Biomaterials 25 (2004) 34533462

Effect of the particle size on the micro and nanostructural features of

a calcium phosphate cement: a kinetic analysis
M.P. Ginebra*, F.C.M. Driessens, J.A. Planell
Research Centre in Biomedical Engineering, Biomaterials Division, Department of Materials Science and Metallurgy, Technical University of Catalonia
(UPC), Av. Diagonal 647, E08028 Barcelona, Spain
Received 21 February 2003; accepted 12 October 2003
Abstract
The aim of this work is to investigate the possibility of controlling the nal micro and nanostructural features of a calcium
phosphate cement by modifying the particle size of the starting powder, and to study the effect of this parameter on the kinetics of
the setting reaction. The development of calcium phosphate materials with tailored structures at the micro and nanoscale levels
could allow the modulation of some specic responses in biologic phenomena such as protein adsorption and cell adhesion, which
strongly depend on the nano-sized roughness of the interface.
It is shown that the higher specic surface, produced by the reduction of the particle size of the powder, strongly accelerates the
hydrolysis of the a-TCP into calcium-decient hydroxyapatite. The higher degree of supersaturation attained in the solution favours
the nucleation of smaller crystals. Thus, by increasing the specic surface of the starting powder in a factor of 5, the size of the
precipitated crystals is strongly reduced, and the specic surface of the set cement increases by a factor of 2. The reduction of the
particle size produces a substantial decrease of the setting time and accelerates the hardening of the cement without signicantly
affecting the nal strength attained. The mechanical strength achieved by the cement cannot be univocally related to the degree of
reaction, without considering the microstructural features.
r 2003 Elsevier Ltd. All rights reserved.
Keywords: Calcium phosphate cement; Bone tissue engineering; Microstructure; Mechanical properties
1. Introduction
Recent studies have put forward the signicant role
that both microstructure and nanostructure of surfaces
and interfaces play in biological phenomena such as
protein adsorption and bone cell adhesion in different
materials [1], and specically in hydroxyapatite ceramics
[2]. These ndings suggest that the capability to process
materials with tailored structures at the micro and
nanoscale level provides the possibility to elicit specic
responses from surrounding cells and tissues. This is
indeed relevant to optimise the in vivo response of
materials, but it is even more crucial to design adequate
substrates and scaffolds for bone tissue engineering,
where the cell response can be more sensitive to the
surface material properties.
The applications of calcium phosphate cements as
materials for bone regeneration have been increasing in
different elds such as orthopaedic surgery, dentistry,
maxillofacial surgery and reconstructive surgery [3,4]
since they were discovered by Brown and Chow in 1983
[5,6]. Moreover, they are also being used as candidate
materials for the development of scaffolds for bone
tissue engineering [7,8], as an alternative to the tradi-
tional porous-sintered ceramic scaffolds. This approach
takes advantage of the fact that the nal product is a low
temperature precipitated hydroxyapatite, more reactive
than the sintered one, and more similar to the biological
apatites.
In this context, the possibility to control the
morphology of the interface of the calcium phosphate
cements both at a microstructural and nanostructural
level is a crucial issue.
The chemical reactions that take place during the
setting of a calcium phosphate cement depend on its
chemical composition. However, in general, it can be
stated that the setting and hardening of the cement paste
is the result of dissolution and precipitation processes,
and the particle size of the powder phase is a key factor
ARTICLE IN PRESS
*Corresponding author. Fax: +34-93-4016706.
E-mail address: maria.pau.ginebra@upc.es (M.P. Ginebra).
0142-9612/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biomaterials.2003.10.049
that must be considered when the properties of the
cement are optimised to comply with the clinical
requirements of a given application.
Indeed, the effect of the particle size of the cement
powder on the setting and hardening properties has been
extensively studied in other hydraulic cements, such as
Portland cement [912]. As a matter of fact, the well-
known fast setting Portland cements have as one of their
characteristics the small particle size of their powder [9].
In the eld of calcium phosphate cements, some
authors have studied the inuence of the particle size of
the reactants on the nal mechanical properties of some
cements, such as dicalcium phosphate (DCP)- and
tetracalcium phosphate (TTCP)-based cements [13] or
dicalcium phosphate dihydrate (DCPD)- and TTCP-
based cements [14]. Also, the effect of this parameter on
the pH, workability and setting times of a DCP-a-
tricalcium phosphate cement has been studied [15].
However, little attention has been paid to the effect of
this processing parameter on the nal micro and
nanostructural features of the cement and on the kinetic
aspects of the setting reaction.
The aim of this work is to verify the possibility to
control the nal micro and nanostructure of a calcium
phosphate cement through the particle size of the initial
powder phase. The cement studied has a-tricalcium
phosphate (a-TCP) as the only active reactant. In
previous studies, it has been shown that the setting
and hardening of this cementing system is the result of
the hydrolysis of a-TCP according to the following
reaction:
3Ca
3
PO
4

2
H
2
O-Ca
9
HPO
4
PO
4

5
OH: 1
The progressive dissolution of the a-TCP particles
and the simultaneous formation of an entangled net-
work of precipitated calcium-decient hydroxyapatite
(CDHA) crystals determine the setting and hardening of
the cement [16,17]. Many of the commercial calcium
phosphate cements developed in the last years have a-
TCP as the main reactant [3].
In this work, the particle size of the powder phase is
introduced as a variable, in order to determine its
inuence on the nal structure of the cement at the
micro and nanoscale range. The reaction kinetics and
the underlying physico-chemical mechanisms responsi-
ble for the nal structure and properties of the cement
are investigated.
2. Materials and methods
2.1. Cement preparation
The cement powder consisted of 85% a-TCP and
15% b-TCP, according to quantitative X-ray diffraction
(XRD) analysis [16]. It was prepared by heating in air at
1300

C for 15 h and then quenching to room tempera-

ture in air an appropriate mixture of CaHPO
4
(Merck
02144) and CaCO
3
(Merck 2076).
Two cement powders with different particle sizes were
studied. To obtain them the a-TCP was milled for two
different lengths of time (30 and 360 min) in an agate
ball mill (Pulverisette 6, Fritsch GmbH) at two different
rotating speeds (525 and 650 rpm, respectively). In both
cements, a 2 wt% of precipitated hydroxyapatite (Merck
2143) was added as a seed in the powder phase.
The liquid phase consisted of an aqueous solution 2.5
wt% of disodium hydrogen phosphate, Na
2
HPO
4
(Merck 6586). The liquid-to-powder ratio used was
0.32 ml/g.
2.2. Powder characterisation
The specic surface of the two TCP powders used was
measured by nitrogen adsorption. The particle size
distribution was measured by means of laser diffraction.
The powder had been previously dispersed in ethanol in
an ultrasonic bath.
Hereafter, the ne powder will be coded as Powder F,
and the coarse powder as Powder C, and the cements
prepared by mixing each powder with the liquid phase
as Cement F and Cement C, respectively.
2.3. Setting times and cohesion time
Powder and liquid were mixed in a mortar for about
1 min, and the initial and nal setting times of the
cement paste were determined with Gillmore needles
according to the C266-ASTM standard [18]. The
cohesion time, dened as the minimum time after which
the cement did not suffer disintegration when immersed
in Ringers solution, was determined by visual inspec-
tion [19].
2.4. Cement hardening
To evaluate the strength development, cylindrical
specimens with a diameter of 6 mm and a height of
12 mm were moulded. After 15 min, they were immersed
in Ringers solution at 37

C for periods of 1, 2, 4, 8, 16,

32, 64 and 360 h. Eight samples were prepared for each
cement and period. After the immersion, the samples
were polished, removed from the mould and tested
under compression at a cross-head speed or 1 mm/min in
a Universal Electromechanical Testing Machine.
2.5. Reaction kinetics
Immediately after they had been tested in compres-
sion, the specimens were quenched in acetone and dried
to stop the reaction. Two of the specimens were kept for
microstructural characterisation, and the others were
ARTICLE IN PRESS
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3454
crushed in order to perform XRD. The extent of
conversion or degree of reaction in function of the
reaction time was calculated as described in a previous
paper from XRD results, on the basis of the peak
intensity variation by means of the external standard
method [16].
2.6. Microstructural development and specic surface
The microstructural development was investigated by
scanning electron microscopy (SEM) on the broken
surfaces of the two specimens previously tested in
compression and kept to this end. In these specimens,
the reaction was stopped after different reaction times
(0, 1, 2, 4, 8, 16, 32, 64 and 360 h) as described above.
The fracture surfaces were gold-covered previous to
SEM observation.
The specic surface of the two cements after 360 h in
Ringers solution was measured by nitrogen adsorption.
3. Results
3.1. Powder characterisation
The particle-size distribution of the two starting
powders, coarse (C) and ne (F), as obtained by laser
diffraction is shown in Fig. 1. The main parameters that
characterise the particle-size distribution and the specic
surface area are shown in Table 1. The powder
morphology as observed by SEM is shown in Fig. 2.
3.2. Workability and setting parameters
Both cements gave a homogeneous paste, with
acceptable wettability of the powder by the liquid.
However, as could be expected, the paste was thicker for
the cement that had a smaller particle size of the two
(Cement F).
The results obtained for the cohesion time (C-T), the
initial setting time (I-T) and the nal setting time (F-T)
are given in Fig. 3. It can be observed that a reduction of
the particle size produced a substantial decrease of both
the setting times and the cohesion time. With a
procedure as simple as a longer grinding of the powder,
the setting of the cement was accelerated in a factor
between 2 (I-T) and 5 (F-T), and a fast setting cement
was obtained.
3.3. Compressive strength
The strength development of both cements is shown
in Fig. 4. In both cases, the evolution of the compressive
strength with the reaction time was tted to an
exponential function. The value for C
s
(compressive
strength at saturation) was slightly higher for the Fine
Cement, that is 41 (71.8) vs. 34.6 (71.0) MPa, the
value for the Coarse Cement. A more signicant
difference was found in the hardening rate of the two
cements, as reected by the time constant t obtained for
the different setting reactions, which indicates the time
when the cement reaches a 63% of the compressive
strength at saturation (C
s
). Thus, whereas the Coarse
Cement had a value of t 28:2 (72.2) h, this value was
ARTICLE IN PRESS
0.1 1 10 100 1000
0
10
20
30
40
50
60
70
80
90
100
size (m)
0
2
4
6
8
10
12
powder F
d
i
f
f

v
o
l
u
m
e

(
%
)
0.1 1 10 100 1000
0
10
20
30
40
50
60
70
80
90
100
c
u
m

v
o
l
u
m
e

(
%
)

c
u
m

v
o
l
u
m
e

(
%
)

0
2
4
6
8
10
12
powder C
d
i
f
f

v
o
l
u
m
e

(
%
)
Fig. 1. Particle-size distribution of the two starting powders: coarse
(Powder C) and ne (Powder F).
Table 1
Some particle-size parameters and specic surface of the two a-TCP powders used
a-TCP D(10) (mm)
a
Median Size,
D(50) (mm)
D(90) (mm)
b
Particle size
volume average,
M
v
(mm)
Particle size
number average,
M
n
(mm)
Specic surface (m
2
/g)
Powder C 2.12 10.88 25.12 12.95 2.07 0.5470.04
Powder F 1.39 2.22 11.69 5.06 1.55 2.7370.08
a
D(10) means 10% of the powder particles are smaller than this value.
b
D(90) means 90% of the powder particles are smaller than this value.
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3455
reduced to t 6:1 (70.9) h for the Fine Cement, which
is a reduction in a factor of 4.5 by reducing the particle
size of the cement.
3.4. Reaction kinetics
The phase evolution during the setting reaction in the
two cements studied is shown in Fig. 5, where the XRD
diagrams obtained after different reaction times are
shown.
The XRD diagrams for t 0 h correspond to the
unreacted powder phase of the two cements. Both could
be identied as a-TCP with approximately a 15 wt% b-
TCP formed during the quenching. A widening of the
peaks was clearly noticed in the powder phase of
Cement F, due to the crystallite size reduction caused
by the more severe grinding.
The evolution of the XRD diagrams with time
showed that the cement setting was the consequence of
a hydrolysis reaction, according to Eq. (1). Indeed, the
a-TCP was progressively transformed into CDHA,
while b-TCP remained unreacted, in agreement with
the results published in previous studies [16]. The XRD
diagrams showed that the disappearance of the a-TCP,
and the simultaneous formation of CDHA was much
faster in the Fine Cement than in the coarse one. Thus,
in the Fine Cement, the presence of an apatitic phase
could be clearly detected already after 2 h, and after 8 h
it was already the predominant phase. In contrast, in the
Coarse Cement, CDHA was hardly detected after 2 h
and after 8 h its presence was scarce. Although the
kinetics of the reaction was different, the transformation
of the a-TCP into CDHA was complete in the two
cements studied. Moreover, it was observed that the
CDHA formed in Cement F had a lower crystallinity
than that formed in Cement C, as indicated by the lower
intensity and widening of the XRD peaks. The average
size of the diffraction domains responsible for the Bragg
reection of the (0 0 2) planes was determined using the
well-known Scherrer formula [20]. The values obtained
for the (0 0 2) diffraction plane were 40 and 54 nm for
the Fine and Coarse cements, respectively.
The quantication of the amounts of the different
phases present, calculated on the basis of the peak
intensity variation using the external standard method,
ARTICLE IN PRESS
C-T I-T F-T
0
10
20
30
40
T
i
m
e

(
m
i
n
)
CementC
CementF
Fig. 3. Cohesion time (C-T), initial setting time (I-T) and nal setting
time (F-T) for the cements with coarse (cement C) and ne (cement F)
particle size of the powder phase.
0 50 100 150 200 250 300 350 400
0
5
10
15
20
25
30
35
40
45
(*standard deviation between brackets)
C(t)= C
s
(1-exp (-t/))
Cement F: C
s
=40.1(1.8)MPa;
Cement C: C
s
=34.6(1.0)MPa: =28.2(2.2)h
=6.1(0.9)h
C
o
m
p
r
e
s
s
i
v
e

s
t
r
e
n
g
t
h

(
M
P
a
)
time (h)
Fig. 4. Evolution of the compressive strength (C) of the two cements
studied with time.
Fig. 2. Morphology of the two starting powders as observed by SEM:
(a) Powder C; (b) Powder F.
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3456
allowed to represent the degree of reaction R; as a
function of the reaction time, as shown in Fig. 6. An
exponential function was tted, being the degree of
reaction at saturation (R
s
) equal to 1 in both cases, since
the transformation of a-TCP into CDHA was complete.
The time constant t was four times smaller for the Fine
Cement (5.570.1 h) than for the Coarse one
(20.570.6 h). Therefore, as put in evidence previously
by the faster cement hardening, the reduction of the
particle size had a strong accelerating effect in the
hydrolysis of the cement powder.
3.5. Microstructural evolution and specic surface
The microstructure of the Cement C and Cement F
after 2, 8, 64 and 360 h of setting as observed by SEM is
shown in Figs. 7 and 8, respectively. It can be observed
that the a-TCP particles are much smaller in Fig. 8 than
in Fig. 7.
After 2 h of reaction (Figs. 7a and 8a), small crystals
surrounding the initial powder could be observed. After
8 h (Figs. 7b and 8b), the amount of precipitated crystals
increased in both cements, and two differential features
were detected between the two cements: the size of the
precipitated crystals was clearly smaller in the Fine
Cement, and also their topological distribution was
different. In the Coarse Cement, the precipitated crystals
grew forming shells which surrounded the a-TCP
particles, and some voids were located between these
shells. In contrast, in the Fine Cement, a denser net of
crystals lled the space between a-TCP particles, but
there was some free space between the reactant particles
and the precipitated crystals. This certainly will be of
relevance when the mechanisms controlling the reaction
are discussed.
After 64 h, the Coarse Cement presented a more
crystalline aspect (Fig. 7c). Some unreacted particles
were still detected. In contrast, the Fine Cement was
nearly transformed into CDHA, and a-TCP particles
were not detected (Fig. 8c).
After complete reaction (360 h, Figs. 7d and 8d), it
could be clearly observed that the crystallite size of the
precipitated CDHA was much smaller for the Fine
Cement than for the Coarse Cement. A micrograph of
the Fine Cement at higher magnication is shown in
Fig. 9. Another signicant difference observed between
the two cements was that, in the Fine Cement, the space
occupied by the a-TCP particles was not completely
lled by the small and compactly packed precipitated
crystals, giving rise to small voids or pores all over the
cement. This did not happen in the Coarse Cement, due
ARTICLE IN PRESS
20 30 40 50
t = 360h
t = 64h
t = 8h
t = 2h
t = 0h
-TCP
-TCP
CDHA
-TCP
-TCP
CDHA
I

(
c
t
s
)
I

(
c
t
s
)
2 2
20 30 40 50
t = 360h
t = 64h
t = 8h
t = 2h
t = 0h
Cement C Cement F
(a) (b)
Fig. 5. XRD diagrams of the two cements studied after different reaction times.
-50 0 50 100 150 200 250 300 350 400
0.0
0.2
0.4
0.6
0.8
1.0
R=R
s
(1-exp(-t/))
Cement C; R
s
= 1; = 20.5(0.6)h
Cement F; R
s
= 1; = 5.5(0.1)h
R

time (h)
Fig. 6. Degree of reaction (R) as a function of the reaction time for the
two cements studied.
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3457
ARTICLE IN PRESS
Fig. 7. Evolution of the microstructure with the time of reaction for the Coarse Cement (Cement C) as observed by SEM: (a) after 2 h of reaction; (b)
8 h of reaction; (c) 64 h of reaction; (d) 360 h of reaction.
Fig. 8. Evolution of the microstructure with the time of reaction for the Fine Cement (Cement F) as observed by SEM: (a) after 2 h of reaction; (b)
8 h of reaction; (c) 64 h of reaction; (d) 360 h of reaction.
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3458
probably to the larger size and low packing density of
the precipitated crystals.
The specic surface of the Coarse and Fine set
Cements, after 360 h reaction was 15.9670.03 and
30.3970.05 m
2
/g, respectively.
4. Discussion
In general, the results obtained showed that the
particle size is a key factor, which modies signicantly
the nal properties of the cement, and that specially
affects the kinetics of the chemical reaction and the
mechanical consolidation of the cement. Therefore, it is
a parameter that can be very useful in order to adjust the
behaviour of the cement to the specic clinical require-
ments of each different application. In many cases, the
optimisation of the particle size could avoid the need to
incorporate other additives to the calcium phosphate
cements, for instance polymeric substances, which in
many cases can worsen the mechanical properties of the
cement and in some instances can compromise the
excellent biocompatibility of the calcium phosphate
cements.
In this study, the powder phase of the cement was
grinded following two different milling protocols. It has
to be noted than, as could be expected in powders
grinded in the same mill, the dispersion of the particle
sizes obtained in the two cases was similar. As the effect
of the more severe milling, the number of particles with
a size under 3 mm was drastically increased. One of the
consequences of this reduction in the particle size was
the fact that the paste was thicker, and therefore less
injectable, for Cement F than for Cement C. This was
explained by the higher specic surface of the Fine
Cement, and the higher amount of water adsorbed on
the solid particles.
As it was expected, the reduction of the particle size
resulted in a cement with a much faster setting (Fig. 3).
This behaviour can be explained as the result of the
different interactions occurring in the cement paste.
Indeed, it has to be kept in mind that, in general, both
the chemical reaction that is taking place, and the
physical attractive forces between the cement particles
contribute to the cement setting. In the case of the Fine
Cement, due to the higher specic surface of the
reactants, the chemical reaction took place at a much
higher rate at the initial stages. Additionally, the smaller
the particles the stronger the electrostatic attractive
forces were [21]. The same arguments can explain also
the fact that the cohesion time was reduced when the
neness of the cement powder increased. Moreover, the
reduction of the cohesion time is in agreement with the
KozenyCarman equation, which relates the perme-
ability coefcient of a porous solid with the inverse of
the square of its specic surface [22].
The hardening of the cement was also accelerated
when the particle size of the cement powder was
reduced. Indeed, by increasing the specic surface of
the powder by a factor of 5, the strengthening rate of the
cement at the initial stages (C
N
=t) increased also by a
factor of 5. However, the nal compressive strength at
saturation was only slightly higher for the Fine Cement
than for the Coarse one. Physically, it is accepted that in
calcium phosphate cements, the cement hardening is due
to the entanglement of the crystals of the product phase,
and therefore the morphology and size of these crystals
should condition the nal strength reached by the
cement.
As reported in the previous section, both the specic
surface measurements and the microstructural study
showed that the CDHA crystals formed in the Fine
Cement were signicantly smaller than those formed in
the Coarse Cement. This can be explained by the higher
degree of supersaturation attained by the dissolution of
the powder with higher specic surface. Indeed, a higher
supersaturation in the solution favours the formation of
crystalline nuclei, giving rise to the precipitation of more
crystals. When the supersaturation is lower, less nuclei
precipitate and crystal growth is favoured. This result
means that the surface properties (roughness, specic
surface, porosity) of the interface between the material
and the biologic environment (in vivo or in vitro) can be
controlled by modifying the particle size of the initial
powder used in the cement. And, as it has been shown in
recent investigations, these surface properties, at the
micro and nanostructural level can have a very strong
inuence on the protein adsorption and therefore on the
osteoblast adhesion [1,2].
The reduction in crystal size was also assessed by
calculating the crystallite dimensions by the Scherrer
formula on the XRD patterns. However, the crystallite
dimensions obtained in this way were much smaller than
ARTICLE IN PRESS
Fig. 9. Final microstructure of the Fine Cement, at a higher
magnication.
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3459
those deduced from the SEM images. Several reasons
can explain this apparent disagreement: (i) in the
cements studied there is a wide distribution of crystal
sizes, and the smaller crystals cannot be observed by
SEM at the magnications employed. The size obtained
by the Scherrer formula is an average value; (ii) the
crystals that contribute to the peak widening are only
those below 200 nm approximately. This indicates that,
in the Fine Cement, the proportion of crystals below this
size increase.
Apparently, one would expect a higher strength the
smaller the entangled crystals are, since in small crystals
there are more contact points, and the porosity is lower.
In this sense, Otsuka et al. reported an increase of the
compressive strength of a calcium phosphate cement
when the specic surface of the precipitated phase
increased [14]. However, in our case, a more careful
observation of the microstructures explained the appar-
ent contradiction. In fact, in the Fine Cement, both the
better packaging of the ner initial powder and the
formation of smaller precipitated crystals should favour
a more compact microstructure. However, the precipi-
tated crystals were unable to ll the space occupied by
the reactant particles of a-TCP, due to the fact that
small crystals can ll the space more efciently, leaving a
smaller fraction of empty space between crystals, and
some cavities appeared, in the places where the a-TCP
particles were located (Fig. 8bd). This gave as a result
the appearance of many voids, as large as 110 mm,
which had a weakening effect. In the case of the Coarse
Cement, the precipitated crystals were larger and they
lled less efciently the space available, being less
densely packed. Therefore, they were able to ll the
space occupied by the reactants. Therefore, the micro-
structure was more homogeneous (Fig. 7c and d). This
could explain the fact that the nal strength did not
increase when the particle size of the cement powder was
reduced.
Therefore, it can be stated that the reduction of the
particle size accelerated the setting and hardening of the
cement without affecting signicantly the nal strength
attained. From a chemical point of view, the XRD
analysis showed that the underlying mechanism that
caused this effect was the acceleration of the hydrolysis
of the a-TCP. Indeed, chemically, the mechanism of
hardening consisted in the dissolution of the cement
powder, and the precipitation of a product phase, in this
case a calcium-decient hydroxyapatite. Both mechan-
isms, dissolution and precipitation, were accelerated
when the specic surface of the powder increased. Since
the specic surface increased, the dissolution was
favoured and the liquid phase around the solid particles
became more supersaturated with respect to phosphate
and calcium ions, favouring a more rapid precipitation
of the product phase. The rate of the reaction at the
initial stages (1/t) increased by a factor of 4 in the Fine
Cement as compared to the Coarse one. This value was
similar but somewhat lower to that found when the
initial strengthening rate of the two cements was
compared.
Indeed, the comparison between Figs. 4 and 6
suggested that a relation could be established between
the kinetics of the mechanical consolidation of the
cement and the chemical reaction taking place. In Fig.
10, the normalised strength (C=C
s
) is represented vs. the
normalised degree of reaction (R=R
s
; in our case is R
because R
s
1). In order to nd the best tting
equation to each group of experimental points, and to
nd the statistic signicance between groups, a dummy
variable multiple linear analysis was performed [23]. The
tted model was the following:
C=C
s
b
0
b
1
R b
2
f b
3
Rf ; 2
where C=C
s
was the normalised strength, R was the
degree of reaction, b
i
are the tted parameters and f is
the dummy variable, which gets 0 for Cement F and 1
for Cement C.
The tted equation was the following:
C=C
s
0:480 1:46 R 0:276 f 0:291Rf : 3
The statistics of the regression is summarised in Table 2.
All parameters were signicant (po0:05), and as
consequence, it could be stated that, although in the
two cements studied the strength acquisition depended
linearly on the degree of reaction, both the slope and the
intersect at the origin were statistically different. There-
fore, for a given chemical composition of the cement, it
is not possible to relate univocally the mechanical
strength achieved by the cement to the degree of
reaction of the same. Indeed, the particle size of the
cement powder has a signicant effect in this depen-
dence, which has to be taken into account.
ARTICLE IN PRESS
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
C/C
s
=K(R-R
0
)
Cement F: K = 1.46; R
0
= 0.33
Cement C: K = 1.17; R
0
= 0.17
C
/
C
s
R
Fig. 10. Normalised strength (C=C
s
) vs. normalised degree of reaction
(R=R
s
R; since R
s
1) for the two cements studied.
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3460
According to the above results, the tted equations
for each cement were:
Cement F : C=C
s
b
0
b
1
R 0:480 1:46 R
1:46R 0:33; 4
Cement C : C=C
s
b
0
b
2
b
1
b
3
R
0:204 1:17 R
1:17R 0:17: 5
When the tted equations were written as C=C
s

kR R
0
; R
0
represented the degree of reaction that the
cement has to surpass to achieve a detectable mechanical
strength. Below this value, the strength of the cement
was negligible. It could be observed that the parameter
R
0
was much higher for the Fine Cement than for the
Coarse one. Also the slope (k) was higher for the Fine
Cement, indicating that once R
0
was attained, the
increase of the compressive strength with the degree of
reaction was more marked than in the Coarse Cement.
These differences can be attributed to the different
microstructure of the precipitated phase in each case as
explained previously (size, shape and distribution of the
crystals formed).
In a previous study, a physico-chemical model was
put forward to explain the kinetics of the hydrolysis of
a-TCP in a cementing system, analogous to the one
studied in this work, but with a different particle-size
distribution [17]. From the temporal evolution of the
depth of reaction, calculated from the degree of reaction
data and the particle-size distribution, two rate-limiting
mechanisms were proposed: initially the surface area of
the reactants and, subsequently, the diffusion of the
liquid phase through the hydrated layer formed around
the reactants. These two stages were related to the
formation of a shell-like structure that separated the
reactants during the reaction [17].
The same calculations were performed with the two
cements studied in this work, in order to elucidate
whether the same mechanisms could be identied. Fig.
11 represents the temporal evolution of the depth of
reaction for the two cements studied. In the Coarse
Cement, the two controlling mechanisms could be
identied, in two reaction steps: in the rst one the
reaction was controlled by the available surface area for
dissolution of the reactants (d K
1
t), and in the second
one it was controlled by diffusion of the liquid through
the hydrated layer (d K
2
t
1=2
). The transition time was
around 20 h. However, in the Fine Cement, only one
mechanism was identied, since the dependence of the
depth of reaction with time was linear throughout the
whole reaction, as shown in Fig. 11. This would suggest
that the reaction was controlled only by the available
surface area for dissolution, and therefore that the layer
of products formed was not efcient to separate the
liquid from the solid reactant. In fact, as it has been
mentioned previously, it could be observed in the SEM
microstructures that the formation of precipitated
CDHA took place not directly on the reactant particles,
but in the volume between particles (Fig. 8b), due to the
higher supersaturation. A shell-like structure was not
observed, and the voids were formed around the a-TCP
particles and not between the shells of precipitated
crystals, as was the case in the Coarse Cement.
Consequently, the liquid and solid reactants would be
in contact along the whole reaction and the specic
surface of reaction would be the only rate-limiting
parameter.
5. Conclusion
The particle-size distribution is a variable that
controls in a very signicant way the kinetics of the
setting reaction of the cement. The crystallite size of the
nal product can be strongly reduced by increasing the
specic surface of the starting powder. This allows
developing calcium phosphate materials with tailored
structures at the micro and nanoscale levels, with the
aim to modulate some specic responses in biologic
phenomena such as protein adsorption and cell adhe-
sion, which strongly depend on the structure of the
interface at the nanometric range.
ARTICLE IN PRESS
Table 2
Statistics of the dummy variable multiple linear regression of C=C
s
vs.
R
Parameter Coefcient St.dev. t-Ratio p-Value
b
0
0.480 0.07 6.47 0.000
b
1
1.460 0.09 15.31 0.000
b
2
0.276 0.09 3.15 0.012
b
3
0.291 0.12 2.45 0.037
0
0 10
2
4
6
8
10
= 0.09t
= 0.14t
= 0.43t
1/2

CEMENT F
CEMENT C

m
)
Time (h)
Fig. 11. Depth of reaction (d) as a function of the reaction time for the
two cements studied.
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3461
Acknowledgements
The authors thank the Science and Technology
Spanish Ministry for funding this work through project
CICYT MAT2002-04297.
References
[1] Boyan BD, Sylvia VL, Liu Y, Sagun R, Cochran DL, Lohmann
CH, Dean DD, Schwartz Z. Surface roughness mediates its effects
on osteoblasts via protein kinase A and phospholipase
A
2
. Biomaterials 1999;20:230510.
[2] Webster TJ, Ergun C, Doremus RH, Siegel RW, Bizios R.
Specic proteins mediate enhanced osteoblast adhesion on
nanophase ceramics. J Biomed Mater Res 2000;51:47583.
[3] Bohner M. Physical and chemical aspects of calcium phosphates
used in spinal surgery. Eur Spine 2001;10:S11421.
[4] Driessens FCM, Boltong MG, Khairoun I, De Maeyer EAP,
Ginebra MP, Wenz R, Planell JA, Verbeeck RMH. Applied
aspects of calcium phosphate bone cement. In: Wise DL, Trantolo
DJ, Lewandrowski KU, Gresser JD, Cattaneo MV, editors.
Biomaterials engineering and devices; human applications, vol. 2:
Orthopaedic, dental and bone graft applications. Totowa (NJ):
Humana Press; 2000. p. 25360.
[5] Brown WE, Chow LC, U.S. patent 4,612,053 (1986).
[6] Chow LC. Development of self-setting calcium phosphate
cements. J Ceram Soc Japan (International Edition)
1991;99:92735.
[7] Bohner M. Calcium phosphate emulsions: possible applications.
Key Eng Mater 2001;192195:7658.
[8] Barralet JE, Grover L, Gaunt T, Wright AJ, Gibson IR.
Preparation of macroporous calcium phosphate cement tissue
engineering scaffold. Biomaterials 2002;23:306372.
[9] Lea FM. The chemistry of cement and concrete, 3rd ed. London:
Edward Arnold Publishers Ltd.; 1970.
[10] Higginson EC. The effect of cement neness on concrete. In:
Verbeck GJ, editor. Fineness of cement. Philadelphia, PA: ASTM
STP 473. American Society for Testing and Materials; 1970,
p. 7181.
[11] Brown PW. Effects of particle size distribution on the kinetics
of hydration of tricalcium silicate. J Am Ceram Soc 1989;72:
182932.
[12] Osbaeck B, Johansen V. Particle size distribution and rate of
strength development of Portland cement. J Am Ceram Soc
1989;72:197201.
[13] Sanin N, Takagi S, Chow LC, Matsuya S. Particle size effects on
pH and strength of calcium phosphate cement. J Dent Res 1991;
70: 567(abstract n. 2411).
[14] Otsuka M, Matsuda Y, Suwa Y, Fox JL, Higuchi WI. Effect of
particle size of metastable calcium phosphates on mechanical
strength of a novel self-setting bioactive calcium phosphate
cement. J Biomed Mater Res 1995;29:2532.
[15] Fernandez E, Gil FJ, Ginebra MP, Driessens FCM, Planell JA,
Best SM. Production and characterisation of new calcium
phosphate bone cements in the CaHPO
4
-a-Ca
3
(PO
4
)
2
system:
pH, workability and setting times. J Mater Sci Mater Med
1999;10:22330.
[16] Ginebra MP, Fern! andez E, Boltong MG, Driessens FCM,
Ginebra J, De Maeyer EAP, Verbeeck RMH, Planell JA. Setting
reaction and hardening of an apatitic calcium phosphate cement.
J Dent Res 1997;76:90512.
[17] Ginebra MP, Fern! andez E, Driessens FCM, Planell JA. Modeling
of the hydrolysis of a-tricalcium phosphate. J Amer Ceram Soc
1999;82:280812.
[18] Standard Test Method for time of setting of hydraulic cement
paste by Gillmore needles. ASTM C266-89. In: Annual book of
ASTM standards, vol. 04.01. Cement, lime, Gypsum. Philadel-
phia, PA: ASTM; 1993. p. 18991.
[19] Fern! andez E, Boltong MG, Ginebra MP, Driessens FCM,
Berm! udez O, Planell JA. Development of a method to measure
the period of swelling of calcium phosphate cements. J Mater Sci
Lett 1996;15:10045.
[20] Cullity BD, Stock SR. Elements of X-ray diffraction, 3rd ed.
Upper Saddle River, NJ, USA: Prentice-Hall; 2001.
[21] Powers TC. Estructura f!sica de la pasta de cemento Portland. In:
Taylor HFW, editor. La qu!mica de los cementos. Bilbao:
Ediciones Urmo; 1967. p. 44472.
[22] Carman PC. Flow of gases through porous media. New York:
Academic Press; 1956.
[23] Draper N, Smith W. Applied linear regression, 2nd ed.
New York: Wiley; 1981.
ARTICLE IN PRESS
M.P. Ginebra et al. / Biomaterials 25 (2004) 34533462 3462