Journal of
Materials Chemistry
C
Dynamic Article Links <
Cite this: J. Mater. Chem., 2011, 21, 12660
PAPER
www.rsc.org/materials
Molecular dynamics simulations of a bioactive glass nanoparticle
Antonio Tilocca*
Received 2nd May 2011, Accepted 20th June 2011
DOI: 10.1039/c1jm11927c
Molecular dynamics simulations of a Bioglass spherical nanoparticle (6 nm diameter) have been
carried out to investigate the main structural features induced by the reduced size, which could play
a specific role in the observed enhanced bioreactivity of these systems, in addition to the higher surface
area. Compared to the bulk glass or to the extended flat surface, the simulations reveal that the most
relevant effects of the reduced size are a further slight reduction in the already low silicate connectivity
on the nanoparticle surface, together with a ring size distribution shifted towards three-membered
rings, and a higher Na+/Ca2+ ratio in close proximity of the surface. The possible ways in which these
effects can translate into higher bioreactivity are discussed.
Introduction
Due to their ability (‘‘bioactivity’’) to interact and rapidly form
stable bonds with hard and soft tissue, modified silicate glass
compositions such as 45S5 Bioglass (46.1SiO2; 24.4Na2O;
26.9CaO; 2.6P2O5 mol%) are widely employed in dental and
orthopaedic clinical applications.1 Compact bioactive glasses (in
solid blocks) are employed in low-load bearing applications, such
as middle-ear implants and maxillofacial reconstruction,2
whereas current clinical applications in orthopaedics and
dentistry mostly involve the use of glass microparticles as fillers
to treat bone and periodontal defects.3–5 The superior performances of bioactive glasses in these applications compared to
other materials have recently been associated with their ability to
promote growth of new tissue, thanks to the osteogenic activity
of the ionic dissolution products of the glass.6 The potential
ability of bioactive glasses to stimulate regeneration of living
tissues, rather than just replacing them, has triggered new
research on these materials, aiming at exploiting their potential
use in tissue engineering.7 A key factor to control both the tissuebonding and tissue-regeneration activity is the size of the glass
particles. Reducing the particle size below the conventional
micrometre scale leads to an enhancement in the bioreactivity of
materials for medical applications. For instance, nanoscale
features introduced on the surface improve the cellular response
of titania implant substrates;8–10 protein adhesion and proliferation were significantly enhanced on the surface of hydroxyapatite nanocrystals compared to the flat surface;11 compared to
microparticles, the presence of nanosized bioactive glass particles
enhances protein adsorption on composites;12 nanometre-sized
bioactive glass particles induce a higher dentin remineralisation
rate, again compared to micrometre-sized particles,13 and it has
Department of Chemistry, University College London, 20 Gordon Street,
London, WC1H 0AJ, UK. E-mail: a.tilocca@ucl.ac.uk
12660 | J. Mater. Chem., 2011, 21, 12660–12667
also been proposed that the reduction of particle size could
enhance the antibacterial properties of bioactive glasses.14 The
biocompatibility and non-toxicity of sub-micron bioactive glass
nanoparticles were also assessed and proven very recently.15
A combination of different factors is likely responsible for the
superior bio-reactivity of sub-micron glass particles: the high
surface area/volume ratio certainly contributes to enhance
protein adsorption and favours release of the ionic products of
glass dissolution; moreover, it has been suggested that the match
between nanoscale features on the glass surface and nanometresized collagen and apatite molecular units improve the contact at
the implant–tissue interface and enhances deposition and mineralisation of new living tissues.10 A deeper understanding of
these factors is often hampered by the intrinsic difficulty to
characterise systems of this size using experimental methods; on
the other hand, atomistic computer simulations can nowadays
approach nanosized systems in a relatively straightforward
fashion.16 Molecular Dynamics (MD) computer simulations
have recently shown their efficacy to reveal structural and
dynamical features of bulk bioglasses (BGs) which are correlated
to the glass bioreactivity.17 In particular, given the established
role of the glass dissolution in starting and sustaining the
bioactive process, the modelling studies of BGs have focused on
identifying links between specific structural features and the
tendency of a glass composition to degrade and dissolve in an
aqueous environment. Classical (employing empirical force
fields) Molecular Dynamics (MD) simulations of the bulk and of
the flat surface have helped to understand the role of several
structural factors in the bioactivity, such as the silicate network
connectivity,18,19 the presence of ring and chain fragments,20,21
and the tendency to form modifier (Na and Ca) cation-rich
aggregates.22 Ab initio (parameter-free) MD simulations have
examined the properties of the melt precursors23,24 and focused
on the reactivity of the flat surface, highlighting the main
adsorption sites and the mechanism of the initial dissolution
This journal is ª The Royal Society of Chemistry 2011
steps of 45S5 Bioglass.25,26 All these studies have involved
extended periodic systems, and no previous simulations have yet
targeted bioactive glass nanoparticles, even though a number of
recent studies27–33 has shown the efficacy of classical MD to
model a wide range of structural and dynamical properties of
crystalline and amorphous oxide particles with sizes up to 10 nm.
It must be remarked that the much heavier computational
demands of ab initio simulations do not currently allow their
applications to model systems of this size, containing several
thousands of atoms, even though the continuous developments
in the performances of electronic structure codes make it now
possible to tackle nanoclusters of smaller size, for instance in
a bottom-up approach.16
In the present work, classical MD simulations are carried out
to obtain the first realistic model of a bioactive glass nanoparticle: the particle diameter (6.5 nm) is representative of the
size of small BG nanoparticles that can be prepared using
experimental methods.34,35 The analysis is mainly focused on
identifying specific features associated with the reduced size and
high active surface area, which differ from the corresponding
bulk-like properties, and could thus represent key factors in the
special bioreactivity of nanosized BGs.
Computational methodology
Classical MD simulations were performed with the DL_POLY
2.20 code,36 using an ionic (full-charges) shell-model interatomic
potential18,37 recently developed to model multicomponent
phosphosilicate glasses. Oxygen atoms were represented as core–
shell units, with a small mass (0.2 atomic units) assigned to the
shells to allow them to follow the core motion adiabatically.38
Besides the harmonic interaction with the corresponding core,
oxygen shells interact with each other and with all cations
through short-range Buckingham potentials. Coulombic forces
act between all species and three-body truncated† harmonic
potentials are used to control the intra-tetrahedral O–Si–O and
O–P–O angles. The MD time step was 0.2 fs; no periodic
boundary conditions were applied, in order to model the isolated
nanoparticle. Electrostatic forces were calculated through direct
large enough to
Coulomb sum, with a distance cut-off of 70 A,
include all atomic pairs in the system. The cut-off for van der
Waals interactions was set to 8 A.
After a short relaxation at 300 K, the random sphere was
heated up and held at 2500 K in a constant-temperature MD run
of 240 ps, during which the total potential energy decreased and
converged to a stable value. A further microcanonical run of 240
ps at an average temperature of 2500 K followed, during which
the potential energy only showed thermal oscillations around the
same constant value. Starting from this equilibrated melt, the
particle was quenched to room temperature in four consecutive
constant-temperature runs at 2000–1500–1000 and 300 K,
respectively, lasting 55 ps each, corresponding to a nominal
cooling rate of 10 K ps1, which is the standard for MD
simulations of melt-derived glasses.41 A final microcanonical run
of 240 ps at an average temperature of 300 K completed the
simulation. The analysis presented in this manuscript was carried
out on all configurations extracted from this last trajectory,
sampled every 0.4 ps.
Results
A snapshot of the final structure of the 45S5 nanoparticle is
shown in Fig. 1. The particle maintained a roughly spherical
shape: the ratios I1/I2 and I1/I3 between the three calculated
principal moments of inertia were 1.05 and 1.08, and the corresponding a, b and c ellipsoid parameters were 33.65, 32.66 and
close to the 32 A
radius of the initial sphere.
30.95 A,
In order to characterise the nanoparticle, we have analysed
several structural parameters, focusing in particular on how the
structure changes moving from the internal to the surface region.
Fig. 2 shows the total number, density and fraction profiles of
each species i (i ¼ O, Si, Na, Ca, P) going from the centre to the
surface of the nanoparticle. Given the approximately spherical
shape of the nanoparticle, the profiles were calculated by dividing
the volume in concentric spherical shells of thickness dR ¼ 1 A,
located at distance (R dR) from the particle centre of mass.
The atomic fraction and density in each shell were calculated as
A standard approach involving melt-and-quench of a random
mixture, which has proven effective to model a wide range of
bulk17 and nanosized28,29,33,39 amorphous materials, was
employed to generate a model of a 45S5 Bioglass nanoparticle of
approximately 6 nm diameter. An initial mixture of 10 017 atoms
at the correct composition (46.1SiO2; 24.4Na2O; 26.9CaO;
2.6P2O5 mol%) was inserted at random positions in a sphere of
radius, yielding the experimental 45S5 Bioglass
31.725 A
density of 2.702 g cm3.40 Besides the density, the only constraints
applied in the initial structure generation were distance cutoffs to
prevent particles to start unphysically close to each other.
† We found that switching from the screened to the equivalent truncated
form of the harmonic 3-body term in the original potential improves the
stability of the trajectory and damping of core–shell motion is no longer
needed in the MD simulations.
This journal is ª The Royal Society of Chemistry 2011
Fig. 1 Structure of the Bioglass nanoparticle, extracted from the MD
trajectory at 300 K. Silicate and phosphate groups are represented as balland-sticks, Na and Ca as spheres. Si, P, Na, Ca, O are coloured yellow,
brown, blue, cyan and red, respectively.
J. Mater. Chem., 2011, 21, 12660–12667 | 12661
Fig. 2 Total number ni(R) (top panel), number density ri(R) (middle)
and fraction ci(R) (bottom panel) of different species as a function of
their distance from the nanoparticle centre of mass. Dashed horizontal
lines in the middle panel mark the corresponding number density of each
species in the bulk. The larger statistical fluctuations seen towards the
core reflect the decreasing number of atoms contained in volume elements
for R / 0.
ci(R) ¼ ni(R)/N(R) and ri(R) ¼ ni(R)/V(R), where ni and N are the
time-averaged number of atoms of species i and of all species
found in the shell, respectively, while V is the volume of the shell.
The middle panel of Fig. 2 reveals significant departures from the
bulk density of all species (bulk values are denoted by the dashed
from the particle centre. These
lines) starting beyond 30 A
deviations can be used to define the surface region of the nano thick) where the distribution
particle, as the external shell (5–6 A
is significant different from the bulk. This same region is also
highlighted by the relative atomic fractions shown in the bottom
panel of the figure: the surface is characterised by a marked
increase in the sodium fraction cNa, while the other fractions
decrease. As a consequence, the nanoparticle surface is
composed almost exclusively of Na and O, and significantly
depleted in Si and Ca compared to the bulk.
The time evolution of the atomic fractions in the core and
surface regions during the full MD trajectory (Fig. 3) shows that
Na enrichment at the surface mainly takes place during the
cooling, after which the converged fractions oscillate around
steady values, confirming that full structural relaxation has been
achieved within the simulation time.
Fig. 3 Time evolution of the atomic fractions in the core (top panel,
from the centre) and surface
including all atoms found at R < 15 A
regions during the melt-and-quench process.
(bottom panel, R > 30 A)
characteristic nearest-neighbours distances) is not significantly
affected by the reduced size. However, some interesting differences emerge in the peak heights. The lower intensity of the peaks
in the surface region (red curves) obviously reflects the empty
external region outside the nanoparticle, compared to a compact
the g(r)’s
bulk-like distribution. Below the surface (R < 30 A)
normally overlap with their bulk counterpart, denoting full
recovery of the bulk-like structure already in the internal regions
directly under the nanoparticle boundary. The only notable
exception is represented by the X–Na (X ¼ Si, Na, Ca) pairs:
unlike the other curves, all the gX–Na(r) functions (central row of
remain below the corresponding bulk values,
Fig. 4) for R < 30 A
Short-range structure
The short range structure was analysed through the pair distri thick
bution functions g(r) calculated separately in four 10 A
(10, 20 A),
(20, 30 A)
and
spherical shells spanning R ¼ (0, 10 A),
where R is the distance from the particle centre of mass
R > 30 A,
(Fig. 4). The good match of the distances corresponding to the
main peak in each of the nanoparticle and the bulk g(r)’s suggests that the short-range structure (as represented by the
12662 | J. Mater. Chem., 2011, 21, 12660–12667
Fig. 4 Radial distribution functions g(r) calculated in four spherical
shells located at different distances from the centre of mass of the BG
represents the surface region); the g(r)’s of the
nanoparticle (R > 30 A
bulk BG19 are also shown.
This journal is ª The Royal Society of Chemistry 2011
with a corresponding higher intensity in the gX–Na(r) functions at
surface. This shows that the Na enrichment at the
the (R > 30 A)
surface highlighted above is balanced by an uniform perturbation which depletes the sodium coordination shells surrounding
Si, Na, or Ca atoms throughout the nanoparticle. This uniform
depletion reflects the homogeneous distribution of Na ions in the
45S5 BG bulk.17,18
Silicate network connectivity
The Qn distributions (where Qn is a Si atom linked to another n
network formers through n bridging oxygens) are shown in
Fig. 5. The decrease in the Si and P density on the surface
results in a reduced silicate network connectivity compared to
the bulk. While the fraction of chain-like Q2 units does not
markedly change there, the Q1 and Q3 curves intersect as
a result of a surface depleted of branched Q3 units and richer in
Q1 chain-end units. The average silicate network connectivity in
the surface region (calculated from the Qn fractions there) turns
out to be 1.83, compared to 2.07 in the bulk:18 on average, on
the nanoparticle surface a Si atom forms 0.25 less Si–O–Si
bridges than in the bulk. This small, but non-negligible difference is also reflected in the Si–Si running coordination numbers
plotted in Fig. 6 (top panel), where again the contributions
from the inner and outer spherical shells of the particle were
distinguished by separately plotting the running coordination
number n(r) of each shell.
Coordination shell of the network-modifier ions
The middle and bottom panels of Fig. 6 focus on another
important feature of the nanoparticle surface: the reduced Ca–O
and (especially) Na–O coordination. In particular, the number of
nearest-neighbours oxygen atoms n(rc) (where rc is the distance of
the first corresponding g(r) minimum) is reduced from 5.8 in the
bulk to 4.0 in the surface for sodium, and from 6.1 to 5 for
calcium. The more marked depletion in the oxygen coordination
shell of exposed sodium atoms denotes that their increased
fraction at the surface is not fully balanced by enough oxygen
atoms, so that the latter are unable to complete the ideal pseudooctahedral coordination of Na in BGs.42
Fig. 5 Distribution of Qn silicate species going from the inner to the
external (surface) regions of the nanoparticle: R is the distance from
the particle centre of mass. The dashed lines denote the Qn distribution of
the corresponding bulk model.19
This journal is ª The Royal Society of Chemistry 2011
Fig. 6 Running (cumulative) coordination number for Na–O, Ca–O and
Si–Si pair interactions, calculated in different regions going from the
to the inner regions (R < 30 A).
The
nanoparticle surface (R > 30 A)
corresponding g(r) functions (dashed) and the bulk curves (blue) are also
shown for reference.
Rings
Additional surface effects emerge from the distributions of intraand inter-tetrahedral angles shown in Fig. 7. The surface is
characterised by significant distortion, especially in the Si–O–Si
angles (Fig. 7, top panel), whose distribution is shifted to lower
values compared to the bulk-like angle around 132 . Whereas
significant deviations in the O–Si–O and Si–O–Si angles can
result from three- and five-fold coordinated Si species,23 only
a negligible amount of these defects (less than 0.1% of all Si
atoms) was found in the nanoparticle.
Another possible source of distortion in these angles is
a different population of silicate rings compared to the bulk: Si
atoms in 3-membered rings form smaller Si–O–Si angles
compared to those found in 4-membered and larger rings which
are dominant in bulk bioglasses.20 This is also suggested by the
Fig. 7 O–Si–O and Si–O–Si angle distributions for Si atoms located in
and inner regions of the nanoparticle.
external (R $ 30 A)
J. Mater. Chem., 2011, 21, 12660–12667 | 12663
Fig. 8 Si–Si–Si angle distributions for Si atoms located in external (R $
and inner regions of the nanoparticle.
30 A)
Si–Si–Si angle distributions in Fig. 8, where a significant increase
in the peak around 60 , characteristic of 3-membered rings,20 is
observed for the Si atoms belonging to the surface of the nanoparticle, accompanied by a reduced intensity of those Si–Si–Si
angles between 90 and 140 associated to 4- and 5-membered
rings.20 The ring size distributions were explicitly calculated and
shown in Fig. 9, confirming that three is indeed the dominant
ring size of surface regions, whereas four-membered rings
dominate the bulk-like core of the nanoparticle.20
Sodium mobility
Having identified the main structural features of the nanoparticle, it is important to examine the dynamical behaviour of
the system. In particular, it is interesting to focus on the
dynamics of Na atoms because their local environment on the
surface shows marked deviations from the bulk. Another
important reason is that the high computational requirements of
the present simulations, due to the large size and non-periodic
nature of the system, limit the time window which can be
explored to an interval where no significant ionic diffusion occurs
at room temperature (atoms do diffuse in the melt, but it is the
room temperature dynamics that is of interest here). Within this
relatively static picture, sodium atoms are the most mobile and
their dynamical behaviour is thus easier to characterise. The flat
evolution of the Mean Square Displacement (MSD) of sodium in
different regions (whose average MSD’s are also shown in Fig. 10
Fig. 9 Ring size distributions (number of rings per tetrahedral atom)
and inner (R < 30 A)
regions of the
calculated in the surface (R > 30 A)
nanoparticle.
12664 | J. Mater. Chem., 2011, 21, 12660–12667
Fig. 10 Mean Square Displacement of individual Na ions belonging to
different regions of the nanoparticle, going from the R ¼ 0 core to the R >
surface. The averaged MSD(Na) in each region (bold coloured lines)
30 A
is also plotted.
as coloured lines) confirms that no long-range diffusion occurs at
room temperature on the simulated time window. However, the
individual MSD curves denote a small but increasing fraction of
more mobile Na ions as we move away from the centre of the
nanoparticle (going from the bottom to the top panels in Fig. 10).
Direct inspection of the MD trajectories of these atoms shows
that their dynamics can be described as either a single jump or
a shuttling motion between two adjacent sites. These local hops
appear more frequently for Na atoms located close to the
surface, as shown by the traces of the individual MD trajectories
plotted in Fig. 11, only for the Na ions located near the centre
and near the edge of the nanoparticle. The figure highlights small
groups of adjacent Na ions on the surface whose red traces are
generally more expanded. Because the migration of mobile Na
atoms in bulk BGs involves correlated hops, in a vacancy-like
Fig. 11 3D traces of the individual MD trajectories of Na cations
located near the centre (blue) and near the surface (red) of the nanoparticle. Atoms found in between these two regions are not shown for
clarity. Note that the red traces which apparently overlap with the blue
ones in the centre actually mark the trajectory of atoms located on the
(longitudinally) opposite external side of the sphere.
This journal is ª The Royal Society of Chemistry 2011
mechanism,42 it is plausible to come across small aggregates of
more mobile ions on the nanoparticle surface.
Discussion
Table 1 reports the concentrations of specific sites found on the
surface of the nanoparticle and the flat surface.21 In general, the
BG nanoparticle shows a rather regular structure, with surface
The equilibrated
effects mostly limited to the external 5–10 A.
nanoparticle does not contain any structural defect: in particular,
neither under-coordinate Si defects (Si3c) nor two-membered
(2M) rings, which were found, albeit in small concentrations, on
the flat 45S5 Bioglass surface,21,25 are present. This suggests
that these specific defects do not contribute to the special bioreactivity of nanosized bioglasses: the main surface effects turn
out to involve: (i) high Na+/Ca2+ and Si(Q1)/Si(Q3) ratios; (ii)
a relevant fraction of three-membered rings; and (iii) enhanced
mobility of Na+.
In common with the flat surface, the dry surface of the ascreated BG nanoparticle is significantly enriched in sodium and
depleted in calcium cation: Fig. 2 denotes a significant population of sodium atoms extending through the boundaries of the
nanoparticle, accompanied by a corresponding slight but
constantly lower sodium density in the internal regions of the
nanoparticle, compared to the bulk glass (Fig. 4). We have
shown that the coordination shell of these exposed sodium atoms
is significantly depleted of oxygen atoms, compared to their
pseudo-octahedral coordination in the bulk (Fig. 6). The marked
under-coordination of exposed Na+ ions will enhance their Lewis
acidity; combined with the marked Ca2+ depletion at the nanoparticle surface (Table 1 and Fig. 2), one can thus expect that
once the glass particle is immersed in an aqueous medium, the
initial interface will be dominated by Na+–water interactions.
Because high bioactivity is associated with (and, to a large extent,
depends on) a fast initial dissolution of the glass once immersed
in a physiological medium,1 these observations suggest that
a high Na+/Ca2+ ratio is an important feature of as-created BG
surfaces. Whereas this feature applies to both nanosized and
larger BG samples,21 Table 1 shows that the Na+/Ca2+ ratio at the
BG nanoparticle surface (4.29) is even larger than that measured
at the flat surface (3.04), with both being much larger than the
bulk ratio of 1.81. This could represent an important effect of the
reduced size, which could contribute to a faster initial dissolution
of the nanoparticle.
Experimental and simulation data have revealed a large excess
of non-bridging oxygen and alkali ions on the freshly created
surface of silicate glasses, compared to the bulk.43–48 It has been
proposed that the increased alkali fraction shields the negative
charge of the excess NBOs created upon exposure of a fresh
surface, which is thus stabilized by the relocation of mobile alkali
cations in the outermost regions.43,48 This appears to be the
driving force behind sodium enrichment of the BG surface as well:
Fig. 12 (bottom panel) highlights a net increase in the NBO : BO
fraction in the surface region, with a corresponding increase in the
number of NBO per Si in the same region, also accompanied by
a decrease in the NBO : Na ratio (top panel of Fig. 12).
In mixed alkali–alkaline earth silicate glasses, sodium and
potassium cations showed a much stronger tendency to transfer
to the surface compared to calcium.46–48 Both thermodynamic
and kinetic factors appear to play a role in this effect: on one
hand, the more mobile alkali cations can react more rapidly to
the sudden change in the local field created by the exposure of the
surface;46 on the other hand, a lower free energy environment has
been shown to favour the preferential migration of sodium to the
surface.48 Similar effects are presumably involved also in this
case: Na cations are much more mobile than Ca in BGs.42
Moreover, despite their similar size, the different charge results in
a higher field strength for Ca2+ than Na+ and in a higher site
selectivity. Previous MD simulations42 have shown that Ca ions
enforce stricter requirements on their local environment in BGs,
whereas Na ions are more flexible in this respect, and can more
easily accept lower coordination numbers than those they normally have in the bulk. Na is thus more stable in a non bulk-like
environment with a decreasing density of O neighbours, such as
that found on the nanoparticle surface.
Even though no defects are present, the silicate network in the
surface region of the nanoparticle appears significantly perturbed
with respect to the bulk, both in terms of inter-tetrahedral angles
(Fig. 7) and of ring sizes (Fig. 8 and 9). The dominance of threemembered rings, in particular, is important: exposed small rings
Table 1 Concentration (nm3) of ionic sites found on the 45S5 BG
nanoparticle surface and on the corresponding flat surfacea
Site
Flat surface21
Nanoparticle
Na+
Ca2+
Si3c
NBO
2MR
3MR
7.0
2.3
0.13
11.75
0.17
0.58
7.3
1.7
0
11.9
0
0.32
a
In order to directly compare the site density with that calculated for the
thick spherical shell located
flat surface,21 we have considered the 4 A
(see Fig. 2).
between R ¼ 30 and R ¼ 34 A
This journal is ª The Royal Society of Chemistry 2011
Fig. 12 (Bottom panel) NBO and BO relative fractions; (top panel)
NBO : Si and NBO : Na ratios, as a function of the distance from the
nanoparticle centre of mass.
J. Mater. Chem., 2011, 21, 12660–12667 | 12665
have been associated with bioactivity, as they can represent
nucleation sites for the deposition of calcium phosphates (CaP)
in an aqueous environment at the conditions (neutral pH) found
in a physiological medium.49,50 The relative stability of these sites
on hydrated BG surfaces, highlighted by previous simulations,17,25,26 could in fact enable a fraction of the small rings
initially present on the dry surface to resist hydrolytic opening
upon contact with moisture and thus remain available to support
CaP nucleation.
Even though the low connectivity of the 45S5 silicate network
reduces the density of rings that can be formed throughout the
structure (compared to higher-silica compositions), the more
marked presence of small rings in the external regions of the
nanoparticle can thus represent another key factors in its bioreactivity.25,21
Previous simulations21 had highlighted how the effective
structural relaxation is able to restore the bulk-like network
connectivity to a large extent on the flat BG surface, which did
not differ significantly from the bulk in this respect. On the other
hand, Fig. 5 and 13 show that the nanoparticle surface, while
maintaining similar amounts of other Qn species, has a higher Q1/
Q3 ratio, and a correspondingly lower network connectivity, than
the bulk and flat surface. This is another potentially critical
factor leading to a faster release of soluble silica fragments from
the nanoparticle into the contact medium and thus accelerating
its dissolution and the release of ionic products which are deemed
essential to trigger tissue regeneration.6
The higher mobility of Na cations at the surface, compared to
the core, reflects the reduced number of oxygen atoms found in
their coordination shell: previous simulations have shown that
the diffusive hops of sodium between two sites in BGs involve
intermediate species where a fraction of the Na–O interactions in
the starting site has been lost.42 Due to the reduced Na–O
coordination, the energetic cost of creating these intermediates
will be reduced on the nanoparticle surface thus reducing the
hopping barrier. Another factor to take into account is the
reduced network density in the surface region: on one hand this
could facilitate the local structural distortions needed to
accommodate cations hopping to a new site,42 but on the other
hand the lower density could also result in a reduced number of
new favourable sites able to receive migrating Na. Additional
simulations sampling much longer time scales would be needed
to quantitatively address this important issue: the present results
only give some indications that cation migration could be
enhanced on the surface of a BG nanoparticle, which could
favourably affect the cation release in solution.
Conclusions
The present simulations have analysed in detail the structure of
an isolated, dry Bioglass nanoparticle. Upon immersion in an
aqueous physiological medium, the surface will adsorb and react
with water, and the hydrolysis will start the dissolution of the
glass.1 Before explicitly studying these key processes, it is vital to
understand the main features of the non-hydrated particle:
because glass degradation will not proceed to any significant
extent before the particle is set in a liquid environment, focusing
on its dry form represents a suitable source of information and
a reproducible reference to start exploring the properties of these
systems (in the same spirit of other simulations of isolated/in
vacuo nanoparticles28–33), which will be extended in the future
with simulations of the nanoparticle/water interface.
The analysis shows that the nanometre size maintains and
enhances some of the properties of the BG bulk and flat surface
most beneficial for the bioactive behaviour. The most important
one is probably the high fragmentation of the silicate network:
the network connectivity further decreases on the nanoparticle
surface, with a Q1/Q3 ratio higher than in the bulk and in the flat
surface, which could promote a faster release of soluble silica
species (a key effect for the tissue-regeneration properties of these
materials6) from nanosized BG particles. An important role is
most likely played by three-membered silicate rings present in
a higher fraction on the nanoparticle surface than in the bulk and
whose association with the tissue-bonding bioreactivity as
nucleation sites for calcium phosphate has been previously
proposed.49 Finally, compared to the flat surface, the nanoparticle denotes an even more marked enrichment in Na+ ions
with reduced oxygen coordination and (consequently) higher
mobility: these features will accelerate the initial stages of the
dissolution process, which involve exchange and release of these
ions in the surrounding physiological medium. In addition and
combined with the higher surface area, these specific effects could
all contribute to the enhanced bioreactivity of sub-micron and
nanosized Bioglass particles. In order to assess their relative
weight, further simulations of the explicit interface between the
nanoparticle and an aqueous medium are in progress.
Acknowledgements
The author would like to acknowledge financial support from
Royal Society (University Research Fellowship). Dr K. Okhotnikov (Stockholm University) is acknowledged for prompting
the modification of the 3-body function in the force field.
Computer resources on the HECToR national high-performance
computing service were provided through the EPSRC-funded
Materials Chemistry Consortium (EP/F067496).
Notes and references
n
19
21
Fig. 13 Comparison of Q distributions of bulk, flat surface
surface region).
nanoparticle surface (averaged over the R > 30 A
12666 | J. Mater. Chem., 2011, 21, 12660–12667
and
1 L. L. Hench, J. Am. Ceram. Soc., 1998, 81, 1705.
2 I. Kinnunen, K. Aitasalo, M. Pollonen and M. Varpula,
J. Craniomaxillofac. Surg., 2000, 28, 229.
3 J. Wilson and S. B. Low, J. Appl. Biomater., 1992, 3, 123.
This journal is ª The Royal Society of Chemistry 2011
4 I. Allan, H. Newsam and M. Wilson, Biomaterials, 2001, 22, 1683.
5 M. J. Peltola, K. M. J. Aitasalo, A. J. Aho, T. Tirri and
J. T. K. Suonp€
a€
a, J. Oral Maxillofacial Surg., 2008, 66, 1699.
6 I. D. Xynos, A. J. Edgar, L. D. K. Buttery, L. L. Hench and
J. M. Polak, J. Biomed. Mater. Res., 2001, 55, 151.
7 L. L. Hench and J. M. Polak, Science, 2002, 295, 1014.
8 R. A. Gittens, T. McLachlan, R. Olivares-Navarrete, Y. Cai,
S. Berner, R. Tannenbaum, Z. Schwartz, K. H. Sandhage and
B. D. Boyan, Biomaterials, 2011, 32, 3395.
9 G. Mendonca, D. B. S. Mendonca, L. G. P. Simoes, A. L. Araujo,
E. R. Leite, W. R. Duarte, F. J. L. Aragao and L. F. Cooper,
Biomaterials, 2009, 30, 4053.
10 E. Palin, H. Liu and T. J. Webster, Nanotechnology, 2005, 16, 1828.
11 S. Okada, A. Nagai, Y. Oaki, J. Komotori and H. Imai, Acta
Biomater., 2011, 7, 1290.
12 S. K. Misra, D. Mohn, T. J. Brunner, W. J. Stark, S. E. Philip, I. Roy,
V. Salih, J. C. Knowles and A. R. Boccaccini, Biomaterials, 2008, 29,
1750.
13 M. Vollenweider, T. J. Brunner, S. Knecht, R. N. Grass, M. Zehnder,
T. Imfeld and W. J. Stark, Acta Biomater., 2007, 3, 936.
14 V. Mortazavi, M. Mehdikhani Nahrkhalaji, M. H. Fathi,
S. B. Mousavi and B. N. Esfahani, J. Biomed. Mater. Res., Part A,
2010, 94, 160.
15 S. Labbaf, J. R. Jones, O. Tsigkou, K. H. M€
uller, M. M. Stevens and
A. E. Porter, Biomaterials, 2011, 32, 1010.
16 S. T. Bromley, I. P. R. Moreira, K. M. Neyman and F. Illas, Chem.
Soc. Rev., 2009, 38, 2657.
17 A. Tilocca, J. Mater. Chem., 2010, 20, 6848.
18 A. Tilocca, A. N. Cormack and N. H. de Leeuw, Chem. Mater., 2007,
19, 95.
19 A. Tilocca, J. Chem. Phys., 2008, 129, 084504.
20 A. Tilocca, A. N. Cormack and N. H. de Leeuw, Faraday Discuss.,
2007, 136, 45.
21 A. Tilocca and A. N. Cormack, Langmuir, 2010, 26, 545.
22 J. K. Christie and A. Tilocca, Chem. Mater., 2010, 22, 3725.
23 A. Tilocca, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76,
224202.
24 J. K. Christie, A. Pedone, M. C. Menziani and A. Tilocca, J. Phys.
Chem. B, 2011, 115, 2038.
25 A. Tilocca and A. N. Cormack, J. Phys. Chem. C, 2008, 112, 11936.
This journal is ª The Royal Society of Chemistry 2011
26 A. Tilocca and A. N. Cormack, ACS Appl. Mater. Interfaces, 2009, 1,
1324.
27 H. Zhang, B. Gilbert, F. Huang and J. F. Banfield, Nature, 2003, 424,
1025.
28 A. Roder, W. Kob and K. Binder, J. Chem. Phys., 2001, 114, 7602.
29 V. Van Hoang, J. Phys. Chem. B, 2007, 111, 12649.
30 T. X. T. Sayle, S. C. Parker and D. C. Sayle, Chem. Commun., 2004,
2438.
31 D. Spagnoli, J. C. Banfield and S. C. Parker, J. Phys. Chem. C, 2008,
112, 14731.
32 V. N. Koparde and P. T. Cummings, J. Phys. Chem. B, 2005, 109, 24280.
33 I. V. Schweigert, K. E. J. Lehtinen, M. J. Carrier and
M. R. Zachariah, Phys. Rev. B: Condens. Matter Mater. Phys.,
2002, 65, 235410.
34 X. Chen, C. Guo and N. Zhao, Appl. Surf. Sci., 2008, 255, 466.
35 S. Lin, C. Ionescu, K. J. Pike, M. E. Smith and J. R. Jones, J. Mater.
Chem., 2009, 19, 1276.
36 W. Smith and T. R. Forester, J. Mol. Graphics, 1996, 14, 136.
37 A. Tilocca, N. H. de Leeuw and A. N. Cormack, Phys. Rev. B:
Condens. Matter Mater. Phys., 2006, 73, 104209.
38 P. J. Mitchell and D. J. Fincham, J. Phys.: Condens. Matter, 1993, 5,
1031.
39 V. Van Hoang and B. T. H. Khanh, J. Phys.: Condens. Matter, 2009,
21, 075103.
40 C. C. Lin, L. C. Huang and P. Shen, J. Non-Cryst. Solids, 2005, 351,
3195.
41 K. Vollmayr, W. Kob and K. Binder, Phys. Rev. B: Condens. Matter
Mater. Phys., 1996, 54, 15808.
42 A. Tilocca, J. Chem. Phys., 2010, 133, 014701.
43 J. J. Kelso, C. G. Pantano and S. H. Garofalini, Surf. Sci., 1983, 134,
L543.
44 S. H. Garofalini and S. M. Levine, J. Am. Ceram. Soc., 1985, 68, 376.
45 D. M. Zirl and S. H. Garofalini, J. Am. Ceram. Soc., 1992, 75, 2353.
46 O. Gedeon and J. Zemek, J. Non-Cryst. Solids, 2003, 320, 177.
47 A. Abbas, J. M. Delaye, D. Chaleb and G. Calas, J. Non-Cryst.
Solids, 2003, 315, 187.
48 L. R. Corrales and J. Ju, J. Am. Ceram. Soc., 2006, 89, 36.
49 N. Sahai and M. Anseau, Biomaterials, 2005, 26, 5763.
50 V. Bolis, C. Busco, V. Aina, C. Morterra and P. Ugliengo, J. Phys.
Chem. C, 2008, 112, 16879.
J. Mater. Chem., 2011, 21, 12660–12667 | 12667