Hydrogen-substituted β-tricalcium phosphate synthesized in organic media

A hydrogen substitution mechanism, previously unknown in pure β-tricalcium phosphate, was discovered in crystals precipitated from ethylene glycol solutions. The structure was described by means of Rietveld refinement of powder X-ray diffraction data and corroborated by chemical analysis and IR spectroscopy.

second heating bath at 150°C for approximately 20 min to allow for precipitation. The resulting platelet suspension was centrifuged at 4000 rpm for 30 min and the precipitate was washed by successive dispersion and centrifugation in ethanol (C2H6O, absolute 99.8%, Grogg, Switzerland), demineralized water and again in ethanol, before drying at 30°C under vacuum for 24 h.
The synthesis conditions are detailed in Error! Reference source not found. namely standard conditions (applied for all analyses unless otherwise indicated) and conditions with individually varied parameters to study the effect of the latter on the crystal structure. Note that, since the pH could not be accurately measured in ethylene glycol, a measure of the acidity was defined as the difference between the H + and OHionic concentrations introduced into the solution, in the form of NaOH, Na2HPO4 and/or H3PO4, normalized by the total precursor concentration, i.e. ([H + ]-[OH -])/([Ca]+[P]). To produce Mg-containing platelets, 0.5 or 1 mol% of the Ca ions were replaced by Mg ions (MgCl2·6H2O, Ph.Eur., MERCK, Germany). Glycerol was used as an alternative solvent (glycerol, anhydr. for synth., MERCK, Germany). Error! Reference source not found. links the numbered synthesis conditions to the corresponding dataset name used in the CIF files.

Table S1
Synthesis parameters corresponding to the numbered conditions in Fig. 4

S1.2. Scanning electron microscopy
The scanning electron microscopy image in Error! Reference source not found. was obtained on a Zeiss EVO MA25 instrument using a working distance of 8 mm and an acceleration tension of 20 kV, following sputter-coating with Pt for 30 s.

S1.3. Attenuated total reflection (ATR)-FTIR
In addition to the transmission FTIR analysis (detailed in section 2.3), the samples were analyzed in the attenuated total reflection (ATR) mode, by means of a diamond crystal on the Bruker Lumos IR spectrometer. Spectra were acquired between 400 and 4000 cm -1 at a resolution of 4 cm -1 and 64 accumulations.

S2.1. Morphology
The hexagonal morphology of a representative β-TCP platelet sample (synthesized under standard conditions; see Error! Reference source not found.) is shown in Error! Reference source not found.. The detailed relationships between synthesis conditions and geometric features have been described previously (Galea et al., 2013;Galea et al., 2014). The full wavenumber range of transmission FTIR spectra of sintered β-TCP and β-TCP platelets is provided Error! Reference source not found.. The broad band at 3700 to 3200 cm -1 and the absorption at 1630 cm -1 are attributable to OHgroups (Ping et al., 2001) and H2O molecules (Cerruti et al., 2003) respectively, resulting from varying degrees of residual moisture in the KBr pellets used for transmission FTIR analysis. Note that this moisture cannot account for the HPO4 2absorption observed at 875 cm -1 , since the latter was also observed in ATR-FTIR spectra which were free of water-related absorptions (see Error! Reference source not found.). The peak at 2350 cm -1 is due to slight variations in the gaseous CO2 concentration between the different analyses (Nuevo et al., 2006).
Finally, the small bands close to 2900 cm -1 were attributed to organic contaminations (C-H bonds) in the KBr pellets. Note that this signal was not detected in ATR-FTIR spectra (Error! Reference source not found.b).

Figure S2
Full wavenumber range of transmission FTIR spectra of sintered β-TCP and β-TCP platelets. (The wavenumber region showing phosphate absorption is detailed in Fig. 2). The absorptions at 3700-3200 cm -1 and 1630 cm -1 are attributable to residual moisture in the KBr pellets while the peak at 2350 cm -1 results from gaseous CO2.
ATR-FTIR spectra of commercially available monetite as well as β-TCP platelets containing variable fractions of monetite platelets are provided in Error! Reference source not found.. These spectra were compared amongst each other in order to ensure that the peak at 875 cm -1 , assigned to HPO4 2groups in the β-TCP structure, does not result from the monetite (CaHPO4) platelets, a by-product of the synthesis obtained especially at lower temperatures. The spectrum of pure monetite reveals a broad band with a maximum at 890 cm -1 , attributable to the P-O(H) stretching mode in HPO4 2groups (Tortet et al., 1997), which overlaps indeed with the band at 875 cm -1 in β-TCP platelets. However, monetite exhibits several other absorptions which are not present in β-TCP, namely between 1300 and 1450 cm -1 (P-O-H in-plane bending) (Tortet et al., 1997) and close to 400 cm -1 (O-P-O bending; partly cut off by the lower spectral range limit) (Tortet et al., 1997). Thus, if the HPO4 2band of the purest β-TCP sample (1 wt% monetite) was due to the monetite platelets, a strong absorption at 400 cm -1 would have to be present, which is not the case. Moreover, the intensity of the absorption at 875 cm -1 does not converge to zero as the monetite fraction approaches zero. Therefore, the band at 875 cm -1 in β-TCP plateletsas well as the side peak at 855 cm -1 which is not observed in monetitemust result predominantly from HPO4 2groups in the β-TCP structure and is thus corroborate the Rietveld refinement findings.

S2.3. Effect of synthesis conditions -statistical analysis
One-way ANOVA followed by Bonferroni means comparison indicated no significant difference (p>0.05) in the β-TCP Ca/P (or (Ca+Mg)/P ratio) between any of the synthesis conditions for which at least two samples were analyzed. Moreover, the differences between the standard condition and conditions with only one analyzed sample were within two times the estimated standard deviation (ESD) calculated by the refinement algorithm (error bars in Fig. 4). Note that the large ESDs obtained in conditions A1 and C2 can be explained by the high monetite fractions in the sample (data not shown).
For some conditions, more than one synthesis parameter was varied simultaneously. Specifically, samples in conditions C2 and D1 to D3 were produced in the tubular reactor instead of the batch reactor (see Error! Reference source not found.). However, a significant effect of the reactor type on the Ca/P ratio can be ruled out by comparing condition C1 to group D. Similarly, while the acidity and reaction time are inconsistent in some conditions, groups D and F demonstrate that these parameters have no significant effect.