14256
J. Phys. Chem. B 2007, 111, 14256-14264
Structural Effects of Phosphorus Inclusion in Bioactive Silicate Glasses Antonio Tilocca*,† and Alastair N. Cormack‡ Department of Chemistry, UniVersity College London, United Kingdom, and New York State College of Ceramics, Alfred UniVersity, Alfred, NY 14802 ReceiVed: July 19, 2007; In Final Form: October 10, 2007
Molecular dynamics simulations of four bioactive silicate glasses containing between 0 (P0) and 12 (P12) mol % P2O5 have been carried out in order to elucidate the structural role of phosphorus in these materials. In particular, we have focused on structural features which can have a direct role in the bioactive mechanism of dissolution and bone bonding. The higher affinity of modifier Na and Ca cations for coordinating phosphate rather than silicate, together with the formation of P-O-Si linkages, lead to increasing repolymerization of the silicate network with increasing P2O5 content, which in principle would represent a negative effect of P inclusion on the glass bioactivity. However, this effect is counterbalanced by the concomitant increase in the amount of free orthophosphate groups, whose fast release is deemed to enhance the bioactivity. The strong affinity of the orthophosphates for calcium ions leads to a clear tendency toward separation of silicate-rich and phosphate-rich phases for the P12 composition. Although this could reduce the bioactivity in the case of P12, in general, the favorable balance between the effects mentioned above should result in a positive effect of partial Si f P substitution on the glass bioactivity.
1. Introduction The use of melt-derived silicate glasses containing Na2O, CaO, and P2O5 for biomedical applications to restore bone/ muscle functionality depends on their special combination of high biocompatibility and ability to induce the formation of a crystalline hydroxyapatite (HA) layer on their surface shortly after implant.1-4 The HA layer, being equivalent to the inorganic portion of the bone, provides an effective bridge between the implanted glass and the living tissues, thus promoting the fast integration of the implant within the body: by interacting with and incorporating biomolecules such as collagen, further cellular steps lead to a strong and stable chemical bond between the glass and human hard tissue (bone) and, for the most active compositions, soft (muscle) tissue.1 Since the formation of the HA layer is a crucial prerequisite for bone bonding, the bioactivity of these glasses is generally measured as the rate of HA formation in vitro5,6 or in vivo.7,8 The high sensitivity of this property to the amount and kind of components present in the glass1,3 in principle provides an excellent means to optimize the rate of HA formation toward the level required in specific applications: in practice, although the bioactivity of many different compositions has been measured and the general mechanism leading to bone bonding roughly identified,3 this optimization is hindered by the lack of fundamental understanding of the effects leading to the different bioactivity of different compositions. Many compositions, including or not including P, Na, and Ca, can form a HA layer on their surface upon contact with a physiological fluid, but the time needed to crystallize HA varies considerably, with important consequences on the long-term stability of the implant.3,5,9,10 It is now clear that, since the initial stages of the HA formation mechanism involve ion release and partial dissolution of the glass network,3 an essential requirement for bioactivity is a rather open silicate network, * Corresponding author. E-mail:
[email protected]. † University College London. ‡ Alfred University.
which can be easily attacked and hydrolyzed in the physiological environment. This process is usually limited to the surface of the glass and leads to the formation of a surface layer rich in silica and hydroxyl groups: a high concentration of SiOH groups on the surface is considered essential to induce calcium phosphate nucleation and apatite crystallization.10,11 The bioinactivity of melt-derived silicate glasses containing more than 60% SiO21 reflects their lower dissolution rate, which results in an excessively slow HA formation rate, such that no bone bonding is produced upon implant. It has also been reported that dissolved silicate ions have a direct role in promoting apatite nucleation.12,13 The general picture, therefore, links the bioactivity level of these glasses to their solubility; a more specific analysis, however, is hindered by the lack of detailed structural information: most experimental data on the bulk structure of melt-derived Na2O-CaO-SiO2-P2O5 glasses concern the silicate/phosphate speciation, based on NMR and IR-Raman spectra.14-19 NMR experiments have revealed that the network of glass compositions close to the highly bioactive 45S5 Bioglass is dominated by Q2 and Q3 silicate species (where Qn denotes a Si atom bonded to n bridging oxygens, BOs).14,15 Notwithstanding the difficulty to assess the presence of lower amounts of Q1 silicates in these glasses based on 29Si NMR data alone,15 Raman data seem to indicate that lower amounts of Qn species other than Q2 and Q3 can also be present in 45S5.16 These data confirm the link between high bioactivity and open network structure, and some effort has been made to rationalize this connection through structural parameters such as the network connectivity.20-22 Another important structural feature revealed by previous experiments is the preferential association of Ca2+ with Q2 and of Na+ with Q3 silicate species, which is deemed to play a central role in modulating the dissolution rate of the glass in a physiological environment.14,16 It is clear that further structural information is needed in order to extend and complete the current partial picture of melt-derived bioactive glasses, if a more fundamental understanding of these materials
10.1021/jp075677o CCC: $37.00 © 2007 American Chemical Society Published on Web 11/30/2007
Structural Effects of Phosphorus Inclusion
J. Phys. Chem. B, Vol. 111, No. 51, 2007 14257
TABLE 1: Composition of the (Na2O)0.24(CaO)0.27[(SiO2)1-x (P2O5)x]0.49 Glasses Where the Label in the First Column Denotes the Total P2O5 Molar Percentage glass
100‚x
SiO2%
Na2O%
CaO%
P2O5%
L (Å)
P0 P2 P5 P12
0.0 5.1 10.8 24.3
48.87 46.14 43.22 36.30
24.34 24.36 24.42 24.36
26.79 26.93 26.93 27.17
0.0 2.57 5.43 12.17
26.8553 26.7339 26.5953 26.3118
is sought. It should be remarked that, although technological advances such as sol-gel processing have considerably extended the range of applicability of bioactive glass compositions,23 structural investigations of melt-derived Na2O-CaO-SiO2P2O5 glasses are still extremely useful: since the bioactivity level of many melt-derived compositions has been measured, these structural investigations can provide direct insight into structure-activity effects, and the resulting knowledge should be transferable to glasses of different composition and/or obtained from different routes. Unlike glasses obtained using sol-gel procedures, no hydroxyl groups are present in bulk meltderived silicate glasses as the ones examined in the present work, which however form silanol surface active sites upon partial dissolution. By investigating bulk structural properties of (OHfree) compositions, which affect their dissolution, one can highlight structure-activity effects which indirectly reflect the tendency of the glass to develop reactive silanol groups on the surface. An efficient and accurate way to investigate the bulk structure of modified silicate glasses is represented by molecular dynamics (MD) simulations: binary silicate glasses have often been modeled using classical24-26 and ab initio27-30 MD, whereas simulations of multicomponent silicate glasses are less frequent. We have recently started a series of computational investigations of the structural properties of bioactive silicate glasses using classical and ab initio molecular dynamics (MD) simulations.31-33 By comparing the structure of typical compositions with different bioactivity, we have highlighted a number of effects which can influence the bioactive behavior. In particular, we have focused on the network connectivity (NC), the coordination environment of network formers and network modifier cations, the clustering behavior and formation of inhomogeneous regions within the bulk,31,32 and the occurrence of chain and ring fragments33 as possible markers of the bioactivity. Another structural issue which deserves further investigation is the role of phosphorus in bioglasses: although P-free compositions can still crystallize HA on their surface through the phosphate ions present in the physiological fluid, in general, the inclusion of a small amount of phosphorus in silicate glasses enhances their ability to crystallize HA.5,34-39 Dissolved phosphate increases the degree of supersaturation in the body fluid and accelerates apatite crystallization;5 moreover, it can play an important role in controlling the pH close to the implant and avoiding excessive surface acidity which prevents bone bonding.40 The calcium phosphate layer formed on the surface of P-free glasses penetrates the silica layer and is less uniform, compared with P-containing glasses, resulting in a lower bonebonding strength.37,41 For sol-gel glasses, P2O5 inclusion promotes the degradation of the silicate network42 leading to faster HA crystallization.38,39 The surface of P-containing bioglasses shows a marked hydrophilic character, which is in part related to the presence of labile P-O-Si bonds, which can be opened through water adsorption and dissociation.43 Understanding these effects requires a deeper knowledge of the structural role of phosphorus in bulk bioglasses: in the most active compositions, containing less than 50% SiO2 and very
TABLE 2: Total Qn Distributions for Si and P and Corresponding Network Connectivities Si Q0 P0 P2 P5 P12
Q1
Q2
P Q3
Q4
1.1 27.0 52.6 18.5 0.8 0.4 20.6 51.9 25.3 1.7 0.5 13.5 47.5 34.1 4.4 0 1.3 32.3 54.9 10.9
Q5
Q0
Q1
Q2 NC(Si) NC(P)
0 0 65.4 34.6 0 0 61.6 38.4 0 0.6 39.8 57.1 3.1
1.91 2.07 2.28 2.77
0 0.35 0.38 0.63
low P2O5 amounts, PO4 groups are predominantly isolated as orthophosphates.14,15,17,31,44 In this form, they can easily be released through the silica-rich layer formed after the glass is soaked in the physiological fluid, thus increasing the rate of formation of calcium phosphate. However, if a fraction of P can be incorporated into the silicate network, this will limit the mobility and reduce the rate of release of orthophosphate (even though, as mentioned before, surface P-O-Si bonds are easily hydrolyzable): it has been reported that, unlike orthophosphates, pyro- and polyphosphate species have an inhibitory effect on biomineralization, because orthophosphates are released from bone-bonding compositions, whereas polyphosphate are found among the ions released from bio-inactive compositions,45 hence, the importance of investigating in great detail the degree of P-Si cross linking and how it is affected by the glass composition. Investigating the structural effect of P T Si substitution in the glass is also relevant to understand the different properties of bioactive glass ceramics, obtained by thermal treatment of bioactive glasses,46 because the residual glassy matrix (after crystallization of P-free silicates) is enriched in phosphorus.47 Finally, the importance of understanding structural features related to the phosphorus incorporated in silicate glasses extends beyond bioactive compositions: for instance, P incorporation in the silicate network can lead to enhanced conductivity in lithium silicophosphate glasses.48 Li ions are found close to the phosphate groups, and their migration involves jumps between consecutive P centers in monodimensional -Si-O-P-O-SiO-P- chains with alternated Si-O-P linkages. In these glasses, Si-O-Si and Si-O-P bonds are favored over P-O-P bonds, which are formed only at low Si:P ratio.48 In this paper, the Bioglass 45S5 composition is used as a reference, and its structure is compared with that of three other compositions with a different amount of P2O5 but the same amount of tetrahedral species (Si + P): this allows us to investigate the effect of the Si T P substitution. Although there are some indications that glasses with substantially higher amounts of P2O5 are bio-inactive,4 to our knowledge, a systematic investigation of the structural role of phosphorus in glass compositions containing up to 12% P2O5 and a silica amount lower than the Bioglass 45S5 has not been previously carried out. The ultimate purpose is to reveal structural trends related to the P content, focusing in particular on those effects which can result in a different bioactivity with respect to the reference Bioglass 45S5 composition. 2. Computational Methods The compositions studied here, reported in Table 1, can be specified by the general molar formula (Na2O)0.24(CaO)0.27 [(SiO2)1-x(P2O5)x]0.49, with 0.0 e x e 0.25. They were selected in such a way to examine the effect of substituting Si with P, without other interferences. Starting from the standard 45S5 Bioglass composition, which contains ∼46 mol % (45 wt %) SiO2 and 2.6 mol % P2O5, P2O5 substitutes for 2SiO2 + Na2O + CaO. In this way, the total number of tetrahedral sites NT )
14258 J. Phys. Chem. B, Vol. 111, No. 51, 2007
Tilocca et al.
Figure 2. Si-P cross-connectivity: concentration of BOs linking any pair of tetrahedra (circle symbols); linking two silicate (square symbols); linking Si to P (triangle symbols); linking two phosphate (diamond symbols).
Figure 1. Concentration of Qn species: (top panels) including all BOs; (central panels) only including BOs connecting Si to Si and P to P; (bottom panels) only including BOs connecting Si to P.
NSi + NP, of oxygen atoms NO, and the fraction of Na2O and of CaO are kept constant; this substitution thus allows us to isolate the effect of P f Si substitution from other effects and to set constant the overall network connectivity, defined as the average number of bridging oxygens NBO per tetrahedral species.20,49 This is especially useful in order to investigate specific substitutional effects reducing other interferences.22 The network connectivity of a specific composition can be predicted as NBO/ NT ) 8 - 2(NO/NT), based on the assumption of an ideal fourfold coordination for all Si and P atoms, which is roughly satisfied in our simulations (see below). We modeled four compositions, labeled P0, P2, P5, and P12, where the end digit(s) correspond to the molar percentage of P2O5; P2 corresponds to the 45S5 Bioglass composition. Approximately 1400 atoms are included in each periodic cubic supercell, whose side, reported in Table 1, yields a density of 2.702 g/cm3, corresponding to the 45S5 Bioglass.16 Although replacement of SiO2 by P2O5 should lead to a density increase, the percent increase corresponding to the compositions studied here is expected to be rather small,50 with a correspondingly negligible effect on the structural properties. Therefore, as the only available experimental density was that of the 45S5 glass, we preferred to use the same 45S5 density for all glasses, than to use approximate densities. In the classical molecular dynamics simulations, which were carried out with the DL_POLY code,51 we used an interatomic potential recently developed in our group to model silicate glasses incorporating Na, Ca, and PO4 ions.31,52 Polarization effects are approximately incorporated in the dynamics through the adiabatic shell-model approach:52-54 this inclusion is useful to accurately describe the asymmetric and distorted ionic environments found in disordered materials, and for bulk silicate glasses, it leads to a better description of the intertetrahedral structure and of the local environment surrounding modifier Na
and Ca cations, as well as of the Qn distribution of modified silicate glasses, compared with rigid-ion potentials.52 In the adiabatic shell model, a small fraction of the core mass (0.2 au for oxygen ions in this work) is shifted to the shells, which move according to conventional equations of motion and follow the ionic motion adiabatically. In order to improve the energy conservation and the stability of the trajectories over relatively long time scales, it is necessary to add a damping term to the core-shell harmonic forces.52,54 Alternatively, one could use a relaxed shell model55 where the shell positions are explicitly optimized to ensure that the net force on every shell is zero at each time step, but for the systems studied here, we found the adiabatic shell dynamics a more efficient way to achieve essentially the same goal. The full list of potential parameters is reported in refs 31 and 33. Formal ionic charges with Ewald summation of the longrange Coulomb interactions are used; short-range Buckingham potentials (cutoff distance of 8 Å) describe the interaction of the oxygen shells with each other and with Si, Na, Ca, and P cations, and three-body screened harmonic potentials are used to control the intra-tetrahedral O-Si-O and O-P-O angles. In all simulations, a small time step (0.2 fs) is used to control the high frequency of the core-shell spring. The melt-derived glass structures were obtained using a standard molecular dynamics cooling procedure:52 starting from a random arrangement of ions in the supercell, the system was heated and held at 3500 K for 100 ps, ensuring a suitable melting of the sample. The liquid was then continuously cooled to 300 K at a nominal cooling rate of 10 K/ps, in 320 ps. The resulting glass structure was then used in a final trajectory of 200 ps, the last 150 ps of which were included in the structural analysis. All MD simulations were performed in the NVT statistical ensemble, at variance with our previous simulations,31 in which the cooling was perfomed in the NPT ensemble. The constantvolume cooling, including small thermal expansions above 1000 K, leads to a final glass structure with a theoretical density closer to the experiment, without affecting the other structural properties. All of the calculations of this work were performed on a Dual Xeon 5160 workstation equipped with 4 GB of memory. 3. Results 3.1. Role of P in the Glass Network. The distributions of Qn species reported in Table 2 were calculated using a T-O
Structural Effects of Phosphorus Inclusion
J. Phys. Chem. B, Vol. 111, No. 51, 2007 14259
TABLE 3: Self-Qns Distributions and Corresponding Network Connectivities: Qns Is a Tetrahedral Atom Linked to n Atoms of the Same Type, That Is, Si-O-P Cross Links Are Not Included Si 0
P0 P2 P5 P12
P
Q
Q1
Q2
Q3
Q4
Q5
Q0
Q1
Q2
NC(Si)s
NC(P)s
NCs
1.1 0.9 0.5 0
27.0 22.3 17.9 14.9
52.6 51.1 47.4 37.9
18.5 24.0 30.8 41.4
0.8 1.7 3.4 5.3
0 0 0 0.5
100 100 94.2
0 0 5.8
0 0 0
1.91 2.03 2.19 2.39
0 0 0.06
1.91 1.83 1.75 1.45
TABLE 4: Cross-Qnc Distributions and Corresponding Network Connectivities: Qnc Is a Tetrahedral Atom Linked to n Atoms of the Other Type, That Is, Only Si-O-P Cross Links Are Included Si P0 P2 P5 P12
P
Q0
Q1
Q2
Q3
Q4
Q5
Q0
Q1
Q2
NC(Si)c
NC(P)c
NCc
100 96.1 90.3 65.7
0 3.9 9.7 29.9
0 0 0 4.2
0 0 0 0
0 0 0 0
0 0 0 0
65.4 61.6 45.6
34.6 38.4 51.3
0.0 0 3.1
0 0.04 0.097 0.38
0.35 0.38 0.57
0 0.07 0.15 0.46
cutoff distance of 2.0 Å; the connectivity of the silicate and phosphate networks, denoted NC(Si) and NC(P) in the table, can be extracted as weighted average of the corresponding Qn(Si) and Qn(P) distributions.60 If a further weighted average of NC(Si) and NC(P) is made, the overall average number of BOs per tetrahedral (Si/P) species is obtained, which exactly matches the theoretical NC of 1.90. This is not surprising, since the latter only depends on the assumption of fourfold coordination for Si and P, which is satisfied for the studied compositions: Table 2 shows that only for P12 a very small amount of Q5 Si is present. A very small fraction of five-coordinated Si had been observed also in our previous MD models of glasses containing 55% (BG55) and 65% (BG65) SiO2;31 as mentioned in ref 31, these small abundances are compatible with NMR data for modified silicate glasses at ambient pressure.56,57 The occurrence of fivecoordinated Si atoms in BG55 and in the present P12 glass seems to indicate that their formation involves an NC(Si) above the values characteristic of highly bioactive compositions: NC(Si) is 2.77 for both BG55 and P12. On the other hand, sixcoordinated Si has been detected in modified phosphosilicate glasses but only for P2O5 amounts above 30%58 (significantly larger than in the present compositions) or in glasses quenched under high pressures;59 therefore, it is not surprising that our model structures are completely free of Si6c. The general message of Table 2 is that the total connectivity of the silicates increases when P2O5 substitutes 2SiO2 + Na2O + CaO: this is a direct consequence of the intrinsic lower network-forming ability of phosphate than silicate groups. Indeed, since the substitution keeps NBO constant and the average number of BOs around P is always smaller than around Si, a corresponding increase in the number of BOs around Si will balance the effect of the substitution. In other words, the overall NC is kept fixed at 1.90 by an increase in NC(Si) which balances the much lower NC(P) shown in Table 2. The higher silicate NC is partitioned between Si-O-Si and Si-O-P bonds, with a net prevalence of the former. The Qn distributions in the table show that the P0, P2, and P5 glasses are predominantly Q2 (for P2, this agrees with the experimental NMR and Raman data14-16); the increase in NC(Si) is mainly due to a continuous decrease in Si(Q1) and Si(Q2) and a concomitant increase in Si(Q3) and Si(Q4), which ultimately results in a predominantly Q3 distribution for the P12 glass. Table 2 also confirms that phosphate groups are much less connected than silicate, as the Qn(P) distribution is dominated by isolated orthophosphates for P2 and P5 and by pyrophosphate Q1 for P12. These trends are better highlighted in the top panels of Figure 1, which show the absolute concentration of the various Qn species.
In order to investigate the extent of Si-P copolymerization, we split the Qn distributions into the contributions from “self” linkages (such as Si-O-Si and P-O-P) and “cross” linkages (Si-O-P): in this way, a Qns (Si) species is bonded to n BOs, which all link it to another Si, whereas the n BOs of a Qnc (Si) are involved in Si-O-P linkages. From the self and cross contributions reported in Tables 3 and 4, we can separately identify the connectivity of Si atoms with each other (NCs(Si)) and with phosphorus (NCc(Si)), and the same is true for P atoms. The last column in Tables 3 and 4 shows the average self-connectivity NCs and cross-connectivity NCc for the various compositions, which satisfy the relation NC ) NCs + NCc. Although the Si connectivity is dominated by “self” Si-O-Si linkages (NCs(Si) > NCc(Si)), the situation is reversed for P, which, besides having a general much lower tendency to polymerize, does not easily form P-O-P bridges, but where it does polymerize, it prefers cross links with Si: NCc(P) > NCs(P). In the central and bottom panels of Figure 1, the concentration of self- and cross-polymerized Qn species is shown; for phosphorus, the absolute prevalence of Q0s species illustrates that almost no self-polymerized P tetrahedra (such as P2O27 ) are present, whereas a moderate amount of Q1c cross-linked to Si are formed. For silicon, the trend of the total connectivity closely reflects the self-connectivity, as Si-O-Si linkages are favored over Si-O-P; although less favored, the concentration of cross-linked Q1c (Si) species increases with the P2O5 content, and the concomitant decrease in Q0c (Si) shows that some Q0c (Si) f Q1c (Si) conversion is taking place upon P inclusion; that is, new Si-O-P linkages are being created. To summarize, substitution of P for Si leads to a repolymerization of the silicate network, through an increasing number of both Si-O-Si and Si-O-P linkages, resulting in less open glass networks compared with the P-free composition. At the same time, despite a significant fraction of the added phosphate getting involved in cross-link with Si, the P coordination is dominated by species with no (Q0 orthophosphate) or limited (Q1 pyrophosphate with only one BO linking P to Si) connectivity, whose concentration increases with the P2O5 content. The result is an increasingly more connected silicate network partially cross-linked to an altogether poorly connected phosphate network. The overall partitioning of T-BO-T bridges in the glasses is quantitatively examined in Figure 2, which shows that most BOs connect two Si and only a negligible fraction is bridging between two P’s. The inclusion of P2O5 leads to an increased number of P-O-Si links, at the expenses of Si-O-Si, with a
14260 J. Phys. Chem. B, Vol. 111, No. 51, 2007
Tilocca et al.
TABLE 5: Si(P)-Na(Ca) Coordination Numbers (Cutoff Radius Rc ) 4.5 Å). CN(T-M) Is the Average Number of M Modifier Cations Found Inside a Sphere of Radius Rc Centered on T and the Numbers between Parentheses Are the Corresponding Density Ratios G(T-M)/G(M) ) (CN(T-M)/R3c )‚L3/NM, Where NM Is the Total Number of Modifier Cations and L Is the Cell Side CN P0 P2 P5 P12
Si-Na
Si-Ca
Si-M
P-Na
P-Ca
P-M
5.58 (4.60) 5.34 (4.55) 4.91 (4.33) 4.27 (4.10)
3.24 (4.85) 3.07 (4.73) 2.90 (4.64) 2.33 (4.01)
8.82 (4.69) 8.41 (4.62) 7.81 (4.44) 6.6 (4.07)
6.21 (5.29) 6.06 (5.35) 5.39 (5.18)
3.82 (5.89) 3.56 (5.70) 3.49 (6.01)
10.03 (5.51) 9.62 (5.47) 8.88 (5.48)
TABLE 6: Si(P)-Si(P) Coordination Numbers (Cutoff Radius Rc ) 6 Å) Where CN(T-T′) Is the Average Number of T′ Atoms Found Inside a Sphere of Radius Rc Centered on T CN P0 P2 P5 P12
Si-Si
P-P
Si-P
P-Si
12.4 11.1 10.6 9.3
1.3 3.2 6.2
1.3 2.4 4.4
11.2 9.5 6.6
relevant 25% fraction of BOs connecting Si to P for the P12 composition. The latter composition is the only one for which a very small amount (0.17 BO/nm3) of BOs are found bridging between two P atoms. 3.2. Coordination of Modifier Cations Around Tetrahedral Sites. In the present and similar compositions, each modifier (M ) Na or Ca) cation is embedded in a distorted octahedral shell of approximately six oxygens, bonded to Si or P.31,32 The coordination numbers (CN) of modifier cations surrounding the tetrahedral Si/P sites, reported in Table 5, have been calculated by integrating the corresponding Si-M or P-M radial distribution function (not shown) up to the first minimum at Rc ) 4.5 Å. The Table clearly highlights the preference of both Na and Ca modifiers for phosphate: CN(P-M) is always greater than CN(Si-M), reflecting the higher number of nonbridging oxygens (NBOs) bonded to P, whose local charge balance will attract more M cations in the phosphate coordination shell. In order to investigate the effect of P inclusion and how Na/ Ca are partitioned between silicate and phosphate tetrahedra, we also report in Table 5 the ratio fT-M ) F(T-M)/F(M) between the local density of M ions in the Si or P second coordination shell and the overall density of M in the whole cell; in this way, the lower number of Ca than Na cations is taken into account and weighted out. In practice, the secondcoordination shell of tetrahedral (T) species will be populated by a significantly higher number of modifier (M) cations than what could be predicted by uniformly distributing the M cations in the cell volume; fT-M is a quantitative measure of how much this number exceeds a uniform distribution. The table shows a steady decrease in fSi-M, whereas fP-M is roughly constant. This shows that, upon phosphate inclusion, the modifier cations are stripped out of the silicate network, in agreement with experimental observations.14,65 In the glasses studied here, this effect is enhanced for higher phosphate content: the relative decrease in modifier concentration around Si reflects the increased amount of phosphate groups, which are able to maintain a constantly higher relative concentration of modifiers around
n
Q Figure 3. R Na/Ca ratios for Si (left) and P (right) Qn species. The Qn species above (below) the R ) 1 dashed horizontal line prefer Na (Ca) coordination. Ratios with n ) 0, 1, 2, 3, and 4 are plotted as star, diamond, triangle, square, and circle symbols, respectively.
them. Increasing the amount of phosphate also affects the relative balance of Na/Ca cations around the network formers: in Table 5, a moderate preference for Ca in the second coordination shell of Si, shown by fSi-Ca > fSi-Na up to the P5 composition, is reversed for P12, for which the preference for Na ions around Si becomes clear. At this relatively high phosphorus composition, a marked difference between the second coordination shells of Si and P is indeed established, with an excess of Ca around phosphate and an excess of Na around silicate groups. The cation environment surrounding the different Qn Si and P sites can be investigated in detail by further separating the Si(P)-M coordination numbers into the various Qn-M, n ) 0, ..., 4 contributions from each Qn. By using the calculated QnNa and Qn-Ca coordination numbers (not shown), the ratios n n RQn Na/Ca ) CN(Q -Na)/CN(Q -Ca)‚NCa/NNa, are a useful measure of the preference of Qn for Na or Ca coordination.31 A unit R ratio denotes a statistical (random) distribution of Na and Ca ions surrounding Qn, whereas deviations from unity denote a nonstatistical environment around the central Qn: R > 1 denotes preference of Qn for coordination of Na ions, whereas R < 1 denotes preference of Qn for coordination of Ca ions. Figure 3 shows that sites with lower connectivity, such as Q1 silicates and Q0 orthophosphates, always favor Ca ions, whereas sites with higher connectivity, such as Q3 and Q4 silicates, as well as the few Q2 phosphates found in P12, have a larger affinity for Na.14 When the higher connectivity sites dominate the silicate coordination, as for P12, the net effect is a stronger aggregation of calcium and phosphates, separated from the silicate network. Summarizing this section, Na and Ca in general favor the coordination of phosphate over silicate, and this leads to the removal of both cations from the silicate network when P is included at low concentration; however, increasing the amount of P2O5, the affinity of Ca for phosphate grows increasingly stronger than that of Na, resulting in a marked aggregation of calcium phosphate-rich regions. As we will discuss in the following, selective association between Ca and phosphate plays an important role in the separation of calcium phosphate-rich regions for the P12 composition, analogous to what we have recently observed for a bio-inactive composition.31 3.3. Silicate- and Phosphate-Rich Domains. In phosphosilicate glasses, glass-in-glass separation between phosphateand silicate-rich phases can be induced by the presence of two
Structural Effects of Phosphorus Inclusion
J. Phys. Chem. B, Vol. 111, No. 51, 2007 14261
Figure 5. Radial distribution function of Si-O and P-O bonds, showing the various contributions from nonbridging and bridging oxygens, with the latter further separated into the contributions from P-BO(-Si), P-BO(-P), Si-BO(-P), and Si-BO(-Si).
Figure 4. Structure of glasses P2 (left panels) and P12 (right panels): (a) only Si (turquoise), P (yellow), and O (red) atoms are displayed; (b) only the silicate network is displayed; (c) only the phosphate groups are displayed, with the Si atoms represented as spheres.
high-valence ions, such as Si4+ and P5+, which will concentrate in different regions.47,61 The phase separation can negatively affect the bioactivity due to the increased viscosity of the inhomogenous phase-separated glass, which reduces the solubility of Si and Ca ions, whereas the effect on the phosphate mobility is less clear.47 As a matter of fact, our recent investigations of the structure of bioactive and bio-inactive glasses have revealed a clear tendency of bio-inactive compositions to form Ca-rich phosphate and Ca-poor silicate regions,31 and another important point is that glass-in-glass separation is considered a prerequisite for crystallization,62 which tends to decrease the kinetics of apatite layer formation in partially crystallized glass ceramics.34,46,47 Besides the analysis of the previous section, a direct way to probe the occurrence of relatively small silicate- and phosphaterich regions in the glasses is to compare their Si-P coordination numbers, within a larger volume than the nearest neighbor shell; we considered a cutoff distance of 6 Å in order to extend the observation space roughly up to the second neighbors. Table 6 shows that, for P2, CN(Si-Si) matches CN(P-Si), and analogously CN(P-P) matches CN(Si-P); therefore, for this composition, the space surrounding P and Si is uniformly (i.e., randomly) occupied by other phosphate and silicates. On the other hand, for P5 and P12 CN(Si-Si) > CN(P-Si), and CN(P-P) > CN(Si-P): these inequalities now mark a tendency to form silicate- and phosphate-rich regions which already starts with P5, although it is more relevant for P12. Indeed, by comparing the structure of P2 and P12 in Figure 4, it is evident that the phosphate groups are uniformly spread over the whole
cell in P2, whereas for P12 the figure shows the formation of voids in the silicate network, filled by phosphates. Therefore, the presence of a significantly larger fraction of P2O5 than the standard 45S5 Bioglass seems to trigger the phase separation in these compositions; since the same effect was observed in compositions containing a higher percent of SiO2, a constant amount of P2O5, and a correspondingly decreasing amount of modifier cations,31 the [Na2O+CaO]/[P2O5] ratio seems to be an important factor ruling the phase-separation properties of these bioglasses, before thermal treatment: the present simulations indicate that a lower ratio, roughly
