ARTICLE IN PRESS

Journal of Physics and Chemistry of Solids 68 (2007) 431–437

www.elsevier.com/locate/jpcs

Computer simulation of stoichiometric hydroxyapatite:

Structure and substitutions
Nasser Y. Mostafaa,, Paul W. Brownb
a
Chemistry Department, Faculty of Science, Suez Canal University, Ismailia, Egypt
b
Materials Research Laboratory, The Pennsylvania State University, University Park, PA 16802
Received 19 February 2006; received in revised form 13 November 2006; accepted 9 December 2006

Abstract

A force field for the modeling of hydroxyapatite; Ca10(PO4)6OH2 (HAP) is established based upon transferable potentials. Ca–O, P–O
and O–O potentials were transported from those previously published for fluorapatite and based on single crystal experimental data. The
interactions of hydroxyl oxygen with calcium and phosphate were re-scaled by fitting to experimental data for CaO and AlPO4,
respectively, to account for the reduction in the oxygen charge from 2.0 to 1.426. Force field accuracy is tested by comparing the
calculated and experimental values for the cell constant and atom positions in the unit cell. The elastic constants and bulk modulus
calculated for HAP are in close agreement with the experimental results. The potentials were also used to calculate the compressibility
data of HAP and fluorapatite, and these results also agree with the published experimental data. Using formal charges for metal cations
allows modeling the complete solid solution of Cd–Ca hydroxyapatite with a good accuracy.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: D. Crystal structure; D. Elastic properties; D. Lattice dynamics; D. Mechanical properties

1. Introduction a substitute material for bone over the past several decades
[10,11]. Biological apatites can be described as a calcium-
Calcium phosphate apatites have the general chemical deficient HAP modified by different ion substitutions.
formula Ca5(PO4)3X, where X is an electronegative Among these, CO2 3 substitute for PO34 and for OH,
+ 2+ 4 3
element, such as a halogen, or a group such as, an Na for Ca , SiO4 for PO4 and F for OH. It is


hydroxyl group. This group of compounds is structurally known that the incorporation of these substituents affects
complex inorganic materials and exhibit a diverse range of the mechanisms of formation, the stabilities, the morphol-
applications. Hydroxyapatite, HAP [Ca10(PO4)6(OH)2] has ogies and the mechanical properties of HAP crystals.
received extensive attention due to its considerable However, the types and extents of most of these substitu-
industrial, technological and biotechnological interest. It tions are incompletely understood due to many practical
is used as a catalyst in various reactions, such as complications. One of these complications is the difficulty
dehydration and decomposition of alcohols [1], methane in establishing the mechanisms of charge compensation
oxidation [2–4] and conversion of benzene to phenol [5]. associated with substitutions in calcium-deficient HAPs.
Further, it is used as adsorbents for separation of proteins This is in part because substitutions may also require the
[6,7] and as an ion exchanger for many metal cations [8,9]. presence vacancies.
In these applications, both bulk and surface structures and Apatites are commonly regarded as crystallizing in
properties are important. the hexagonal P63/m space group [12]. Fluorapatite
Because of its crystallographic similarity to calcified [Ca10(PO4)6F2]; FAP is the archetype apatite with hexago-
tissue of vertebrates, HAP has attracted much attention as nal symmetry and the space group P63/m wherein the unit
cell contains two formula units. The 10 calcium ions are
Corresponding author. Tel.: +20 64 382216; fax: +20 64 322381. distributed among two nonequivalent crystallographic
E-mail address: nmost69@yahoo.com (N.Y. Mostafa). sites. Six CaII(6h) ions are symmetrically located about

0022-3697/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jpcs.2006.12.011
 ARTICLE IN PRESS
432 N.Y. Mostafa, P.W. Brown / Journal of Physics and Chemistry of Solids 68 (2007) 431–437

the sixfold screw axes and the other four CaI(4f) ions are model substitution in HAP crystal structure. The set force
positioned along the threefold axes. Surrounding CaI are fields are based on the traditional Buckingham function,
three oxygen triangles, two of which are located one above which are widely used to model inorganic materials. In the
and one below CaI and the third larger triangle almost at present work we report the results of simulations of
the same height along the c-axis as CaI. Surrounding CaII structural compressibility and substitution for HAP.
are six oxygens and one fluoride ion. The fluoride ions are
located on columns parallel to the crystallographic c-axis 2. Computational methods
on the hexad (63) at (0, 0, 14) and (0, 0, 34) surrounded by a
triangle of CaII. All calculations were performed using a general utility
In HAP, the larger OH ions cannot fit within the CaII lattice program developed by Gale and called GULP [34].
triangle so they are displaced along the z direction. Kay These calculations are based on interatomic potentials and
et al. [13], using neutron diffraction refinement, found the lattice energy minimization, allowed by the simultaneous
displacement to be 0.3 Å from the mirror planes containing relaxation of the cell constants and the internal coordi-
CaII, Thus, two different hydroxyl arrangements may nates.
occur in HAP resulting in hexagonal and monoclinic In general, inorganic, ionic and semionic crystal struc-
structures. Hexagonal HAP has a disordered hydroxyl tures are modeled by considering two-body ionic short-
arrangement along the c-axis and the hydroxyl ions are range interaction potentials of the Buckingham type plus a
pointed in upper and lower directions (in statistical Coulombic interaction term [35]. This form is composed of
fashion) against the mirror planes at z ¼ 14 and 34. This a Born–Mayer repulsive exponential term and an r6
leaves the space group (P63/m) unchanged. In contrast, attractive term:
monoclinic HAP (space group P21/b) [14] has an ordered
E ij ¼ Aij expðrij =rij Þ  C ij =r6ij þ ðqi qj Þ=rij ,
hydroxyl arrangement along the c-axis, and hydroxyl ions
are aligned in the same direction on the hydroxyl columns where rij is the distance between ions i and j, with charges qi
and are anti-parallel in adjacent columns. This changes the and qj, and Aij, rij and Cij are fitted parameters.
mirror planes into glide planes and doubles the unit cell In the present study, we considered the application of
dimension along the b hexagonal axis. formal charges to ions, as this enables the transferability
Although the application of computer simulation studies of the potential. Doing so simplifies the description of
have made substantial contributions to establishing struc- different oxidation states and the defect chemistry.
tures in complex crystals, such as zeolites and super- Harmonic three-body short-range interaction potentials
conductors, surprisingly few have been done for apatites. are introduced in order to better represent the tetrahedral
Lattice energy minimization based on empirical potentials configuration of oxygen ions around phosphorus. This
is a powerful technique in studying crystal structures and three-body potential is of the type
defects. Many high quality sets of potentials for inorganic
E ijk ¼ 12K ijk ðY  Y0 Þ2 .
materials are available in the literature but they tend to be
restricted to classes of compounds including simple oxides It is a simple harmonic potential about the equilibrium
[15–21], zeolites [22–27], or superconductors [28,29]. These bond angle, where Kijk is the force constant and (Y–Y0) is
force fields are based on fitting to experimental data [18–20] the deviation from the equilibrium angle, which is taken to
or quantum mechanical calculations [21,26,27]. be the ideal tetrahedral angle, 109.471.
Application of computer simulation to investigate The electronic polarizability of the ions is taken into
apatites, especially HAP, was anticipated to provide a account by using the classical model of Dick and Over-
deeper understanding of stability, mechanisms of substitu- hauser [36]. According to this model polarizable ion, the
tion and many surface phenomena. There are previous oxygen ions in the present case is represented by a core and
investigators used force field to model HAPs [30–33]. a massless shell, connected by a spring. The electronic
Hauptmann et al. [30] force field has been used successively polarizability of the model ion, EP, is then determined by
to model interaction between HAP and water [32] and the spring constant and the charges of the core and shell:
disorder phase transition of HAP. Nevertheless, Lee et al.
EP ¼ 12Kr2 ,
[31] completely ignored the disorder of hydroxyl ion, which
causes a great division of the cell constants from those of where r is the separation distance between the centers of the
the experimental one. All of the other [32,33] force fields core and the shell. During the energy minimization
were based on the partial charge model, which made them calculations, the shell positions are allowed to relax about
useless in investigation substitution and solid solutions. the ionic core resulting in a dipole that mimics the
Also, this prevents transporting potential parameters by electronic polarization. In this model, all intermolecular
fitting to simple oxides. potentials act on the shell of the atom if one exists or on the
Thus the aims of the present paper are to derive core if one does not.
consistent high quality force field values permitting The O–H interaction was simulated using an attractive
modeling apatites in general but with emphasis on HAP. Coulomb-subtracted Morse potential. The parameters for
This force field based on formal charge model in order to this potential were developed by Saul et al. [37] using ab
 ARTICLE IN PRESS
N.Y. Mostafa, P.W. Brown / Journal of Physics and Chemistry of Solids 68 (2007) 431–437 433

initio quantum mechanical cluster calculations with a point Table 1

charge representation of the surrounding lattice. These Potential parameters for hydroxyapatite and fluorapatite
functions take the form Buckingham A (eV) r (Å) C (eV Å6) rcut (Å)
2
V ij ¼ Df1  exp½bðr  r0 Þg . Os–Os 22764 0.1490 32.58 12
This potential is appropriate for representing covalent Os–Pc 836 0.3500 13 10
Os–Cac 1288 0.3316 0 10
bonding [38]; D may then be interpreted as the covalent
Os–Fc 198 0.1110 0 10
bond dissociation energy and r0 the equilibrium bond Fc–Pc 697 0.2389 0 10
length. In this approach fractional charges are used for the Fc–Cac 2031 0.2711 0 10
hydroxyl oxygen and hydrogen (O1.426–H+0.426). Charges Os–Cdc 876.5 0.350 0 10
on the hydroxyl species were distributed across both ions to Cdc–O1.42 627.7 0.350 0 10
Fc–Fc 1128 0.2753 0 10
give an overall charge of 1 with O assigned 1.426 and H
Os–O1.42 22764 0.1490 13.940 12
+0.426, which reproduces the calculated dipole moment of Cac–O1.42 850.8 0.3316 0 10
the O–H group. Additional Buckingham parameters Pc–O1.42 518.47 0.3510 0 10
describe the interactions of the lattice ions with hydroxyl O1.42–O1.42 22764 0.1490 0 12
oxygens. H–Os 311.97 0.2500 0 10
Three body k (eV Å2) y0 (1)
3. Results and discussion Os–Pc–Os 2.840 109.5 1.8
1
Morse D (eV) b (Å ) r0 (Å)
3.1. Potential parameters O1.42–H+0.42 7.0525 3.1749 0.9485 2.0
2
Spring k (eV Å )
The reliabilities of the values for interatomic potentials Oc–Os 98.67 0.7
are of central importance in any computer simulation study
using empirically determined potentials. Hence we sought (a) Charges: Oc ¼ +0.860; Os ¼ 2.860; Fc ¼ 1; Pc ¼ +5; Cac ¼ +2;
Cdc ¼ +2; O1.42 ¼ 1.426; H+0.42 ¼ +0.426.
to derive a potential model for HAP that was compatible
(b) rcut ¼ truncation potential cutoff distance, c ¼ core, s ¼ shell.
with existing models for FAP [39] and use the formal
charges for the cations, so that the effects of substitution by
different metal cations could be readily modeled. The
potential model used describes the calcium ion in an Mg(OH)2 [44]. Table 1 lists the potential parameters
appropriate way; Mulliken analysis employing the ab initio derived in this work as well as those transported from
Hartree–Fock method [40] predicts total charges of metal other published literature.
ions located either at type I or II sites which are very close
to the formal charges. 3.2. Hydroxyapatite structure
Because adequately accurate single crystal data for HAP
could not be found, we modeled its structure using The unit cell for HAP crystal derived from the X-ray
transferable empirical potentials fitted for the structure data [45] is shown in Fig. 1. The hydroxyl ions are
and physical properties of single crystal FAP [39]. positioned in the columnar arrangements parallel to c-axis
Furthermore, we transferred the previously published of the unit cell. This unit cell exhibits disorder in the
O–H Coulomb-subtracted Morse potential from Sauer hydroxyl ion locations in relation to the two mirror planes
[27,41]. This potential considers that O and H interact only at 14c and 34c. This disorder results in doubling the number of
through covalent bonding. hydroxyl ions per unit cell, with partial occupancy 0.5 for
We used the approach described by Schröder et al. [42] each (Fig. 1).
to modify the A parameter of the Buckingham potential Although the GULP code can handle partial occupancy
which describes the repulsion forces of the cation–hydroxyl using mean field theory [33], it cannot be used to describe
oxygen interaction to compensate for the reduction of the structure of HAP. According to mean field theory,
the attractive forces when O2 is replaced by O1.426. each site experiences a potential that is the mean of all
AlPO4 and CaO experimental structure and properties are possible disordered configurations. In the case of HAP the
used to derive these new A Buckingham parameters for two partially occupied (OH) sites overlap each other (see,
P+5–O1.426 and Ca+2–O1.426, respectively. Considering Fig. 1). Thus, in HAP occupancy of these two (OH) sites
the fractional charge model, which maintains the stoichio- is mutually exclusive. To overcome this difficulty we
metric relationship (qO ¼ qCa for CaO and 4qO+qAl ¼ 3 used the super cell of HAP (44 atoms) without applying
for AlPO4). The fit is based on a least squares procedure any symmetry (P1), and removed the excess OH group,
minimizing the sum of the squares of the differences see Fig. 2.
between the experimental and calculated values [30]. This The super cell of HAP was reproduced using two
method succeeds in improving the structure predictions for approaches, both of which produced the same results with
bridging hydroxyl groups in zeolite [42,43] and for the respect to the relaxed HAP crystal structure. The first
hydroxide ions in inorganic solid such as Ca(OH)2 and approach employed the experimentally established HAP
 ARTICLE IN PRESS
434 N.Y. Mostafa, P.W. Brown / Journal of Physics and Chemistry of Solids 68 (2007) 431–437

super cell [45]; the second employed the super cell of the experimental values [45], see Table 3. Using the
FAP [46] wherein F ions were replaced by OH groups. aforementioned potentials we reproduced the structure
The unit cell dimensions and the atomic positions were parameters of HAP with more accurate cell constants than
allowed to relax without any constrains during the energy those obtained previously using other force fields with
minimization procedures. Both approaches produced partial charge model [30–32] or those obtained using
the same result for the final relaxed super cell as shown density function (DFT) calculation of de Leeuw [47], see
in Fig. 2. Table 2.
The cell constants calculated for HAP crystal are listed This model is also capable of reproducing the distortion
in Table 2 and show good agreement with the experimental of the O–H group from the reflection planes, which in our
values [45]. The atomic positions in the unit cell of the calculation is 0.294 Å compared to 0.3 Å experimentally
relaxed structure, are also in satisfactory agreement with determined using neutron diffraction [45].

3.3. Elastic constant matrix and bulk modulus

Katz and Ukraincik [48] produced a set of pseudo-single

crystal elastic constants for HAP extrapolated from single
crystal elastic constant determined FAP. This extrapola-
tion was based on the similarities between the HAP and
FAP crystal structures, and the comparison of the isotropic
elastic moduli determined for polycrystalline specimens of
HAP and FAP. No direct experimental data based on
measurements of elastic constants of HAP single crystals
were found.
Calculation of accurate elastic constants of HAP crystals
is important in understanding the anisotropic elastic
behavior of calcified tissues. Biomechanical investigations
[49–51] consider the anisotropy in the elastic behavior of
Fig. 1. The unit cell of hydroxyapatite in P63/m space group.
bone as the result of its two constituents, HAP crystals and
collagen matrix. The orientation of HAP crystals within
the collagen matrix influences the anisotropic mechanical
characteristics of bone. These analyses are even more
relevant to enamel because of the greater proportion of
HAP mineral present, 96% vs. 60 in bone.
Table 4, shows the calculated elastic constant matrices
and bulk moduli for HAP and FAP along with the
experimental data for FAP [52] and the pseudo-
single crystal data for HAP [48]. The relationship between
structure and anisotropic elasticity of HAP is illustrated in
the values of the compressional moduli (C11, C22 and C33).
Although the analysis reproduces the bulk modulus for
both HAP and FAP with a reasonable accuracy, it gives
high values for the C11 and low values for the C33
anisotropic elastic constants for both HAP and FAP.
Fig. 2. The super cell of hydroxyapatite in P1 space group after removing These variations reflect the variability in the experimental
excess OH groups. data reported in the literature [46,52], which differ by more

Table 2
Cell constants of hydroxyapatite crystal

Parameter Experimental [45] Present DFT [47] Lee et al. [31] Rabone et al. [33] Hauptmann et al. [30]
calculations

a 9.418 9.412 9.563 9.528 9.35 9.455

c 6.875 6.853 6.832 6.6077 6.86 6.901
a 90 90 90 90 90 90
b 90 90 90 90 90 90
g 120 120 120 120 120 120
 ARTICLE IN PRESS
N.Y. Mostafa, P.W. Brown / Journal of Physics and Chemistry of Solids 68 (2007) 431–437 435

Table 3 1.02
Calculated and experimental atomic positions of hydroxyapatite unit cell
1.00 Experimental
Site x y z
Calculated
0.98
Exp. [45] Cal. Exp. [45] Cal. Exp. [45] Cal.
0.96
CaI 0.6666 0.6667 0.3333 0.3333 0.0017 0.0011
CaII 0.2467 0.2441 0.9933 0.9933 0.25 0.2447
0.94
P 0.3979 0.3962 0.3678 0.3746 0.25 0.2469

V/Vo
O1 0.3275 0.3367 0.4844 0.4878 0.25 0.2478
0.92
O2 0.5860 0.5709 0.4649 0.4574 0.25 0.2427
O3 0.3424 0.3339 0.2577 0.2709 0.0685 0.0828
0.90
OH 0.000 0.000 0.000 0.000 0.1999 0.2072
H 0.000 0.000 0.000 0.000 0.0608 0.0762
0.88

0.86

Table 4 0.84
Calculated and experimental elastic constants of hydroxyapatite and 0 2 4 6 8 10 12 14 16 18 20 22
fluorapatite Pressure (GPa)
Constant Hydroxyapatite Fluorapatite Fig. 3. Experimental [53] and calculated HAP cell volume variations with
pressure normalized by room conditions.
(MPa) Exp. [48] Cal. Exp. [52] Cal.

C11 13.7 15.75 (15) 14.34 16.48 (15)

C13 5.49 5.97 (9) 5.75 6.03 (4.9) 1.02 Calculated
C33 17.2 14.73 (14) 18.05 14.54 (19.4) Experimental
C44 3.96 4.39 (10.8) 4.15 4.02 (3) 1.00
Bulk 89.0 90.66 (1.9) 93.0 91.76 (1.3)
modulus 0.98

0.96

than 6%. This deviation may be also due to an unrealistic 0.94

V/Vo

description of the phosphate tetrahedral, with respect to 0.92

charge and geometry. An ab initio study predicted a charge
on phosphate of +1.35 [40], and this is far from the formal 0.90
charge used in the present model. Additionally, the
tetrahedral exhibited a slightly disordered geometry. 0.88

0.86
3.4. Compressibility of HAP and FAP
0.84
To check further the quality of the force field analysis, -2 0 2 4 6 8 10 12 14 16 18 20 22
we calculated the changes in the cell volumes of FAP and Pressure (GPa)
HAP with pressure up to 15 GPa. These variations are
Fig. 4. Experimental [46] and calculated FAP cell volume variations with
shown in Figs. 3 and 4, respectively, with the values pressure normalized by room conditions.
normalized by atmospheric pressure. The experimental
values [46,53] are also provided for comparison. The
reductions in the lattice volume of both apatites are linear complete Cd–Ca solid solution of HAP in the full range
with pressure and there is no evidence of phase transitions from Ca10(PO4)6(OH)2 to Cd10(PO4)6(OH)2. The variation
in this pressure range. The results are in excellent of unit cell dimension with substitution of Ca2+ ions with
agreement with those obtained by Brunet et al. [53] using Cd2+ ions in the unit cell is shown in Fig. 5, in comparison
synchrotron X-ray powder diffraction data for synthetic with experimental data [54].
HAP and those obtained by Comodi et al. [46] using
spectroscopic studies for FAP. 4. Conclusions

3.5. Cd–Ca HAP solid solution Based on lattice dynamics approximations we have
established a force field that satisfactorily describes the
Cd–O and Cd–OH interaction potential parameters was hydroxyapatite structure and single crystal elastic proper-
obtained by fitting to crystal structure of CdO. These ties. This force field is based on and compatible with those
parameters, as shown in Table 1, were used to model the of Meis et al. [39]. The interatomic potentials used are
 ARTICLE IN PRESS
436 N.Y. Mostafa, P.W. Brown / Journal of Physics and Chemistry of Solids 68 (2007) 431–437

9.9 6.90 References

a (calculated)
a (Experimntal) [1] S. Sugiyama, J.B. Moffat, Catal. Lett. 76 (2001) 1.
9.8 [2] S. Sugiyama, E. Nitta, H. Hayashi, J.B. Moffa, Catal. Lett. 59 (1999)
6.85
c (calculated) 67.
c (Experimental) [3] S. Sugiyama, K. Abe, H. Hayashi, J.B. Moffat, Catal. Lett. 57 (1999)
9.7 161.
6.80 [4] S. Sugiyama, Y. Iguchi, T. Minami, H. Hayashi, Catal. Lett. 46
(1997) 279.
9.6 [5] B. Liptáková, M. Hronec, Z. Cvengrošová, Catal. Today 61 (2000)

c (A)
a (A)

6.75 143.
[6] I.D. Smiciklas, S.K. Milonji, C.S. Zec, J. Mater. Sci. 3 (2000) 2825.
9.5 [7] R. Giovannini, R. Freitag, Bioseparation 9 (2001) 359.
[8] X.G. Chen, J.V. Wright, J.L. Conca, L.M. Peurrung, Water Air Soil
6.70 Pollut. 98 (1997) 57.
9.4 [9] C.K. Danny, J.F. Porter, G. McKay, Water Res. 35 (2001) 3876.
[10] Z. Shi, L.E. Erickson, J. Hazardous Mater. 87 (2001) 99.
[11] F.C.M. Driessens, R.M.H. Verbeek, Biominerals, CRC Press, Boca
6.65
9.3 Raton, FL, 1990.
[12] J.C. Elliott, Studies in Inorganic Chemistry, Elsevier, Amsterdam,
1994.
9.2 6.60 [13] M.I. Kay, R.A. Young, A.S. Posner, Nature 12 (1964) 1050.
0 2 4 6 8 10 [14] J.C. Elliott, P.E. Mackie, R.A. Young, Science 180 (1973) 1055.
Cd-substituted [15] M.S. Islam, J. Mater. Chem. 4 (2000) 1027.
[16] L. Minervini, R.W. Grimes, J.A. Kilner, K.E. Sickafus, J. Mater.
Fig. 5. Variation of the cell constants of cadmium–calcium hydroxyapa- Chem. 10 (2000) 2349.
tite solid solutions against number of cadmium atoms (experimental [17] S.M. Woodley, C.R.A. Catlow, J.D. Gale, P.D. Battle, Chem.
results from [54]). Commun. 19 (2000) 1879.
[18] J.D. Gale, Philos. Mag. B 73 (1996) 3.
[19] M. Cherry, M.S. Islam, C.R.A. Catlow, J. Phys. Chem. 99 (1995)
14614.
[20] S.M. Woodley, P.D. Battle, J.D. Gale, C.R.A. Catlow, Phys. Chem.
based on the shell model for oxygen in phosphate Chem. Phys. 1 (1999) 2535.
tetrahedra and on the rigid ion model for oxygen in the [21] T.S. Bush, J.D. Gale, C.R.A. Catlow, P.D. Battle, J. Mater. Chem. 4
tunnels along the c-axis (OH groups). The rigid ion model (1994) 831.
with formal charges is also used to describe calcium and [22] A.R. Ruiz-Salvador, A. Gómez, D.W. Lewis, C.R.A. Catlow, L.M.
Rodrı́guez-Albelo, L. Montero, G. Rodrı́guez-Fuentes, Phys. Chem.
phosphorus. During, submission of the revised manuscript
Chem. Phys. 2 (2000) 1803.
of this article, we find a model developed by Calderin et al. [23] L. Whitmore, B. Slater, C.R.A. Catlow, Phys. Chem. Chem. Phys. 2
[55]. This model is the closest to the present model (2000) 5354.
potential. However, Calderin’s model is more complicated [24] S. Girard, C.M. Draznieks, J.D. Gale, G. Férey, Chem. Commun.
because it contains different potential parameters for each (2000) 1161.
calcium site and represents O–H interaction with different [25] S. Djordje, R. Nevenka, J. Porous Mater. 8 (2001) 239.
[26] M. Sierka, J. Sauer, Faraday Discuss. 106 (1997) 41.
models, which is not compatible with Bush et al. [21] and [27] J. Sauer, J. Phys. Chem. 100 (1995) 11043.
Sauer [27] potential libraries of simple oxides. The [28] S.A. French, C.R.A. Catlow, J. Mater. Chem. 11 (2001) 102.
modeling of HAP structure and elastic constants derived [29] M.S. Islam, M.S.D. Read, S. D’Arco, Faraday Discuss. 106 (1997)
using empirical force field are consistent with observed 367.
experimental data and successively reproduce the distor- [30] S. Hauptmann, H. Dufner, J. Brickmann, S.M. Kast, S.R. Berry,
Phys. Chem. Chem. Phy. 5 (2003) 635.
tion of OH groups along c-axis. The present force field is
[31] W.T. Lee, M.T. Dove, E.K.H. Salje, J. Phys.: Condens. Matter 12
more accurate than the pervious force field found in (2000) 9829.
literature [30,31] in representing the crystal structure. The [32] N.H. de Leeuw, Phys. Chem. Chem. Phys. 6 (2004) 1860.
advantage of this force field is its ability to model [33] J.A.L. Rabone, N.H. De Leeuw, J. Comput. Chem. 27 (2) (2006)
substitution and solid solution of hydroxyapatite, because 253.
it uses the formal charge model for both metal cations [34] J.D. Gale, J. Chem. Soc. Faraday Trans. (1997) 629.
[35] P.P. Ewald, Ann. Phys. (Leipzig) 64 (1921) 253.
(+2) and phosphate ions (+5). This was examined by [36] B.G. Dick, A.W. Overhauser, Phys. Rev. 112 (1958) 90.
modeling Ca–Cd hydroxyapatite solid solutions in the full [37] P. Saul, C.R.A. Catlow, J. Kendrick, Philos. Mag. B 51 (1985) 107.
range from Ca10(PO4)6(OH)2 to Cd10(PO4)6(OH)2 given a [38] C.R.A. Catlow, Computer Modeling in Inorganic Crystallography,
satisfactory match with experimental data. Academic Press, 1997.
[39] C. Meis, J.D. Gale, B.J. Carpena, D. Gosset, J. Phys. Chem. A 104
(2000) 5380.
Acknowledgment
[40] A. Chartier, C. Meis, J.D. Gale, Phys. Rev. B 64 (2001) 85110.
[41] J. Sauer, J. Phys. Chem. 91 (1987) 2315.
We thank Dr. J.D. Gale for providing the GULP code [42] K.P. Schröder, J.M. SauerLeslie, C.R.A. Catlow, J.M. Thomas,
and for useful discussions. Chem. Phys. Lett. 188 (1992) 320.
 ARTICLE IN PRESS
N.Y. Mostafa, P.W. Brown / Journal of Physics and Chemistry of Solids 68 (2007) 431–437 437

[43] R. Grau-Crespo, A.G. Peralta, A.R. Ruiz-Salvador, A. Gómez, [50] D.T. Reilly, A.H. Burstein, J. Biomech. 8 (1975) 393.
R. López-Cordero, Phys. Chem. Chem. Phys. 2 (2000) 5716. [51] T.N. Gardner, J.C. Ellioitt, Z. Sklar, G.A.D. Briggs, J. Biomech. 25
[44] P.S. Baram, S.C. Parker, Philos. Mag. 73 (1996) 49. (1992) 1265.
[45] J.M. Hughes, M. Calmeron, K.D. Crowley, Am. Mineral. 74 (1989) 870. [52] M.C. Sha, Z. Li, R.C. Bradt, J. Appl. Phys. 75 (1994) 7784.
[46] P. Comodi, Y. Liu, P.F. Zanazzi, M. Montagnoli, Phys. Chem. [53] F. Brunet, D.R. Allan, S.A.T. Redfern, R.J. Angel, R. Miletich, H.J.
Miner. 28 (2001) 219. Reicmann, J. Sergent, M. Hanfland, Eur. J. Mineral. 11 (1999) 1023.
[47] N.H. de Leeuw, Chem. Commun. 17 (2001) 1646. [54] A. Bigi, M. Gazzano, A. Ripamonti, J. Chem. Soc. Dalton Trans.
[48] J.L. Katz, K. Ukraincik, J. Biomech. 4 (1971) 221. (1986) 241.
[49] S. Lees, F.R. Rollins Jr., J. Biomech. 5 (1972) 557. [55] L. Calderin, D. Dunfield, M.J. Stott, Phys. Rev. B 72 (2005) 224304.