gf load were made
EBM fabricated Ti–6Al–4V and Co–29Cr–6Mo alloy mesh and for mounted and polished mesh struts and foam ligaments.
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1399

Fig. 2 – Reticulated mesh (a) and stochastic foam (b) models for EBM building of open cellular structures. In (a), the
dode-thin (Materialise/Magics) build element is shown on the left in cube and diamond orientations (0 and 45◦ ) which
correspond to the viewing arrows in the 3D model on the right. In (b), the foam pore and ligament structure is shown on the
left with a 3D model on the right.

A total of 25 Vickers indentations were made on each of two 2.5. Characterization of materials and microstructures
mounted sections from each sample were made to produce
100 HV values. Twenty-five HRC indentation values were Microstructures for the solid and open-cellular EBM fabri-
made for the solid sample. cated components were initially observed by optical metallog-
Tensile specimens were machined from solid Ti–6Al–4V raphy (OM), followed by scanning electron microscopy (SEM)
and transmission electron microscopy (TEM). The OM was
and Co–29Cr–6Mo specimens and tested in an upgraded
performed on sectioned, mounted, polished, and etched sam-
Tinius-Olson Universal Testing Machine (SIN 175118). The
ples using a digital imaging Reichert MEF4 A/M metallograph.
Co–29Cr–6Mo alloy samples (solid cylinders) were also HIP-
The etching protocols for the Ti–6Al–4V samples have been
annealed at 0.1 GPa for 4 h in Ar at 1200 ◦ C, and tensile
described in detail by Murr et al. (2009) while those for the
tested. All tensile tests were conducted at room temperature
Co–29Cr–6Mo alloy samples have been described by Gaytan
(∼20◦ C) utilizing a strain rate of ∼10−3 s−1 . Yield stress
et al. (2010).
measurements utilized the standard 0.2% engineering strain SEM analysis, including fractography for failed tensile
offset. Fracture surfaces for failed specimens were also specimens, employed a Hitachi S-4800 field emission (FE) SEM
examined by scanning electron microscopy (SEM). fitted with an EDAX energy-dispersive (X-ray) spectrometer
1400 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

Fig. 3 – Examples of solid and open-cellular, monolithic EBM built components of Ti–6Al–4V. (a) shows vertical built cylinder
(z) in build direction along z-axis and horizontal built cylinder (x, y) along x or y axis perpendicular to the build direction
(heavy arrow). (b) Foam and dode-thin reticulated mesh examples compared to solid cylinder (z). (c) Non-parallel view in (b)
showing 2 different density foam cubes at left, a solid cylindrical foam, an open core mesh cylinder, a similar mesh cylinder
with thin tube, and solid cylinder at right. Rockwell (HRC) values for solid vertical and horizontal cylinders are shown in (a).

(EDS) system; operated primarily at 20 kV in the secondary system operated at 300 kV, employing a goniometer-tilt stage,
electron emission mode. an EDS attachment, and a digital imaging system.
Solid, full-density samples were prepared for TEM analysis
by sectioning EBM-fabricated geometries (mostly cylinders)
3. Results and discussion
to prepare coupons which could be ground to a thickness
of ∼200 µm punched into 3 mm standard TEM samples 3.1. Comparison of solid Ti–6Al–4V and Co–29Cr–6Mo
which were mechanically dimpled and electropolished using alloy components
protocols described in Murr et al. (2009) for the Ti–6Al–4V
samples, and in Gaytan et al. (2010) for the Co–29Cr–6Mo Fig. 3 illustrates some examples of EBM fabricated solid
samples. The TEM was a Hitachi H-9500 high-resolution (full density) components (Fig. 3(a)) along with open-cellular
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1401

mesh and foam product prototypes, and complex functional


(solid/cellular) monolithic prototypes (Fig. 3(b) and (c)). From
the left in Fig. 3(b) and (c): 2 foam example rectangular
forms, foam open-core cylinder, 2 dode thin mesh cylinders
(one with open core and one with a thin solid cylinder
and open core), and a solid cylinder (right). Fig. 3(a) shows
simple solid cylinders built in vertical (z) and horizontal (x, y)
axial reference systems for Ti–6Al–4V components. Similar
reference builds were also fabricated for the Co–29Cr–6Mo
alloy as well, along with rectangular plates built in the
vertical direction (z-axis). These prototypes provided a solid
component reference for the microstructures and mechanical
behavior of the open-cellular systems. Fig. 4 compares
the as-fabricated (EBM) Ti–6Al–4V microstructures for the
vertical (z) and horizontal (x) cylinders shown in Fig. 3(a).
The classical α-acicular (hcp phase) Widmanstätten platelet
structure surrounded by interfacial β (bcc) phase is shown in
Fig. 4(a), while in Fig. 4(b) the microstructure is dominated by
thinner and smaller plates of α phase and small β grains as a
consequence of more rapid cooling in the building cylinder in
the horizontal (x) orientation. These comparisons are made in
the horizontal reference plane perpendicular to the vertical
cylinder axis and the EBM build direction (Fig. 3(a)). The
corresponding hardness (HV and HRC) values illustrate the
relationship of these microstructures to mechanical behavior
as it is represented by hardness. For example, the simple Fig. 4 – Ti–6Al–4V microstructures observed by optical
rule of thumb indicates the yield stress (YS—0.2% offset) ∼ = metallography. (a) Vertical cylinder (z) build as in Fig. 3(a).
HV/3. The HV values for Fig. 3(a) were 3.7 GPa and 4.5 GPa, (b) Horizontal cylinder (x) build as in Fig. 3(a).
respectively, for the corresponding vertical (z) and horizontal
(x, y) cylinder builds. These values reflect the microstructural
refinement illustrated on comparing Fig. 4(a) and (b). These
HV values also correlate with the HRC values shown in
Fig. 3(a).
The microstructure variation between Fig. 4(a) and (b)
(α→ fine α/β) is a consequence of the different thermal
environments and the effect on the cooling (solidification)
rate. More rapid cooling occurs for the horizontal (x, y) builds
in contrast to the vertical (z) builds shown in Fig. 3(a). There
is also some apparent fine α′ -martensite plate structure in
Fig. 4(b). These features (α→ α′ ) have been shown for EBM
in contrast to SLM where the melt scan is a factor of
∼30 greater for laser scanning in contrast to electron beam
scanning (Murr et al., 2009). These different microstructures
have also been studied in more detail elsewhere (Murr et al.,
2009), while Fig. 5 shows the coarse Ti–6Al–4V α-martensite
(Widmanstätten) acicular platelet structure with a magnified,
TEM image insert showing the α − β regions, and some
dislocation structure within the α-grains.
In contrast to the Ti–6Al–4V microstructures shown in
Figs. 4–6 shows a complex, microstructural architecture
composed of Cr23 C6 (cubic) precipitates within a (111)
textured fcc Co–Cr matrix; a variance of directional
solidification (Chalmers, 1964) as discussed for elongated
grain growth in SLM of Ti–6Al–4V by Thijs et al. (2010). The
Cr23 C6 microstructural architecture illustrated in the OM and
TEM image insert in Fig. 6 has been described in detail in
Gaytan et al. (2010), to result from orthogonal EBM beam Fig. 5 – Acicular α-phase grain structure in Ti–6Al–4V. The
scanning. Similar columnar, precipitate architecture has been insert shows a TEM image portion. Note α and β phases,
created by EBM built nickel-based superalloy 718 (Strondl and dislocation structure in α grains in the insert.
et al., 2008). Fig. 7 compares the as-fabricated Co–29Cr–6Mo
1402 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

Fig. 7 – OM (a) and TEM views of microstructures for


Fig. 6 – OM 3D composite illustrating columnar carbides EBM-built Co–29Cr–6Mo cylinder (z) (b), following
(Cr23 C6 ) formed in the (111) direction in EBM fabricated HIP-anneal according to ASTM-F75 Standard. (a) shows
cylinders (z) of Co–29Cr–6Mo as in Fig. 3(a) with TEM 3D equiaxed grains with annealing twins while (b) shows
composite insert showing individual cubic Cr23 C6 intrinsic (fcc) stacking faults.
precipitates (from Gaytan et al., 2010). The arrow on the
right illustrates the EBM build direction.

alloy microstructures after HIP and anneal (4 h @ 1130 ◦ C) in


contrast to Fig. 6. Here the columnar Cr23 C6 architecture has
dissolved, forming a classical, low-stacking fault-annealing
twin-related, equiaxed grain structure (Murr, 1975) containing
frequent annealing twin boundaries. Stacking faults are also
present in the as-fabricated microstructure/microstructural
architecture but are only weakly evident in contrast to the
TEM 3D insert in Fig. 6 (arrows). However, the stacking
fault density increased noticeably in the HIP-annealed
components (Fig. 7(b)).
Fig. 8 compares bar-graph summaries of mechanical
(tensile) behavior for the Ti–6Al–4V and Co–29Cr–6Mo alloy
fabricated and thermo-mechanically processed in nominal,
wrought and cast-HIPed form, EBM as-fabricated and HIP- Fig. 8 – Tensile properties for Ti–6Al–4V and Co–29Cr–6Mo.
annealed form, etc. Tensile specimens machined and tested Note yield stress (YS) represents 0.2% engineering offset.
from vertical (z) and horizontal (x, y) cylindrical builds UTS represents the ultimate tensile strength.
illustrated in Fig. 3(a) along with commercial wrought
materials correspond to the microstructures illustrated in
Figs. 4–7. While the EBM as-fabricated Ti–6Al–4V components
perform as well as the nominal wrought and coarse α-phase α-phase wrought material, exhibiting prominent ductile-
wrought products, the EBM fabricated and HIP-annealed dimple fracture as illustrated in comparing Fig. 9(a) and
Co–29Cr–6Mo alloy components are far superior to ASTM- (b), respectively. The Co–29Cr–6Mo alloy EBM-fabricated
F75 standards for biomedical implant fabrication; exceeding fracture structure for the vertical (z-axis) build exhibited an
the nominal yield stress by 30%, the UTS by 75%, and the orthogonal, ductile dimple array in many fracture surface
elongation by more than 200%. areas corresponding to the Cr23 C6 precipitate array geometry
The corresponding fracture surface structures for the EBM- illustrated in the horizontal (surface) reference plane for the
fabricated Ti–6Al–4V are indistinguishable from the coarse OM-3D composition shown in Fig. 6. This feature is shown
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1403

similar to the Ti–6Al–4V alloy fracture surface, ductile dimple


structure.

3.2. Open-cellular structure fabrication; dynamic stiffness


versus density measurements

Figs. 11–13 show typical examples for EBM fabricated


Ti–6Al–4V (Fig. 11) and Co–29Cr–6Mo alloy (Figs. 12 and
13) open cellular structures: reticulated mesh (Fig. 12) and
stochastic foam (Fig. 13) densities as indicated. These open-
cellular structures were based on the CAD-element models
illustrated in Fig. 2. In addition to the structure densities
noted in Fig. 11 for Ti–6Al–4V, additional densities were
fabricated, and the corresponding stiffness (Young’s modulus)
measured. These measured densities versus stiffness were
as follows (ρ in g/cm3 : E in GPa); with the bold densities—
represented in Fig. 11: foam (Fig. 11(a) and (b)) −0.58 (0.48),
0.68 (1.35), 0.83 (1.29); mesh (Fig. 11(c)) −0.78 (0.92), 0.86
(1.53), 1.22 (2.70), 1.59 (6.15). Correspondingly, the structure
densities noted in Figs. 12 and 13 for the Co–26Cr–6Mo
alloy and their stiffnesses along with additional ρ : (E)
measurements were as follows: foam (Fig. 13) −0.77(0.70),
0.69 (0.35), 0.66 (0.28), 0.63 (0.25); mesh (Fig. 12) −1.85 (3.68),
1.25 (1.16), 0.98 (0.49). These measured values were plotted
as relative stiffness (E/Es ) versus relative density (ρ/ρs ) as
Fig. 9 – Fracture surface structure for (a) EBM—Ti–6Al–4V shown in Fig. 14 for Es = 110 GPa and ρs = 4.43 g/cm3 for
and (b) wrought (forged) Ti–6Al–4V. (See Fig. 8.) the Ti–6Al–4V, and Es = 210 GPa and ρs = 8.44 g/cm3 for
the Co–29Cr–6Mo alloy. The complete field of data in Fig. 14
are interpolated along a fitted line with a slope, n, of 2.1,
corresponding to the general scaling for open-cellular solids
established by Gibson and Ashby (1997) in Eq. (1). While α1
in Eq. (1) has been observed to vary between 0.1 and 4,
depending upon foam structure (Gibson and Ashby, 1997;
Ashby and Tianjian, 2003), the relative modulus in Fig. 14 is
within a factor of roughly half an order of magnitude for the
range of variation of α1 which at the intercept of the fitted
line is ∼0.2 in Fig. 14. Indeed, the fact that the fitted line in
Fig. 14 for both the mesh and foam structures approximates
the ideal foam model, where n = 2 in Eq. (1), is an indication
that the mesh structure behaves essentially the same as the
foam structure.
Figs. 15 and 16 compare selected mesh and foam strut
and ligament structures, respectively, for the Ti–6Al–4V and
Co–29Cr–6Mo alloys as illustrated in Figs. 11–13, respectively.
While these strut and ligament structures show surface
roughness as a consequence of incomplete melt-sintering,
this can enhance bone cell ingrowth efficiency so long as the
effective pore sizes or open cell sizes allow for optimal bone
cell ingrowth; >100 µm (Singh et al., 2010). This feature can
limit the upper end of the density range required to achieve
requisite stiffness to reduce or eliminate stress shielding.
However, adjustments can be made in the fabrication of open
cellular structures as shown in Fig. 11 to 13 by adjusting the
Fig. 10 – Fracture surface structure for (a) build element (or unit cell) size together with the strut or
EBM—Co–29Cr–6Mo and (b) EBM-HIP/annealed ligament diameter in order to maintain the requisite pore
Co–29Cr–6Mo. Magnification for (a) and (b) is shown in (a). size while optimizing the density. Foam ligament dimension
variations as contrasted with strut dimensions are illustrated
in Figs. 15 and 16.
in Fig. 10. Correspondingly, the fracture surface features for Figs. 17 and 18 compare the mesh-strut and foam-
the as-fabricated and HIP-annealed Co–29Cr–6Mo alloy where ligament microstructures for the EBM-fabricated Ti–6Al–4V
1404 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

Fig. 11 – EBM-fabricated Ti–6Al–4V foam (a) and (b) and mesh (c) examples. (a) and (b) show views along diagonal and
normal to face (45◦ relative to (a)). Densities are noted.

and Co–29Cr–6Mo alloy open cellular structures represented It should be noted on comparing Figs. 17 and 18 that the
in Figs. 11–13. The Ti–6Al–4V mesh-strut microstructures Co–29Cr–6Mo alloy in Figs. 17(b) and 18(b) exhibit a similar
(Fig. 17(a)) are dominated by α′ -martensite plates corre- directional carbide microstructure to that shown for the EBM
sponding to the rapidly solidified (cooled) horizontal solid fabricated solid samples in Fig. 6, although not as extensive.
builds illustrated in Fig. 4(b), while the foam ligament ex- Note that the directional carbide columns in Figs. 17(b) and
ample shown in Fig. 18(a) exhibits a mixture of mainly α′ 18(b) form in build direction (arrow on the left in Figs. 17(b)
martensite, with some α-phase acicular platelets because and 18(b)).
of slightly different solidification rates. The α′ -martensite The Co–29Cr–6Mo alloy mesh-strut and foam-ligament
platelets are especially notable in contrast to the regular α- microstructures exhibit the same Cr23 C6 columnar precip-
phase acicular platelets for slower cooling solids as shown in itate architecture characteristic of the build process (beam
Fig. 5. scan geometry and corresponding, melt pool directional so-
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1405

Fig. 12 – EBM-fabricated Co–29Cr–6Mo mesh examples in 2 views at 45◦ , (a) and (b) respectively. Densities are noted.

Fig. 13 – EBM-fabricated Co–29Cr–6Mo foam examples in 2 views at 45◦ , (a) and (b) respectively. Densities are noted.
1406 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

Fig. 14 – Relative stiffness versus relative density plots for


Ti–6Al–4V and Co–29Cr–6Mo mesh and foam samples
Fig. 16 – SEM image comparisons for EBM mesh and foam
(Figs. 11–13). The data is fitted to a single line with a slope
samples. (a) Co–29Cr–6Mo mesh. (b) Co–29Cr–6Mo foam.
n = 2.1.

Fig. 15 – SEM image comparisons for EBM mesh and foam


samples. (a) Ti–6Al–4V mesh. (b) Ti–6Al–4V foam.

Fig. 17 – 3D composite OM comparisons for Ti–6Al–4V (a)


and Co–29Cr–6Mo (b) mesh strut microstructures. The
lidification for solid cylindrical components) illustrated in arrow on the left denotes the build direction.
Fig. 6. Correspondingly, the associated Vickers microinden-
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1407

Fig. 18 – 3D composite OM comparisons for Ti–6Al–4V (a)


and Co–29Cr–6Mo (b) foam ligament microstructures.
Magnifications of (a) and (b) are the same as shown in (a).
The arrow on the left denotes the build direction.

Ti-6Al-4V
5

4.1 4
4
3.4 3.5
HV (GPa)

2 Fig. 20 – X-ray images illustrating total knee replacements


for both knees. (a) Right knee replacement. (b) Left knee of
1
same patient prior to replacement. (c) Left knee (b)
0 replacement X-ray slightly magnified from (b). X-rays
SOLID (Z) SOLID (X, Y) MESH FOAM
courtesy of Patricia Murr.
Co-29Cr-6Mo
6
5.2
5
4.5 4.5
4
strength structures when compared to the solid EBM fabri-
3.5
HV (GPa)

cated components. Fig. 19 illustrates these microindentation


3
(HV) comparisons for the solid and open-cellular structures of
2
Ti–6Al–4V and Co–29Cr–6Mo alloy.
1

0
SOLID (Z) SOLID (Z) HIP MESH FOAM 3.3. Knee replacement components and stress shielding
design strategies
Fig. 19 – Vickers microindentation hardness (HV)
comparisons for Ti–6Al–4V and Co–29Cr–6Mo solid and Fig. 20 illustrates two examples of contemporary, commer-
open-cellular (mesh and foam) components fabricated by cial, orthopaedic total knee replacements consisting of a
EBM. Co–29Cr–6Mo alloy femoral appliance (f ) and a Ti–6Al–4V tib-
ial stem appliance (t). These components are cemented using
acrylic cement onto the shaped bone segments with no stress
tation hardness (HV) measurements for the 3D reconstruc- shielding surface structure porosity. The left X-ray (Fig. 20(a))
tions in Figs. 17 and 18 are indicative of relatively high shows a right knee replacement with a slightly longer tibial
1408 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

Fig. 22 – Stress shielding design graph. Shaded region


corresponds to the Ti–6Al–4V and Co–29Cr–6Mo data field
in Fig. 14. Fitted lines are indicated by 1–4 shown in the key
box.

Fig. 21 – Co–29Cr–6Mo femoral prototype with mesh


structure fabricated by EBM and HIP-annealed using with those commercial appliances shown in Fig. 20(a) and (c).
ASTM-F75 standard, and finish machined and partially This prototype has a solid internal core surrounded by a mesh
polished as shown in (a). (b) Shows a magnified view of structure corresponding to Fig. 2(a). The design assumes
mesh structure section. a transition between the central intramedulary/trabecular
bone region and the outer cortical bone, having an assumed
stiffness of ∼5 GPa. This corresponds to a relative stiffness
stem than the right knee (X-ray image in Fig. 20(c)). The dam- (E/Es ) in Figs. 14 and 23 of 0.045, corresponding to a relative
aged (multiply-torn meniscus) left knee prior to replacement density of ∼0.36, or an approximate density of 1.6 g/cm3
(Fig. 20(c)) is shown in the center X-ray image in Fig. 20(b). For characteristic of the mesh structure shown in Figs. 2(a), 3(b)
commercial implants, the Co–Cr component is usually cast and (c), and 11(c).
while the Ti–6Al–4V component is usually fabricated from Porous implants which rely upon bone cell ingrowth for
wrought stock. stability, and especially the tibial stem in Fig. 24, must be
Fig. 21 shows a Co–29Cr–6Mo alloy prototype fabricated stabilized initially with a press fit which is not easily obtained,
by EBM, HIP-annealed as illustrated in Fig. 7, and machined and requires meticulous surgical technique. The use of a
and partially polished. The prototype is oriented as illustrated longer stem prosthesis as illustrated in Fig. 24 anchored in
in the side-view X-ray image of Fig. 20(a) corresponding to the intramedulary isthmus would seem desirable. Since shear
the femoral component (f ). The inside of the implant has strength is influenced (reduced) by porosity, this can become
been built as a monolith having a mesh array similar to that an issue along with adequate fatigue strength, perhaps more
fabricated in Fig. 12. This array is roughly 2 mesh openings so in other prostheses, such as the hip stem. In addition, for
with a thickness of ∼600 µm, and a pore or mesh opening severe knee damage or fracture of the femor, the femoral
size of ∼400 µm; with an effective density of ∼4.2 g/cm3 , as component shown in Figs. 20 and 21 can be connected (or
illustrated in the magnified view in Fig. 21(b). This density, or fabricated as a monolith by EBM) to a stem as illustrated in
its equivalent relative density (ρ/ρs ∼ = 0.5) corresponds to a Fig. 23.
relative stiffness (E/Es ) of 0.09 in Fig. 14. This is represented A low modulus, porous system having adequate stress
in the stress shielding design graph reproduced (from Fig. 14) shielding can also be developed using foam stem designs
in Fig. 22 for both the Ti–6Al–4V and Co–29Cr–6Mo alloy such as those suggested in Fig. 3(b) and (c), and in the CAD-
data. This corresponds to an assumed femoral cortical bone model designs illustrated in Fig. 24. These features have been
stiffness of ∼19 GPa. described in previous work addressing EBM implant design
Fig. 23 shows the EBM fabrication of a Ti–6Al–4V tibial stem and fabrication (Murr et al., 2010) for Ti–6Al–4V. Of course
prototype having an exaggerated stem length in comparison the functional porosity, monolithic stem concepts illustrated
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1409

Fig. 23 – Ti–6Al–4V EBM tibial knee stem monolithic prototype. (a) Dode-thin build element (Fig. 2(a)) creating a CAD model
with a solid rod core surrounded by a mesh. Note the enlarged mesh section insert on the left. (b) EBM prototype product
with a mesh density of ∼0.93 g/cm3 . (c) Magnified section in (b).

in Fig. 24 can be incorporated in femoral stem designs as these columnar precipitates dissolved and were replaced
described above, as well as the tibial stem concepts illustrated by a higher density of intrinsic stacking faults which
in Fig. 23. kept the hardness essentially constant, while there was
a slight drop in the yield stress. However, the UTS
increased by 20% while the elongation increased by 900%.
4. Summary and conclusions In contrast to the ASTM-F75 Co–29Cr–6Mo transplant
standard, the yield stress for the EBM-fabricated and HIPed
We have demonstrated that additive manufacturing using component increased by 30% while the elongation increased
electron beam melting can fabricate compatible, patient by ∼200%.
specific total knee implants: Co–29Cr–6Mo femoral appliances We measured the dynamic stiffness (Young’s modulus)
and Ti–6Al–4V tibial appliances consisting of monolithic, and density for Ti–6Al–4V and Co–29Cr–6Mo alloy mesh and
functional solid-mesh-foam components. In this study, we foam samples fabricated by EBM, and plotted these data
have compared characteristic microstructures for solid, as relative stiffness (E/Es ) versus relative density (ρ/ρs ) on
mesh, and foam Ti–6Al–4V prototypes. The Ti–6Al–4V log–log axes. A fitted line to the data resulted in a slope, n
solid EBM prototypes exhibit the α-phase, acicular platelet in Eq. (1) of 2.1. From this data plot, a stress shielding design
microstructures similar to commercial, wrought products, graph was constructed which permits Co–29Cr–6Mo and
while the mesh and foam prototypes exhibit α′ -martensitic Ti–6Al–4V mesh and foam implant prototypes to be fabricated
platelets or mixtures of α and α′ platelets which are for compatible bone stiffness in contrast to requisite
spatially smaller than the solid α-phase platelets, giving rise density.
to harder structures and correspondingly higher strengths.
In contrast, the EBM fabricated solid, mesh, and foam
Co–29Cr–6Mo prototypes all exhibited a directional, columnar Acknowledgments
Cr23 C6 precipitate architecture parallel to the EBM build
direction intermixed with some stacking faults in the fcc We are pleased to acknowledge the support of John Wooten
matrix. These columnar precipitates were spaced ∼2 µm of CalRam, Inc., Simi Valley, CA, and Ulf Ackelid of Arcam
within textured, directional grains. Following HIP-annealing AB, Sweden in specimen preparation and mechanical testing.
1410 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411

Gibson, L.J., Ashby, M.F., 1997. Cellular Solids: Structure and


Properties. Cambridge University Press, New York, NY.
Hacking, S.A., et al., 2002. Fibrous tissue ingrowth and
attachment to porous tantalum. J. Biomed. Mater. Res. 52,
631–638.
Harryson, O., Cormier, D.R., 2005. Direct fabrication, of custom or-
thopaedic implants using electron beam melting technology.
In: Gibson, I. (Ed.), Advanced Manufacturing Technology for
Medical Applications: Reverse Engineering, Software Conver-
sion and Rapid Prototyping. John Wiley & Sons Ltd., London,
pp. 191–206 (Chapter 9).
Heinl, P., et al., 2007. Cellular titanium by selective electron beam
melting. Adv. Eng. Mater. 9 (5), 360–364.
Heinl, P., et al., 2008. Cellular Ti–6Al–4V structures with
interconnected macro-porosity for bone implants fabricated
by selective electron beam melting. Acta Biomater. 4,
1536–1544.
Hollander, D.A., et al., 2006. Structural, mechanical and in-
vitro characterization of individually structured Ti–6Al–4V
produced by direct laser forming. Biomaterials 27 (7),
955–963.
Holy, C.E., Shoichet, M.S., Davies, J.E., 2000. Engineering three-
dimensional bone tissue in vitro using biodegradable scaffolds:
investigating initial cell-seeding density and culture period.
J. Biomed. Mater. Res. 51 (3), 376–382.
Long, M., Rack, H.J., 1998. Titanium alloys in total joint
replacement—a materials science perspective. Biomaterials 19
(18), 1621–1639.
Marin, E., et al., 2010. Characterization of cellular solids in
Ti–6Al–4V for orthopaedic implant applications: trabecular
titanium. J. Mech. Behav. Biomed. Mater. 3, 373–381.
Fig. 24 – CAD model foam stem having inner low density Multen, L., et al., 2009. Selective laser melting: a regular unit
foam and outer higher density foam. (a) and (b) show cell approach for the manufacture of porous, titanium, bone
in-growth constructs suitable for orthopaedic applications.
different foam density models. (b) Shows a cut-away
J. Biomed. Mater. Res. Part B 89B (2), 325–334.
(half-section) view.
Murr, L.E., 1975. Interfacial Phenomena in Metals and Alloys.
Addison-Wesley Publishing Co., Inc., Reading, MA.
Murr, L.E., et al., 2009. Microstructure and mechanical properties
This research was supported by Mr. & Mrs. MacIntosh of electron beam-rapid manufactured Ti–6Al–4V biomedical
Murchison Chair endowments at The University of Texas at prototypes compared to wrought Ti–6Al–4V. Mater. Charact.
60, 96–105.
El Paso.
Murr, L.E., et al., 2010. Next generation biomedical implants using
additive manufacturing of complex cellular and functional
REFERENCES mesh arrays. Phil. Trans. R. Soc. A 368, 1999–2032.
Niinomi, M., 2001. Recent metallic materials for biomedical
applications. Metall. Mater. Trans. A 32A, 477–486.
Niinomi, M., 2007. Recent research and development in metallic
Ashby, M.F., et al., 2000. Metal Foams: A Design Guide.
materials for biomedical, dental, and healthcare products.
Butterworth-Heinemann, Boston, Mass.
Mater. Sci. Forum 539–543, 193–200.
Ashby, M.F., Tianjian, L., 2003. Metal foams: a survey. Sci. China
Niinomi, M., 2008. Mechanical biocompatibilities of titanium
Ser. B 46 (6), 521–532.
alloys for biomedical applications. J. Mech. Behav. Biomed.
Bobyn, J.D., et al., 2005. The effect of proximally and fully porous-
Mater. 1, 30–42.
coated canine hip stem design in bone modeling. J. Orthop.
Park, J.B., 2000. In: Bronzino, J.D. (Ed.), The Biomedical Engineering
Res. 5, 393–408.
Handbook Vol. 1. CRC Press LLC, Boca Raton, FL.
Chalmers, B., 1964. The Principles of Solidification. Wiley, New Ryan, G., Pandit, A., Apatsidis, D.P., 2006. Fabrication methods
York. of porous metals for use in orthopaedic applications.
Chuna, C.K., Leong, K.F., Lim, C.S., 2003. Rapid Prototyping: Princi- Biomaterials 27 (13), 2651–2670.
ples and Applications, 2nd ed. World Scientific, Singapore. Sanan, A., Haines, S., 1997. Reporing holes in the head: a history
Dunand, D.C., 2004. Processing of titanium foams. Adv. Eng. Mater. of cranioplasty. Neurosurgery 40 (3), 588–603.
6, 369–376. Singh, R., et al., 2010. Titanium foams for biomedical applications:
Freyman, T.M., Yannas, I.V., Gibson, L.J., 2001. Cellular materials a review. Mater. Tech. 25, 127–136.
as porous scaffolds for tissue engineering. Prog. Mater. Sci. 46 Stamp, R., et al., 2009. The development of a scanning strategy
(3–4), 273–282. for the manufacture of porous biometerials by selective laser
Gaytan, S.M., et al., 2010. Comparison of microstructures and melting. J. Mater. Sci. Mater. Med. 20 (9), 1839–1848.
mechanical properties for solid and mesh cobalt-base alloy Strondl, A., et al., 2008. Investigations of MX and γ ′ /γ ′′ precipitates
prototypes fabricated by electron beam melting. Metall. Mater. in the nickel-based superalloy 718 produced by electron beam
Trans. A 41A, 3216–3227. melting. Mater. Sci. Engr. A 480, 138–147.
J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M AT E R I A L S 4 (2011) 1396–1411 1411

Thijs, L., et al., 2010. A study of the microstructural evolution Yang, S., et al., 2001. The design of scaffolds for use in tissue
during selective laser melting of Ti–6Al–4V. Acta Mater. 58, engineering, part I. Traditional factors. Tissue Eng. 7 (6),
3303–3312. 679–689.