Biomaterials 19 (1998) 2015 — 2029

Plasma sprayed hydroxyapatite coatings on titanium substrates

Part 1: Mechanical properties and residual stress levels
Y.C. Tsui!, C. Doyle", T.W. Clyne!,*
!Department of Materials Science and Metallurgy University of Cambridge, Pembroke Street, Cambridge CB2 3QZ, UK
"Howmedica, Pfizer Hospital Products, Ash House, Fairfield Avenue, Staines, Middlesex TW18 4AN, UK
Received 1 September 1997; accepted 3 March 1998

Abstract

Hydroxyapatite (HA) coatings have been sprayed on to substrates of Ti—6Al—4V, using a range of input power levels and plasma gas
mixtures. Coatings have also been produced on substrates of mild steel and tungsten, in order to explore certain aspects of the
mechanical behaviour of HA without the complication of yielding or creep in the substrate. Studies have been made of the phase
constitution, porosity, degree of crystallinity, OH~ ion content, microstructure and surface roughness of the HA coatings. The
Young’s moduli in tension and in compression were evaluated by the cantilever beam bend test using a tungsten/HA composite beam.
The flexural Young’s modulus was determined using a free-standing deposit under the same test. Adhesion was characterised using the
single-edge notch-bend test; this is considered superior to the tensile bond strength test in common use. Measured interfacial fracture
energies were of the order 1—10 J m~2. Stress levels were investigated using specimen curvature measurements in conjunction with
a numerical process model. The quenching stress for HA was measured to be about 10—25 MPa and the residual stress level in HA
coatings at room temperature are predicted to lie in the approximate range of 20—40 MPa (tensile). These residual stresses could be
reduced in magnitude by maintaining the substrate at a low temperature (possibly below room temperature) during spraying and it
may be worthwhile to explore this. Ideally, the HA coating should have low porosity, high cohesive strength, good adhesion to the
substrate, a high degree of crystallinity and high chemical purity and phase stability. In practice, such combinations are rather difficult
to achieve by just varying the spraying parameters. ( 1998 Published by Elsevier Science Ltd. All rights reserved

Keywords: Plasma spraying; Hydroxyapatite; Young’s modulus; Adhesion; Residual stresses

1. Introduction faces strongly affect its performance. While the former

HA-coated implants have been widely used in ortho-
paedics [1, 2] and dentistry [3, 4]. This cementless fix-
ation technique combines the strength, ductility and ease
of fabrication of metallic implants with the increased
biocompatibility associated with HA. Once HA is im-
planted, it has the ability to bond directly to the bone [1],
to achieve earlier and greater fixation strength [2, 5, 6]
and to reduce healing time [7] and pain levels. The
reason for its acceptability lies in its having a composi-
tion similar to the mineral phase of bone and tooth
enamel [8]. In an implanted prosthesis, the stability and
the adherence of implant/coating and coating/bone inter-
* Corresponding author. Tel.: 01223 334332; fax: 01223 334567;
e-mail: twe 10@cam.ac.uk
bond is largely mechanical and the latter is mainly
physiochemical in nature, the performance of both inter-
faces is closely related to the coating properties.
HA coatings have been applied by a variety of
methods: dip coating [9, 10], electrophoretic deposition
[11, 12], hot isostatic pressing [9], ion-beam sputtering
[13], ion beam dynamic mixing [14], plasma spraying
[15], conventional flame spraying [16, 17] and high-
velocity oxy-fuel (HVOF) combustion spraying [18, 19].
Among these, plasma spraying appears to be the most
favourable one in terms of chemical control, biocorro-
sion resistance [20], process efficiency [15] and the de-
gree to which the substrate fatigue resistance is reduced
[21].
The ideal HA coating for orthopaedic implants would
be one with low porosity, strong cohesive strength, good
adhesion to the substrate, a high degree of crystallinity

0142-9612/98/$—See front matter ( 1998 Published by Elsevier Science Ltd. All rights reserved.
PII S 0 1 4 2 - 9 6 1 2 ( 9 8 ) 0 0 1 0 3 - 3
2016 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029
and high chemical purity and phase stability. Amorphous
HA tends to dissolve rapidly in the physiological envi-
ronment, so that coatings with low crystallinity quickly
become weak and may promote inflammatory responses.
There may be benefits in tailoring the chemistry of the
coating surface in some way so as to promote bone
growth. It is possible that a thin reprecipitated amorph-
ous HA layer on the surface could thus be advantageous,
but the details of this are not yet clear. While a number of
studies [19, 22—29] have been devoted to characterising
the changes induced by plasma spraying in chemical
composition and crystallinity, studies on the mechanical
behaviour of HA coatings are rare. It is well known that
the Young’s modulus of a plasma sprayed coating is
usually much lower than its corresponding bulk value.
This can be attributed to the presence of pores and
microcracks inside the coating. There are virtually no
data available on the Young’s modulus of sprayed HA,
which is essential in predicting the residual stress levels
presented inside the coatings and in determining their
fatigue behaviour under cyclic loading.
A number of in vivo studies [1, 2, 7, 30, 31] have in-
dicated that failure mainly occurs at the metal/coating
interface. The longer the period of implantation, the
higher is the probability of failure at this interface (since
the strength of the bone/HA interface tends to increase
with time during the early stages of post-operative recov-
ery). Therefore, any anticipated long term benefit is ex-
pected to depend on the adhesive and cohesive integrity
of the coating, which are strongly dependent on micro-
structure. To optimise the adhesion of HA coatings on
metallic implants, a reliable method is needed for charac-
terisation. The adhesion of HA coatings on metal sub-
strates is frequently determined by the tensile adhesion
test. This test has long been regarded as semi-quantita-
tive at best and useful only for ranking purposes. The
main problem associated with this test is that failure
depends on the distribution of the flaws present at the
specimen edge, which results in a wide scatter for the
strength values obtained. In addition, there is a danger of
significant penetration of adhesive (usually epoxy) into
the coatings or even down to the interface if they are thin.
To characterise the interfacial adhesion in a systematic
way, a fracture mechanics approach should be adopted.
The interfacial fracture toughness, K (or the closely
*#
associated critical strain energy release rate, G ), should
*#
be considered, along with the strength, in the overall
design of an implant system. There has been very little
research in this area concerning biomedical materials.
Filiaggi et al. [32] used the short bar chevron notch test
and obtained values of K equal to 0.60—1.41 MPa m1@2.
*#
Evan et al [33] used the double-cantilever beam test and
obtained values of G equal to 1 J m~2 and 4 J m~2 for
*#
bead-blasted and grit-blasted substrates, respectively. It
should be noted that both of these sets of values represent
relatively brittle interfaces.
In this study, coatings plasma sprayed with different
input power levels and different plasma gas mixtures (Ar
with H or He) have been examined. Free-standing de-
2
posits were obtained by pre-spraying a layer of salt
before coating deposition, which was followed by immer-
sion in water. Young’s moduli were measured by a canti-
lever beam bending method. Adhesion was measured by
a single-edge notch bend test. A numerical model was
used to evaluate the quenching stress of the HA coatings
in conjunction with the use of an in situ curvature
monitoring technique. Some predictions of the residual
stress levels in sprayed HA coatings produced under
various conditions are presented.
2. Experimental procedure
2.1. Substrate preparation and plasma spraying
Substrates of Ti—6wt%Al—4wt%V were prepared by
pickling in acid (8% HF and 40% HNO ) for 1 min, to
3
remove surface oxide, and degreased by rinsing in
acetone. Substrates were then grit-blasted with brown
Al O (!80 mesh) under a pressure of 6 bar for &30 s.
2 3
The surfaces were air-blasted to remove any residual grit
and finally cleaned with alcohol. Similar substrate prep-
aration procedures were employed with mild steel and
tungsten. The substrates were then coated with HA, using
the spraying conditions given in Table 1. The spraying
equipment employed was a PT VPS system with a F4-V
gun. The spraying pattern consisted of a number of
cycles, depending on the thickness of the coating re-
quired, each consisting of six vertical passes of the gun,
followed by an inter-cycle cooling period. This was done
to ensure that the substrate temperature remained within
a specified range. For spraying onto tungsten
(&100!125 lm thick), a mild steel strip (&2 mm thick)
with many holes in it was placed at the back of the
substrate for support. Argon cooling was applied at the
back. For spraying onto mild steel and Ti—6Al—4V, the
pre-set minimum temperature between each cycle was
&400°C and that for spraying on tungsten was &100°C.
The input power levels used were between 30 and 42 kW.
Hydrogen or helium was used as the secondary plasma
gas. The flow rates are shown in Table 1. Coatings of
thickness in the range 100 lm—1 mm were obtained.
Table 1
Spraying conditions used for all the specimens
Chamber pressure (mbar) 200
Spraying stand-off distance (mm) 270
Nozzle internal diameter (mm) 8
Plasma gas flow rates (slpm) Ar"50#H "4—9 or
2
Ar"35—40#He"50
 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029 2017

2.2. Coating characterisation ferometric (Wyko RS Plus) surface profilometers were

used to characterise the topography and to measure the
Particle size analysis was performed for the HA pow- surface roughness parameter, R . A Camscan S2 scann-
!
ders using a Malvern MasterSizer E machine. Powders ing electron microscope was used to examine the
and sprayed coatings were examined by X-ray diffraction morphology of the powders prior to spraying, the micro-
using a computer controlled Philips PW 1710 diffrac- structures of the coatings, the transverse section of the
tometer. Diffraction scans were run with a generator specimens and the fracture surfaces on the deposits and
tension of 40 kW and a generator current of 25 mA. substrates after mechanical testing.
A divergence slit angle of 0.5° and a receiving slit width of
0.2 mm were used. For both HA coatings and powders,
each diffraction scan was run from 20° to 40° (2h), with 2.3. Young+s modulus measurement
a step of 0.02° and a scan time of 1 s, using Cu K
a
radiation. The peaks were identified by comparing with The Young’s modulus of the coating was first esti-
the JCPDS. Curve fitting was performed on all the HA mated by bending a composite (substrate#coating)
crystalline peaks in the range 28—35° and on the amorph- beam. A trial three-point bend test, with &1 mm thick
ous peak centred on 31° (2h). Crystallinity was then coating on &1.5 mm thick Ti—6Al—4V substrate, showed
calculated using the following equation: that the load/displacement plot was very insensitive to
the presence of the HA coating, i.e. the slope of the plot
+A
Percentage of crystallinity" # ]100% (1) was almost identical to that when testing the substrate
+A #+A alone. This indicated that the Young’s modulus of the
# !
coating was much less than that of the substrate. There-
where +A is the sum of the areas under all the HA
# fore, in order to obtain accurate results, a very thin
crystalline peaks and +A is the sum of the area under
! substrate was used.
the amorphous peak.
Thin tungsten strips with thickness 0.1 and 0.125 mm
The infrared spectra of HA coatings and powders
were used. Tungsten was chosen, in view of its high yield
were recorded on an infrared Fourier transform spec-
stress, to eliminate the possibility of any inelastic pro-
trometer (Perkin-Elmer 1720-X). The HA powders or
cesses occurring during spraying and subsequent cooling.
the debonded HA coatings, after grinding into powder
Argon cooling was applied at the back to keep the
form, were mixed with potassium bromide (KBr) in
substrate temperature as low as possible. This was done
the weight ratio of 1 to 100, respectively. The mixed
powders were then subjected to compression under
a pressure of 10 bar to form a transparent disc. Infrared Table 2
light was allowed to pass through the disc and the trans- Thermo-physical and mechanical properties of HA used in the numer-
mittance at various wave numbers from 400—4000 cm~1 ical process model
was recorded.
The density of a deposit, o , was determined by a hy- Property Temperature (K) Value Ref.
$
drostatic weighing technique [34]. After spraying, HA Thermal conductivity 283 0.72 [62]
coatings were removed from tungsten substrates using (W m~1 K~1) 1352 2.16 [62]
a scalpel. Pieces of HA deposit were ultrasonically Specific heat capacity 293—1300 766 [63]
cleaned in acetone and then dried at 50°C, before the (J kg~1 K~1)
density measurements were performed. The porosity Latent heat of fusion — 15.4 [64]
(kJ kg~1)
level, in percentage by volume, of the deposit (P ) was
& Melting point (K) — 1923 [65]
determined by evaluation of the deposit density, o , and !
$ Droplet temperature (K) — 1923
comparison with theoretical bulk density, o , Density (Mg m~3) see Table 3
5) Thermal expansivity 293 13.3 [66]

A A BB
o (MK~1) 1093 13.3 [66]
P " 1! $ ]100% (2)
& o Young’s modulus (GPa) see Table 3"
5) Poisson’s ratio — 0.30 [33]
To check the theoretical density of HA, the powder Elastic limit (MPa) 50 #
Quenching stress (MPa) see Table 6
density was also measured by a similar technique. The
powder was put in a glass container before weighing and ! Droplet superheat was assumed negligible in all cases; this is prob-
an ultrasonic bath was used to remove the trapped bub- ably not accurate, but the final predictions are relatively insensitive to
bles inside the liquid. the superheat.
" Temperature dependence taken as that of sprayed ZrO —8%Y O
A Leitz M8 optical stereo microscope was used to 2 2 3
coatings [67], i.e. falls to 56% of the room temperature value on heating
examine the coating surfaces and the fracture surfaces by 1000 K.
obtained after mechanical testing. Both mechanical # Assumed to be the same as the bond strength value reported [20]
(Rank Taylor Hobson Talysurf 6) and optical inter- for sprayed HA coatings.
2018 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029

so as to prevent macrocracking, which would tend to coating on a rectangular section substrate, to that of the
occur if the stress in the coating exceeded its flexural unknown value of the deposit Young’s modulus, E ,
$
strength. Argon cooling is very important, since thick HA

C D C D
h2 h2
coatings are needed for measurement and in such cases & "E h b $1#h d#d2 #E h b 4 !h d#d2
# $ $1 3 $1 4 4 3 4
the stress would otherwise build up to a high level,

C
because the thermal expansivity mismatch between
W and HA is large (see Table 2). The composite beams, !E b 1(h2 #h2 )(h #h )#d(h2 #h h #h2 )
$ 4 $1 $2 $1 $2 $1 $1 $2 $2
which were less than 1 mm thick, had very low beam
stiffnesses. This precluded use of a conventional mechan-
ical testing machine. A cantilever beam bending method
was found to be suitable. One end of the specimen was
$2 $ $2D
#3 d2(h #h )#d3 #E b(h #d) d(h #h )
2 $1 $1 $2 C
clamped and an increasing load was applied at the other
end, while its displacement was measured using a travel-
#1(h2 #h h #h2 )#d2
3 $1 $2 $2 $2 D (3)

ling microscope (see Fig. 1a). The Young moduli in ten- where E is the Young’s modulus of the substrate and the
sion and in compression were measured by arranging for 4
definition of other variables can be found in Fig. 1b. The
the HA coating to be on the top or the bottom, respec- distance from the interface to the neutral axis, d (positive
tively, during loading. The maximum load that could be into the substrate), is defined by
applied, without causing through-thickness macrocracks
3E h2!E h h !E h2 !E h2
to form in the coating (with HA in tension) or delamina- d" 4 s $ $1 $2 $ $1 $ $2 (4)
tion (with HA in compression), was less than 1 N. A check 3[E (h #h )#2E h ]
$ $1 $2 4 4
was made as to whether the loading was elastic by noting and & can be calculated from the experimental data,
whether the beam returned to the same vertical position #
using the following equation:
on unloading.

A BA B
The following equation [35] was used to relate the ¸3 P
&" (5)
composite beam stiffness, & , for a tapered section HA # 3 u
#
where ¸ is the length of the specimen between the clamp
and the loading point, P is the applied load and u is the
displacement of the loading point.
The Young’s moduli of free-standing deposits were
also measured. The Young’s modulus value was cal-
culated by using

A BA B
1 ¸3 P
E" (6)
2 I u
where I is the moment of inertia. Some of the specimens
with free-standing deposits had a tapered section; in this
case, the following expression was used for the moment of
inertia:
bh3 b(Dh )3
I" $1#A (d!y )2# $ #A (y !d)2 (7)
12 1 1 36 2 2
and the position of the neutral axis, d, can be expressed as
A y #A y
d" 1 1 2 2 (8)
A #A
1 2
where the definition of the variables can be found in
Fig. 1c. Bending about the y-axis could in principle oc-
cur. However, since the width of the deposit was signifi-
cantly greater than *h , this effect could be ignored.
$
2.4. Adhesion measurement
Fig. 1. (a) Loading arrangement for the cantilever beam bending test,
(b) specimen geometry for calculating the Young’s modulus from this 2.4.1. Single-edge notch-bend test
test, assuming the coating is tapered in section, and (c) location of the THE single-edge notch-bend (SENB) test was applied
neutral axes for parts of a tapered section coating. to measure the interfacial fracture toughness of the
 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029
Fig. 2. Loading configuration and specimen geometry for (a) single-
edge notch, bend test and (b) tensile bond strength test (c) Optical
micrograph of the pre-crack formed after spraying.
HA/Ti—6Al—4V system under pure opening mode of
loading. Relatively thick layers of HA, about
400$50 lm, were applied on Ti—6Al—4V bars, to avoid
penetration of adhesive down to the interface. The speci-
men geometry and loading configuration are shown in
Fig. 2a. Before spraying, an area was masked off on the
substrate surface, about 5 mm in height and over the
complete width (b), prior to grit blasting. A pre-crack was
thus formed after spraying (see Fig. 2c). The coated half
was then joined to the other half using an adhesive epoxy
resin. At least 24 h was allowed for the adhesive to set;
during this period, the specimens were clamped. Excess-
ive adhesive was removed using a scalpel. The load, P, at
which the crack propagated (and a load drop occurred),
was recorded and the interfacial fracture toughness, K ,
*#
was calculated from [36]
A AB AB
3PsJa a a 2
K " 1.99!2.47 #12.97
*# bh2 h h
AB A BB
a 3 a 4
!23.17 #24.80
h h
where the conditions required are
¸/h'2; (¸#2s)/h"4; g*4.2h; 0(a/h(0.6;
2)h/b)8
In this study, a"5 mm; h"20 mm; b"6 mm or 8 mm;
¸"60 mm; s"27 mm and g"180 mm, so that all of
these conditions were satisfied.
2019
2.4.2. Tensile bond strength test
For this test, HA coatings of thickness 150$20 lm
were sprayed on Ti—6Al—4V rods of 25.4 mm in diameter.
Tests were conducted in accordance with ASTM C633,
with the exception that the coating thickness was below
the lower limit of 380 lm set by the standard. This was
done to allow direct comparison with results reported in
the literature. The coated half was joined to another half
by epoxy resin under the same procedure as described in
Section 2.4.1. The loading configuration is shown in
Fig. 2b. The tensile strength was obtained by dividing the
critical load at failure by the coated area.
2.5. Prediction of residual stresses
The quenching stress [37—39] for HA was measured
using an in situ curvature monitoring technique [40], in
conjunction with a numerical model [41—43]. Thermal
histories during spraying were recorded from thermal-
couples spot welded to the rear of the thin strips. Curva-
ture histories were obtained using a video recording
technique [40]. This involves recording the image of the
end of the strip, followed by image analysis of successive
frames in order to establish the changing specimen curva-
ture. This allows the rapid assimilation of data from
a large number of frames, carrying detailed information
about stress development.
The thermophysical and mechanical properties of HA
used in the numerical process model are summarised in
Table 2. Since the mechanical behaviour of sprayed HA
coatings is very similar to sprayed ZrO —8 wt%Y O
2 2 3
coatings, it was assumed that the Young’s modulus tem-
perature dependency of the former is the same as that of
the latter. Moreover, the values of thermal conductivity
at various temperatures of sprayed HA coatings were
assumed to change in the same way. These values repres-
ent reasonable estimates, since other researchers reported
the values of thermal conductivity to be 1.1—
1.6 W m~1 K~1 in the temperature range of 200 to
1000 K for sintered HA [44]. From a knowledge of
material properties and the values of quenching stress,
the residual stress levels for spraying at various condi-
tions with different substrate thicknesses or materials can
be predicted easily.
3. Results and discussion
(9)
3.1. Coating characterization
The size distribution of the as-received HA powder
used in this study is shown in Fig. 3a. The SEM micro-
(10) graph of the powder, shown in Fig. 3b, indicates that the
particles are mainly angular in shape. Figure 4a shows
the free surface of a 36 kW coating and Fig. 4b shows the
corresponding polished transverse section. The coatings
2020 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029

sprayed at different power levels and different secondary
plasma gas look very similar to those shown in Fig. 4.
Pores and microcracks are clearly visible. The pores form
as a result of poor bonding between adjacent splats,
whereas microcracking arises from shrinkage of the splat
during quenching and subsequent differential thermal
contraction between substrate and coating.
Measured coating properties are summarised in
Table 3. In general, the properties of coatings sprayed at
30 kW with H are similar to those of coatings sprayed at
2
36 kW with He as the secondary plasma gas. Using
helium as the secondary gas is less efficient at heating up
the HA particles, for a given power level, because the gas
flow rates needed are much higher (see Table 1), so that

Fig. 4. SEM micrographs from an HA coating produced with a plasma

Fig. 3. (a) Particle size distribution and (b) SEM micrograph of the power of 36 kW, showing (a) the free surface and (b) a transverse
as-received HA powder. section.

Table 3
Measured property data for HA coatings sprayed with different power levels and plasma gas mixtures

Power (kW) Density Porosity# Young’s modulus (GPa) Crystallinity

(Mg m~3) (%) (%)
Tension Compression Flexural

30 2.82$0.01 10.7 0.50$0.32 3.17$1.32 — 93.6$2.5

33 2.84$0.02 10.0 2.69$1.19 5.34$1.83 4.31$2.01 89.3$1.8
36 (He)! 2.82$0.02 10.6 0.49$0.28 3.96$1.24 — 93.2$2.0
36 2.89$0.01 8.6 2.92$0.97 5.30$1.15 3.72$1.53 85.0$1.9
36 (Ar)" — — — — — 83.5$2.4
40 (He)! — — 1.55$0.54 3.90$1.06 — 86.7$2.1
42 2.89$0.01 8.6 3.42$0.87 5.20$1.34 — 84.0$1.8
Powder 2.99$0.03 5.2 — — — 98.0$1.5

! Helium secondary gas.

" Argon cooling on the rear surface.
# Based on a theoretical density [52, 68, 69] of 3.156 Mg m~3.
 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029 2021

Fig. 5. Amorphous contents and porosity levels of HA coatings

sprayed with different power levels and plasma gas mixtures.

more cold gas is injected into the plasma. (Hydrogen

absorbs and subsequently releases heat more efficiently,
since it can dissociate as well as ionise.) However, the
presence of hydrogen in the plasma may promote more
chemical changes in the HA—see below.
As shown in Fig. 5, the porosity level decreases with
increasing input power level. This is expected, since more
complete particle melting usually results in a lower por-
osity content. At low spraying powers, the HA particles
were poorly melted. When they impacted on the substra-
te (or the already formed coating), they were not able to
spread out completely to form splats and therefore, could
not conform to the surface. The crystalline cores did not
pack well and left pores in the coating. The measured
porosity levels (&10%) are relatively high compared
with many VPS coatings. This can be attributed to the
large particle size used (average size &80 lm), which
inhibits complete melting of the particles. After spraying, Fig. 6. X-ray diffraction spectra, showing the peaks of HA and other
the coatings often contained small amounts of calcium calcium phosphate compounds, plus the curve fitting profiles, for speci-
phosphate compounds other than Ca (PO ) (OH) , mens sprayed at (a) 33 kW, (b) 36 kW and (c) 36 kW with He gas. The
10 46 2 arrows indicate non-HA peaks, which are less prominent at lower
which have different bulk density values. This was not
plasma power and with He gas.
taken into account in calculating the porosity levels.
Moreover, a lower value of 3.08 Mg m~3 for pure HA has
also been used previously [24, 45], so that there must be All the peaks in the XRD spectrum of the HA powder
some uncertainty about the absolute values in Table 3. match the JCPDS 9-432. It can be seen, from Table 3 and
The trends shown in Fig. 5 should not, however, be Fig. 5, that the degree of crystallinity decreases with
affected by this. increasing input power, as the degree of melting of HA
The density of the HA powder was measured as particles increases. This indicates that melted portions of
2.992 Mg m~3. The fact that this is slightly less than the HA particles solidified into an amorphous phase after
theoretical value indicates that the powder might have spraying. Crystalline peaks became less sharp and
contained some pores after the powder production pro- broadened after spraying and these effects are more
cess (the details of which have not been released by the prominent for coatings sprayed at higher powers. Again,
manufacturer); this is common, for example, after sinter- the spectrum of 30 kW coating is similar to that of 36 kW
ing. Moreover, the powder might not have been pure (He) coating. The degree of crystallinity was determined
stoichiometric HA and ion substitutions could have oc- using Eq. (1), with the area under each peak being cal-
curred. For example, (CO )2~ can substitute for (PO )3~ culated after performing curve fitting on the spectrum
3 4
in the HA lattice. (see Fig. 6).
2022 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029
Fig. 7. Fourier transform infrared spectra for (a) HA powder and (b)
(1) the HA powder and the specimens sprayed at (2) 30 kW, (3) 33 kW,
(4) 36 kW with He, (5) 36 kW and (6) 42 kW.
Ideally, all the peaks within the XRD spectrum should
be fitted or, even better, a calibration curve should be set
up by mixing pure amorphous HA and pure crystalline
HA in different weight ratios and then scanned within the
required region. However, the presence of other calcium
phosphate compounds, often having peaks which over-
lap with those of HA (see Fig. 6), makes it difficult to use
a calibration curve. Moreover, pure amorphous HA is
very difficult to prepare. The method used here gives
a quick and reasonable estimate of the degree of crystal-
linity. It is preferable to the method based on comparing
peak heights with those from the HA powder, since other
features (e.g. crystal size) can also affect the height of
a crystalline peak.
As shown in Fig. 6, the amount of other chemical
compounds present increases as the power input in-
creases (from 33 to 36 kW). There is also an increase, at
a given power level, when hydrogen, rather than helium,
is present in the plasma gas mixture. This may be due to
the greater chemical activity of the hydrogen, although
the increased chemical change in the HA when hydrogen
is present appears to be no greater than might have been
expected from the greater degree of heating of the par-
ticles. The coatings contain significant quantities of CaO
and Ca P O at high powers. Difficulties were encoun-
4 2 9
tered in identifying these compounds, since most of their
peaks overlap with each other in this 20—40° 2h region.
Construction of a calibration curve would be very com-
plicated. In one study by Toth et al. [46], it was shown
that no linear relationship exists between the HA/tri-
calcium phosphate (TCP) intensity ratio and the
HA/TCP mass ratio, even though their mass absorption
coefficients are similar. Therefore, no attempt was made
in the present study to estimate the percentage of various
compounds in the sprayed coatings.
The FTIR spectrum of the HA powder is shown in
Fig. 7a. The peaks corresponding to the OH~ stretching
mode at 3572 cm~1 and the OH~ flexural mode at
630 cm~1 can be seen easily. Moreover, absorption
bands corresponding to the aliphatic C—H bond [47] are
present, which suggests that the HA powder contained
some organic material. Also, bands corresponding to the
carbonate ion [47—49] can be seen, which suggests that
the powder may have a carbonate-substituted apatite
structure. In Fig. 7b, bands corresponding to the OH~
ion gradually disappear as the power level increases. This
reflects different degrees of dehydroxylation during
spraying, in which HA is converted [24, 25, 44, 50] to
oxyhydroxyapatite (OHA), with the formula Ca (PO )
10 46
(OH) O * (*"vacancy, x(1). This cannot be de-
2~2x x x
tected easily in the XRD spectrum. A broader peak in
sprayed coatings around 800—1200 cm~1 can be ex-
plained by the presence of other Ca—P compounds and
amorphous HA in that region [13, 27, 48].
The surface roughnesses of grit-blasted substrates and
HA coatings are shown in Table 4. For the same power
level (36 kW), HA coatings sprayed with H as the sec-
2
ondary plasma gas are rougher than those sprayed with
He. Thin coatings have lower surface roughnesses than
thick coatings, as expected.
3.2. Young+s modulus measurement
A calculation, using Eq. (4), for the geometry of the
composite beams in this study, showed that the neutral
axis was always within the tungsten substrate. It follows
that the entire HA coating would either be in compres-
sion or in tension during bending, while the tungsten
substrate would be partially in compression and partially
in tension. It was assumed that the Young’s modulus of
tungsten in tension is the same as that in compression. As
more data are available for coatings sprayed with the
Ar/H gas mixture, discussion will be focused on this type
2
of coating. From the data presented in Fig. 8, the Young’s
modulus in compression is more than six times higher
than that in tension for the coating sprayed at 30 kW, but
is only 50% higher for the coating sprayed at 42 kW. The
 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029 2023

Table 4
Surface roughness data for grit-blasted substrates and HA coatings sprayed with different input power levels and plasma gas mixtures

Specimen Plasma power (kW) Secondary plasma gas Coating thickness (lm) R (lm)
!
Ti—6Al—4V — — — 1.1$0.1
Mild steel — — — 1.2$0.1
HA coating 30 H 390 6.8$0.1
2
HA coating 33 H 400 7.0$0.1
2
HA coating 33 H 100 6.0$0.2
2
HA coating 36 H 440 8.1$0.1
2
HA coating 36 H 100 6.2$0.1
2
HA coating! 36 H 500 8.2$0.4
2
HA coating 36 He 400 6.7$0.1
HA coating 36 He 90 5.7$0.2
HA coating 40 He 400 6.1$0.2
HA coating 40 He 110 5.8$0.1
HA coating 42 H 500 6.1$0.2
2
! Argon cooling on the rear surface.

tension during loading. The larger difference between the

Fig. 8. Experimental data for Young’s moduli in tension, compression
and flexure, for HA coatings sprayed with different power levels and
plasma gas mixtures.
Young’s modulus of the coatings in compression shows
a sharp rise with power level initially and then levels off.
This correlates with the trend in porosity content. How-
ever, the Young’s modulus of the coating in tension
increases rapidly at first and then slowly with rise in
power level.
Two microstructural features account for the changes
in Young’s modulus: microcracks and pores. When the
coating is in compression, closure of microcracks occurs
quickly, so that pores have a dominant effect on the the
Young’s modulus. When the coating is in tension, how-
ever, both microcracks and pores affect Young’s
modulus. The flexural Young’s moduli (measured on
free-standing deposits sprayed at 33 and 36 kW) lie some-
what between these two curves. This is consistent, since
part of the coating is under compression and part under
Young’s moduli in tension and in compression for coat-
ings sprayed at lower powers might imply that such
coatings contain more microcracks. Since microcracking
is a major stress relaxation mechanism in ceramics [37]
during the splat solidification, this finding correlates with
the finding from residual stress modelling (see Section
3.4) that the quenching stress increases as the power level
increases.
No data are available in the literature for the Young’s
modulus of plasma sprayed HA coatings. However,
Eberhardt et al. [51] showed that only a value of less
than &5.5 GPa would be consistent with their results.
The use of much higher values (55 and 110 GPa) in their
prediction of residual stresses that resulted led to values
for the coating exceeding the tensile strength of sprayed
HA (190 MPa) and the interfacial shear strength
(6—7 MPa) of the HA/Ti system. Brittle fracture of the
coating and/or interfacial sliding would have occurred
had these residual stress levels been correct. All the
Young’s modulus values obtained in this study are less
than 5.5 GPa, and are much lower than the values for
bulk HA. (For example, reported [52, 53] values for
sintered HA are &35—90 GPa). The results reported here
are therefore consistent with the findings of Eberhardt
et al. [51].
3.3. Adhesion measurement
The single-edge notch-bend (SENB) test, which is
a pure mode I test, was used to measure the interfacial
fracture toughness in the HA/Ti—6Al—4V system. A rela-
tively thick layer of HA (&400 lm) was applied to the
Ti—6Al—4V bar, to avoid the penetration of adhesive
down to the interface. The results are shown in Table 5
and Fig. 9a. Since only a few specimens of each type were
available, a t-test was used to determine the statistical
2024 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029

Table 5
Interfacial adhesion data for various HA coatings on Ti—6Al—4V substrates, obtained from the SENB and tensile bond strength tests

Plasma power (kW) Secondary SENB test Tensile bond strength test
plasma gas
No. of specimens K (MPa m1@2) G (J m~2) No. of specimens p (MPa)
*# *# *
30 H 7 0.28$0.05 0.69$0.24 — —
2
33 H 12 0.50$0.08 2.20$0.68 4 5.97$0.78
2
36 H 9 0.83$0.10 5.97$1.42 4 6.91$1.06
2
36 He 6 0.51$0.08 2.28$0.70 4 5.85$0.90
36! H — — — 3 13.92$1.70
2
40 He — — — 3 4.15$0.63
42 H 7 1.08$0.11 10.07$2.03 — —
2
!After heat treatment at 600°C for 1 h.

energy, G ) increases with plasma power level. However,

Fig. 9. (a) Interfacial fracture toughnesses and (b) tensile bond strengths
of HA coatings on Ti—6Al—4V substrates, sprayed at different power
levels and plasma gas mixtures.
significance of the data. Statistical significance is implied
if P(0.05. Marginal significance (MS) is defined at
0.05(P(0.1 and non-significance (NS) is defined at
P'0.1.
In general, the interfacial toughness, K (and the close-
*#
ly related critical strain energy release rate, or fracture
*#
specimens sprayed at 36 and 42 kW showed no signifi-
cant difference in adhesion. Also, specimens sprayed at
36 kW with He had better adhesion than those sprayed
at 30 kW with H , despite other properties of these two
2
coatings being very similar. For all of the specimens,
fracture occurred mainly at the interface, with retention
of some HA splats on the substrate surface.
Only Filiaggi et al. [32] have reported in detail on the
interfacial fracture toughness of the HA/Ti—6Al—4V sys-
tem. They used a modified short bar (SB) chevron notch
specimen and obtained K in the range of
*#
0.60—1.41 MPa m1@2 — slightly higher than the values re-
ported here. (Both the SB and SENB tests are mode
I tests.) However, their substrate surface roughness was
two to four times greater than those in this study. The
HA powder particle size is thought to have a strong effect
on adhesion. Neither powder particle size nor the crystal-
linity of sprayed coatings were reported in the work of
Filiaggi et al. [32], which makes direct comparison diffi-
cult. The failure crack path in their specimens was the
same as that observed in this study. One advantage of
using the SENB test, compared with the SB test, is that
SENB specimens are easier to prepare. It would be very
useful to evaluate the interfacial fracture toughness under
mixed mode conditions, which is closer to the situation
commonly encountered in service. For this purpose,
a four-point bend delamination test [54, 55] was per-
formed. However, this was not successful, since the elastic
mismatch between the substrate and the coating is too
large. The stored elastic energy in the coating was insuffi-
cient to cause debonding and through-thickness cracking
was found to occur; even when the metal started to yield,
no debonding was observed.
The measured tensile bond strengths are shown in
Table 5 and Fig. 9b. The data suggest that there is no
difference in adhesive strength between specimens
sprayed at 36, 36 (He) and 33 kW. This contrasts with
the SENB test results. This has probably arisen because
the SENB test (with a pre-crack) is more sensitive than
the tensile test. The bond strengths obtained are similar
 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029 2025

Table 6
Quenching stress values and average residual stresses in the coatings at room temperature, predicted using the numerical process model

Power level (kW) Secondary plasma gas Quenching stress Average coating stress (MPa)
p (MPa)
2
Ti—6Al—4V substrate Mild steel substrate

30 H 15 21$2 19$1
2
33 H 20 36$3 26$3
2
36 He 13 18$1 17$1
36 H 22 38$4 28$4
2
42 H 23 41$6 29$4
2

Fig. 10. Quenching stress values found to give the best agreement
between the process model and experimental curvature history data, for
HA coatings sprayed with different power levels and plasma gas mix-
tures.

to those measured by Filiaggi et al. [32] and Gross and

Berndt [56]. However, these values are significantly
lower than those reported elsewhere [19, 21, 24, 57—59].
A possible explanation for this is that misleadingly high
strengths can be obtained if the adhesive used in prepar-
ing specimens for this test inadvertently penetrates to the
interface being studied. The danger of this is especially
marked when a thin coating is used. This might explain
the apparent increase in bond strength after heat treat-
ment, shown in Table 5. Since more cracks are usually
found after heat treatment, the adhesive penetrates to the Fig. 11. Comparison of experimental and predicted variations in
(a) substrate temperature and (b) curvature, for HA sprayed on to
interface more easily with such specimens. Ti—6Al—4V substrates at 30 kW.

3.4. Prediction of residual stresses

The quenching stress (obtained by fitting of experi-
mental and predicted curvature histories) was found to
increase with power level; this is shown in Table 6 and
Fig. 10. This trend is consistent with the differences
between the Young’s moduli in tension and in compres-
sion, which show fewer microcracks for those specimens
sprayed at higher powers. Since microcracking is be-
lieved to be the main stress relaxation mechanism for
ceramics when splats cool during spraying, the lower the
extent of microcracking, the higher the quenching stress.
Although the Young’s modulus of sprayed HA is similar
to that [60] of sprayed ZrO —8%Y O , the quenching
2 2 3
stress for the latter is much lower than that of the former.
Experimental and predicted temperature histories for
a specimen sprayed at 30 kW are shown in Fig. 11a and
the corresponding curvature histories are shown in
Fig. 11b. These plots show excellent agreement between
2026 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029
Fig. 12. Predicted residual stress distributions for specimens sprayed at
(a) 30 kW on mild steel and (b) 36 kW on Ti—6Al—4V.
predicted and observed plots. The residual stress distri-
bution predicted by the numerical process model is
shown in Fig. 12a. The stress level in the coating is very
uniform across the thickness in this case. As the power
level increases, the agreement becomes inferior for those
specimens sprayed on Ti—6Al—4V. This is mainly due to
the plasma momentum effect at high temperatures, caus-
ing the Ti—6Al—4V to creep (preferentially near the base
of the specimens, where the stresses from this source are
the highest), since maximum temperatures are higher for
specimens sprayed at higher powers. This plasma mo-
mentum effect cannot readily be taken into account in
the process model. Therefore, only those specimens
sprayed on mild steel were used at high powers to predict
the values of the quenching stress, which should be inde-
pendent of the substrate material [37].
Figure 12b shows that creep has occurred in the sub-
strate for the specimen sprayed at 36 kW. Coating stress
levels would have been higher if creep were not allowed
in the model, since the thermal expansivity of the coating
is greater than that of the Ti—6Al—4V. At higher powers,
the variation of stress in the through-thickness direction
is slightly larger, since the Young’s modulus of the coat-
ing becomes higher. Average coating stresses are listed in
Table 6, with the$sign indicating the range between
maximum and minimum stresses in the coatings.
These coating stress values are much lower than those
reported by Brown et al. [61] using X-ray diffraction.
They reported the residual stress level for air plasma
sprayed HA to be greater than 200 MPa. This seems
unlikely, since the coating has such a low Young’s
modulus. Moreover, the flexural strength of pure sintered
HA reported elsewhere [45] (&39 MPa) is much less
than 200 MPa. This implies that the HA coating simply
cannot bear such a high stress. It is widely recognised
that it is very difficult to obtain accurate residual stress
data for coatings of this type using X-ray diffraction,
partly because of the high surface roughness.
4. Conclusions
The following conclusions can be drawn from this
work.
(1) HA coatings have been deposited by vacuum plasma
spraying on to Ti—6Al—4V, tungsten and mild steel
substrates, in order to study mechanical character-
istics. The HA powder used for this work was rela-
tively coarse.
(2) As the plasma power level is increased, the crystal-
linity and OH~ ion content of the coatings decrease,
while the amount of non-HA calcium phosphate
compounds present increases. These changes are at-
tributable to an increased degree of particle melting.
The melted material freezes to an amorphous phase.
(3) As the plasma power level is increased, the porosity
level and extent of microcracking in the coatings
decrease. These are associated increases in the
Young’s modulus, quenching stress and residual
stress in the coating and an increase in interfacial
adhesion.
(4) For a given plasma power level, melting of the HA
powder is more efficient if hydrogen is used as a sec-
ondary plasma gas, rather than helium. The presence
of the hydrogen may promote rather more chemical
reaction in the HA than occurs with only inert
plasma gases, although the effect is apparently not
very strong.
(5) The absolute values of Young’s modulus obtained in
this work are relatively low ((6 GPa). This is consis-
tent with the observed presence of extensive micro-
cracking and high-porosity levels. Furthermore, the
Young’s modulus is considerably lower in tension
than in compression. This is attributed to the opening
up of microcracks under tensile loading.
(6) Predicted residual stress levels in HA coatings on
massive Ti—6Al—4V substrates at room temperature
 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029
are also relatively low (&20—40 MPa tensile). This is
largely attributable to the low-coating stiffness. These
stresses will always be tensile (unless the substrate is
held below room temperature during spraying), since
the thermal expansivity of HA is above that of
Ti—6Al—4V.
(7) While the residual stress levels are low, they could
well be significant in practical terms. For example,
they may help to promote through-thickness crack-
ing. Furthermore, in view of the relatively low inter-
facial toughness commonly observed in this system
(see below), the energy released when they are relaxed
may be sufficient to promote interfacial debonding,
particularly if the coating is thick. Therefore, it may
be worthwhile to minimize the substrate temperature
during spraying, which will tend to reduce the resid-
ual stress level.
(8) Interfacial adhesion has been characterised using the
single-edge notch-bend (SENB) test and the tensile
bond strength test. The latter is considered much less
reliable than the former, since it is sensitive to the
presence of flaws at the edge of the specimen and to
slight misalignment of the loading system. The SENB
test results indicated that the interfacial toughness
increased significantly with increasing power level,
whereas tensile bond strength data were inconclusive.
An apparent increase in bond strength on heat treat-
ing the specimen may have been due in large part to
penetration of epoxy adhesive to the interface, as
a consequence of a higher level of cracking in the
specimen. The interfacial toughness values obtained
are all relatively low (up to &10 J m~2). The tough-
ness is limited by a tendency for debonding to occur
within the HA coating, close to the interface, which is
facilitated by the network of microcracks.
Acknowledgements
Financial support for one of us (YCT) has been sup-
plied by Commonwealth Scholarship Commission in the
United Kingdom and by Howmedica.
References
[1] Geesink RGT, Groot KD, Klein CPAT. Bonding of bone to
apatite-coated implants. J Bone Joint Surg 1988;70-B:17—22.
[2] Oonishi H, Yamamoto M, Ishimaru H, Tsuji E, Kushitani S,
Aono M, Ukon Y. The effect of hydroxyapatite coating on bone
growth into porous titanium alloy implants. J Bone Joint Surg
1989;71-B:213—6.
[3] Poulmaire D, Ducos M, Setti Y, Hypolite MP. Development
of dental implants in titanium with HA coatings. In: Blum-
Sandmeier S, Eschnauer H, Huber P, Nicoll AR, editors.
Proc 2nd Plasma-Technik-Symp. Plasma-Technik AG 1991;3:
191—202.
2027
[4] Matejka D, Palka V, Brezovsky M, Sebo P, Infner I, Pastorok M.
Plasma coated composite intraosseous implants. In: Blum-
Sandmeier S, Eschnauer H, Huber P, Nicoll AR, editors.
Proc 2nd Plasma-Technik-Symp. Plasma-Technik AG 1991;3:
171—7.
[5] LeGeros RZ. Review paper: biodegradation and bioresorption of
calcium phosphate ceramics. Clinical Mater 1993;14:65—88.
[6] Dalton JE, Cook SD. In vivo mechanical and histological charac-
teristics of HA-coated implants vary with coating vendors. J Bio-
med Mater Res 1995;29:239—45.
[7] Klein CPAT, Patsa P, Wolke JGC, Blieck-Hogervorst JMAd,
Groot Kd. Long-term in vivo study of plasma-sprayed coatings
on titanium alloys of tetra-calcium phosphate, hydroxyapatite
and a-tricalcium phosphate. Biomaterials 1994;15:146—50.
[8] Joshi SV, Srivastava MP, Pal A, Pal S. Plasma spraying of
biologically derived hydroxyapatite on implantable materials.
J Mater Sci Mater Med 1993;4:25—255.
[9] Lacefield WR. Hydroxyapatite coating. Ann NY Acad Sci
1988;523:72—80.
[10] Li TT, Lee JH, Kobayashi T, Aoki H. Hydroxyapatite coating by
dipping method, and bone bonding strength. J Mater Sci Mater
Med 1996;7:355—7.
[11] Ducheyne P, Raemdonck WV, Heughebaert JC, Heughebaert M.
Structural analysis of hydroxyapatite coating on titanium. Bio-
materials 1986;1986:97—103.
[12] Raemdonck WV, Ducheyne P, Meester PD. Auger electron spec-
troscopic analysis of hydroxyapatite coating on titanium. J Amer
Ceram Soc 1984;63:381—4.
[13] Ong JL, Lucas LC. Post-deposition heat treatments for ion beam
sputter deposited calcium phosphate coatings. Biomaterials
1994;15:337—41.
[14] Yoshinari M, Ohtsuka Y, Derand T. Thin Hydroxyapatite coat-
ing produced by the ion beam dynamic mixing method. Bio-
materials 1994;15:529—35.
[15] Berndt CC, Haddad GN, Farmer AJD, Gross KA. Thermal
spraying for bioceramic applications. Mater Forum
1990;14:161—73.
[16] Bortz SA, Onesto EJ. Flame-sprayed bioceramics. Amer Ceram
Soc Bull 1975;52:898.
[17] Weinlaender MJB III, Kenney EB, Moy PK. Raman microprobe
investigation of the calcium phosphate phases of three commeri-
cally available plasma-flame-sprayed hydroxyapatite-coated den-
tal implants. J Mater Sci Mater Med 1992;3:397—401.
[18] Haman JD, Lucas LC, Crawmer D. Characterization of high
velocity oxy-fuel combustion sprayed hydroxyapatite. Bio-
materials 1995;16:226—37.
[19] Wolke JGC, Groot Kd, Kraak TG, Herlaar W, Blieck-Hoger-
vorst JMAd. The characterization of hydroxylapatite coatings
sprayed with VPS, APS and DJ systems. In: Bernecki TF, editor.
Proc Thermal Spray Conf ASM Int 1991:481—90.
[20] Lin JHC, Liu ML, Ju CP. Structure and properties of hy-
droxyapatite-bioactive glass composites plasma sprayed on
Ti—6Al—4V. J Mater Sci Mater Med 1994;5:279—83.
[21] Brossa F, Cigada A, Chiesa R, Paracchini L, Consonni C. Ad-
hesion properties of plasma sprayed hydroxyapatite coatings for
orthopaedic prostheses. Biomed Mater Eng 1993;3:127—36.
[22] Khur KA, Cheang P. Characterization of plasma sprayed hy-
droxyapatite powders and coatings. In: Berndt CC, Bernecki TF,
editors. National Thermal Spray Conf ASM Int 1993:347—52.
[23] Gross KA, Berndt CC. In vitro testing of plasma-sprayed hy-
droxyapatite coatings. J Mater Sci Mater Med 1994;5:219—24.
[24] Wang BC, Chang E, Yang CY, Tu D, Tsai CH. Characteristics
and osteoconductivity of three different plasma-sprayed hy-
droxyapatite coatings. Surf Coat Technol 1993;58:107—17.
[25] Radin SR, Ducheyne P. Plasma spraying induced changes of
calcium phosphate ceramic characteristics and the effect on
in vitro stability. J Mater Sci Mater Med 1992;3:33—42.
2028 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029
[26] Salsbury RL. Quality control of hydroxyapatite coatings: purity
and crystallinity determinations. In: Bernecki TF, editor. Nation-
al Thermal Spray Conf ASM Int 1991:471—3.
[27] Weng J, Liu XG, Li XD, Zhang XD. Intrinsic factors of apatite
influencing its amorphization during plasma-spray coating. Bio-
materials 1995;16:39—44.
[28] Berndt CC, Gross KA. Characteristics of hydroxylapatite bio-
coatings. In: Berndt CC, editor. Proc Int Thermal Spray Conf
Exposition ASM 1992:465—70.
[29] McPherson R, Gane N, Bastow TJ. Structural characterisation of
plasma sprayed hydroxylapatite coatings. J Mater Sci Mater Med
1995;6:327—34.
[30] Inadome T, Hayashi K, Nakashima Y, Tsumura H, Sugioka Y.
Comparison of bone-implant interface shear strength of hy-
droxyapatite-coated and alumina-coated metal implants. J Bio-
med Mater Res 1995;29:19—24.
[31] Hayashi K, Inadome T, Tsumura H, Nakashima Y, Sugioka Y.
Effect of surface roughness of hydroxyapatite-coated titanium on
the bone-implant interface shear strength. Biomaterials 1994;
15:1187—91.
[32] Filliaggi MJ, Coombs NA, Pilliar RM. Characterization of the
interface in the plasma-sprayed HA coating/Ti—6Al—4V implant
System. J Biomed Mater Res 1991;25:1211—29.
[33] Evans SL, Lawes KR, Gregson PJ. Layered, adhesively bonded
hydroxyapatite coatings for orthopaedic implants. J Mater Sci
Mater Med 1994;5:495—9.
[34] Furness JAG. Thermal cycling creep of aluminium-based com-
posites. PhD Thesis. University of Cambridge, 1991.
[35] Timoshenko SP, Gere, JM. Mechanics of materials. D. Van Nos-
trand Company, 1972.
[36] Kleer G, Schonholz R, Doll W, Sturlese S, Zacchetti N. Interface
crack resistance of zirconia base thermal barrier coatings. In:
Vincenzini P, editor. 7th CIMTEC World Ceramics Congress.
Elsevier, 1990:329—38.
[37] Kuroda S, Clyne TW. The quenching stress in thermally sprayed
coatings. Thin Solid Films 1991;200:49—66.
[38] Kuroda S, Clyne TW. The origin and quantification of the
quenching stress associated with splat cooling during spray de-
position. In: Eschenauer H, Huber P, Nicoll AR, Sandmeier S,
edotors. 2nd Plasma Technik Symposium. Plasma Technik 1991;
3:273—84.
[39] Kuroda S, Fukushima T, Kitahara S. Generation mechanisms
of residual-stresses in plasma-sprayed coatings. Vacuum 1990;41:
1297—9.
[40] Gill SC, Clyne TW. Investigation of residual stress generation
during thermal spraying by continuous curvature measurement.
Thin Solid Films 1994;250:172—80.
[41] Gill SC, Clyne TW. Thermomechanical modelling of the develop-
ment of residual stress during thermal spraying. In: Eschenauer H,
Huber P, Nicoll AR, Sandmeier S, editors. 2nd Plasma Technik
Sym. Plasma Technik 1991;3:227—38.
[42] Gill SC, Clyne TW. Residual stress modelling and characterisa-
tion of thermally sprayed ceramic coatings. In: Vincenzini P edi-
tor. High performance ceramic films and coatings. Amsterdam:
Elsevier, 1991:339—52.
[43] Gill SC, Clyne TW. Property data evaluation for the modelling
of residual stress development during vacuum plasma spray
deposition. In: Exner H, editor. 1st European Conf on Adv
Mats Procs (Euromat ’89). Deutsch Gesell f Metallk 1990;1:
1221—30.
[44] Kijima T, Tsutsumi M. Preparation and thermal properties of
dense polycrystalline oxyhydroxyapatite. J Amer Ceram Soc
1979;62:455—60.
[45] Chaki TK, Wang PE. Densification and strengthening of
silver-reinforced hydroxyapatite-matrix composite pre-
pared by sintering. J Mater Sci Mater Mater Med 1994;5:
533—42.
[46] Toth JM, Hirthe WM, Hubbard WG, Brantley WA, Lynch KL.
Determination of the ratio of HA/TCP mixtures by X-ray diffrac-
tion. J App Biomater 1991;2:37—40.
[47] Socrates G. Infrared characteristic group frequencies: table and
charts. New York: Wiley, 1994.
[48] Whitehead RY, Lucas LC, Lacefield WR. The effect of dissolution
on plasma sprayed hydroxylapatite coatings on titanium. Clin
Mater 1993;12:31—9.
[49] Osaka A, Miura Y, Takeuchi K, Asada M, Takahashi, K. Calcium
apatite prepared from calcium hydroxide and orthophosphoric
acid. J Mater Sci Mater Med 1991;2:51—5.
[50] Brendel T, Engel A, Russel C. Hydroxyapatite coatings by a poly-
meric route. J Mater Sci Mater Med 1992;3:175—9.
[51] Eberhardt AW, Zhou C, Rigney ED. Bending and thermal stres-
ses in a fatigue experiment of hydroxyapatite coated titanium
rods. In: Berndt CC, Sampath S, editors. Proc 7th National
Thermal Spray Conf ASM Int 1994:165—9.
[52] Jarcho M, Bolen CH, Thomas MB, Bobick J, Kay JF, Doremus
RH. Hydroxylapatite synthesis and characterization in dense
polycrystalline form. J Mater Sci 1976;11:2027—35.
[53] Akao M, Aoki H, Kato K. Mechanical properties of sintered
hydroxyapatite for prosthetic applications. J Mater Sci 1981;
16:809—12.
[54] Howard SJ, Tsui YC, Clyne TW. The effect of residual stresses
on the debonding of coatings. Part I. A model for delamin-
ation at a bimaterial interface. Acta Metall Mater 1994;42:
2823—36.
[55] Charalambides PG, Cao HC, Lund J, Evans AG. Development of
a test method for measuring the mixed mode fracture resistance of
bimaterial interfaces. Mech Mater 1990;8:269—83.
[56] Gross KA, Berndt CC. Optimization of spraying parameters for
hydroxyapatite. In: Blum-Sandmeier S, Eschnauer H, Huber P,
Nicoll AR, editors. Proc 2nd Plasma-Technik-Symp. Plasma-
Technik AG 1991;3:159—70.
[57] Brandon JR, Taylor R. Phase stability of zirconia-based thermal
barrier coatings. Part I. Zirconia-yttria alloys. Surf Coat Technol
1991;46:75—90.
[58] Yankee SJ, Pletka BJ, Salsbury RL. Quality control of hy-
droxylapatite coatings: the surface preparation stage. In:
Bernecki TF, editor. Proc Int Thermal Spray Conf ASM 1991:
475—9.
[59] Rouquet N, Frayssinet P, Bonel G, Hanker JS, Hardy D,
Guilhem A. Biochemical effects of calcium phosphate hy-
droxylapatite (HA) coatings on orthopaedic biomaterials. In:
Blum-Sandmeier S, Eschnauer H, Huber P, Nicoll AR, editors.
Proceedings of the 2nd Plasma-Technik-Symposium, Plasma-
Technik AG 1991;3:203—9.
[60] Tsui YC, Clyne TW. Adhesion of thermal barrier coating
systems and incorporation of an oxidation barrier layer. In:
Berndt CC, editor. Thermal spray: practical solutions for
engineering problems. ASM, Ohio: Materials Park, 1996:
275—84.
[61] Brown SR, Turner IG, Reiter H. Residual stress measurement in
thermal sprayed hydroxyapatite coatings. J Mater Sci Mater Med
1994;5:756—9.
[62] Clyne TW, Tsui YC. The effect of intermediate layers on residual
stress distributions and debonding of sprayed thermal barrier
coatings. In: Ilschner B, editor. 3rd Int Symp on Structural and
Functional Gradient Materials. PPUR 1994:129—36.
[63] Karapet’yants MK, Karapet’yants ML. Handbook of thermo-
dynamic constants of inorganic and organic compounds. London:
Ann Arbor — Humphrey Science Publishers, 1970.
[64] Parker VB, Wagman DD, Evans WH. Selected values of chemical
thermodynamic properties: table for the alkaline earth elements.
United States Department of Commerce, National Bureau of
Standards, 1971.
[65] Willmann G. Mat Werkstofftechn 1992;23:107.
 Y.C. Tsui et al. / Biomaterials 19 (1998) 2015—2029 2029

[66] Maruno S, Hayashi K, Sumi Y, Wang YF, Iwata H. CRC Hand-
book of bioactive ceramics. Yamamuro T, Hench LL, Wilson J,
editors. Boca Raton: CRC Press, 1991:187—93.
[67] Tsui YC, Howard SJ, Clyne TW. Application of a model for the
effect of residual stresses on debonding of coatings under applied
loads. In: Vincenzini P, editor. Advances in inorganic films and
coatings. Techna Srl 1995:19—26.
[68] Halouani R, Bernache-Assolant D, Champion E. Ababou A.
Microstructure and related mechanical properties of hot
pressed hydroxyapatite ceramics. J Mater Sci Mater Med 1994;
5:563—8.
[69] Best S, Bonfield W. Processing behaviour of hydroxyapatite pow-
ders with contrasting morphology. J Mater Sci Mater Med
1994;5:516—21.