New insights into the bioactivity of SiO2–CaO and SiO2–CaO–P2O5 sol–gel glasses by molecular dynamics simulations

1 23

Journal of Sol-Gel Science andTechnology ISSN 0928-0707Volume 67Number 1 J Sol-Gel Sci Technol (2013) 67:208-219DOI 10.1007/s10971-011-2453-4

New insights into the bioactivity of SiO2–CaO and SiO2–CaO–P2O5 sol–gel glassesby molecular dynamics simulations

G. Malavasi, L. Menabue,M. C. Menziani, A. Pedone, A. J. Salinas& M. Vallet-Regí

1 23

Your article is protected by copyright and

all rights are held exclusively by Springer

Science+Business Media, LLC. This e-offprint

is for personal use only and shall not be self-

archived in electronic repositories. If you wish

to self-archive your article, please use the

accepted manuscript version for posting on

your own website. You may further deposit

the accepted manuscript version in any

repository, provided it is only made publicly

available 12 months after official publication

or later and provided acknowledgement is

given to the original source of publication

and a link is inserted to the published article

on Springer's website. The link must be

accompanied by the following text: "The final

publication is available at link.springer.com”.

ORIGINAL PAPER

New insights into the bioactivity of SiO2–CaO and SiO2–CaO–P2O5 sol–gel glasses by molecular dynamics simulations

G. Malavasi • L. Menabue • M. C. Menziani •

A. Pedone • A. J. Salinas • M. Vallet-Regı

Received: 12 November 2010 / Accepted: 22 March 2011 / Published online: 1 April 2011

� Springer Science+Business Media, LLC 2011

Abstract The structures of binary xCaO � (100 - x)SiO2

glasses with x = 10, 20 and 30 mol-% and ternary

(20 - x)CaO � xP2O5 � 80SiO2 glasses with x = 3, 10, 15,

17 and 20 mol-% have been studied by means of classical

molecular dynamics simulations using both the melt-

quenched and the sol–gel protocols. The structural picture

derived correlates the bioactive behaviour to the combined

effects of the connectivity of the extended silicate network

and to the tendency to form (or not to form) non-homo-

geneous domains. In this context, a mathematical rela-

tionship that relates the Ca/P ratio in the Ca phosphate

micro-segregation zones to the P2O5 content in ternary

glasses has been developed and this has been used to fine-

tuning the optimum amount of P in a glass for its highest in

vitro bioactivity. The composition with optimal Ca/P ratio,

80Si–14.8Ca–5.2P, has been synthesized and the results of

bioactivity tests have confirmed the prediction.

Keywords Molecular dynamic � Sol–gel processes �Bioactive glasses � In vitro bioactivity tests

1 Introduction

Sol–gel glasses in the CaO–SiO2 and CaO–SiO2–P2O5

systems were widely investigated as biomaterials because

some compositions bond with the living tissues, i.e. are

bioactive [1–4].

These investigations extensively explored the SiO2–

CaO–P2O5 phase diagram modifying the proportions of the

chemical components, the bioactivity window was then

determined [5] by means of in vitro studies in a Simulated

Body Fluid (SBF) [6]. The results reported differences in

the mechanism of formation of the hydroxycarbonate

apatite (HCA) layer, indicative of bioactivity, depending

on the glass components.

Thus, for binary CaO–SiO2 glasses, the bioactive

response increased with the CaO content [3]. This was

explained by considering that the increase in the network

disorder produced by calcium favors a quick leaching of

Ca2? from glass to solution and an increasing of the silanol

(Si–OH) concentration on the glass surface [7]. Both facts

yielded a very quick formation of an amorphous calcium

phosphate (a-CaP) layer, precursor of HCA, after soaking

binary glasses in SBF. Then, the HCA nanocrystals layer

was formed after maintaining glasses few days more in

SBF. Although a complete characterization by using sev-

eral experimental techniques is required to clearly distin-

guish between a-CaP and HCA, perhaps the most

convenient way to do that is by using FTIR spectroscopy

[7]: for a-CaP one band at 563 cm-1 appears in the spec-

trum, but for nanocrystalline HCA two bands at 563 cm-1

and 602 cm-1 are present. For CaO–SiO2 glasses it was

concluded that the bioactive response was quick enough for

CaO higher than 10 mol-%. However, CaO contents higher

than 30 mol-% provoked the formation of calcite (CaCO3)

in the layer together with HCA [3].

G. Malavasi � L. Menabue (&) � M. C. Menziani � A. Pedone

Department of Chemistry, University of Modena and Reggio

Emilia, Via Campi 183, 41125 Modena, Italy

e-mail: [email protected]

A. J. Salinas � M. Vallet-Regı

Departamento de Quimica Inorganica y Bioinorganica, Facultad

de Farmacia Universidad Complutense, 28040 Madrid, Spain

A. J. Salinas � M. Vallet-Regı

Network Center of Biomedical Research on Bioengineering,

Biomaterials and Nanomedicine (CIBER-BBN), Madrid, Spain

123

J Sol-Gel Sci Technol (2013) 67:208–219

DOI 10.1007/s10971-011-2453-4

Author's personal copy

Regarding ternary CaO–SiO2–P2O5 glasses, it was

reported that small amounts of P2O5 modified the forma-

tion mechanism of HCA described for CaO–SiO2 glasses

[4, 8]. This was attributed to the formation of calcium

phosphate nanocrystals during the ternary glasses synthesis

that were visualized by high resolution transmission elec-

tron microscopy [9]. That way, a substantial decrease of the

calcium in the glass network compared with the nominal

one took place decreasing the amount of Ca2? able to be

leached to SBF. Thus, few new Si–OH groups were formed

and the obtaining of a-CaP was significantly retarded

compared to analogous P-free gel glasses. However, cal-

cium phosphate nanocrystals in ternary glasses behave as

crystallization nuclei accelerating the HCA formation. In

summary, small amounts of P2O5 in the glasses of 3 mol-%

accelerated their bioactive response compared to analogous

P-free glasses, since shorter times were required for the

HCA formation although longer times had been necessary

for the formation of the initial a-CaP [7].

With the aim of obtaining further details on the struc-

ture-bioactivity relationships of this class of glasses, the

structures of binary xCaO � (100 - x)SiO2 glasses with

x = 10, 20 and 30 mol-% and the ternary (20 - x)CaO �xP2O5 � 80SiO2 glasses with x = 3, 10, 15, 17 and 20 mol-%

will be studied in this paper by computational simulations

and experimental techniques.

The availability of increasingly powerful computational

methods and resources nowadays makes computer simu-

lations an attractive complementary tool to experimental

techniques, to obtain an atomistic view into the bioactive

behaviour of bioglasses and fill the gap in fundamental

knowledge. In particular, molecular dynamics (MD) sim-

ulations have now shown to be able to tackle effectively

complex multicomponent amorphous systems and provide

an unprecedented high resolution view into their atomistic

bulk and surface structure and properties [10–17]. Whereas

the computational simulation of glasses by quenching a

melt from high to room temperature with different com-

putational protocols [18–23] is a well-established exercise,

several issues arise when the structure of a sol–gel prepared

glass has to be simulated. In fact, the time-scales charac-

teristic of the full sol–gel processes are yet not accessible

by MD simulations, moreover the hydrolysis and conden-

sation reactions involved are not described in purely clas-

sical methodologies.

Therefore, previous computer simulations studies of

sol–gel derived materials have focused mainly on the

modeling of some specific aspects of the process [24–30]

or have assumed such materials to be the same as melt-

quenched glasses [25, 26].

Interesting are the results obtained by Mead et al.

[27] who have simulated a sol–gel calcium-silicate glass

by using a modified melt-quench approach in which

under-density and hydroxylation, the two most important

differences of sol–gel glasses with respect to melt-derived

glasses, were added.

Very recently Bhattacharya et al. [26] reported on the

modelling of the sol–gel synthesis process of nano-porous

silica gel in an aqueous environment by means of MD

simulations making use of a reactive three body inter-

atomic potential. This approach allows for charge transfer

upon the rupture or formation of chemical bonds making it

possible to simulate the elementary reaction processes for

sufficiently large size- and time-scales, generating more

accurate gel structures.

Unfortunately, the extension of reactive force-fields to

complex binary or ternary amorphous systems containing

chemical bonds ranging a large part of the Pauling scale is

not obvious and requires further substantial refinement of

the whole methodology and parameters. Moreover, further

insights into mechanism of the reactions occurring during

the sol–gel process of Si(OCH3)4 units could be gained by

exploiting ab initio molecular dynamics (AIMD) simula-

tions [31]. However, at present, the size and time scales

which can be probed by ab initio molecular dynamics are

limited to few hundreds of atoms and tens of picoseconds.

The computational approach followed in this paper

ignores the particulars of the sol–gel synthesis route

focusing instead on grasping some important peculiarities

of the final sol–gel product obtained with respect to the

melt-quenched one. In fact, the classical melt-quenched

protocol [18] and the one modified according to Mead et al.

[2, 27] to account for sol–gel characteristics will be

exploited for the elaboration of structure-bioactivity rela-

tionships useful in the fine-tuning of the optimal glass

composition with optimized bioactivity. In particular the

main goal of the present work is to define one or more

structural descriptors able to predict HCA formation as a

consequence of SBF soaking during in vitro tests. More-

over, confirmation or confutation of the computational

results will be provided by experimental tests.

2 MD simulations

2.1 Melt-quenched computational procedure

The structures of binary calcium-silicate glasses of compo-

sition xCaO � (100 - x)SiO2 with x = 10, 20 and 30 mol-%

and the ternary calcium-phospho-silicate glass with com-

position (20 - x)CaO � xP2O5 � 80SiO2 with x = 3, 10, 15,

17 and 20 mol-% have been generated through the compu-

tational protocol described in Refs. [18, 19].

The initial configurations were generated by placing

randomly 2,000–2,800 atoms in a cubic box whose

dimensions and atomic compositions in Table 1. The glass

J Sol-Gel Sci Technol (2013) 67:208–219 209

123


densities have been calculated according to the Priven’s

empirical method [32] implemented in the SciGlass�

software [33].

In order to asses the effects of box size on the medium

range order of ternary glasses the ternary system with

x = 3 was simulated by using 5,200 atoms (sample called

80Si–17Ca–3P-big hereafter).

The DLPOLY [34] package has been employed for

NVT MD simulations by using an empirical partial charge

rigid ionic model [18], which has been shown to reproduce

well the structure, transport and mechanical properties of

oxides, silicates and silica based glasses [18]. Integration of

the equations of motion has been performed using the

Verlet Leap-frog algorithm with a time step of 2 fs. Cou-

lombic interactions have been calculated by the Ewald

summation method with a cut-off of 12 A and an accuracy

of 10-4. The short range interaction cut-off was set to

5.5 A. The structural parameters reported for each glass are

averages over 3 independent simulations.

2.2 Sol–gel computational procedure

One of the main differences between a melt-quench and a

sol–gel glass is due to the smaller density caused by the

introduction of OH groups in the latter case. In NVT MD

simulations the density of the system must be known

a priori. The determination of the experimental density for

sol gel glasses is hampered by the hygroscopic nature of

the glass and the variation of its value as consequence of

different thermal treatments. In this work we have deter-

mined the experimental density of the sol gel glass with

composition 80SiO2 � 15CaO � 5P2O5 after thermal treat-

ment at 700 �C for 3 h. The density of 2.360 g/cm3 is only

1.0% less than that determined by using the Priven

empirical method (see Table 1) and several calculations by

using the two densities did not provide appreciable dif-

ferences on the structure of the simulated glasses. For these

reasons, the densities calculated by the empirical method

were used for the other compositions.

The other important characteristic of sol–gel glasses is

that they contain a certain concentration of hydroxyl

groups. Because of the thermal treatment usually employed

to dehydrate the glass, the content of H atoms is only due to

hydroxyl groups. As reported in previous papers [35–37]

the experimental concentration of hydroxyl groups in sil-

ica-based gel glasses is around 5 groups/nm2. This value is

reported in a recent study on the surface properties of sol–

gel 80SiO2 � 15CaO � 5P2O5 glass [36]. In this work, some

of us employed thermogravimetic techniques to follow the

de-hydroxylation process. This process, which causes a

drastic reduction of the glass surface area takes place from

around 550� to 850 �C and the concentration of hydroxyl

groups on the surface changes from 18 groups/nm2 at

600 �C to 0.050 groups/nm2 at 900 �C, passing to 5 groups/

nm2 at 700 �C. [36] Therefore, starting configurations

containing 5 OH groups per nm2 were generated by con-

sidering the OH group as a single site with bond length of

1 A. The short range O–H and the hydrogen bonds

O–H��O potential parameters used are reported in Table 2

[38]. The final structural models were generated by using

the simulation procedure described above in the previous

section.

3 Synthesis of the sol–gel-glasses

The glass compositions modeled by MD and the one that

will be predicted (see the following) to be the most

Table 1 Gel glass compositions, in vitro activity, number of atoms, cell size and density used in the MD simulations of the glasses studied

Formulation % SiO2 % CaO % P2O5 In vitro activitya nSi nO/nH nCa nP Density (g/cm3) Cell size (A)

90Si–10Ca 90 10 0 B 868 1,832 96 0 2.318 34.54

80Si–20Ca 80 20 0 B 800 1,800 200 0 2.445 34.27

70Si–30Ca 70 30 0 B 726 1,762 310 0 2.580 33.98

80Si–17Ca–3P 80 17 3 B 560 1,344 119 42 2.409 31.02

80Si–17Ca–3P-big 80 17 3 B 1,416 3,386 294 104 2.409 42.18

80Si–15Ca–5P 80 15 5 NS 560 1,400 105 70 2.389 31.39

80Si–15Ca–5P–OH 80 15 5 NS 560 1,432/64 105 70 2.360b 31.65

80Si–14.8Ca–5.2P 80 14.8 5.2 NS Not modeled

80Si–10Ca–10P 80 10 10 NS 560 1,540 70 140 2.351 32.25

80Si–5Ca–15P 80 5 15 NS 560 1,680 35 210 2.340 32.97

80Si–3Ca–17P 80 3 17 NS 560 1,736 21 238 2.337 33.24

80Si–20P 80 0 20 NS 560 1,820 0 280 2.333 33.65

a B bioactive, NS not studied beforeb Experimental value determined in the present work

210 J Sol-Gel Sci Technol (2013) 67:208–219

123


bioactive, i.e. 80Si–14.8Ca–5.2P, were synthesized fol-

lowing a sol–gel procedure previously described [3, 4]. In

Fig. 1 the ten glasses investigated are indicated and the

bioactivity window previously described is highlighted. Six

compositions were here synthesized for the first time (open

stars), thus their bioactive response in SBF was still

unknown. The other four compositions (dark stars) had

been previously reported as bioactive, [3, 4] but they were

here synthesised and evaluated in identical conditions that

the other six compositions for comparative purposes.

Tetraethylorthosilicate (TEOS), triethylphosphate (TEP)

and calcium nitrate tetrahydrate, Ca(NO3)2 � 4H2O (the three

reactants of Aldrich) were used as sources of SiO2, P2O5

and CaO, respectively. The corresponding stoichiometric

amounts to obtain the compositions included in Table 1

were sequentially added to a solution of HNO3 in water at

pH 0.5. In all the cases, the HNO3 ? H2O/TEOS ? TEP

molar ratio was 8. After that solutions were stirred for 1 h

and cast in Teflon� containers that were hermetically

closed and kept 3 days at room temperature for gelation.

Then, the gels were aged 3 days at 70 �C and dried at

150 �C for 2 days. For drying, the lids of containers were

replaced by other ones featuring a 1 mm hole. Xerogels

obtained were grounded and sieved taking 0.5 g of the

fraction between 32 and 63 lm that were compacted at

50 MPa uniaxial and 150 MPa isostatic pressures to obtain

disks (13 mm in diameter, 2 mm in height). These disks

were heated at 700 �C for 3 h for stabilization and nitrate

removal. That way, disks of glass used for the in vitro

bioactivity studies in SBF were obtained.

3.1 In vitro bioactivity evaluation of the

sol–gel-glasses

Assessment of the in vitro bioactivity response was carried

out by soaking the glass disks for 3, 9, 24 h, 3 and 7 days in

the Simulated Body Fluid (SBF) proposed by Kokubo et al.

[6] at 37 �C. SBF is an acellular aqueous solution with an

inorganic ion composition almost equal to human plasma:

Na?: 142, K?: 5.0, Mg2?: 1.5, Ca2?: 2.5, Cl-: 147.8,

HCO3-: 4.2, HPO4

2-: 1.0 and SO42-: 0.5 mM, and buf-

fered at physiological pH 7.40 with tris(hydroxymethyl)

aminomethane/HCl. After the SBF treatments the speci-

mens were removed from solution, gently rinsed with water

and acetone and allowed to air-dry for 24 h. To avoid

contamination by microorganisms, SBF was filtered with a

0.22 lm Millipore� system before the in vitro assays.

Moreover, all the manipulations/operations were carried

out in a laminar flux cabinet Telstar� AV-100.

The surface of the glasses were analyzed before and

after the SBF treatments by Fourier Transform Infrared

(FTIR) spectroscopy, in a Nicolet Nexus Spectrometer

equipped with a diamond ATR Goldengate�, Scanning

Electron Microscopy (SEM) and Energy Dispersive X-ray

Spectroscopy (EDS), in a JEOL 6400 Microscope equipped

with an Oxford-LINK Pentafet System.

The bioactive response of glasses was evaluated in terms

of the formation of a HCA layer in the surface of the

specimens after being soaked in SBF at 37 �C. Those

materials coated by this layer were considered bioactive.

When the glasses surface remained unchanged after the

SBF treatment, they were considered non bioactive. Fur-

thermore, the in vitro studies allowed the comparison of the

relative kinetics of the glass bioactive responses. Thus, as

shorter was the time required for the HCA layer formation,

higher was the bioactivity of material.

Table 2 Interatomic potential: analytical functions and parameters

Morse� U rð Þ ¼ Dij � 1� eð�aijÞ�ðr�r0Þ2� �

� 1� �

þ Cij=r12

Dij (eV) aij (A-2) r0 (A) Cij (eV A12)

Si2.4–O-1.2 0.3405540 2.0067 2.10000 1.0

P3.0–O-1.2 0.8313260 2.5858 1.80079 1.0

Ca1.2–O-1.2 0.0302110 2.2413 2.92324 5.0

H0.6–O-1.2 0.4371660 3.0643 1.24513 0.0

O-1.2–O-1.2 0.0423950 1.3793 3.61870 22.0

H-bond–U(r, h) = [(A/r12) - (B/r10)] � (cos h)4

A (eV A12) B (eV A10)

H–O��O 15000.0 0

Fig. 1 Ternary CaO–SiO2–P2O5 phase diagram including the glass

compositions studied. Open stars indicate new glasses prepared in

this paper. Filled stars correspond to glasses already described but

investigated here for comparative purposes. Shadowed: the in vitro

bioactivity window reported [5]


123


Practically from their beginnings, in vitro bioactivity

tests were controversial with respect to the validity of

extrapolating their results to the in vivo conditions. In that

sense a recent article of Bohner and Lemaitre [39] argues

that in spite of the very extended which they are, mainly

using Kokubo’s SBF solution [6] in many occasions these

tests produced false positives and false negatives and that

this method needs to be improved. Nevertheless, the in

vitro evaluation of glasses in SBF is perfectly valid to

evaluate their bioactivity in a comparative way as it was

carried out in this study.

4 Results and discussion

4.1 MD simulations

The simulation of sol–gel glasses requires the a priori

knowledge of the OH content in the structure and a time-

consuming manual procedure for the preparation of the input

structure. Therefore, in this work a standard melt-quench

protocol [18, 19] has been applied to the complete series of

glasses considered and the modified sol–gel procedure

recently described by Mead et al. [27] has been carried out to

simulate the 80SiO2 � 15CaO � 5P2O5 glass. The comparison

between the structural features of this last glass simulated

using a standard melt-quench procedure (80Si–15Ca–5P)

and modified sol–gel procedure (80Si–15Ca–5P–OH)

highlights the similarities and the differences between the

adopted procedures. The structural model of the sol–gel

glass at room temperature has an internal pressure very close

to 0 kbar suggesting that the introduction of the OH groups

do not lead to stresses in the glass.

The similarities in the total distribution functions (TDF),

that represent the relative probability of finding any atom at a

distance r from an atom placed at the origin, of 80Si–15Ca–

5P and 80Si–15Ca–5P–OH, reported in Fig. 2, indicate that,

overall, comparable structures are obtained, the main dif-

ference being a peak detected around 1 A in the 80Si–15Ca–

5P–OH distribution due to the presence of O–H bonds. A

small difference between the curves, due to modifications of

the Si–Si interatomic distance [27] and the introduction of

the Ca–H pairs which lie in the range 2.8–3.4 A, in agree-

ment with previous experimental works (Ca–H = 3.0–3.1 A

in Ref. [27] and Ca–H = 2.95 A in Ref. [40]).

A more detailed description of the structure of the glass

studied is summarized in the following.

The local structure around cations is characterized by

the cations-oxygens bond lengths, coordination number

(CN) and oxygens-cations-oxygens bond angle distribu-

tions (BAD). Whereas, information on the medium range

order can be obtained from the distribution of non-bridging

(NBO), bridging (BO) and three-bridging oxygens (TBO),

cations-cations pair distribution functions (PDF), distribu-

tion of Qn species (where n is the number of bridging

oxygens bonded to the network former cations Si or P),

self- and cross-connectivity between TO4 tetrahedra

(T = Si and P) and ring size distribution.

4.1.1 Bond length and bond angle distributions

The X–O bond length (where X = Si, P, H and Ca),

obtained from the first peak of the corresponding PDFs and

the bond angle distributions for the different atom groups

in the studied glasses are summarized in Table 3. The O–H

bond distance found for the modeled sol–gel glass is

1.07 A (Fig. 3, Table 3) in good agreement with the find-

ing of Mead and Mountjoy (1.05 A) [27].

The second maximum at 1.9–2.1 A can be attributed to

the H-bond between the H and O atoms in a different

hydroxyl group. The short range order of Si and P ions is

rather rigid and hardly affected by the composition, con-

versely, the Ca–O bond length increases from 2.30 A for

90Si–10Ca glass to 2.37 A for Ca-rich glasses (70Si–

30Ca). These values fit very well with neutron diffraction

data of iso-compositional gel glasses [40] confirming the

goodness of computational procedure to reproduce the

local environment of modifier cations. A similar trend is

observed in ternary glasses where Ca–O bond length

increases from 2.32 A for 80Si–3Ca–17P to 2.37 A for

80Si–17Ca–3P.

The O–Ca–O bond angle distribution (Fig. 4) shows a

peak close to 90� which results from the Ca ions in octa-

hedral geometry connecting two NBOs belonging to dif-

ferent Si/P-tetrahedra, and a second peak centered at 60�(less pronounced with respect to the previous one) derived

from the Ca coordination with two NBOs (or one NBO and

one BO) belonging to the same Si/P-tetrahedron [11].

0 2 4 6 80

2

4

6

8

10

12 80Si-15Ca-5P 80Si-15Ca-5P-OH

TD

F (

r)

r [Å]

Fig. 2 Total distribution function (TDF) for the 80Si–15Ca–5P (solid

line) and 80Si–15Ca–5P–OH (dotted line) glasses


123


The decreasing of the Si–O–Si bond angle with CaO

content is the consequence of the higher CaO/SiO2 ratio:

when the BO of a Si–O–Si linkage also interacts with Ca,

the Si–O–Si angle will decrease to accommodate it.

4.1.2 Coordination number (CN) distribution

and clustering effects

Table 4 reports the coordination number of Si, P and Ca to

Oxygens as well as the number of Ca ions coordinated to Si

and P. Each CN has been obtained by integration to the first

minimum in the corresponding pair distribution function

curves with cutoffs of 2.0, 2.0, 3.0, 4.2 and 4.2 A for Si–O,

P–O, Ca–O, Si–Ca and P–Ca, respectively. Si and P display

the classical tetrahedral coordination characteristic of net-

work former ions. Ca exhibits the typical behaviour of

network modifier ions or of charge compensator: it prefers

to occupy coordination sites with a highly distorted

geometry. In the binary SiO2–CaO systems the mean

coordination number of Ca increases as a function of Ca

content, consistently with the increasing of the Ca–O dis-

tance (Table 3). The mean Ca CNs slightly differ from that

reported in Ref. [27] (4.5–5.3), this is due to the different

potential utilized in the present work and more probably by

the cut-off used in the coordination number computation.

The introduction of hydroxyl groups changes signifi-

cantly the oxygen type contribution to the Ca coordination:

in fact, the CN of Ca in the OH free system (80Si–15Ca–

5P) is due to 3.3 BO, 1.7 NBO and 0 TBO, whereas in the

80Si–15Ca–5P–OH glass system this distribution shows an

increment of BO (3.5), a reduction of NBO (1.1), and the

appearance of 0.3 hydroxyl groups (OH). These results are

consistent with that reported in Ref. [27], where segrega-

tion of hydroxyl groups at the Ca sites with consequent

reduction of NBO near the Ca ions is highlighted.

The coordination numbers of Si–Ca and P–Ca found

in the simulated structure can be compared with those

Table 3 Mean bond distances and mean bond angles of the modelled glasses

Distances [A] Bond angles [degrees]

Si–O P–O Ca–O (O–H) O–Si–O O–P–O Si–O–Si (P–O–Si)

90Si–10Ca 1.616 – 2.304 108 – 154

80Si–20Ca 1.616 – 2.352 109 – 150

70Si–30Ca 1.616 – 2.368 108 – 148

80Si–17Ca–3P 1.616 1.520 2.368 108 109 149 (160)

80Si–17Ca–3P-big 1.616 1.536 2.336 108 108 150 (159)

80Si–15Ca–5P 1.616 1.536 2.336 108 108 157 (157)

80Si–15Ca–5P–OH 1.616 1.536 2.320 (1.070) 108 109 156 (158)

80Si–10Ca–10P 1.616 1.536 2.336 108 109 156 (160)

80Si–5Ca–15P 1.616 1.536 2.320 108 109 156 (161)

80Si–3Ca–17P 1.616 1.536 2.320 108 108 157 (160)

80Si–20P 1.616 1.536 – 108 109 158 (162)

0 1 2 3 4 50

1

2

3

4

5

6

7

8

PD

F g

O-H

(r)

r [Å]

O-H

Fig. 3 Pair Distribution Function of O–H pair for the 80Si–15Ca–

5P–OH glass

0 20 40 60 80 100 120 140 160 1800,0

0,5

1,0

1,5

2,0

2,5 90Si-10Ca 80Si-20Ca 80Si-15Ca-5P 80Si-15Ca-5P-OH 80Si-3Ca-17P

BA

D ρ

(θ)

Bond Angle θ [degrees]

Fig. 4 O–Ca–O Bond Angle Distribution (BAD) of modeled glasses


123


expected for a homogeneous distribution of ions given by

the formula CNT–Ca = 4/3 p Rc3 qCa, where Rc is the cut-off

distance (Si–Ca = P–Ca = 4.2 A) to the first minimum in

the T-Ca pair distribution function and qCa is the density

number of Ca ions in the system [10]. The analysis of these

data shows (Table 4) that in the binary systems Ca cations

are homogeneously distributed around the Si, while, in the

ternary systems, the Ca2? cations prefer to surround P

with respect to Si ones; this leads to the formation of Ca

phosphate rich regions. This clustering effect can be

quantified by the ratio fP–Ca between the P–Ca coordination

number found in the simulated model and the P–Ca2?

coordination number calculated for a homogenous distri-

bution also reported in Table 4. In particular, the clustering

effect is remarkable in the ternary system in the range

17–10% CaO, with a mean ratio value of 1.6. The very

similar ratio fP–Ca obtained for the glass systems 80Si–

17Ca–3P-big and 80Si–17Ca–3P seems to indicate that the

rich P–Ca domains found in these glasses is not an artificial

result due to the finite size of the box.

It is worth noting that in the 80Si–15Ca–5P glass 1.70 Ca

ions around each P ion are found, therefore, the composition

of the phosphate environment in this glass is similar to that

of HA for which the Ca/P atomic ratio is 1.6667. The

coordination numbers and the clustering effect determined

for the modeled glass using the melt-quenched and sol–gel

procedure are very similar and the presence of –OH groups

does not affect the preference of Ca ions to surround pref-

erentially the PO4 tetrahedra with respect to SiO4 ones.

4.1.3 Qn species distributions

The distributions of Qn species of the glasses studied

are reported in Table 5, together with the connectivity of

the silicate and phosphate networks, denoted NC(Si) and

NC(P), computed as weighted average of the correspond-

ing Qn(Si) and Qn(P) distributions.

This index highlights the major difference between the

structural model obtained by melt-quenched and sol–gel

procedure. In fact, the –OH groups destroy the silicate and

phosphate network connectivity leading to a lower value of

NC (Si) and (P) for the 80Si–15Ca–5P–OH with respect to

80Si–15Ca–5P (Table 5). This is also confirmed by the

lower percentage of BO showed in Fig. 5 for the 80Si–

15Ca–5P–OH with respect to the 80Si–15Ca–5P glass.

Thus, as expected, the sol–gel glass presents a network

connectivity lower with respect to that determined using

the melt-quenched model and this has to be taken into

account in the subsequent comparative analysis of the

series of glasses considered. No significant differences

were observed in the Qn distribution of the various chem-

ical species in the 80Si–17Ca–3P-big glass.

Table 5 shows that Si is predominantly Q4 for the 90Si–

10Ca and 80Si–20Ca glasses and Q3 for the 70Si–30Ca

glass with an overall decrement of the network connec-

tivity from 3.78 to 3.16. In Table 5 are reported the con-

tributions of Si–O–Si, Si–O–P and P–O–P units to the glass

network connectivity. It is interesting to note that in all

cases the number of P–O–P units are very low as confirmed

by the results of 31P MAS NMR study on similar glasses

[41]. Nevertheless, the Qn distribution obtained in Ref. [41]

showed only Q0 and Q1 species for P and this discrepancy

with MD results showed above (Table 5). This is a well

known drawback of structural models obtained by classical

MD simulations which overestimate the concentration of

Si–O–P units in the glass because of the force field, the

high cooling rates and the small simulation box size

employed [42].

In the ternary glasses Si is predominantly Q4 for all the

compositions although the network connectivity slightly

Table 4 Coordination numbers and clustering ratio (fT–Ca) of the modelled glasses

Coordination number (CN) Number of Ca atoms around Si and P atoms

Si–O P–O Ca–O O–Si O–P O–Ca Si–Caa fSi–Ca P–Caa fP–Ca

90Si–10Ca 4.0 – 5.0 1.90 – 0.26 0.73 (0.72) 1.01 –

80Si–20Ca 4.0 – 5.3 1.78 – 0.59 1.59 (1.54) 1.03 –

70Si–30Ca 4.0 – 5.5 1.65 – 0.97 2.67 (2.45) 1.09 –

80Si–17Ca–3P 4.0 4.0 5.2 1.67 0.13 0.46 1.19 (1.24) 0.96 1.99 (1.24) 1.61

80Si–17Ca–3P–big 4.0 4.0 5.2 1.68 0.12 0.47 1.19 (1.24) 0.96 2.02 (1.24) 1.63

80Si–15Ca–5P 4.0 4.0 5.0 1.60 0.20 0.40 0.97 (1.05) 0.92 1.70 (1.05) 1.62

80Si–15Ca–5P–OH 4.0 4.0 4.9 1.48 0.18 0.39 0.93 (1.03) 0.90 1.65 (1.03) 1.60

80Si–10Ca–10P 4.0 4.0 5.3 1.46 0.37 0.34 0.56 (0.65) 0.86 1.04 (0.65) 1.60

80Si–5Ca–15P 4.0 4.0 4.8 1.34 0.50 0.24 0.25 (0.30) 0.83 0.41 (0.30) 1.37

80Si–3Ca–17P 4.0 4.0 5.0 1.30 0.56 0.06 0.13 (0.18) 0.72 0.25 (0.18) 1.39

80Si–20P 4.0 4.0 – 1.24 0.62 – – –

a The corresponding values estimated for a homogeneous distribution are reported in parentheses


123


decreases from 3.94 to 3.61 when P2O5 is substituted by

CaO (Table 5). Phosphorous is predominantly Q3 for all

the compositions except for the 80Si–17Ca–3P for which it

is predominantly Q2. As for silicon, the network connec-

tivity of P decreases with CaO content. The overall parti-

tioning of the T–BO–T (T = Si and P) bridges in the

glasses is quantitatively examined in Fig. 5, which shows

that while the Si connectivity is dominated by ‘self’ Si–O–

Si linkages the situation is reversed for P, which does not

form P–O–P bridges, but prefers to cross-link with silicon.

This trend is very similar to that report by Tilocca et al.

[43] for phospho-silicate glasses obtained by a melt-

quenched computational procedure where the authors

investigated the effect of substitution of P2O5 for SiO2: at

lower % of P2O5 the PO4 tetrahedra are mainly Q0, whereas

at high % of P2O5 the PO4 units are linked with SiO4

tetrahedra with formation of Si–O–P bridges. In that paper,

the increment of glass network connectivity is correlated to

a decrement of the glass bioactivity. On the basis of this

consideration, among the glasses studied in the present

work higher bioactivity should be expected for the 80Si–

17Ca–3P composition, however other considerations have

to be done for the determination of Structure-Bioactivity

relationships, as reported in the following paragraph.

4.1.4 Structure-bioactivity relationships

Previous works showed that the 80Si–20Ca glass, presents

higher initial reactivity in SBF with respect to the ternary

Table 5 Qn species distributions and corresponding connectivity

Si P Si–O–Si %a Si–O–P %a P–O–P %a NC(Si) NC(P)

Q0 Q1 Q2 Q3 Q4 Q0 Q1 Q2 Q3 Q4

90Si–10Ca 0 0 2.1 18.1 79.8 89.6 3.78

80Si–20Ca 0 0.3 6.3 36.3 57.1 77.9 3.51

70Si–30Ca 0 3.5 15.8 42.8 39.9 65.1 3.16

80Si–17Ca–3P 0 0 4.3 30.3 65.4 0 9.6 42.8 32.8 14.8 71.4 7.7 0.1 3.61 2.53

80Si–17Ca–3P–big 0 0 4.0 30.9 65.1 0 9.4 42.7 33.3 14.6 71.6 7.7 0.0 3.61 2.53

80Si–15Ca–5P 0 0.2 2.9 26.0 70.9 0 5.7 38.9 44.6 10.8 67.2 12.7 0.1 3.68 2.60

80Si–15Ca–5P–OH 0 0.1 6.5 30.4 63.0 0 5.8 41.9 44.0 8.3 63.3 12.3 0.3 3.56 2.55

80Si–10Ca–10P 0 0 0.9 17.7 81.4 0 5.5 30.2 44.7 19.6 57.5 23.3 1.0 3.80 2.79

80Si–5Ca–15P 0 0 0.3 11.8 87.9 0 2.4 22.6 47.3 26.7 47.4 37.7 1.6 3.88 3.00

80Si–3Ca–17P 0 0 0 8.7 91.3 0 1.4 1.5 52.3 27.8 44.2 37.8 2.5 3.92 3.06

80Si–20P 0 0 0 6.4 93.6 0 1.6 15.5 50.5 32.4 39.8 41.8 3.4 3.94 3.14

a %Y–O–X = (no. of Y–O–X/no. of O) 9 100, where X,Y = Si and P

Fig. 5 Percentage of BOs linking any pair of tetrahedra (dark blue

diamond), linking two Si (magenta square), Si and P (light blue upper

triangle) and two P (green square) tetrahedra. The corresponding data

for the sol–gel modelled glass (80Si–15Ca–5P–OH) and bigger box

glass (80Si–17Ca–3P-big) are reported with filled circle and filled

triangle symbols, respectively (Color figure online)

Fig. 6 Ca/P ratio in the clustered zone, as computed from MD

simulations, versus molar % P2O5. The corresponding data for the

sol–gel modelled glass (80Si–15Ca–5P–OH) and bigger box glass

(80Si–17Ca–3P-big) are reported with filled circle and filled triangle

symbols, respectively


123


system 80Si–17Ca–3P: in few hours the glass is coated by

amorphous calcium phosphate (aCaP), whereas the con-

version of aCaP to crystalline carbonate hydroxyapatite

(HCA) nanocrystals similar to the biological ones needs

2 days. Whereas, the formation of aCaP on the surface of

the 80Si–17Ca–3P glass needs 2 days because of its higher

network connectivity, but the HCA crystallization is faster

because of the micro-segregation of Ca phosphate rich

regions with Ca/P ratio similar to that of hydroxyapatite.

Therefore, it might be asserted that the bioactive

response of binary glasses (80Si–20Ca) is controlled by a

quick Ca2?/H? interchange yielding to the formation of

new silanol groups (Si–OH) able to attract calcium and

phosphate ions in solution accelerating the formation of an

aCaP layer, while in the ternary glasses the higher con-

nectivity and lower Ca leaching cause a slower formation

of new silanol groups (Si–OH) able to promote the for-

mation of an aCaP.

From the results of the computational simulation study a

structural model of binary and ternary silicate glasses

emerges that correlates the bioactive behaviour to the

combined effects of the connectivity of the extended sili-

cate network and to the tendency to form (or not to form)

non-homogeneous domains. This offers the opportunity to

find a mathematical relationship (equation reported in

Fig. 6) between the Ca/P ratio in the Ca phosphate micro-

segregation zones and the P2O5 content in ternary glasses

(see Fig. 6); it can be used to predict the bioactive response

of new compositions of glasses still not synthesized or the

fine-tune of the optimum amount of P in a glass for its

highest in vitro bioactivity, by assuming that the glasses for

which the Ca/P ratio in the Ca phosphate micro-segregation

zones is closer to that of HA (Ca/P = 1.667) will promote

the fastest crystallization of aCaP to HCA. Therefore, the

composition with optimal Ca/P ratio of 1.6667, 80Si–

14.8Ca–5.2P, has been synthesized and bioactivity tests

have been carried out to confirm or confute the structural

models derived from computational simulations.

4.2 Bioactivity tests and model validation

Figure 7 reports the FTIR spectra of four glasses previ-

ously described [3, 4] as bioactive after different soaking

times in SBF. After 7 days, the spectra of all samples show

bands of phosphate and carbonate characteristic of a

positive in vitro bioactive response.

The FTIR spectra of the six glasses prepared for the first

time in this paper are shown in Fig. 8. As it can be

observed for the samples containing 5 and 5.2 mol-% of

P2O5 respectively, the characteristic bands of crystalline

phosphate at 563 and 602 cm-1 are present after 1 day of

soaking, indicative of a quicker bioactive response of these

samples. It must be considered that at t = 0 low intensity

bands are also observed in this region. This was explained

for the presence of some calcium phosphate nanocrystals in

the initial glasses that were detected by Transmission

Electron Microscopy [9] for a glass containing 3 mol-%

of P2O5. In addition, the glass with 10 mol-% P2O5, i.e.

Fig. 7 FTIR spectra of

previously studied glasses [3, 4]

before and after being soaked

for different times in SBF. As

observed, after 7 days of

soaking in all cases are visible

bands of phosphate and

carbonate groups indicative of a

positive (?) in vitro bioactive

response


123


80Si–10Ca–10P, seems to present the phosphate bands

after 7 days of soaking.

To confirm the bioactive response of glasses, they were

studied by SEM and EDS. In Fig. 9 the SEM micrographs of

80Si–14.8Ca–5.2P, 80Si–15Ca–5P and 80Si–10Ca–10P

before and after being soaked 7 days in SBF are presented.

The corresponding EDS spectra are also included. As it can

be observed at 7 days, only the surface of 80Si–14.8Ca–5.2P

Fig. 8 FTIR spectra of glasses prepared here for the first time before

and after being soaked for different times in SBF. As observed, the

first two glasses exhibit a very positive (??) in vitro bioactive

response, whereas for the last three glasses is negative (-). For 80Si–

10Ca–10P glass the situation is not conclusive by using this

characterization technique

Fig. 9 SEM micrographs and the corresponding EDS spectra of three

glasses with possible in vitro bioactive response after the studies

by FTIR spectroscopy. As observed these techniques confirm that

80Si–10Ca–10P glass is not bioactive because no new material was

formed on the glass surface


123


and 80Si–15Ca–5P appears covered by a layer of newly

formed material. In addition, the EDS analysis confirms

that the newly formed layer is mainly composed of cal-

cium and phosphorous. On the other hand, the 80Si–

10Ca–10P glass surface was not covered by new material

and the EDS spectrum indicated the same chemical

composition as in the initial glass. Any glass of this series

with P2O5 content higher than 10 mol-% did not pro-

duce a bioactive response. As an example in Fig. 10 the

unchanged surfaces of 80Si–5Ca–15P and 80Si–20P

before and after 7 days of soaking in SBF are presented.

In summary, the in vitro bioactivity tests confirm that the

glasses with P2O5 content of 5.2 and 5 mol % are the ones

presenting the quickest bioactive response as it was pre-

dicted by the molecular dynamics simulations.

5 Conclusions

The efficacy of molecular dynamics simulations to predict

the chemical composition of a sol–gel glass optimal for the

quickest positive bio-response in osseous tissue regenera-

tion has been demonstrated. In fact, the mathematical

relationship derived by means of a MD-derived structural

descriptor (Ca/P ratio in the calcium-phosphate rich

regions) determined that the appropriate glass composition

is 80% SiO2–14.8% CaO–5.2% P2O5 (80Si–14.8Ca–5.2P)

and the experiments have confirmed this prediction.

Acknowledgments Financial support from the Italian Ministry

MIUR (Project COFIN2006, Prot. 2006032335_005 and Project

COFIN2006, Prot. 2006033728 is gratefully acknowledged as well as

the research project MAT2007-61927 MAT2008-736 of CICYT

Spain. A.P. would like to thank the ‘Fondazione Cassa di Risparmio

di Modena’ for financial support.

References

1. Li R, Clark AE, Hench LL (1991) J Appl Biomat 2:231–239

2. Pereira MM, Clark AE, Hench LL (1994) J Biomed Mater Res

28:693–698

3. Martinez A, Izquierdo-Barba I, Vallet-Regi M (2000) Chem

Mater 12:3080–3088

4. Vallet-Regi M, Izquierdo-Barba I, Salinas AJ (1999) J Biomed

Mater Res 46:560–565

5. Vallet-Regi M, Ragel CV, Salinas AJ (2003) Eur J Inorg Chem

103:1029–1043

6. Kokubo T, Kushitani H, Sakka S, Kitsugi T, Yamamuro T (1990)

J Biomed Mater Res 24:721–734

7. Salinas AJ, Vallet-Regı M, Izquierdo-Barba I (2001) J Sol-Gel

Sci Technol 21:13–25

8. Salinas AJ, Martin AI, Vallet-Regi M (2002) J Biomed Mater Res

61:524–532

9. Vallet-Regi M, Salinas AJ, Ramirez-Castellanos J, Gonzalez-

Calbet JM (2005) Chem Mater 17:1874–1879

10. Tilocca A (2009) Proc R Soc A 465:1003–1027

11. Tilocca A, Cormack AN, de Leeuw NH (2007) Chem Mater

19:95–103

12. Mead RN, Mountjoy G (2006) J Phys Chem B 110:14273–14278

13. Pedone A, Malavasi G, Cormack AN, Segre U, Menziani MC

(2007) Chem Mater 19:3144–3154

14. Pedone A, Malavasi G, Menziani MC, Segre U, Cormack AN

(2008) J Phys Chem C 112:11034–11041

Fig. 10 SEM micrographs of

the glasses with high P2O5

contents before and after being

soaked 7 days in SBF. Any new

material is observed in the glass

surface confirming that both

glasses are non bioactive


123


15. Lusvardi G, Malavasi G, Cortada M, Menabue L, Menziani MC,

Pedone A, Segre U (2008) J Phys Chem B 112:12730–12739

16. Tilocca A, Cormack AN (2010) Langmuir 26:545–551

17. Garofalini SH (2001) Molecular modelling theory: application in

the geosciences. In: Cygan RT, Kubiki JD (eds) Reviews in

mineralogy geochemistry. Geochemical Society, Mineralogical

Society of America, Washington, DC

18. Pedone A, Malavasi G, Menziani MC, Cormack AN, Segre U

(2006) J Phys Chem B 110:11780–11795

19. Pedone A, Malavasi G, Menziani MC, Segre U, Cormack AN

(2008) Chem Mater 20:4356–4366

20. Huff NT, Demiralp E, Cagin T, Goddard WA III (1999) J Non-

Cryst Solids 253:133–142

21. Vollmayr K, Kob W, Binder K (1996) Phys Rev B 54:15808–

15827

22. Pota M, Pedone A, Malavasi G, Durante C, Cocchi M, Menziani

MC (2010) Comput Mater Sci 47:739–751

23. Lee BM, Baik HK, Seong BS, Munetoh S, Motooka T (2006)

Comput Mater Sci 37:203–208

24. Hanneman A, Schon JC, Jansen M (2005) J Mater Chem 15:

1167–1178

25. Petkov V, Holzhuter G, Troge U, Gerber T, Himmel B (1998)

J Non-Cryst Solids 231:17–30

26. Bhattacharya S, Kieffer J (2008) J Phys Chem C 112:1764–1771

27. Mead RN, Mountjoy G (2006) Chem Mater 18:3956–3964

28. Kolb M, Herrmann J (1985) J Phys A 18:L435–L441

29. Jullien R, Hasmy A (1995) Phys Rev Lett 74:4003–4006

30. Pereira JCG, Catlow CRA, Price GD (2001) J Phys Chem A

106:130–148

31. Elanany M, Selvam P, Yokosuka T, Takami S, Kubo M, Imamura

A, Miyamoto A (2003) J Phys Chem B 107:1518–1524

32. Priven A (2004) Glass Technol 45:244–254

33. SciGlass 3.5 (1997) SciVision, Burlington

34. Smith W, Forrester TR (1996) J Mol Graph 14:136–141

35. Zhravlev LT (1987) Langmuir 3:316–318

36. Lusvardi G, Malavasi G, Menabue L, Aina V, Bertinetti L,

Cerrato G, Morterra C (2010) Langmuir 26:10303–10314

37. Saravanapavan P, Hench LL (2003) J Non Cryst Solids 318:1–13

38. Pedone A (2009) J Phys Chem C 113:20773–20784

39. Bohner M, Lemaitre J (2009) Biomaterials 30:2175–2179

40. Skipper LJ, Sowrey FE, Pickup DM, Drake KO, Smith ME,

Saravanapavan P, Hench LL, Newport RJ (2005) J Mater Chem

15:2369–2374

41. Lenova E, Izquierdo-Barba I, Arcos D, Lopez-Noriega A, Hedin

N, Vallet-Regı M, Eden M (2008) J Phys Chem C 112:5552–

5562

42. Pedone A, Charpentier T, Malavasi G, Menziani MC (2010)

Chem Mater 22:5644–5652

43. Tilocca A, Cormack AN (2008) Il Nuovo Cimento B 123:1415–

1423


123


New insights into the bioactivity of SiO2–CaO and SiO2–CaO–P2O5 sol–gel glasses by molecular dynamics simulations

Documents