Laura J.
Skipper
*a,
Frank E.
Sowrey
a,
David M.
Pickup
a,
Kieran O.
Drake
bc,
Mark E.
Smith
b,
Priya
Saravanapavan
d,
Larry L.
Hench
d and
Robert J.
Newport
a
aSchool of Physical Sciences, University of Kent, Canterbury, Kent, UK CT2 7NR. E-mail: ljs32@kent.ac.uk; Fax: +44 1227 827558; Tel: +44 1227 823776
bDepartment of Physic, University of Warwick, Coventry, UK CV4 7AL. Fax: +44 2476 692016; Tel: +44 2476 522380
cESRF, 6 Rue Jules Horowitz, BP 220, 38043, Grenoble, Cedex 9, France. Tel: +33 4388 81904
dDepartment of Materials, Imperial College London, London, UK SW7 2AZ. Fax: +44 2075 946767; Tel: +44 2075 843194
First published on 29th April 2005
We have used neutron diffraction with isotopic substitution to gain new insights into the nature of the atomic scale calcium environment in bioactive sol–gel glasses, and also used high energy X-ray total diffraction to probe the nature of the processes initiated when bioactive glass is immersed in vitro in simulated body fluid (SBF). Recent work has highlighted the potential of sol–gel derived calcium silicate glasses for the regeneration or replacement of damaged bone tissue. The mechanism of bioactivity and the requirements for optimisation of the properties of these materials are as yet only partially understood but have been strongly linked to calcium dissolution from the glass matrix. The data obtained point to a complex calcium environment in which calcium is loosely bound within the glass network and may therefore be regarded as facile. Complex multi-stage dissolution and mineral growth phases were observed as a function of reaction time between 1 min and 30 days, leading eventually to the formation of a disordered hydroxyapatite (HA) layer on the glass surface, which is similar to the polycrystalline bone mineral hydroxyapatite. This methodology provides insight into the structure of key sites in these materials and key stages involved in their reactions, and thereby more generally into the behaviour of bone-regenerative materials that may facilitate improvements in tissue engineering applications.
The local calcium environment in crystalline calcium silicate minerals and apatites is extremely diverse; in most minerals there are several crystallographically distinct Ca environments. This diversity, and thus the complexity, of environments is associated with a wide range of possible calcium–oxygen coordination numbers. Amorphous materials are intrinsically even more structurally complex in the sense that the long-ranged order of the crystalline form is lost. Analysis of the short-range environment in calcium silicate glasses has been attempted using X-ray absorption fine structure and near edge structure (EXAFS and XANES) and X-ray powder diffraction,9–11 however the complexity of the calcium environment limits the useful information available from these techniques. For instance, conventional diffraction measurements show the Ca–O correlations as a broad feature centred around 2.35 Å, but this feature is overlapped by the strong O–O correlation making a quantitative measure of Ca environment impossible based solely on such data.
The Q-space simulation of the structure factor is generated using the following equation (eqn. 1):
p(Q)ij = (Nijωij/Cj)(sinQRij/QRij)exp(−Q2σ2/2) | (1) |
The analysis of the data to provide an interference function, i(Q), follows the method outlined by Warren20 and implemented by Pickup et al.21 The data were normalised to the measured incident beam intensity, corrected for X-ray beam polarisation, and the small contribution of the empty sample cell was subtracted. Further corrections were made to account for absorption of the incident beam by the sample and for Compton (inelastic) scattering—noting that this could only be carried out approximately in the case of those samples reacted with SBF, which were necessarily inhomogeneous as a result. Real-space pairwise atomic correlations may be generated by Fourier transformation of the interference function using the following generic relationship (eqn. 2):
D(r) = 2/π ∫Qi(Q) sinQr dr | (2) |
It is important to note that, due to the inhomogeneous nature of the reacted samples and the fact that the correlation function represents an average over all pairwise terms, weighted by concentration and element-dependent X-ray form factors, the D(r) may only be analysed quantitatively in the case of an unreacted sample; relative trends are however more easily derived.
The two neutron diffraction patterns were collected on the GEM diffractometer at the ISIS spallation neutron source at the Rutherford Appleton Laboratory, UK. The resultant interference function, the structure factor i(Q), can reveal real-space structural information by Fourier transformation to give the corresponding total pair correlation function (eqn. 3):18
T(r) = T0(r) + 2/π∫Qi(Q)M(Q)sin(Qr)dQ | (3) |
Hence, if two experiments are performed in which the scattering length of element A is varied by isotopic substitution, the difference between the experimental correlation functions is of the form (eqn. 4):23
ΔT(r) = T(r) − T′(r) = c2A(b2A − b2A′)t′AA′ + 2ΣcAcjbj(bA − bA′)t′Aj(r) | (4) |
The i(Q) of the natCa and 44Ca-enriched (CaO)0.3(SiO2)0.7 samples are shown in Fig. 1, together with the corresponding weighted difference between the two. The structural parameters derived from fitting to the difference function are provided in Table 1 with additional information on the ‘host’ silica network derived from the original diffraction data and from complementary X-ray diffraction. The Fourier transform of the difference function, and the associated numerical fit using the parameters presented in Table 1, are shown in Fig. 2; this provides information exclusively on the Ca–X pair correlations present in the material. The slight ripple through the data is a residual artefact of the data analysis process.
![]() | ||
Fig. 1 i(Q) data for natCa (top) and 44Ca (middle) (CaO)0.3(SiO2)0.7 and difference between i(Q) functions from the 44Ca and natCa samples. natCa and difference have been offset by +0.2 and −0.2 respectively for clarity. |
![]() | ||
Fig. 2 Difference correlation function, tCa–X(r), obtained by Fourier transformation of the difference between i(Q) functions from the 44Ca and natCa samples (solid line) and simulation of the tCa–X(r) function (dotted line). |
Correlation | R/Å (±0.01) | N (±0.25) | S/Å (±0.01) |
---|---|---|---|
D–D | 1.28 | 1.02 | 0.05 |
Si–O | 1.61 | 3.8 | 0.05 |
Si–D | 2.20 | 0.7 | 0.15 |
Ca–O | 2.32 | 2.3 | 0.08 |
Ca–O | 2.51 | 1.65 | 0.07 |
O–O | 2.64 | 4.64 | 0.09 |
Ca–O | 2.75 | 1.05 | 0.08 |
Ca–D | 2.95 | 0.6 | 0.08 |
The parameters obtained for Si–O and O–O are as one would expect for porous sol–gel derived silica21 and this lends confidence to the fitting of the other correlations observed. In a fully condensed silica network the O–O coordination number would be 6, in these materials it is reduced to ∼4.6 due to the presence of non-bridging oxygen atoms (NBO) which are abundant in high surface area materials. It is important also to note that the parameters obtained from the fit to the difference correlation function, tCa–X(r), may be carried over to the fitting of the original datasets unchanged: we take this as further evidence that the fits generated to tCa–X(r) are robust.
The fit parameters obtained uniquely from the NDIS results show clearly that the Ca–O environment actually consists of distinct, but partially overlapping, correlation shells centred at 2.3 Å, 2.5 Å and 2.75 Å. This is the first time a Ca–O environment of this complexity has been discerned in the context of contemporary bioactive glass materials: it is a key observation given the central role that Ca dissolution plays in the material's ability to promote bone growth.
It is helpful at this point to draw comparisons with crystalline calcium silicate materials where Ca–O distances ranging from 2.3 to 2.8 Å have been observed, indicating that the values obtained here are certainly physically reasonable per se. In theory it should be possible to extract information from the NDIS data relating to higher correlations such as those from Ca–Si and Ca–Ca. However in practice, even using this method, the residual overlaps between the large range of possible correlations above ∼3 Å renders this unreliable as a quantitative exercise. What the NDIS results do offer is the uniquely detailed quantitative data required for computer simulation studies of bioactive glasses, and in particular towards the full understanding of the Ca dissolution and subsequent mineral deposition processes.
![]() | ||
Fig. 3 Relative mol% of P2O5 and CaO calculated from XRF measurements of (CaO)0.3(SiO2)0.7 before and after exposure to SBF. (Corroborative evidence regarding calcium content measured using X-ray absorbance has been previously reported.10) |
Conventional powder diffraction experiments typically focus on the observed Bragg peaks (in Q-space), and the analogous studies of amorphous materials focus on the diffuse pair correlation function (in real-, r-, space). Here, a total diffraction strategy is adopted26 in which Bragg analysis in Q-space and Fourier analysis of the r-space pair distribution, to provide both crystallographic and local order information, are combined. The experimentally determined interference functions, i(Q), for the unreacted and reacted samples are shown in Fig. 4. It is immediately apparent that, after only 30 min, the shape of the i(Q) has altered significantly, most noticeably on the high-Q side of the first principal diffraction peak at ∼2.1 Å−1; our earlier published work on the structure of these glasses as a function of calcium content enables it to be concluded that this change is associated with the initial rapid dissolution of Ca.5 After 1 h, whilst the underlying amorphous pattern remains, there is clear evidence for the formation of Bragg diffraction peaks: i.e. of the growth of crystallites within the sample. The peaks observed at 5 happear very strong (although the intensity may in part be the result of an inhomogeneity of crystallite concentration within the powder sample) and may be attributed to the presence of calcium phosphate. In particular, the two peaks at 2.8 and 9.4 Å−1, seen only in the 1 and the 5 h spectra, are associated with the (412) and (200) reflections from octacalcium phosphate respectively (and not with hydroxyapatite).16 However, by 10 h the polycrystalline calcium phosphate appears to have been largely replaced by a more dominant disordered phase which, in its turn becomes increasingly prominent and well-defined thereafter. By 25 h, and even more clearly by 72 h, the i(Q) has the appearance of a poorly crystalline hydroxyapatite overlaying a (Ca-depleted) silicate glass interference pattern. Given that hydroxyapatite recrystallises at ∼600 °C, the 7 day sample was annealed in order to help verify its presence: the diffraction features seen in Fig. 4 readily sharpened to reveal the presence of polycrystalline hydroxyapatite.
![]() | ||
Fig. 4 i(Q) data for SBF treated samples: from top to bottom, 7 days, 72 h, 25 h, 10 h, 5 h, 1 h, 30 min, unreacted. |
The corresponding real-space pair correlation function, D(r), allows us to consider the amorphous (short-range) component in more detail. Fig. 5 shows the D(r) curves for the untreated and the reacted samples. In order to bring out the important qualitative changes taking place, we have chosen to scale the various D(r) such that the area of the Si–O first neighbour correlation, at 1.61 Å, is made the same for each. It is possible to discern four distinct changes that occurred with increasing exposure to SBF solution. The correlation at 2.02 Å, labelled as peak (a) in Fig. 5, was found as a major feature only in the SBF-reacted (CaO)0.3(SiO2)0.7 sol–gel glasses and corresponds to the second neighbour O–H bond lengths found in hydrated calcium phosphates.27 The correlation labelled peak (b) at 2.3 Å appears only in the unreacted sample: after exposure to SBF it disappears. Peak (c), at 2.54 Å, begins to change position on exposure to SBF, reaching a maximum correlation distance of 2.63 Å after 5 h, labelled peak (d). Similarly, peaks (e) and (f), at 3.6 and 3.75 Å respectively, also show a shift in their position; the peak at shorter r being replaced on SBF treatment by a feature at a longer correlation distance. Peak (g) at 4.08 Å narrows as the SBF reaction progresses. From a comparison with distances observed in calcium silicate minerals,28–30 peaks (b) and (c) may be assigned as Ca–O, and peak (e) with the corresponding higher order Ca–O, Ca–Ca and Ca–Si correlations. Peak (d) is within the range reported in the literature for the Ca–O bond in both calcium phosphates and hydroxyapatite31 whereas (f) is most probably due to a combination of Ca–O–Ca, inter-tetrahedral P–O and intra-tetrahedral P–O–P peaks in hydroxyapatite. The decrease in intensity of peaks (b), (c) and (e) in the D(r), and the associated increase in the intensity of peaks (d), (f) and (g), mirror in real-space the changes already identified in the interference patterns, i(Q): that is to say the removal of Ca from the glass network and the growth of a calcium phosphate layer followed by an increasingly ordered form of hydroxyapatite. Given that the pair correlation function, D(r), is an average over the volume of the sample in the X-ray beam—in contrast to the Bragg features observed in the i(Q), which may derive from small crystallites at the glass surface—we surmise that the real-space data reveal a change in the underlying glass network structure, which may be correlated with the measured changes in the overall sample composition discussed above and shown in Fig. 3. The XRF analysis also revealed a decrease in phosphorus content at ∼5 h, to a value less than that observed at 30 min, which is reflected in the relative intensities of the inter-tetrahedral P–O and intra-tetrahedral P–O–P correlations observed at ∼3.7 Å. This provides additional supportive evidence for a dynamic reaction mechanism driven by the interplay between the pH of the SBF and the surface charge of the glass.
![]() | ||
Fig. 5 D(r) data for SBF treated samples: from top to bottom, 7 days, 72 h, 25 h, 10 h, 5 h, 1 h, 30 min, unreacted. In all cases the correlation functions have been normalised to the Si–O first-neighbour peak as it appears in the unreacted sample. |
We have also demonstrated the nature of the principal stages to the generation of HA on the surface of a bioactive glass. Further, a link has been demonstrated between these distinct structural stages and the macroscopically determined elemental analysis of the glass as a function of time; that is to say, the atomic-scale structure to the surface charge- and pH-driven dissolution/growth processes have been related. Following immersion of a sol–gel derived bioactive glass in body fluid there is rapid initial dissolution of Ca, there is a growth of an amorphous, or very poorly crystalline, calcium phosphate layer within 1 h of immersion, which becomes increasingly evident and/or ordered by 5 h. However, by 10 h this polycrystalline layer is apparently replaced by a predominantly disordered material, now more akin to an amorphous hydroxyapatite; it is this layer that then develops steadily into a more ordered, bone mineral-like, hydroxyapatite.
Although these observations pertain to a particular bioactive material, it is reasonable to suggest that the underlying mechanisms may have a high degree of commonality with those associated with other bioactive, tissue-regenerative, solids—and certainly that the experimental methodology for their investigation requires the generic breadth employed here. It is only by such a detailed knowledge of the sites of importance in complex materials that systematic improvements in their application to tissue engineering applications can be achieved.
This journal is © The Royal Society of Chemistry 2005 |