Structure–composition relationships of bioactive borophosphosilicate glasses probed by multinuclear ¹¹B, ²⁹Si, and ³¹P solid state NMR

Abstract

By combining ¹¹B, ²⁹Si, and ³¹P nuclear magnetic resonance (NMR) experimental results, we present a comprehensive structural investigation of 15 borophosphosilicate (BPS) glasses of the Na₂O–CaO–B₂O₃–SiO₂–P₂O₅ system: in two base compositions comprising 46 mol% (“S46”) and 49 mol% (“S49”) SiO₂, progressive replacements of SiO₂ by B₂O₃ were performed at a constant total Na₂O and CaO content. The S46 glass members constitute B-bearing analogs of “45S5 Bioglass” that is utilized extensively for bone grafting in periodontal and orthopedic surgery. Orthophosphate ions prevail throughout all structures, while the silicate network polymerization increases slightly with a growing amount of B₂O₃ in the glass. ¹¹B NMR revealed continuous BO₃ → BO₄ conversions for increasing B₂O₃ content, with asymptotic fractions of 34% and 43% of B^[4] coordinations out of the borate speciation observed for the series of S46 and S49 glasses, respectively. While all BPS glasses are homogeneous across a μm-scale, strong preferences for B^[3]–O–B^[3] and B^[4]–O–Si^[4] bond formation lead to structures comprising (sub)nm-sized domains of BO₃ groups in boroxol rings and borosilicate networks built by SiO₄ and BO₄ tetrahedra. These borate/borosilicate networks are merged mainly by B^[4](3Si) and B^[3](1Si) moieties in Si-rich BPS glasses (where each value in parentheses specifies the number of bonds to Si atoms), while B^[4](3Si) and B^[4](2Si) groups are the dominant network contact points in the B-rich glasses. We discuss the partitioning of non-bridging oxygen ions between the BO₃ and SiO₄ groups, the relative propensities for B^[4]–O–Si^[4] and B^[4]–O–B^[3] bond formation, as well as the expected bearings of our proposed BPS structural model for the glass degradation in aqueous media, where we identify the fractional population of B^[4] coordinations and the silicate network connectivity to constitute the dissolution-controlling parameters.

---

1 Introduction

The elements Al, B, P, and Si build the networks of a vast majority of all oxide-based glasses.^1–3 Most of them involve one or two network-forming species, while the development and structural characterization of glasses incorporating three (or more) network formers are much less explored. Yet, they are of interest for the fundamental understanding of glass chemistry, where an improved insight into such mixed-former glass networks will assist the tailoring of their physical, chemical, and mechanical glass properties.

Our study concerns borophosphosilicate (BPS) glasses that show great promise for bone-grafting applications.^4–14 On their contact with body fluids, such bioactive glasses (BGs)^9,15–17 integrate with the bone/tooth by forming a surface layer of hydroxy-carbonate apatite (HCA), whose composition closely mimics bone mineral. Some of the alkali/alkaline-earth-rich borate and borosilicate glasses involved B for Si substitutions of already existing phosphosilicate BGs,^{4–6,11–13} such as the “19-93”¹⁸ and the “45S5 Bioglass®”^15,17 compositions; the latter is soda-lime-silica based with molar equivalents 24.6Na₂O–26.7CaO–46.1SiO₂–2.6P₂O₅ and has been in clinical use for decades.¹⁷ Phosphosilicate BGs feature rings and chains of interconnected silicate tetrahedra (SiO₄).^3,17 The incorporation of BO₃/BO₄ borate groups modifies the glass network by the formation of Si–O–B and B–O–B bonds. Compared with phosphosilicate BGs, the borate-based counterparts degrade faster in aqueous media^19,20 and are claimed to convert more completely into HCA.^4–9,14 The glass degradation is tunable by varying the n_B/n_Si molar ratio.^4,5 Moreover, the dissolution products of B-bearing BGs may promote bone growth and angiogenesis,^9,13 as well as stimulating RNA synthesis in fibroblast cells²¹ to further aid the bone regeneration.

The present report targets two series of Na₂O–CaO–B₂O₃–SiO₂–P₂O₅ glasses (one based on the “45S5” composition¹⁵), where SiO₂ is replaced by B₂O₃ at a fixed total Na₂O and CaO content. These network modifier-rich (boro)phosphosilicate glasses exhibit fragmented glass networks, mainly built by ring/chain motifs of interlinked SiO₄, BO₄, and BO₃ polyhedra. The local glass structures were probed by a combination of magic-angle spinning (MAS) ¹¹B, ²⁹Si, and ³¹P NMR experiments. Besides providing the first comprehensive structural account on B-substituted BGs (such as the 45S5 based glasses^4,6,13), we demonstrate that P-richer bioactive BPS glasses are feasible to prepare: increasing the P content of a silicate-based BG is believed to promote its HCA formation in vitro^17,22,23 and P-rich BPS glass analogs should consequently also be beneficial.

This work also complements the few existing structural reports on BPS glasses that generally targeted the compositional space of low modifier contents relative to P and Si,^24–26 encompassing M₂O–B₂O₃–SiO₂–P₂O₅ glasses of low P and alkali metal M = {Na, K} contents^24,26 or P-rich compositions of the Na₂O–(CaO)–B₂O₃–SiO₂–P₂O₅ system.^25,27 For the latter, P assumes its usual role as a network former in the guise of phosphate chains that are cross-linked to the silicate/borate networks.^24,25 This structural scenario is also shared by P-rich M₍₂₎O–SiO₂–P₂O₅ glasses,^28,29 but contrasts markedly with that for the fragmented networks of bioactive phosphosilicate glasses, where P is typically a minor component (≲6 mol% P₂O₅) and predominantly forms orthophosphate (PO₄³⁻) anions^{22,23,30–34} that are charge-balanced by Na⁺/Ca²⁺ and distributed randomly at interstitial positions around the silicate network.^32,35 This property ensures a rapid release of phosphate moieties when the BG is subjected to (simulated) body fluids.^23,33,36 One of the goals of the present work is the verification that this favorable structural role of P remains when substantial amounts of borate groups are incorporated into the glass structure. Previous work on Na₂O–B₂O₃–SiO₂–P₂O₅ glasses²⁶ suggests this not to apply for compositions featuring (significantly) lower Na contents than the total modifier (Na⁺, Ca²⁺) reservoirs of the present glasses.

2 Structural features of borophosphosilicate glasses

2.1 Basic building blocks

The present soda-lime BPS glasses involve three traditional glass network-forming species: B, P, and Si. Regarding the structural role of P, earlier reports from Na–B–Si–P–O glasses of relatively low amounts of Na but with comparable Si and P contents as our current glasses revealed a dominance of P₂O₇⁴⁻ diphosphate groups and P–O–B structural fragments,^24,26 suggesting a traditional network-forming role of P. However, as demonstrated in Section 4.1, very few phosphate tetrahedra in the present modifier-rich BPS structures bind to network-forming groups and the P speciation is dominated by orthophosphate anions, denoted Q⁰_P. Henceforth, Qⁿ_P (Qⁿ_Si) labels a PO₄ (SiO₄) tetrahedron with n bridging oxygen (BO) species^1–3 that share corners with neighboring tetrahedra. The remaining 4 − n tetrahedral corners represent non-bridging oxygen (NBO) ions that are charge-balanced by electropositive metal ions (herein, Na⁺ and Ca²⁺).

In contrast with the structure of P-rich (boro)phosphosilicate glasses that also feature SiO₅/SiO₆ polyhedra,^25,28,29 Si is exclusively present in tetrahedral coordination (Si^[4]) in modifier-rich/P-poor BPS glasses (see Section 4.2),^24–26 where it assumes a set of interconnected {Qⁿ_Si} groups. Si may also bond to the two B coordinations, B^[3] and B^[4], which represent trigonal BO₃ and tetrahedral [BO₄]⁻ moieties, respectively.^3,37–39 Here B^[p](mSi) labels a B^[p] species with m B–O–Si bonds.

The Na⁺/Ca²⁺ cations play a dual structural role, where they partially balance the negatively charged [BO₄]⁻ and {Q⁰_P, Q¹_P} moieties, whereas the remaining cations are termed network modifiers because they reduce the glass network polymerization by breaking Si/B^[3]–O–Si/B^[3] bonds, thereby driving BO → NBO conversions.^1–3 The polymerization degree associated with a network-forming species E may be described by its average number of BO atoms per polyhedron (N̄^E_BO),^3,23,33,40 which is often referred to as the network connectivity.^22,41,42 For silicate-based glasses, the silicate network connectivity (N̄^Si_BO) may be derived from the set {xⁿ_Si} accessible by ²⁹Si NMR.^1–3 Herein, the symbol “x” is reserved for fractions, where xⁿ_E labels the fractional population of Qⁿ_E out of the entire {Qⁿ_E} speciation (0 ≤ n ≤ 4).

Concerning the B partitioning among B^[3]/B^[4] coordinations and their precise structural relationship relative to the silicate tetrahedra, no clear consensus is reached for BPS glasses^24–27 and the compositional space relevant for bioactive glasses remains essentially unexplored. Yet, these aspects have been intensively investigated for the limiting borosilicate systems,^{2,3,6,37,39,43–54} notably so for Na₂O–B₂O₃–SiO₂ glasses. For the latter stoichiometric compositions parametrized as RNa₂O–B₂O₃–KSiO₂, Bray and co-workers^37,38 proposed a structural model that provides the relative {B^[4], B^[3]} populations for variable molar ratios K and R; it is henceforth referred to as the Yun–Dell–Bray–Xiao (YDBX) model. It predicts that for low Na₂O contents (i.e., small R-values), B^[3] → B^[4] conversions occur up to R ≤ R_max = K/16 + 1/2,³⁸ meaning that Na⁺ solely acts to charge-balance the [BO₄]⁻ tetrahedra. Consequently, for all compositions with R < R_max, the borosilicate structure only comprises B^[3]–O–B^[3], Si–O–Si, and B^[4]–O–Si bonds, the latter forming B^[4](OSi)₄ groups. NBO species only exist when R > R_max, where they initially enter solely at the SiO₄ tetrahedra, whereas the BO₃ groups starts accommodating NBO ions for Na-richer compositions with R > (K + 1)/4.³⁸

The YDBX structural picture provides reasonable predictions of the B speciation in Na–B–Si–O glasses, and sometimes also for some other M⁺/M²⁺ cation-bearing amorphous borosilicates. However, the model was formulated solely with insight from ¹¹B NMR, and subsequent studies indicate its limitations: overall, the borosilicate glass structure is less ordered than suggested by the YDBX model and the NBO ions are more evenly distributed among SiO₄ and BO₃ groups^{45,46,48–53} compared with the YDBX scenario. Moreover, the {BO₃, BO₄, SiO₄} network building blocks are markedly more intermixed, with all types of B^[3]/B^[4]–O–B^[3]/Si linkages encountered,^39,46–54 albeit the preferences of B–O–Si bond formation decreases in the order B^[4] > B^[3](non-ring) > B^[3](ring).^50–52 Here “ring” specifies BO₃ groups that exclusively form boroxol (B₃O₆) rings, as present in vitreous B₂O₃ and borate glasses,^2,3,39,55 whereas “non-ring” implies linkages to BO₄/SiO₄ tetrahedra.^50–54,56 To avoid local negative charge accumulation, [BO₄]⁻ tetrahedra involve solely BO (as opposed to NBO) species,^{3,37,38,40,53} while B^[4]–O–B^[4] bridges are strongly disfavored, or even absent.^{1–3,37,38,50–53} Herein, we explore the composition-dependence of the B speciation in the more complex soda-lime BPS glass system, targeting NBO-richer networks than those thoroughly investigated in the ternary/quaternary M₍₂₎O–(M′O)–B₂O₃–SiO₂ systems.

2.2 Our borophosphosilicate glass design

Table 1 lists our targeted Na₂O–CaO–B₂O₃–SiO₂–P₂O₅ glass compositions. Each is labeled SN_p(q), where N is the sum of SiO₂ and B₂O₃ contents in mol%, while p and q represent the mol% of P₂O₅ and B₂O₃, respectively. Note that S46_2.6(0) corresponds to the widely utilized “45S5 Bioglass” composition.^15,17 When collectively referring to all its members, we employ the nomenclature “S46” and “S49” for the two series, which mainly differ in (i) the Si contents of the two base compositions, S46_2.6(0) and S49_4.0(0), and thereby their glass-network polymerization degrees reflected in the corresponding silicate network connectivities^22,23,33 N̄^Si_BO = 2.11 and N̄^Si_BO = 2.54; (ii) the P₂O₅ contents that are constant at 2.6 mol% and 4.0 mol% within each respective set; (iii) the relative Na/Ca amounts, with the molar ratios n_Na/n_Ca = x_Na/x_Ca being 1.84 and 2.08, respectively, while the sum x(Na₂O) + x(CaO) remains constant within each S46 and S49 family, where x_E or x(E) denotes the molar fraction of species E.

Table 1 Glass compositions^a

Label	f^b/%	x(Na2O)	x(CaO)	x(SiO2)	x(B2O3)	x(P2O5)	Stoichiometric formula^c	xNa/xCa^d	xB/xSi^d	R′ ^e	K ^f
a Each glass is labeled SNp(q), where N = 100[x(SiO2) + x(B2O3)] is the sum of mol% of SiO2 and B2O3, whereas p and q are the mol% of P2O5 and B2O3, respectively.b f = 100x(B2O3)/[x(SiO2) + x(B2O3)] is the percentage of B2O3 substitution for SiO2.c Stoichiometric formula with coefficients specified as atom fractions that sum to unity.d Ratios of the atomic fractions of Na and Ca (xNa/xCa), and B and Si (xB/xSi), respectively.e Ratio , with calculated from eqn (1).f K = x(SiO2)/x(B2O3).
S462.6(0)	0	0.246	0.267	0.461	0.000	0.026	Na0.173Ca0.094Si0.162B0.000P0.018O0.552	1.84	0.00	—	—
S462.6(5)	10	0.246	0.267	0.415	0.046	0.026	Na0.168Ca0.091Si0.142B0.031P0.018O0.550	1.84	0.22	9.46	9.00
S462.6(9)	20	0.246	0.267	0.369	0.092	0.026	Na0.163Ca0.088Si0.122B0.061P0.017O0.548	1.84	0.50	4.73	4.00
S462.6(14)	30	0.246	0.267	0.322	0.138	0.026	Na0.158Ca0.086Si0.104B0.089P0.017O0.547	1.84	0.86	3.15	2.33
S462.6(18)	40	0.246	0.267	0.277	0.184	0.026	Na0.153Ca0.083Si0.086B0.115P0.016O0.546	1.84	1.33	2.37	1.50
S462.6(28)	60	0.246	0.267	0.184	0.277	0.026	Na0.145Ca0.079Si0.054B0.163P0.015O0.543	1.84	3.00	1.58	0.67
S462.6(37)	80	0.246	0.267	0.092	0.369	0.026	Na0.138Ca0.075Si0.026B0.206P0.015O0.541	1.84	8.00	1.18	0.25
S494.0(0)	0	0.241	0.233	0.486	0.000	0.040	Na0.165Ca0.079Si0.166B0.000P0.027O0.562	2.08	0.00	—	—
S494.0(2)	5	0.241	0.233	0.462	0.024	0.040	Na0.162Ca0.078Si0.155B0.016P0.027O0.561	2.08	0.11	14.7	19.0
S494.0(5)	10	0.241	0.233	0.437	0.049	0.040	Na0.160Ca0.077Si0.145B0.032P0.026O0.560	2.08	0.22	7.37	9.00
S494.0(7)	15	0.241	0.233	0.413	0.073	0.040	Na0.157Ca0.076Si0.134B0.047P0.026O0.559	2.08	0.35	4.91	5.67
S494.0(10)	20	0.241	0.233	0.389	0.097	0.040	Na0.155Ca0.075Si0.125B0.062P0.026O0.558	2.08	0.50	3.69	4.00
S494.0(15)	30	0.241	0.233	0.340	0.146	0.040	Na0.150Ca0.072Si0.106B0.091P0.025O0.557	2.08	0.86	2.46	2.33
S494.0(19)	40	0.241	0.233	0.292	0.194	0.040	Na0.146Ca0.070Si0.088B0.117P0.024O0.555	2.08	1.33	1.85	1.50
S494.0(24)	50	0.241	0.233	0.243	0.243	0.040	Na0.141Ca0.068Si0.071B0.142P0.023O0.553	2.08	2.00	1.48	1.00

---

The precise glass forming region of the complex Na₂O–CaO–B₂O₃–SiO₂–P₂O₅ system is unknown and is currently being explored by us. Preliminary results indicate that the S49 series may allow higher B₂O₃ substitution levels than the S49_4.0(24) glass included in the present NMR study. However, it remains unclear how much further the amount of P may be increased for BPS glasses with significant B contents.

As demonstrated in Section 4.1, the phosphate speciation in our Na₂O–CaO–B₂O₃–SiO₂–P₂O₅ glasses is dominated by orthophosphate (Q⁰_P) groups (0.86 ≤ x⁰_P ≤ 0.96), with the remaining constituting Q¹_P moieties. To assist comparisons with the YDBX^37,38 prediction of the B^[3]/B^[4] partitioning, we introduce the ratio that accounts for the consumption of Na⁺/Ca²⁺ ions required to charge-balance all {Q⁰_P, Q¹_P} phosphate groups by considering the fictive monovalent ion , whose content relates to the net amount of positive charges available for the B/Si species of the borosilicate network, according to

Consequently, each Na₂O–CaO–B₂O₃–SiO₂–P₂O₅ glass composition SN_p(q) maps onto a –B₂O₃–SiO₂ counterpart associated with the parameter R′ that may be contrasted directly with the result of a Na₂O–B₂O₃–SiO₂ analog with R = x(Na₂O)/x(B₂O₃) and K = x(SiO₂)/x(B₂O₃) from the YDBX model.^37,38 Table 1 lists the resulting values of R′, K, and other relevant glass composition factors, such as x_B/x_Si.

Note that this simplistic picture only accounts for the total positive charge reservoir provided from the monovalent Na⁺ and divalent Ca²⁺ cations. In our previous NMR studies of ³¹P and ²⁹Si environments in phosphosilicate glasses,^33–35 no significant differences were observed from slight variations in the relative Na/Ca contents among different glasses, where the main glass-network alterations are captured well by solely considering the total glass modifier content, the P content, and the silicate network connectivity. Also for the present BPS glasses, the minor difference among the x_Na/x_Ca ratios of the S46 and S49 families is not expected to alter the glass structures significantly; indeed, we verified for two S49 members that the two x_Na/x_Ca values of 1.84 and 2.08 yield B^[4] fractional populations that agree within 97% [for S49_4.0(15)] and 99% [for S49_4.0(24)]. While such minor differences are immaterial for any result discussed herein, further studies are required to assess the precise structural bearings from the relative amounts of Na⁺ and Ca²⁺ (if any), as well as their influences on the in vitro glass degradation and bioactivity.

3 Materials and methods

3.1 Glass preparation and characterization

6.0 g batches of the BPS glasses were prepared by a standard melt-quench method, using precursors of SiO₂ (99.99%), Na₂CO₃ (99.99%), and CaCO₃ (99.9%) from ChemPur, and NaH₂PO₄ (99.99%, Merck), and H₃BO₃ (99.9%, Sigma). Prior to its weighing, the SiO₂ powder was heated at 950 °C for 24 h to remove potential OH/H₂O contaminations. The precursors were ground, weighted, and transferred to a bottle that was shaken thoroughly before another mixing stage in a mortar for 1 hour to achieve homogeneous powders. The mixture was then placed in a Pt crucible and heated in an electric furnace at 950 °C for 120 min to allow for complete CO₂ removal. Depending on the P₂O₅ and SiO₂ contents of each batch, the temperature was then increased to 1200–1250 °C (S46 series) and 1300–1400 °C (S49 series), held for 20 min (S46 specimens) or 30 min (S49), before quenching the melt by immersing the bottom of the crucible in water.

The evaporation losses during synthesis remained ≲1.5 wt% throughout. We verified by ¹¹B MAS NMR on two specimens [S46_2.6(28) and S46_2.6(37)] that insignificant alterations in the peakshapes and NMR intensities resulted after repeating the melt-quench process on the initially prepared glasses. Besides confirming sufficient precursor mixing, it strongly suggests very limited evaporation losses of the most volatile glass ingredient (i.e., B). Moreover, the relative integrated ³¹P NMR intensities (Section 4.1) among the samples were in excellent agreement with the corresponding nominal x(P₂O₅) values, thereby further suggesting intact batch compositions.

Polished glass samples were coated by a 10–20 nm thick carbon layer and examined by scanning electron microcopy (SEM), using a JEOL JSM-7000F microscope in backscatter electron mode at a 15 kV acceleration voltage. None of the specimens manifested signatures of phase separation down to the shortest accessible length-scale of 1 μm.

3.2 Solid-state NMR

All NMR experiments utilized Bruker Avance-III spectrometers. The ³¹P and ²⁹Si MAS NMR spectra were recorded at a magnetic field (B₀) of 9.4 T, which corresponds to ³¹P and ²⁹Si Larmor frequencies of −162.0 MHz and 79.5 MHz, respectively. Glass powder were filled in 7 mm zirconia rotors that were spun at 7.00 kHz during the NMR signal acquisitions. All ¹¹B NMR experimentation was performed at B₀ = 14.1 T (−192.5 MHz ¹¹B Larmor frequency) with 3.2 mm rotors spinning at 24.00 kHz.

The ³¹P and ²⁹Si NMR acquisitions were performed in blocks, using single-pulse excitation with flip angles of 70° (³¹P) or 60° (²⁹Si) at radio-frequency (rf) nutation frequencies ν_nut = |γ|B₁/2π ≈ 60 kHz, where B₁ is the rf amplitude. The ³¹P NMR acquisitions involved ≈48 accumulated signal transients, with 5000 s “equilibration” delays before the start of each block of 8 transients, and relaxation delays of 2500 s between each rf pulse within the block. Depending on the Si content of the glass, between 92–156 transients were recorded with 3600 s relaxation delays and equilibration delays of 3.5 h between each block of 4–8 transients. 300 Hz and 200 Hz full width at half maximum (fwhm) height Gaussian broadening were employed in the ²⁹Si data processing of the S46 and S49 glasses, respectively, whereas no signal apodization was employed for the ³¹P NMR data.

¹¹B MAS NMR spectra were recorded using short (0.5 μs; ≈15°) rf pulses at ν_nut ≈ 75 kHz, relaxation delays of 15 s, and 512–3072 accumulated transients, depending on the B content of the sample. ¹¹B background signals from the MAS NMR probehead were removed by subtracting the results from an empty rotor. No signal apodization was used in the processing, except for the S49_4.0(2) glass that employed 225 Hz fwhm Gaussian broadening.

Triple quantum MAS (3QMAS)^{57 11}B NMR acquisitions employed the Z-filter scheme^58,59 utilizing the symmetric 0 → ±3 → ±1 → 0 → −1 quantum coherence pathways. All rf pulses for 3Q coherence (3QC) excitation and conversion operated at ν_nut = 115 kHz. The 3QC excitation pulse lasted for 5.8 μs. Two FAM⁶⁰ blocks with equal durations of the pulses and delays of 0.90 μs were used for the ±3 → ±1 coherence transfer. The two central transition (CT) selective 90° pulses of duration 17.0 μs (ν_nut = 7.4 kHz) following the 3QC conversion were interleaved by a Z-filter delay of one rotational period. The 2D NMR acquisitions involved the States procedure⁶¹ for providing absorptive peaks with frequency sign-discrimination along the indirect spectral dimension, together with dwell times {Δt₁, Δt₂} = {41.7, 6.1} μs, 2 s relaxation delays, and 288–3456 co-added transients. Typically, 100(t₁) × 2428(t₂) time-points were recorded and zero-filled to 512 × 8192 points. The indirect 3QMAS dimension was processed according to the “C_z” convention.^57,62

These experimental conditions—i.e., the flip angles, “pre-equilibration” and “relaxation” (or “interpulse”) delays—were selected for each nucleus from separate saturation-recovery T₁ measurements to ensure NMR data quantitatively reflecting the populations of the various B, Si and P structural sites. ³¹P, ¹¹B, and ²⁹Si shifts are quoted relative to 85% H₃PO₄(aq.), neat BF₃·OEt₂, and neat tetramethylsilane (TMS), respectively; the liquids were also exploited for estimating rf nutation frequencies, except for the ³¹P rf pulses that were calibrated directly on the glass samples.

4 Results

4.1 ³¹P MAS NMR

The ³¹P MAS NMR spectra recorded from the S46 and S49 glass families are shown in Fig. 1. The two base glasses, S46_2.6(0) and S49_4.0(0), reveal peak maxima at 8.7 ppm and 8.3 ppm, respectively, corresponding to signals from orthophosphate groups charge-balanced by Na⁺/Ca²⁺ cations.^{22,30–34,63} The marginally lower shift of the NMR peak from the S49_4.0(0) glass, which features a slightly higher x_Na/x_Ca ratio and N̄^Si_BO value, stems from two counteracting factors: the deshielding effect of Na⁺ (relative to Ca²⁺) that increases the shift,^{22,30,33,34,63} and the shielding (shift reduction) resulting from emphasized Ca²⁺–PO₄³⁻ associations when N̄^Si_BO grows.³⁴ The large fwhm of the NMR peaks ≈7.5 ppm reflects the disordered glass structures. The ³¹P resonances of S46_2.6(0) are characterized by a nearly Gaussian peakshape, while that for S49_4.0(0) appears more asymmetric, featuring a “tail” towards lower shifts: it stems from Q¹_P moieties, i.e., phosphate groups with one P–O–Si (as opposed to P–O–P) bond.^{32–34,36,64,65} For both S46 and S49 series, Fig. 1 reveals progressively more asymmetric ³¹P resonances for increasing x(B₂O₃) in the glass. While the growing amount of Q¹_P species may potentially involve either of P–O–Si/B bonding scenarios, additional heteronuclear ³¹P{¹¹B} NMR experiments (to be published elsewhere) suggest that few P–O–B bonds form and that P–O–Si bridges prevail, as in the parent B-free phosphosilicate glasses.

Fig. 1 Experimental ³¹P MAS NMR spectra recorded at 9.4 T from the as-indicated SN_p(q) glasses, with members from the S46 and S49 series shown in the left and right panels, respectively, and listed according to increasing B₂O₃ content from top to bottom. The ³¹P resonances from the Q⁰_P (main peak) and Q¹_p (minor) groups are indicated, with the fractional populations of the latter specified in %. The curve beneath each NMR spectrum represents the difference between experimental and best-fit data.

To quantify the influence of the increased B₂O₃ content on the phosphate speciation, we deconvoluted each ³¹P NMR peak into two Gaussian components by a constrained iterative fitting, considering both the centerband and first-order spinning sidebands stemming from each Q⁰_P/Q¹_P group (see Mathew et al.³³). Each peak component was represented by three parameters, {δⁿ_P, Wⁿ_P, xⁿ_P}, corresponding to the mean ³¹P chemical shift (δⁿ_P), the fwhm (Wⁿ_P), and the fractional population (xⁿ_P). Table 2 lists the best-fit results. For both the S46/S49 glass series, the chemical shift of the orthophosphate groups (δ⁰_P) decreases congruently with the respective maximum of the net NMR signal, suggesting a more pronounced propensity for the Q⁰_P groups to associate with the Ca²⁺ cations relative to Na⁺.^33,34,36 The more than twice as large x¹_P population (≈10%) of the S49_4.0(0) structure relative to its S46_2.6(0) counterpart is consistent with its larger N̄^Si_BO value and the linear N̄^Si_BO/x¹_P relationship established recently by Mathew et al.³³ When the B₂O₃ content grows, a weak increase of x¹_P is observed between 0.041–0.065 (for S46) and 0.100–0.136 (for S49), which reflect minor increases of the silicate network polymerization, as discussed in Section 4.2.

Table 2 ³¹P and ²⁹Si NMR results

Label	^31P NMR data^a								^29Si NMR data^b
	δ^Pmax (ppm)	WP (ppm)	δ^0P (ppm)	W^0P (ppm)	x^0P	δ^1P (ppm)	W^1P (ppm)	x^1P	δ^Simax (ppm)	δ^SiCG (ppm)	WSi (ppm)
a Shift at peak maximum (δ^Pmax; uncertainty ±0.08 ppm) and full width at half maximum (fwhm) height (WP; ±0.15 ppm) of the net ^31P NMR signal, as well as the chemical shift (δ^nP; ±0.3 ppm), fwhm height (W^nP; ±0.5 ppm), and fractional population (x^nP; ±0.01) of each Q^nP contribution extracted by NMR spectra deconvolution.b Peak maximum (δ^Simax; uncertainty ±0.15 ppm), the center-of-gravity shift (δ^SiCG; ±0.25 ppm), and fwhm height (WSi; ±0.25 ppm) of the net ^29Si NMR signal.
S462.6(0)	8.72	7.50	8.7	7.4	0.959	0.4	7.3	0.041	−79.5	−81.0	11.1
S462.6(5)	8.60	7.35	8.6	7.2	0.952	0.4	7.1	0.048	−80.0	−80.7	11.9
S462.6(9)	8.46	7.37	8.5	7.2	0.952	0.4	7.1	0.048
S462.6(14)	8.35	7.41	8.4	7.3	0.952	0.4	6.8	0.048	−79.9	−80.5	12.4
S462.6(18)	8.27	7.44	8.3	7.3	0.949	0.4	6.8	0.051
S462.6(28)	8.07	7.53	8.1	7.4	0.941	0.4	6.8	0.059	−79.5	−80.1	12.7
S462.6(37)	7.79	7.57	7.8	7.4	0.935	0.4	6.6	0.065	−78.6	−79.6	12.5
S494.0(0)	8.34	7.46	8.3	7.3	0.900	0.0	7.8	0.100	−82.5	−84.9	16.0
S494.0(2)	8.28	7.49	8.3	7.3	0.897	0.0	7.8	0.103	−82.8	−84.4	15.7
S494.0(5)	8.14	7.52	8.1	7.3	0.897	0.0	7.6	0.103
S494.0(7)	8.08	7.55	8.1	7.3	0.893	0.0	7.6	0.107	−83.2	−84.2	15.0
S494.0(10)	7.96	7.59	8.0	7.4	0.890	0.0	7.5	0.110
S494.0(15)	7.97	7.66	7.9	7.4	0.887	0.0	7.2	0.113
S494.0(19)	7.64	7.76	7.6	7.5	0.871	0.0	7.3	0.129
S494.0(24)	7.51	7.85	7.5	7.5	0.864	0.0	7.3	0.136	−82.9	−84.0	14.3

---

4.2 ²⁹Si MAS NMR

The ²⁹Si chemical shift of a Qⁿ_Si group in a silicate glass depends essentially on the same structural factors as ³¹P in phosphate glasses (see Section 4.1), i.e., on the relative number of BO/NBO species at the SiO₄ tetrahedron (n) and the precise Na⁺/Ca²⁺ constellation that provides local charge balance.^{1–3,40,66,67} However, in a multicomponent glass—such as the present Na₂O–CaO–SiO₂–P₂O₅ specimens S46_2.6(0) and S49_4.0(0)—the BO/NBO distribution, variations in the relative numbers of Na⁺/Ca²⁺ cations around each SiO₄ group, and distributions in Si–O bond lengths and Si–O–Si bond angles give a plethora of co-existing local ²⁹Si environments with similar chemical shifts that preclude the identification of the responses from individual Qⁿ_Si species.^1–3,40

Fig. 2(a) and (b) presents the ²⁹Si MAS NMR spectra of the two B-free S46_2.6(0) and S49_4.0(0) glasses. Consistent with the expected domination of Q²_Si groups in the S46_2.6(0) structure (N̄^Si_BO = 2.11), its ²⁹Si NMR peakshape is nearly Gaussian. However, a “tail” towards lower shifts is discernible, signifying the presence of Q³_Si groups resonating at lower chemical shifts ≈−88 ppm. Their larger contributions to the more polymerized silicate network of the S49_4.0(0) glass is evident from its NMR peakshape that corresponds to a superposition of (mainly) two Gaussian peaks centered around −80 ppm (Q²_Si) and −88 ppm (Q³_Si).

Fig. 2 Experimental ²⁹Si MAS NMR spectra obtained at 9.4 T (black traces) from the (a) S46_2.6(0) and (b) S49_4.0(0) glasses, shown together with the set of {Qⁿ_Si} component peaks (gray traces) obtained by spectra deconvolution. Each gray number within (or on top) of each NMR peak specifies its relative population in %. The curve beneath each NMR spectrum represents the difference between the experiment and the corresponding best-fit. (c and d) NMR spectra obtained from glasses with as-specified molar fractions of B₂O₃ from each (c) S46 and (d) S49 series. Note that the results for S46_2.6(0)—i.e., the “45S5 Bioglass”¹⁵—are reproduced from Mathew et al.³³

Fig. 2(a) and (b) also displays the deconvolution of each NMR spectrum into its underlying {Qⁿ_Si} peak components, where the results for S46_2.6(0) are reproduced from Mathew et al.³³ Q²_Si groups (≈72%) dominate the silicate speciation of the S46_2.6(0) structure, as expected from its nominal silicate network connectivity of 2.11. Indeed, the best-fit {x¹_Si, x²_Si, x³_Si} populations together with the expression

provided the experimental estimate N̄^Si_BO = 2.14;³³ see Fig. 2(a). The respective {x¹_Si, x²_Si, x³_Si, x⁴_Si} values of S49_4.0(0) are {0.015, 0.403, 0.571, 0.011}, corresponding to N̄^Si_BO = 2.58 and in excellent agreement with the nominal value N̄^Si_BO = 2.54. These populations accord well with those of the B-free Na–Ca–Si–P–O “BG_4.0(2.5)” glass discussed in ref. 33 that is associated with N̄^Si_BO = 2.50. The present S49_4.0(0) composition features very close SiO₂ and P₂O₅ contents to BG_4.0(2.5), with the main distinction being their different Na/Ca contents, where S49_4.0(0) is richer in Na (x_Na/x_Ca = 2.08) relative to BG_4.0(2.5) with x_Na/x_Ca = 1.54.

Fig. 2(c) and (d) displays the ²⁹Si NMR spectra of the B-bearing members of each S46/S49 glass family. Only minor resonance-alterations are observed when SiO₂ is substituted by B₂O₃. Notably, the NMR peak maxima from all S46 glasses remain ≈−80 ppm for x(B₂O₃) ≤ 0.28, where only the B-richest glass reveals a noticeable peak displacement to −78.6 ppm. Likewise, all S49-deriving glasses manifest equal peak maxima ≈−83 ppm, with their main distinctions being a reduced signal intensity in the low-ppm region ≈−90 ppm as the B content is increased. Two factors may account for the minor increase of the ²⁹Si chemical shifts when B is included in the glass structure:

(i) The silicate network polymerization decreases slightly, such that lower-n Qⁿ_Si populations grow, thereby inducing a ²⁹Si deshielding with a typical shift displacement of 7–12 ppm per Qⁿ_Si → Qⁿ⁻¹_Si transformation.^1–3,66,67 However, this implies an increase of the average number of NBO species at the SiO₄ groups. This scenario is precluded by the constant Na⁺/Ca²⁺ content of the glasses of each S46/S49 series: after accounting for the modifier consumption by the phosphate and [BO₄]⁻ species, the BO₃ groups may also accommodate NBO ions and therefore an increase (rather than a decrease) of the silicate network polymerization is anticipated.

(ii) Substitutions of Si by B^[4] species in the second coordination shell of ²⁹Si typically induce ≈5 ppm ²⁹Si deshielding per ²⁹Si–O–Si → ²⁹Si–O–B^[4] conversion,^{1–3,39,45–47} whereas the ²⁹Si–O–Si → ²⁹Si–O–B^[3] replacements preserve the net charge of the O atom, thereby leaving the ²⁹Si chemical shift value essentially invariant.^43,45–47 While it is then tempting to interpret the minor alterations of the ²⁹Si NMR spectra in Fig. 2 to reflect a predominance of Si–O–B^[3] bonds,³⁹ this scenario is excluded from the ¹¹B NMR results discussed in Sections 4.3 and 5.1.

Most likely, the weak but clearly discernible deshielding of the ²⁹Si nuclei reflects the net effect of the two opposing factors (i) and (ii), i.e., the shift elevation from the neighboring B^[4] atoms is partially offset by a slight shift reduction from a higher N̄^Si_BO value. The enhanced Si/B intermixing leads to a larger shift dispersion stemming from a plethora of ²⁹Si environments featuring variable number of Si–O–B^[4] bonds;^40,45–47 see Fig. 2(d). The ²⁹Si resonance-broadening and overall loss of peakshape features for increasing B₂O₃ content is well documented: for instance, see Fig. 3 of El-Damrawi et al.⁴⁶ and Fig. 4 of Edén et al.⁴⁰ Owing to the multitude of potentially co-existing but unresolved ²⁹Si NMR responses, spectra deconvolutions are unfortunately not warranted for the B-bearing glasses. However, given the absence of significant ³¹P–O–B contacts (Section 4.1) coupled with the linear dependence of the Q¹_P population on the average silicate network connectivity,³³ we may estimate N̄^Si_BO from the x¹_P-values of Table 2 and the quantitative x¹_P/N̄^Si_BO relationship established by Mathew et al.³³ Across the S46 glass series, the growth of x¹_P from 0.041 to 0.065 then translates into a minor increase of N̄^Si_BO from 2.11 to ≈2.24, while the S49 glasses yield the corresponding N̄^Si_BO growth from 2.54 to 2.73. Hence, when the B₂O₃ content of the glass elevates, the silicate network polymerization increases slightly, while the ²⁹Si sites become slightly deshielded by the formation of Si–O–B^[4] bonds. The observation of a nearly constant silicate network polymerization on B ↔ Si substitutions at a constant modifier content accords with the recent findings by Smedskjaer et al.⁵³ for more condensed borosilicate networks.

Fig. 3 ¹¹B MAS NMR spectra obtained at 14.1 T from the (a) S46 and (b) S49 glasses for the as-indicated x(B₂O₃) values. The total integrated signal intensity of each NMR spectrum is normalized to unity.

Fig. 4 The fractional population of B^[4] groups (out of the total B speciation) plotted against the molar fraction of B₂O₃ for the S46 (solid squares) and S49 (solid circles) glass series. Open symbols represent predictions from the Yun–Dell–Bray–Xiao (YDBX) model.³⁸

4.3 ¹¹B NMR

4.3.1 MAS NMR. Fig. 3 shows a selection of ¹¹B MAS NMR spectra from each S46 and S49 glass family. The spectra are zoomed around the CT signal region and reveal two groups of resonances: a broad signal from ¹¹BO₃ groups, and a narrow asymmetric Gaussian peak from the ¹¹BO₄ tetrahedra. The latter exhibits lower average quadrupolar coupling constants (C̄_Q, with C_Q = e²qQ/h) owing to the higher symmetry of the charge distribution around the ¹¹B nucleus.^37,38,46,47 The following general trends are observed in Fig. 3 when SiO₂ is substituted by B₂O₃ at a constant network modifier content: (i) the intensity of the ¹¹B^[4] resonance increases due to B^[3] → B^[4] conversions,^37,38,45,47 while its position displaces monotonically towards higher shifts. As analyzed further in Section 5.1, this reflects a concurrent decrease in the number of Si atoms in the second coordination shell of ¹¹B.^{44,45,49–52} (ii) The ¹¹B^[3] resonance alters from an essentially featureless peak, centered at 13.9 ppm (S46) and 13.3 ppm (S49) from the Si-rich glass compositions, to a broad signal revealing two maxima ≈15.8 ppm and ≈12.5 ppm; see Fig. 3. This suggests that the ¹¹B NMR signals in the ≈8–20 ppm spectral region stem from at least two distinct ¹¹BO₃ environments in all Si-dominated glasses, as discussed further below.

The {x^[3]_B, x^[4]_B} populations were determined from the integrated CT NMR signal intensities of the respective {¹¹BO₃, ¹¹BO₄} groups, including the weak first-order ¹¹BO₃ CT spinning sidebands. Each ¹¹BO₄ intensity was corrected for the contributions of the ST centerband (which overlap with the main CT resonance) according to the procedure of Massiot et al.;⁶⁸ the correction yielded 0.02–0.03 lower x^[4]_B values relative to the corresponding uncorrected populations. The results are listed in Table 3, together with the mean isotropic chemical shifts and average quadrupolar products C̄_Qη of the ¹¹BO₄ moieties (C_Qη relates to C_Q according to , where η is the asymmetry parameter of the quadrupolar tensor^1–3). These parameters were obtained by analyzing the CT/ST signals of the ¹¹B MAS NMR spectra as described in ref. 69. We stress that all {[small delta, Greek, macron]_iso, C̄_Qη} parameters listed in Table 3 are mean values of (up to) three co-existing ¹¹BO₄ environments with variable numbers of Si/B^[3] neighbors (see Section 5.1), whose individual NMR parameters cannot be determined.

Table 3 ¹¹B NMR results^a

Label	B^[3]			B^[4]			Populations^b		BO3 populations^c
	δCG (ppm)	[small delta, Greek, macron]iso (ppm)	C̄Qη (MHz)	δCG (ppm)	[small delta, Greek, macron]iso (ppm)	C̄Qη (MHz)	x^[3]B	x^[4]B	x^[3]B(0Si)	x^[3]B(1Si)
a δCG(B^[3]) and δCG(B^[4]) with uncertainties of ±0.08 ppm and ±0.04 ppm, respectively, correspond to the center-of-gravity shifts of the ^11BO3 and ^11BO4 resonances in the MAS NMR spectra. All other NMR parameters for the ^11BO3 and ^11BO4 groups were obtained from 3QMAS and MAS ^11B NMR spectra, respectively, with uncertainties as follows: [small delta, Greek, macron]iso(B^[3])±0.20 ppm; [small delta, Greek, macron]iso(B^[4])±0.06 ppm; C̄Qη(B^[3])±0.10 MHz; C̄Qη(B^[4])±0.10 MHz.b The fractional populations (uncertainty ±0.01) were obtained by integrating the MAS NMR intensities and then correcting for the satellite-transition contributions to the CT centerband peak.^68c Relative populations of BO3 groups without [x^[3]B(0Si)] and with one [x^[3]B(1Si)] bond to Si (uncertainty ±0.03); they are normalized according to x^[3]B(0Si) + x^[3]B(1Si) = 1.
S462.6(5)	14.00			−0.04	0.12	0.39	0.831	0.169
S462.6(9)	13.98	18.6	2.72	0.31	0.43	0.35	0.740	0.260	0.76	0.24
S462.6(14)	13.98			0.49	0.59	0.36	0.705	0.295
S462.6(18)	13.96	18.7	2.74	0.58	0.72	0.35	0.684	0.316	0.86	0.14
S462.6(28)	14.01			0.83	0.94	0.38	0.665	0.335
S462.6(37)	14.02	18.8	2.70	1.02	1.12	0.39	0.660	0.340	1.00	0.00
S494.0(2)	13.21			−0.47	−0.31	0.37	0.635	0.365
S494.0(5)	13.22			−0.29	−0.19	0.35	0.612	0.388
S494.0(7)	13.29			−0.15	−0.01	0.36	0.589	0.411
S494.0(10)	13.40	18.2	2.70	0.09	0.20	0.37	0.576	0.424	0.70	0.30
S494.0(15)	13.62			0.39	0.49	0.37	0.568	0.432
S494.0(19)	13.65			0.51	0.61	0.38	0.567	0.433
S494.0(24)	13.72	18.4	2.72	0.59	0.72	0.40	0.567	0.433	0.83	0.17

---

Fig. 4 plots the B^[4] fractional population (x^[4]_B) against x(B₂O₃). Both S46/S49 series reveal a strong growth of x^[4]_B when x(B₂O₃) is increased at low substitution levels [x(B₂O₃) ≲ 0.07]. When the amount of B is elevated further, x^[4]_B tends towards asymptotic values of ≈0.34 and ≈0.43 for the S46 and S49 glass families, respectively. However, while the plateau is reached at already ≈30% B₂O₃ for SiO₂ substitution of the S49 glasses, the B^[4] population manifests a slow increase throughout the entire range of B₂O₃ contents in the S46 series. Noteworthy, the increase of x^[4]_B (by ≈0.17) across the S46 glasses is more than twice that (≈0.07) observed for the S49 members. As expected from the ¹¹B^[4] peak intensities of Fig. 3, the overall Si-richer S49 glasses exhibit higher relative B^[4] populations than their S46 counterparts. While this trend is predicted by the YDBX model,^37,38 it accounts only qualitatively for the de facto observed alterations of the experimental B^[4] populations for variable amounts of B; see Fig. 4.

4.3.2 3QMAS NMR. Since it removes the anisotropic second-order quadrupolar broadenings, the 3QMAS 2D NMR technique⁵⁷ improves the ¹¹B NMR spectral resolution by producing a high resolution “isotropic” (indirect) dimension. The direct (“MAS”) dimension essentially retains the MAS NMR spectrum. In the isotropic projection of 3QMAS data from glasses, the residual peak broadening stems mainly from isotropic chemical-shift distributions due to the structural disorder.^3,59,70 Fig. 5(a) displays the 3QMAS ¹¹B NMR spectrum together with its projections along each isotropic/MAS dimension from the Si-rich S46_2.6(9) glass. The NMR spectrum reveals two well-separated 2D “ridges” of resonances from the ¹¹B^[4] and ¹¹B^[3] structural sites that extend along both spectral dimensions: the two arrows shown in Fig. 5(a) represent the directions of signal dispersion stemming from the respective spreads of isotropic chemical shifts and C_Q values.⁷⁰

Fig. 5 (a) 3QMAS ¹¹B NMR spectrum recorded at 14.1 T from the S46_2.6(9) glass. (b and c) Zooms around the ¹¹B^[3] signal region of 3QMAS NMR spectra obtained from (b) S46_2.6(9) (black contours); S46_2.6(37) (red contours), and (c) S49_4.0(10) (black contours); S49_4.0(24) (red contours). The arrows mark the directions of the resonance spreads stemming from distributions of isotropic chemical shifts (marked by “CS”) and quadrupolar coupling constants (labeled “Q”).

Fig. 5(b) and (c) shows zoomed areas around the ¹¹B^[3] 2D NMR ridge, where pairs of spectra from glasses with low and high amounts of B are superimposed from the (b) S46 and (c) S49 glass series. Each Si-dominated glass reveals a higher NMR signal-dispersion towards lower shifts relative to its B-richer counterparts: it is manifested by an enhanced NMR intensity ≈74 ppm in the corresponding projection along the isotropic dimension and suggests that the net ¹¹B^[3] resonance stems from two (or more) distinct BO₃ structural environments. This is most transparent for the two S46 samples that exhibit the largest span of B contents; see Fig. 5(b).

The correlation between the signal intensity ≈74 ppm and the relative SiO₂ content of the glass is evident from the isotropic projections shown in the left panel of Fig. 6. The ¹¹B NMR peakshape from the almost Si-free S46_2.6(37) glass fitted well to a single Gaussian peak centered ≈82 ppm [Fig. 6(a)], indicating that its structure comprises primarily one type of BO₃ structural group. Given the domination of B relative to Si in the glass network (x_B/x_Si = 8.00; see Table 1), we attribute it to a BO₃ environment participating in boroxol rings^50–52,55 and devoid of bonds to Si atoms; this moiety is consequently labeled B^[3](0Si). The lower-ppm signal grows for increasing SiO₂ content, indicating it is associated with a “non-ring” BO₃ environment connected to one SiO₄ group, which we denote by ¹¹B^[3](1Si).

Fig. 6 Projections along the isotropic dimension of 3QMAS ¹¹B NMR spectra (left panel) and MAS spectra (right panel) from the as-indicated S46 and S49 glasses. The black and red traces represent experimental and best-fit peakshapes, respectively: they comprise two signal components at ≈80–82 ppm (gray lines; ¹¹B^[3](0Si) groups in boroxol rings) and ≈74 ppm (solid areas; ¹¹B^[3](1Si) moieties). Gray numbers specify the fractional populations (in %) of the B^[3](1Si) groups. (k) Ranges of center-of-gravity (CG) shifts along the isotropic projection for ¹¹B^[3] moieties with 0, 1, 2, and 3 NBO ions, compiled from literature data:^{51,55,56,72–79} the white bar marks [small delta, Greek, macron]_CG, whereas the gray and black rectangles depict the total shift-span and [small delta, Greek, macron]_CG ± σ, respectively. (l) As in (k), but illustrating the range of {[small delta, Greek, macron]_iso}, which is to be compared with the span of experimentally observed extreme [small delta, Greek, macron]_iso-values (dotted lines) at 18.2 ppm and 18.8 ppm; see Table 3.

The relative ¹¹B^[3](0Si) [“ring”] and ¹¹B^[3](1Si) [“non-ring”] contributions were extracted by deconvoluting each NMR spectrum into two components; see Fig. 6. The variable ¹¹B^[3](0Si) and ¹¹B^[3](1Si) peak intensities are responsible for the lineshape alterations of the MAS spectra in Fig. 3 when the x_B/x_Si ratio of the glass is changed. The B^[3](1Si) population grows up to 24% when x(B₂O₃) is decreased from 0.37 to 0.09 in the S46 series; it amounts to 30% out of the total B^[3] speciation in the Si-richest glass [S49_4.0(10)] that was examined by 3QMAS ¹¹B NMR; see Fig. 6(i). As follows by comparing the results for S46_2.6(9) and S49_4.0(10) that feature a constant B₂O₃ for SiO₂ substitution degree of 20%, the B^[3](1Si) contribution is larger for the Si-richer BPS glass with lower NBO content. These trends of growing “non-ring” fraction for increasing (decreasing) amount of Si (NBO) accord with the results of Du and Stebbins for alkali-poor borosilicate glasses.^50,51

5 Discussion

5.1 BO₄ environments

5.1.1 B^[4]–O–B/Si bond statistics. The ¹¹BO₄ NMR peak displacements towards higher chemical shifts for increasing B₂O₃ content (Fig. 3; Table 3) reflect progressive B^[4]–O–Si → B^[4]–O–B^[3]/B^[4] substitutions. Net NMR peak movements of ≈1 ppm are observed across both glass series, while the present ¹¹B CG shifts are overall higher than those of the mainstream borosilicate studies that generally targeted Si-richer glasses with relatively low modifier contents, x(M₍₂₎O) ≲ 0.3.^{39,43,46,47,50–54,69}

Given that the total Na⁺/Ca²⁺ content remains constant within each S46 and S49 series, the ¹¹B^[4] resonance-displacement may be accounted for by a fixed set of groups, {¹¹B^[4](mSi)}, each featuring m bonds to Si atoms (and thereby 4 − m bonds to B^[3]) and resonating at the CG shift δ^[4]_B(mSi). Several investigations report a propensity for BO₄ groups to connect to SiO₄ tetrahedra rather than the planar BO₃ moieties,^{37,38,50–52} whereas direct B^[4]–O–B^[4] linkages are often assumed to be absent in the borosilicate structure.^{1–3,37,38,50–52} Hence, B^[4](4Si) environments are expected to dominate the BO₄ speciation in the Si-richest BPS glasses. Yet, considering the wide ranges of x_B/x_Si ratios sampled across the S46/S49 glass series (notably so for the S46 set; see Table 1), more than two structural moieties should co-exist: with the constraints of combining physically reasonable {δ^[4]_B(mSi)} values with decent “best fits”, three {¹¹B^[4](4Si), ¹¹B^[4](3Si), ¹¹B^[4](2Si)} groups were selected and each ¹¹BO₄ NMR signal of Fig. 3 was deconvoluted into Gaussian-shaped peaks centered at the nearly constant (±0.1 ppm) shift-values {−0.45, 0.65, 1.85} ppm for all glasses. We stress that these NMR spectra deconvolutions are somewhat arbitrary because the set of precise {δ^[4]_B(mSi)} values are a priori unknown. The proposed shift alteration associated with a B^[4](mSi) → B^[4]([m − 1]Si) conversion varies between various studies, where (for instance) shift differences of 0.7–1.0 ppm,^47,71 and 1.7–2.0 ppm (ref. 51 and 52) are suggested.

Fig. 7 presents the set {x^[4]_B(mSi)} of best-fit fractional populations plotted against each ratio

y_B = x_B/(x_B + x_Si), (3)

and

y^[3]_B = x_Bx^[3]_B/(x_Bx^[3]_B + x_Si), (4)

in the left and right panels, respectively. When employing y_B, no distinction is made between the two B^[3]/B^[4] coordinations, whereas the ratio y^[3]_B is relevant for the scenario of B^[4]–O–B^[4] avoidance, where y^[3]_B represents the fraction of BO₃ groups out of the total number of BO₃/SiO₄ species available for bonding to the BO₄ tetrahedra. Note that depending on whether eqn (3) or (4) is considered, y_Si obeys y_Si = 1 − y_B and y_Si = 1 − y^[3]_B, respectively. At low B substitution levels, x(B₂O₃) ≲ 0.05, the BO₄ groups enter the glass structure predominantly as B^[4](4Si) environments, while the B^[4](3Si) population accounts for ≲20% out of the total B^[4] speciation and the fraction of B^[4](2Si) moieties remains x^[4]_B(2Si) < 0.1; see Fig. 7(a)–(d). For increasing B₂O₃ content, the progressive B^[4](4Si) → B^[4](3Si) → B^[4](2Si) conversions imply that BO₄ groups with three bonds to Si prevail for x(B₂O₃) ≳ 0.14, i.e., when ≳30% of SiO₂ is replaced by B₂O₃. In the B-richest S46_2.6(37) structure, similar populations of B^[4](3Si) and B^[4](2Si) environments co-exist, whereas the B^[4](4Si) species only represent ≈9% of all BO₄ groups; see Fig. 7(a) and (b).

Fig. 7 (a–d) NMR-derived best-fit populations [x^[4]_B(mSi)] of BO₄ groups with m B–O–Si bonds observed from the series of (a, b) S46 and (c, d) S49 glasses. The data are plotted against the fractions (a, c, e) y_B = x_B/(x_B + x_Si) and (b, d, f) y^[3]_B = x_Bx^[3]_B/(x_Bx^[3]_B + x_Si). y_B and y^[3]_B are relevant for the scenarios of unrestricted B^[4]–O–B^[3]/B^[4]/Si bond formation and B^[4]–O–B^[4] avoidance, respectively. (e, f) Average number of Si atoms per BO₄ group (N̄_Si) obtained from the fractional populations in (a, c) and (b, d), respectively. The experimental data are compared with the as-indicated statistical models, where P_Si = 0 corresponds to a random bond formation between B^[4] and atoms from the set (e) {Si, B^[3], B^[4]} or (f) {Si, B^[3]}, whereas P_Si > 0 implies a preference for B^[4]–O–Si linkages [see eqn (5)]. To improve visualization in (e, f), the error bars (±0.04) are generally suppressed.

5.1.2 Quantifying the preference for B^[4]–O–Si bonds. Fig. 7(e) and (f) displays the x(B₂O₃)-dependence of the average number of Si atoms in the second coordination sphere of B^[4]: . The corresponding average number of B atoms (N̄_B) is given from the constraint N̄_B = 4 − N̄_Si, where “B” represents the total B speciation in Fig. 7(e) but solely the B^[3] coordinations in (f). The mean number of Si neighbors around the BO₄ groups decreases from N̄_Si = 3.62 in the Si-rich S46_2.6(5) structure to N̄_Si = 2.70 in the B-rich S46_2.3(37) glass. The corresponding data observed for the S49 series are N̄_Si = 3.94 [S49_4.0(2)] and N̄_Si = 2.95 [S49_4.0(24)]. These overall high N̄_Si-values underscore the preference for B^[4]–O–Si contacts relative to B^[4]–O–B. This inference is obvious when contrasting the experimentally observed {N̄_Si} data with the scenario of random distributions of either the Si/B^[3] species [Fig. 7(f)] or of all Si/B^[3]/B^[4] moieties [Fig. 7(e)] around the BO₄ groups. Besides that the statistical (binomial) distribution reveal a markedly faster drop of N̄_Si than that observed experimentally, it leads to substantial {x^[4]_B(0Si), x^[4]_B(1Si)} populations (data not shown) that are incommensurate with the experimental observation of (up to) three co-existing B^[4](mSi) structural groups.

To asses the degree of preference for B^[4]–O–Si bond formation relative to B^[4]–O–B, we employed the procedure introduced recently by Mathew et al.³⁴ It applies to a scenario where two distinct atom species V and W may be coordinated by U: when there is a preference for U–V contacts, the preference factor P_V—where the subscript “V” denotes the preferred species—encodes the deviation from a statistical V/W distribution around U (for which P_V = P_W = 0), where P_V = 1 represents the case of sole U–V bonding. Within this framework, the preference factor P_Si dictates the average number of Si (N̄_Si) and B (N̄_B) atoms around the B^[4] coordinations as follows:³⁴

N̄_Si = 4[P_Si + y_Si(1 − P_Si)], (5)

N̄_B = 4y_B(1 − P_Si) for equal B^[3]/B^[4] probabilities (6)

= 4y^[3]_B(1 − P_Si) for B^[4]–O–B^[4] avoidance, (7)

where y_B and y^[3]_B are given by eqn (3) and (4), respectively. Depending on whether B^[4]–O–B^[4] contacts are allowed [eqn (6)] or forbidden [eqn (7)], the total BO₃/BO₄ ensemble or solely the BO₃ groups contribute to N̄_B.

For each glass, the preference for B^[4]–O–Si bonds was calculated from the experimental data of Fig. 7(a)–(d) by using eqn (5) and solving for P_Si. Onwards disregarding the result for the Si-richest S49_4.0(2) glass that is considered to be an outlier, no striking differences are observed if B^[4]–O–B^[4] linkages would be absent or present, with the respective {P_Si} data typically scattering around the values 0.45 and 0.55. Consequently, the scenario of B^[4]–O–B^[4] avoidance is associated with a somewhat less pronounced propensity for B^[4]–O–Si bond formation; this is evident from Fig. 7(e) and (f), where the experimental N̄_Si values of nearly all members from the S46 and S49 families are bracketed by the results stemming from distribution models associated with P_Si = 0.60 and the corresponding lower limit of P_Si = 0.40 if B^[4]–O–B^[4] bonds are allowed [Fig. 7(e)] and P_Si = 0.20 when B^[4]–O–B^[4] links are absent [Fig. 7(f)].

To conclude, the BO₄ moieties in both S46 and S49 glass families display a clear preference for connecting to SiO₄ tetrahedra rather than to BO₃/BO₄ groups. The observed typical values of P_Si ≈ 0.5 are equivalent to a Si/B distribution where Si atoms first occupy two out of four neighboring positions around each BO₄ tetrahedron, whereas Si and B atoms are statistically distributed around the remaining two tetrahedral corners. However, considering the uncertainties and possible systematic errors of the experimental N̄_Si data, the precise values should not be taken too literally and further studies are required to better confine the relative preferences, as well as if “B^[4] avoidance” applies to modifier-rich BPS glass networks.

5.2 NBO partitioning among SiO₄ and BO₃ groups

The affinity for coordinating NBO ions among the three {P, Si, B} network formers decreases in the order P ≫ Si > B^[3] ≫ B^[4].^3,40 Because B^[4]–NBO bonds are assumed to be absent^3,40 and the NBO accommodation among the Q⁰_P (PO₄³⁻) and Q¹_P (PO_3.5²⁻) moieties are known accurately (see Table 2), we henceforth focus on the NBO partitioning among the SiO₄ and BO₃ groups.

The total NBO population available for depolymerizing the borate/silicate networks may be estimated for each glass after accounting for the consumption of Na⁺/Ca²⁺ cations by the phosphate groups [see eqn (1)] and the [BO₄]⁻ tetrahedra. By using the average silicate network connectivity (N̄^Si_BO) available from the Q¹_P population (Section 4.2), the mean numbers of BO (N̄^B_BO) and NBO (N̄^B_NBO) species at the BO₃ groups may be estimated. The S46 and S49 glasses exhibit relatively narrow N̄^Si_NBO ranges of 1.76–1.89 and 1.27–1.46, respectively, yielding the corresponding approximate N̄^B_NBO-values of 1 and 0.7–0.8. Hence, the {N̄^B_NBO, N̄^Si_NBO} parameter-pair only alters slightly across each S46/S49 glass family when the x_B/x_Si ratio varies. It is instructive to compare the results conveying the relative propensities for Si–NBO and B^[3]–NBO bond formation between the two series, which reveal the pairs {N̄^B_NBO, N̄^Si_NBO} ≈ {1, 2} for S46 and {N̄^B_NBO, N̄^Si_NBO} ≈ {0.75, 1.5} for S49. For the S46 glass members, this implies roughly two and one NBO species per SiO₄ and BO₃ group, respectively. In contrast, the more polymerized S49 networks involve SiO₄ tetrahedra that on the average coordinate ≈1.5 NBO (consistent with similar Q²_Si and Q³_Si populations; see Fig. 2), while ≈25% of the {BO₃} ensemble is devoid of NBO species.

Given that the S46 glass networks involve an NBO distribution peaked around ≈1 NBO per BO₃ moiety, whereas those of the S49 structures comprise BO₃ groups that accommodate either 0 or 1 NBO, one may ask if our ¹¹B NMR data of Fig. 3 and 6 may resolve these resonances. The shift-spans compiled in Fig. 6(k) from literature data^{51,55,56,72–79} (relevant for the peak maxima along the isotropic projection of 3QMAS spectra) suggest an overall minor deshielding of the ¹¹BO₃ environments for an increasing number of NBO anions, but also manifests strongly overlapping shift ranges from ¹¹BO₃ moieties with 0 and 1 NBO ions. Hence, these resonances cannot be separated with the 3QMAS NMR spectra resolution offered from our glasses. Nevertheless, significant contributions from ¹¹BO₃ groups with two or three NBO species may be excluded [Fig. 6(k)], in full accordance with our estimates above and further corroborated by the “typical” isotropic chemical-shift ranges displayed in Fig. 6(l) that are relevant for the ¹¹B MAS NMR spectra of the right panel of Fig. 6.

5.3 BPS structural model

5.3.1 Glass building blocks and bond preferences. Here we summarize our results for the various phosphate, silicate, and borate building blocks and discuss their (expected) preferences for interlinking in the BPS glass structure. For both S46/S49 BPS families, the phosphate ensemble constitutes a main component of isolated orthophosphate (Q⁰_P) tetrahedra (86–96%) that in direct analogy of the B-free silicate glasses³⁵ assume interstitial positions around the borosilicate network. The remaining portion (Q¹_P) form predominantly P–O–Si bonds (as opposed to P–O–B; to be discussed elsewhere), where they terminate segments of the (boro)silicate network. In the BPS glasses, B mainly exists as planar BO₃ groups that constitute 66–83% and 57–64% of the total B speciation in the S46 and S49 glass series, respectively, with BO₄ species accounting for the remaining. The x^[4]_B fraction increases concomitantly with the B₂O₃ content.

The networks of the parent B-free S46_2.6(0) and S49_4.0(0) glasses comprise mainly inter-connected Q²_Si and {Q²_Si, Q³_Si} silicate groups, respectively (see Fig. 2). A slight increase in N̄^Si_BO by up to ≈0.2 results when B is introduced, because some NBO ions relocate to the BO₃ groups, which on the average accommodate ≈1 and ≈0.7–0.8 NBO species in the B-bearing S46 and S49 glasses, respectively. Yet, the {Qⁿ_Si} speciation is expected to remain overall intact, except that Si–O–Si bonds are gradually replaced by Si–O–B^[4] (major) and Si–O–B^[3] (minor) linkages in the borosilicate network. The tetrahedral SiO₄ and BO₄ species strongly prefer to interconnect. This implies that B^[4](4Si) groups solely involving bonds to Si account for >70% of all BO₄ units in the Si-richest S49_4.0(2), S49_4.0(5), and S46_2.6(5) glasses; see Fig. 7. The average number of bonds to Si remains ≳3 until ≈40% of the SiO₂ reservoir is replaced by B₂O₃. Yet, a relatively high mean number N̄_Si ≈ 2.7 of B^[4]–O–Si bonds persists even in the B-richest S46_2.6(37) structure. The estimated preference factor P_Si ≈ 0.5 is equivalent to a scenario with two Si neighbors around each BO₄ tetrahedron, while the remaining two positions are filled statistically by Si and B^[3] atoms according to their relative abundances in the glass. The B^[3] coordinations prefer connecting to other species according to B^[3] > B^[4] > Si. Hence, despite that the total number of B^[3]–O–B^[4] bonds increases concurrently with the B₂O₃ content at the expense of the B^[3](1Si) [i.e. “non-ring”] fraction, the B^[3](0Si) [“ring”] groups mainly associate with themselves as boroxol rings, regardless of the precise x_B/x_Si ratio and total NBO content of the BPS structure.

5.3.2 Medium-range structural organization. The pronounced propensities for B^[3]–O–B^[3] and B^[4]–O–Si bond formation imply that the BPS glass structure likely comprises (sub)nm-sized “domains” of borosilicate (involving Si–O–Si and Si–O–B^[4] bonds) and borate (B^[3]–O–B^[3] fragments in boroxol rings) networks, similar to the organization of the Pyrex® structure discussed in ref. 80–82. These borate/borosilicate networks are connected through B^[3]–O–B^[4]–O–Si and B^[3]–O–Si fragments, with the former constellation generally prevailing. Some of these moieties likely also participate in “superstructural” borate units.^{38,54,74,79,81} The “direct” contact points between the BO₃ and SiO₄ groups, i.e., the B^[3](1Si) (“non-ring”) moieties, are only abundant in the Si-rich glasses and they are gradually replaced by B^[3](0Si) species when the B content of the glass is increased (see Table 3).

The relative proportions of these borate/borosilicate domains depends on the x_B/x_Si ratio of the glass. For instance, the Si-rich S46_2.6(5) structure associated with x_B/x_Si = 0.22 and x^[4]_B = 0.17 (see Tables 1 and 3) is dominated by a silicate network of Q²_Si groups cross-linked by B^[4](4Si) species. This Si-rich borosilicate network co-exist with a domain of boroxol rings featuring ≈1 NBO per BO₃ moiety. The two networks are merged by B^[4](3Si) and B^[3](1Si) groups. When the B₂O₃ content is increased, the number of BO₄ groups increases in the borosilicate network, while B^[4](3Si), B^[4](2Si)—and to a lesser extent B^[3](1Si)—species constitute bridges to the boroxol rings. The B-richest S46_2.6(37) structure comprises mainly boroxol rings that are cross-linked by B^[4](3Si) and B^[4](2Si) groups to a borosilicate network dominated by BO₄ groups; the latter is overall more condensed than those of the Si-rich S46 glasses, due to the absence of NBO species at the BO₄ tetrahedra and because the Q²_Si groups are minor components (x_B/x_Si = 8.00). Since the BO₃ groups now grossly outnumber the SiO₄ tetrahedra, most of the NBO reservoir resides at the boroxol rings (Section 5.2).

5.4 Implications for glass dissolution

Here we discuss our structural model in conjunction with general glass degradation trends reported in the literature for borosilicate/BPS glasses. We identify two structural descriptors that we expect to primarily control the dissolution rate of BPS glasses in aqueous solutions, and highlight some remaining structural questions that must be addressed for achieving bioactive BPS glasses with tailored solubility in (simulated) body fluids.

On their immersion in aqueous solutions, borate-based BGs are claimed to (A) degrade faster and (B) convert more completely to HCA relative to their silicate-based counterparts.^4–6,11,12 However, few (if any) direct experimental comparisons of the HCA formation among borate-bearing/free BGs are reported. The hitherto lacking detailed structural understanding of BPS glasses hampers a rational design for tuning the glass degradability (“solubility”) in aqueous media.

Concerning (B), it remains unclear if the HCA formation rate or amount thereof formed are obviously higher than those observed from highly bioactive silicate-based BGs such as the “45S5” composition,¹⁵ i.e., S46_2.6(0). As commented by Lepry and Nazhat,¹⁴ essentially all in vitro evaluations of HCA formation from B-bearing BGs employed solutions with 20–250 times higher phosphate concentrations than human plasma or its acellular simulated body fluid (SBF)⁸³ counterpart that has become the standard medium for in vitro testing. For a melt-prepared B-based 45S5 composition^4,5 (S46_2.6(46) in our notation), an induction time of 3 days for HCA formation in SBF was reported,¹⁴ which is significantly longer than that ≲12 h (ref. 84) observed for the original 45S5 formulation. More direct in vitro comparisons between silicate/borate-based BGs in SBF are required to evaluate their relative merits for HCA formation.

We next consider the glass degradability in aqueous media, i.e., (A). For (phospho)silicate glasses, it is well-known that a decrease in the silicate network connectivity (N̄^Si_BO) accelerates the glass dissolution.^20,23,42 Solubility data are not gathered for our BPS glasses, but one expects from the more condensed S49_4.0(0) network (N̄^Si_BO = 2.54) that it degrades slower than its S46_2.6(0) counterpart (N̄^Si_BO = 2.11). The literature inarguably shows that the incorporation of B into a silicate-based glass increases its dissolution in aqueous media,^19,20 as well as that bioactive BPS glasses dissolve more rapidly than their B-free counterparts.^4–9 However, that same literature on BG applications is vague regarding the underlying reasons: by assuming that B is exclusively present as BO₃ groups in the glass structure, Yao et al. attributed the comparatively faster degradation of borate/borosilicate glasses to the lower connectivity of the B^[3] coordinations relative to Si^[4].⁵ Yet, the silicate network connectivity of bioactive silicate-based glasses remains N̄^Si_BO < 3 (ref. 23, 41 and 42) and is consequently lower than that of vitreous B₂O₃. Moreover, as shown herein, the networks of Na/Ca-modified BPS glasses also comprise fully polymerized—and thereby highly cross-linking—BO₄ tetrahedra. The higher degradability of B-bearing phosphosilicate glasses merely stems from the higher hydrolysis rate of B–O bonds relative to Si–O, with the reactivity increasing in the order^19,20

Si–O–Si < Si–O–B^[4] < B^[3]–O–B^[3]. (8)

This reactivity order coupled with the progressive replacements of Si–O–Si by Si–O–B and (particularly) B–O–B bonds rationalize the elevated glass dissolution observed^4,5 when B replaces Si in the (boro)silicate networks. Moreover, the combined effects from the higher cross-linking of B^[4] coordinations and a slower hydrolysis of the B^[4]–O–Si bonds relative to the B^[3]–O–B^[3] linkages^19,20 make the borosilicate network portions of the BPS glass more durable. At a constant cation composition and total NBO content, it follows that the higher the x^[4]_B fraction in the BG structure, the more resistant to aqueous attack the glass structure becomes. Hence, the parameter-pair {N̄^Si_BO, x^[4]_B} should primarily govern the solubility of BPS glasses, where the following trends are predicted for the S46/S49 members: (i) at a fixed x_B/x_Si ratio, the lower {N̄^Si_BO, x^[4]_B} values of an S46-based glass implies a faster degradation than its S49 analog. (ii) However, relative to the S46 networks, the S49-based counterparts should exhibit less variations in solubility for increasing B₂O₃ substitution degree, since the alteration of the B^[4] fraction is smaller among the S49 members (see Table 3). We note that all arguments above draw from well-established dissolution trends of (P-free) borate/borosilicate glasses,^19,20 which become applicable to the present BPS glasses with low P contents thanks to the herein established BPS structural model of borate/borosilicate networks surrounded by orthophosphate ions (see Sections 4.1 and 5.3.1). Hence, when ignoring the small Q¹_P population, any given BPS glass network is an approximation of a borosilicate counterpart, as argued in Section 2.2.

Noteworthy, x^[4]_B increases concurrently with x(SiO₂) at constant modifier content in the glass; see Table 3. Hence, lowering the SiO₂ content improves the solubility because both the numbers of Si–O–Si bonds and network-condensing BO₄ moieties are reduced. The YDBX model^37,38 predicts that a further increase in the total Na⁺/Ca²⁺ content of these modifier-rich bioactive BPS glasses drives B^[4] → B^[3] conversions at a constant x_B/x_Si ratio. Consequently, increasing the modifier content of a borosilicate/BPS glass will accelerate its degradability because either/both of the {N̄^Si_BO, x^[4]_B} parameters is/are decreased. Future work will need to settle which structural role the surplus modifiers predominantly assumes—as charge compensators of BO₄ groups or of NBO ions? Such insight combined with the herein reported near invariance of the silicate network polymerization when Si is replaced by B at a constant modifier content (Section 4.2) should allow for tuning each {N̄^Si_BO, x^[4]_B} parameter almost independently for a given {x_B, x_Si, x_P} glass composition. We stress that our proposed role of {N̄^Si_BO, x^[4]_B} to (primarily) govern the BPS glass dissolution is independent on how these values were arranged; while N̄^Si_BO and x^[4]_B depend foremost on the BPS glass composition, particularly the B speciation may display a weak dependence on the thermal history of the glass.

6 Conclusions

We have proposed a structural model for modifier-rich Na₂O–CaO–B₂O₃–SiO₂–P₂O₅ glasses with relatively low P contents [x(P₂O₅) ≤ 0.04] and variable x(SiO₂)/x(B₂O₃) ratios. The following structural alterations are observed when SiO₂ is gradually replaced by B₂O₃ at a constant glass-modifier content:

(i) The phosphate speciation and the average silicate network polymerization only change marginally, where the minor increase of the Q¹_P population (reflecting the extents of Si–O–P bonding) stems from the concurrent silicate-network condensation.³³ Both these trends originate from (ii) a redistribution of NBO species from the SiO₄ tetrahedra to the BO₃ groups, which on the average accommodate 0.7–1 NBO ions per B atom. (iii) Notwithstanding that B^[3] coordinations dominate the B speciation throughout all glasses, a progressive B^[3] → B^[4] conversion is observed for increasing x(B₂O₃). Yet, regardless of the B₂O₃ content, the more condensed and Si-rich S49 glass networks manifest consistently higher fractional populations of BO₄ tetrahedra (37–43%) than their S46 counterparts (17–34%). The x^[4]_B fraction in the S49 glasses reaches a plateau at ≈43% already when 30% of the SiO₂ reservoir is replaced by B₂O₃, whereas minor transformations from trigonal to tetrahedral B coordinations are observed throughout the entire range of B-for-Si substitutions in the S46 series.

The medium-range BPS glass structure is mainly governed by the strong preferences for B^[3]–O–B^[3] and B^[4]–O–Si bonding scenarios relative to other options, such as B^[3]–O–B^[4] and B^[3]–O–Si. Along previous reports on borosilicate glasses of low modifier content,^50–52 substantial amounts of B^[3]–O–Si contacts are only observed in Si-rich BPS networks, whereas they are sparse in structures exhibiting comparable amounts of B and Si. From these bonding preferences, we inferred a BPS structural model that comprises (sub)nm-sized “domains” of a borate network (BO₃ groups in boroxol rings) interlinked with a borosilicate network through B^[4](3Si), B^[4](2Si), and B^[3](1Si) moieties.

We proposed that the parameter-pair {N̄^Si_BO, x^[4]_B} primarily dictates the solubility of alkali/alkaline-earth bearing BPS glasses, whose degradation should decrease concurrently with an increase in either the silicate network polymerization or the B^[4] population. The improved insight about the alterations of these parameters when Si is replaced by B gained from our study should allow for tailoring the glass solubility in (simulated) body fluids. Moreover, for biomedical applications, a high amount of readily leached orthophosphate ions is believed to improve the HCA formation.^23,33,36 We are currently exploring the glass-formation limits of P-richer BPS glasses to verify that the beneficial structural feature of a PO₄³⁻-dominated phosphate speciation remains also in this composition regime, as well as for locating the permissible range of x(SiO₂)/x(B₂O₃) and x(Na₂O)/x(CaO) molar ratios that yield homogeneous glasses. Further work is required to verify the overall validity of the herein proposed borate/borosilicate BPS network organization, which was conjectured from standard ²⁹Si, ³¹P, and ¹¹B (3Q)MAS NMR experimentation. Hence, our future investigations will merely target a more direct probing of the medium-range glass structure by utilizing advanced homo/heteronuclear NMR techniques.^3,85

Note added after first publication

This article replaces the version published on 25th October 2016, which contained errors in Table 3.

Acknowledgements

This work was supported by the Swedish Research Council (contract VR-NT 2014-4667). We thank Baltzar Stevensson for help, and Di Zhang and Renny Mathew for glass synthesis input at an early stage of this project. We also thank R. M. for recording the ²⁹Si NMR spectrum from the S49_4.0(0) glass.

References

1. G. Engelhardt and D. Michel, High-Resolution Solid-State NMR Spectroscopy of Silicates and Zeolites, Wiley, Chicester, 1987.
2. G. N. Greaves and S. Sen, Adv. Phys., 2007, 56, 1–166.
3. M. Edén, Annu. Rep. Prog. Chem., Sect. C: Phys. Chem., 2012, 108, 177–221.
4. W. Huang, D. E. Day, K. Kittiratanapiboon and M. N. Rahaman, J. Mater. Sci.: Mater. Med., 2006, 17, 583–596.
5. A. Yao, D. Wang, W. Huang, Q. Fu, M. N. Rahaman and D. E. Day, J. Am. Ceram. Soc., 2007, 90, 303–306.
6. R. F. Brown, M. N. Rahaman, A. B. Dwilewicz, W. Huang, D. E. Day, Y. Li and B. S. Bal, J. Biomed. Mater. Res., Part A, 2009, 88, 392–400.
7. X. Liu, W. Huang, H. Fu, A. Yao, D. Wang, H. Pan, W. W. Lu, X. Jiang and X. Zhang, J. Mater. Sci.: Mater. Med., 2009, 20, 1237–1243.
8. H. B. Pan, X. L. Zhao, X. Zhang, K. B. Zhang, L. C. Li, Z. Y. Li, W. M. Lam, W. W. Lu, D. P. Wang, W. H. Huang, K. L. Lin and J. Chang, J. R. Soc., Interface, 2010, 7, 1025–1031.
9. M. N. Rahaman, D. E. Day, B. S. Bal, Q. Fu, S. B. Jung, L. F. Bonewald and A. P. Tomsia, Acta Biomater., 2011, 7, 2355–2373.
10. X. Yang, L. Zhang, X. Chen, X. Sun, G. Yang, X. Guo, H. Yang, C. Gao and Z. Gou, J. Non-Cryst. Solids, 2012, 358, 1171–1179.
11. Q. Fu, M. N. Rahaman and D. E. Day, J. Am. Ceram. Soc., 2009, 92, 1551–2916.
12. X. Liu, M. N. Rahaman and D. E. Day, J. Mater. Sci.: Mater. Med., 2013, 24, 583–595.
13. L. A. Haro Durand, G. E. Vargas, N. M. Romero, R. Vera-Mesones, J. M. Porto-López, A. R. Boccaccini, M. P. Zago, A. Baldi and A. Gorustovich, J. Mater. Chem. B, 2015, 3, 1142–1148.
14. W. C. Lepry and S. N. Nazhat, Chem. Mater., 2015, 27, 4821–4831.
15. L. L. Hench, J. Am. Ceram. Soc., 1991, 74, 1487–1510.
16. L. L. Hench and J. M. Polak, Science, 2002, 295, 1014–1017.
17. J. R. Jones, Acta Biomater., 2013, 9, 4457–4486.
18. M. Brink, T. Turunen, R.-P. Happonen and A. Yli-Urpo, J. Mater. Sci.: Mater. Med., 1997, 37, 114–121.
19. B. C. Bunker, G. W. Arnold, D. E. Day and P. J. Bray, J. Non-Cryst. Solids, 1986, 87, 226–253.
20. B. C. Bunker and W. H. Casey, The Aqueous Chemistry of Oxides, Oxford University Press, Oxford, 2016.
21. A. Hoppe, N. S. Güldal and A. R. Boccaccini, Biomaterials, 2012, 32, 2757–2774.
22. M. D. O'Donnell, S. J. Watts, R. V. Law and R. G. Hill, J. Non-Cryst. Solids, 2008, 354, 3554–3560.
23. M. Edén, J. Non-Cryst. Solids, 2011, 357, 1595–1602.
24. H. Gan, P. C. Hess and R. J. Kirkpatrick, Geochim. Cosmochim. Acta, 1994, 58, 4633–4647.
25. H. Yamashita, K. Nagata, H. Yoshino, K. Ono and T. Maekawa, J. Non-Cryst. Solids, 1999, 248, 115–126.
26. F. Muñoz, L. Montagne, L. Delevoye, A. Durán, L. Pascual, S. Cristol and J.-F. Paul, J. Non-Cryst. Solids, 2006, 352, 28–29.
27. M. Sitarz, K. Bulat and Z. Olejniczak, Vib. Spectrosc., 2012, 61, 72–77.
28. R. Dupree, D. Holland, M. G. Mortuza, J. A. Collins and M. W. G. Lockyer, J. Non-Cryst. Solids, 1988, 106, 403–407.
29. M. W. G. Lockyer, D. Holland and R. Dupree, Phys. Chem. Glasses, 1995, 36, 22–30.
30. M. W. G. Lockyer, D. Holland and R. Dupree, J. Non-Cryst. Solids, 1995, 188, 207–219.
31. V. FitzGerald, D. M. Pickup, D. Greenspan, G. Sarkar, J. J. Fitzgerald, K. M. Wetherall, R. M. Moss, J. R. Jones and R. J. Newport, Adv. Funct. Mater., 2007, 17, 3746–3753.
32. R. Mathew, C. Turdean-Ionescu, B. Stevensson, I. Izquierdo-Barba, A. García, D. Arcos, M. Vallet-Regí and M. Edén, Chem. Mater., 2013, 25, 1877–1885.
33. R. Mathew, B. Stevensson, A. Tilocca and M. Edén, J. Phys. Chem. B, 2014, 118, 833–844.
34. R. Mathew, B. Stevensson and M. Edén, J. Phys. Chem. B, 2015, 119, 5701–5715.
35. B. Stevensson, R. Mathew and M. Edén, J. Phys. Chem. B, 2014, 118, 8863–8876.
36. A. Tilocca and A. N. Cormack, J. Phys. Chem. B, 2007, 111, 14256–14264.
37. Y. H. Yun and P. J. Bray, J. Non-Cryst. Solids, 1978, 27, 363–380.
38. W. J. Dell, P. J. Bray and S. Z. Xiao, J. Non-Cryst. Solids, 1983, 58, 1–16.
39. B. C. Bunker, D. R. Tallant, R. J. Kirkpatrick and G. L. Turner, Phys. Chem. Glasses, 1990, 31, 30–41.
40. M. Edén, P. Sundberg and C. Stålhandske, J. Non-Cryst. Solids, 2011, 357, 1587–1594.
41. Z. Strnad, Biomaterials, 1992, 13, 317–321.
42. R. Hill, J. Mater. Sci. Lett., 1996, 15, 1122–1125.
43. K. L. Geisinger, R. Oestrike, A. Navrotsky, G. L. Turner and R. J. Kirkpatrick, Geochim. Cosmochim. Acta, 1988, 52, 2405–2414.
44. A. D. Irwin, J. S. Holmgren and J. Jonas, J. Non-Cryst. Solids, 1988, 101, 249–254.
45. S. W. Martin, D. Bain, K. Budhwani and S. Feller, J. Am. Ceram. Soc., 1992, 75, 1117–1122.
46. G. El-Damrawi and W. Müller-Warmuth, J. Non-Cryst. Solids, 1992, 146, 137–144.
47. L. van Wüllen, W. Müller-Warmuth, D. Papageorgiou and H. J. Pentinghaus, J. Non-Cryst. Solids, 1994, 171, 53–67.
48. S. Wang and J. F. Stebbins, J. Am. Ceram. Soc., 1999, 82, 1519–1528.
49. R. Martens and W. Müller-Warmuth, J. Non-Cryst. Solids, 2000, 265, 167–175.
50. L.-S. Du and J. F. Stebbins, J. Non-Cryst. Solids, 2003, 315, 239–255.
51. L.-S. Du and J. F. Stebbins, J. Phys. Chem. B, 2003, 107, 10063–10076.
52. L.-S. Du and J. F. Stebbins, Chem. Mater., 2003, 15, 3913–3921.
53. M. M. Smedskjaer, J. C. Mauro, R. E. Youngman, C. L. Hogue, M. Potuzak and Y. Yue, J. Phys. Chem. B, 2011, 115, 12930–12946.
54. D. Möncke, G. Tricot, A. Winterstein-Beckmann, L. Wondraczek and E. I. Kamitsos, Phys. Chem. Glasses: Eur. J. Glass Sci. Technol., Part B, 2015, 56, 203–211.
55. R. E. Youngman and J. W. Zwanziger, J. Non-Cryst. Solids, 1994, 168, 293–297.
56. F. Angeli, T. Charpentier, D. De Ligny and C. Cailleteau, J. Am. Ceram. Soc., 2010, 93, 2693–2704.
57. L. Frydman and J. S. Harwood, J. Am. Chem. Soc., 1995, 117, 5367–5368.
58. J.-P. Amoureux, C. Fernandez and S. Steurnagel, J. Magn. Reson., Ser. A, 1996, 123, 116–118.
59. T. Vosegaard, P. Florian, P. J. Grandinetti and D. Massiot, J. Magn. Reson., 2000, 143, 217–222.
60. P. K. Madhu, A. Goldbourt, L. Frydman and S. Vega, Chem. Phys. Lett., 1999, 307, 41–47.
61. D. J. States, R. A. Haberkorn and D. J. Ruben, J. Magn. Reson., 1982, 48, 286–292.
62. Y. Millot and P. P. Man, Solid State Nucl. Magn. Reson., 2002, 21, 21–43.
63. H. Grussaute, L. Montagne, G. Palavit and J. L. Bernard, J. Non-Cryst. Solids, 2000, 263, 312–317.
64. E. Leonova, I. Izquierdo-Barba, D. Arcos, A. López-Noriega, N. Hedin, M. Vallet-Regí and M. Edén, J. Phys. Chem. C, 2008, 112, 5552–5562.
65. F. Fayon, C. Duée, T. Poumeyrol, M. Allix and D. Massiot, J. Phys. Chem. C, 2013, 117, 2283–2288.
66. A.-R. Grimmer, M. Mägi, M. Hähnert, H. Stade, A. Samoson, W. Weiker and E. Lippmaa, Phys. Chem. Glasses, 1984, 25, 105–109.
67. H. Maekawa, T. Maekawa, K. Kawamura and T. Yokokawa, J. Non-Cryst. Solids, 1991, 127, 53–64.
68. D. Massiot, C. Bessada, J. P. Coutures and F. Taulelle, J. Magn. Reson., 1990, 90, 231–242.
69. L. van Wüllen and W. Müller-Warmuth, Solid State Nucl. Magn. Reson., 1993, 2, 279–284.
70. P. R. Bodart, J. Magn. Reson., 1998, 133, 207–209.
71. S. Wegner, L. van Wüllen and G. Tricot, Solid State Sci., 2010, 12, 428–439.
72. J. F. Stebbins, P. Zhao and S. Kroeker, Solid State Nucl. Magn. Reson., 2000, 16, 9–19.
73. S. Kroeker and J. F. Stebbins, Inorg. Chem., 2001, 40, 6239–6246.
74. P. M. Aguiar and S. Kroeker, Solid State Nucl. Magn. Reson., 2005, 27, 10–15.
75. P. M. Aguiar and S. Kroeker, J. Non-Cryst. Solids, 2007, 353, 1834–1839.
76. M. R. Hansen, T. Vosegaard, H. J. Jakobsen and J. Skibsted, J. Phys. Chem. A, 2004, 108, 586–594.
77. J. Banerjee, G. Ongie, J. Harder, T. Edwards, C. Larson, S. Sutton, A. Moeller, A. Basu, M. Affatigato, S. Feller, M. Kodama, P. M. Aguiar and S. Kroeker, J. Non-Cryst. Solids, 2006, 352, 674–678.
78. D. Holland, S. A. Feller, T. F. Kemp, M. E. Smith, A. P. Howes, D. Winslow and M. Kodama, Phys. Chem. Glasses: Eur. J. Glass Sci. Technol., Part B, 2007, 48, 1–8.
79. O. L. G. Alderman, D. Iuga, A. P. Howes, K. J. Pike, D. Holland and R. Dupree, Phys. Chem. Chem. Phys., 2013, 15, 8208–8221.
80. J. C. Phillips and R. Kerner, J. Chem. Phys., 2008, 128, 174506.
81. A. P. Howes, N. M. Vedishcheva, A. Samoson, J. V. Hanna, M. E. Smith, D. Holland and R. Dupree, Phys. Chem. Chem. Phys., 2011, 13, 11919–11928.
82. M. M. Smedskjaer, R. E. Youngman and J. C. Mauro, Appl. Phys. A, 2014, 116, 491–504.
83. T. Kokubo, H. Kushitani, S. Sakka, T. Kitsugi and T. Yamamuro, J. Biomed. Mater. Res., 1990, 24, 721–734.
84. I. Lebecq, F. Désanglois, A. Leriche and C. Follet-Houttemane, J. Biomed. Mater. Res., 2007, 83, 156–168.
85. M. Edén, Solid State Nucl. Magn. Reson., 2009, 36, 1–10.

---