Atomic scale modeling of iron-doped biphasic calcium phosphate bioceramics

Biphasic Calcium Phosphates (BCP) are bioceramics composed of hydroxyapatite (HAp, Ca10(PO4)6(OH)2) and beta-Tricalcium Phosphate (-TCP, Ca3(PO4)2). Because their chemical and mineral composition closely resembles that of the mineral component of bone, they are potentially interesting candidates for bone repair surgery, and doping can advantageously be used to improve their biological behavior. However, it is important to describe the doping mechanism of BCP thoroughly in order to be able to master its synthesis and then to fully appraise the benefit of the doping process. In the present paper we describe the ferric doping mechanism: the crystallographic description of our samples, sintered at between 500°C and 1100°C, was provided by Rietveld analyses on X-ray powder diffraction, and the results were confirmed using X-ray absorption spectroscopy and 57 Fe Mössbauer spectrometry. The mechanism is temperature-dependent, like the previously reported zinc doping mechanism. Doping was performed on the HAp phase, at high temperature only, by an insertion mechanism. The Fe 3+ interstitial site is located in the HAp hexagonal channel, shifted from its centre to form a triangular three-fold coordination. At lower temperatures, the Fe 3+ are located at the centre of the channel, forming linear two-fold coordinated O-Fe-O entities. The knowledge of the doping mechanism is a prerequisite for a correct synthesis of the targeted bioceramic with the adapted (Ca=Fe)/P ratio, and so to be able to correctly predict its potential iron release or magnetic properties.


1-Introduction
The mineral mass of bone is dominated by nanocrystalline non-stoichiometric hydroxyapatite (HAp, chemical formula Ca 10 (PO 4 ) 6 (OH) 2 , Ca/P ratio of 1.67) [1][2][3][4]. Non-stoichiometry is mainly assumed by few weight percent of carbonate substitution and also calcium deficiency, nevertheless many trace elements participate to the non-stoichiometry. Tricalcium phosphate (-TCP, chemical composition Ca 3 (PO 4 ) 2 , Ca/P ratio of 1.5) has a Ca/P ratio close to that of HAp and presents higher solubility under biological conditions [5][6]. HAp and BCP (biphasic calcium phosphates composed of a mixture of HAp and -TCP) have been investigated for biomedical applications in reconstructive surgery (hard tissue replacement implants and bone prosthesis coating) due to their excellent bioactivities, biocompatibility and osteoconductivity [5][6][7][8][9][10]. In addition, the doping effect can advantageously be used, among other levers, to improve the biomedical properties of HAp-based ceramics [11]. Nanocrystalline bone mineral contains numerous essential trace elements [4,9]. The role of many of these ionic species in hard tissues is not fully understood, because of the difficulties encountered in monitoring and quantifying their proportions, which vary according to dietary alteration and to physiological and to pathological causes. However, it is commonly accepted that these various ions play a major role in the biochemistry of bones, enamel and dentine by a substitution process [12].
Our previous results on Zn-doped BCP samples highlighted that a fine description of the incorporation mechanism of the doping element remains a significantfactorin correctly interpreting biological behavior, namely due to the different solubility of the two HAp and -TCPphases [13][14][15]. The incorporation of the doping element in one or the other phases will deeply modify its potential release in biological fluid. The HAp structure is known to accept various ionic substitutions, as has been demonstrated for carbonate [16,17], silicate [18][19][20][21], borate [22,23] and alkaline earth cations Mg 2+ [24][25][26] and Sr 2+ [27,28]. Nevertheless our recent results on the Zn-doping mechanism demonstrated that substitution is not the only mechanism to be considered: insertion into an interstitial site [13], as also described for bevolite (the Sr equivalent with composition Sr 10 (PO 4 ) 6 (OH) 2 ) [29], has been established.
This doping mechanism is temperature-dependent [14]. Zn-doping elements can be locatedin drastically different local environments. The transfer fromthe six-fold coordinated calcium site substitution in -TCPat moderate temperature to the two-fold coordinated insertion in HAp at higher temperature is a significant phenomenon.
Following our Zn-doping insertion mechanism studies we undertook a systematic study of BCP doping by the 3d-metal cation series from manganese to zinc. The present paper is devoted to the specific case of ferric cations. Iron is an essential trace element in bones and teeth, is a micronutrient essential for various biological processes and is an important component of several metalloproteins. Iron represents approximately 35 and 45 mg/kg of body weight in adult women and men, respectively. In the intestinal lumen, iron exists in the form of ferrous and ferric salts, although most dietary inorganic iron is in the form of ferric salts [30]. Recent studies have shown that the presence of Fe 3+ affects the crystallinity and solubility of HAp [31][32][33][34], while small amounts of iron were found to have a positiveimpact on the biomedical properties of HAp [35][36][37]. The blood compatibility, and more generally the biocompatibility and non-cytotoxicity, of Fe 3+ -doped HAp has recently been demonstrated, withimproved bactericidal and mineralizing properties compared to undoped HAp [38][39][40].
Biomagnetic calcium phosphate ceramics, incorporating magnetic ions and exhibiting ferromagnetic properties, play an important role in medicine. DopedmagneticHAp could be useful for biological applications such as magnetic resonance imaging (MRI), cell separation, drug delivery and heat mediation for the hyperthermia treatment of cancers [39,41].
Despite the recently-described insertion mechanism for Zn 2+ [13][14][15] and despite the cationic size difference between Fe 3+ (0.64 Å, CN6) and Ca 2+ (1.00 Å, CN 6) [42], a substitution mechanism at calcium crystallographic sites is commonly considered in the literatureas for all cations. In the present study, a detailed structural description of Fe-doped HAp is investigated to clarify the situation. Series of BCP samples (HAp being the main phase) are synthesized using the sol-gel method with different iron doping levels and with thermal treatments between 500°C and 1100°C. In addition to a long-range order investigation performed using Rietveld refinement on X-ray powder diffraction patterns, the local order is finely described thanks to X-ray absorption spectroscopy and 57 Fe Mössbauer spectrometry.

Sol-gel elaboration of Fe-substituted BCP samples
The sol-gel method previously proposed by the authors was used to synthesize one undoped and four Fe-doped series of BCP samples [14]. Briefly, to produce 2 g of undoped BCP powder, 4.7 g of Ca(NO 3

X-Ray Powder diffraction (XRPD) and Rietveld analyses
XRPD patterns were recorded on a X'Pert Pro Philips diffractometer, with θ-θ geometry, equipped with a solid X-Celerator detector and using Cu K radiation ( = 1.54184 Å).
XRPD patterns were recorded at room temperature in the interval 3° < 2θ < 120°, with a step size of 2θ = 0.0167° and a counting time of 200 s for each data value. A total counting time of about 200 min was used for each sample (some raw data are showing in Figure SI1a from supplementary information). A XRPD pattern was collected from pure NIST standard LaB 6 using the same experimental conditions in order to extract the instrumental resolution function to improve peak profile fitting during Rietveld refinements.
Rietveld refinements of X-ray powder patterns were performed for each sample usingFullProf.2k software [43]. The procedure used (both data-collection and refinement strategy) corresponds to the general guidelines for structure refinement using the Rietveld (whole-profile) method formulated by the International Union of Crystallography Commission on Powder Diffraction [44][45][46]. The Rietveld refinement strategy was detailed in previous related work [14]. Examples of Rietveld plots are showing in Figure SI1b.

X-ray Absorption Spectroscopy (XAS)
Fe  XAS spectra were obtained in fluorescence mode using a 13-element Ge solid-state detector.
The size of the beam was determined by a set of slits (200 μm x 500 μm).
Data processing was performed using the Athena and Artemis programs from the IFFEFIT software package [47] by merging 6 successively-recorded absorption spectra. Single scattering theory was used here. Following Lengeler-Eisenberg normalization, EXAFS oscillations were Fourier Transformed (FT) using a Hanning window between 3.0 and 9.0 Å -1 .
The χ(k) function was Fourier transformed using k 3 weighting, and all shell-by-shell fitting was done in R-space. Theoretical backscattering paths were calculated using successively ATOMS [48] and FEFF6 [49]. This data processing strategy was detailed in the Zn-doping study [15].

57 Fe Mössbauer spectrometry
The Fe-containing samples were analyzed by means of Mössbauer spectrometry using a

Mineralogical analysis of the samples
To correctly interpret the behavior of our samples, their mineral compositions were estimated from the Rietveld refinements performed on Laboratory X-ray diffraction patterns. Mineral compositions for the five series of samples are indicated in Table 1, and the weight percent (wt %) of the two main phases, HAp and β-TCP, are represented in Figure SI2

Lattice parameters of HAp and β-TCP
The variations in the lattice parameters of HAp and β-TCP give initial indications about the mechanism of iron incorporation into their structures. combined with the contraction of thea lattice parameter evince an interstitial mechanism of insertion of iron atoms into the HAp structure, as already described for the Zn-doping mechanism [14]. The increase in unit cell volume ( Figure 1c) between 1000°C and 1100°C is not consistent with a substitution mechanism (as the Fe 3+ cation is smaller than Ca 2+ ). The drastic variations in the lattice parameters and unit cell volume are also directly dependent on the Fe quantity. We can note, for lower sintering temperatures, the slight decrease in the basal a lattice parameter for the Fe-doped series, which could be attributed to a calcium substitution mechanism at 500°C and 600°C.
The case of β-TCP is different. Table SI2 shows the refined lattice parameters of β-TCP as a function of the iron amount introduced for sintering temperatures which stabilize the β-TCP phase (800°C for the undoped series and 700°C for the iron-containing series). The values of both the a and c lattice parameters (and consequently that of unit cell volume) diminish with the increase in the iron doping level, in agreement with a calcium substitution mechanism.
Substitution is first performedat the Ca4 site, and then is carried to the Ca5 site; i.e. calcium substitution from the low density column habitually encountered with this β-TCP phase [13-15, 24, 27, 28].

Location of Fe 3+ in the HAp structure
During Rietveld refinements the occupancy factors of all calcium and phosphorus crystallographic sites were systematically tested. Whatever the temperature and the Fe-doping level, no vacancies were evinced up to 1000°C (Table SI3), unlike iron-containing samples which had been heat treated at 1100°C (Table 2) Table 2and Table SI1).
The large increase in the refined quantity of incorporated iron correlates with the sharp variations in HAp lattice parameters between 1000°C and 1100°C ( Figure 1 and Table SI1).
Such splitting around the special 2b site in apatite-type structures has already been described for Co-doped belovite Co:Sr 10 (PO 4 ) 6 (OH) 2 [29]. Figure 1d shows the refined quantity of iron atoms inserted into the hexagonal channel of the apatite structure and Figure  respectively. Here the maximum amount of iron incorporation into the HAp structure seems to form the Ca 9.75 Fe 0.50 (PO 4 ) 6 (OH) 1

3.2XAS analyses
Three samples of the 15Fe-T series (15Fe-500, 15Fe-800 and 15Fe-1100 samples), as well as two reference samples (-Fe 2 O 3 and Fe 3 O 4 ), were investigated by XAS spectroscopy in fluorescence mode. The 15Fe-T series was chosen because of the single phase feature of the last 15Fe-1100 sample (Table1 and Figure SI2) and because of the absence of the-Fe 2 O 3 phase for all the samples of the series. Reference materials were used to illustrate the wellknown tetrahedral and octahedral coordination for iron cations (considering Fe 2+ and Fe 3+ ).

Temperature dependency of the spectra
Raw data are reported in Figure SI3 showing the normalized EXAFS spectra, with the XANES part of the spectra and the EXAFS modulations for the 15Fe-T series and for the reference compounds. The temperature variation of the Fourier transformed amplitudes (not corrected for phase shift) is represented in Figure 3a in the R-space. The first peak in the radial distribution is observed at an R value of 1.4 Å for samples 15Fe-500 and 15Fe-800 (same value for Fe 3 O 4 with iron in tetrahedral and octahedral coordination, compared to 1.45 for -Fe 2 O 3 with iron in octahedral coordination only). This first peak then shifts closer to R = 1.35 Å for sample 15Fe-1100, when a large amount of iron is inserted into the HAp network. In agreement with the XRPD long-range order analysis, local order considerations indicate a temperature-dependent iron incorporation mechanism in our BCP samples with low coordination at 1100°C.

Fe 3+ insertion into HAp at 1100°C
EXAFS spectra from the single phase 15Fe-1100 sample were used to explore the local Fe 3+ environment determined by Rietveld refinement: insertion into a 12i crystallographic site shifted from the 2b site of the HAp structure (i.e. eccentric to the centre of the hexagonal column) leading to the three-fold coordination. The k 3 -weighted fitted Fe K-edge EXAFS data of 15Fe-1100 is illustrated in Figure 3b (Fitting was performed using Artemis software in the range 1 Å <R< 3.8 Å, not corrected for phase shifts), and fit results are listed in Table 3 (  Table 3 indicates that the long-range order crystallographic model used for Rietveld refinement is coherent.

3.3Results: 57 Fe Mössbauer spectrometry
Spectra obtained at 300K with quadrupolar hyperfine structure resulting from asymmetrical and non-lorentzian profile lines are illustrated in  The proportions of each Fe species (Table 4)

Temperature dependency of the BCP composition
The present characterization of the Fe 3+ -doping mechanism of BCP bioceramics is highly similar to that previously described in detail for Zn 2+ [13][14][15]. With lower sintering temperatures, BCP samples are mainly composed of the undoped HAp phase. Figure SI4 shows temperature dependency of the total refined iron content in the samples for the four Fecontaining series when considering Fe inserted into the two main calcium phosphates (HAp and -TCP) and iron from hematite. The lack of metal transition for lower temperatures was already observed in our previous Zn-BCP doping study [14]. EXAFS characterization of the  (Figure 3a).
Variation of the assigned tetrahedral signal from Mössbauer spectroscopy ( demonstrated by an increase in the total amount of refined iron ( Figure SI4), and by the Rietveld refinement of the -TCP structure (Table SI2) Table   3).

Iron location into the HAp structure
The iron doping mechanism is dependent on sintering temperature, with an interstitial mechanism for iron insertion into the HAp structure that is realized for the higher temperature. An important specific characteristic of this iron-doping mechanism, compared to the previously reported zinc doping studies [13][14][15], is the splitting of the ferric cation out of the centre of the hexagonal channel. It clearly appears that iron insertion in the shifted 2b site at 1100°C (Table 2), correlated to the sharp variation of the HAp lattice parameters (Figure   1), enabled a considerable increase of the iron quantity in the hexagonal HAp channel (from about 0.5 wt % up to 1000°C to a value higher than 2.5 wt % at 1100°C).
The low IS signals,from Mössbauer spectroscopy, we have attributed to Fe 3+ inserted into the HAp structurecorresponding to the linear HAp and triang. HAp assignments in Table 4 have never been observed previously. This should be correlated with the synthesis process; samples were simply heated at 100°C in the study of Jiang et al. [50], and calcinations were performed under nitrogen in the study of Low et al. [51]. It is surprising to observe so great a difference between our samples sintered under air, showing one or two components with IS < 0.2, which was not the case when heating under nitrogen,where the smallest IS valuewas 0.27 [50]. XAS and Mössbauer spectroscopycorroborates the refutation ofa modification of the oxidation state of ferric cations during the sintering process; the whole population of iron cations is trivalent Fe 3+ whatever our doped samples. Previously reported studies on irondoped hydroxyapatites highlight the importance of the synthesis approach. Chandra et al. [38] prepared Fe-doped HAp using a combination of hydrothermal and microwave techniques, and observed a decrease in both the a and chexagonal lattice parameters of HAp with an increase in Fe doping. This does not correspond with our results (anisotropic variation, particularly at a high sintering temperature), and seems to reflect a calcium substitution mechanism. It should be noted that for thelower sintering temperature we also observe a decrease in the basal a lattice parameter combined with an absence of variation in the hexagonal c lattice parameter (Table SI1 and Figure 1). Thus a weak calcium substitution cannot be excluded, even if the Rietveld refinements did not enable us to detect it. The X-ray electronic contrast between Fe 3+ and Ca 2+ is not very marked, and a possible combination of iron to calcium substitution with a calcium site deficiency (as reported for the 75Fe-series in Table 2) could mask such a calcium substitution mechanism for the lower sintering temperatures. This hypothesis should be excluded for higher sintering temperatures, as indicated by 1/ lattice parameter variation, 2/ presence of electronic density at the interstitial site and 3/ the low coordination revealed by XAS and Mössbauer spectroscopy for iron. In the same way, the freeze-drying process used by Tampieri et al. [39] led to the preparation of Fe-doped HAP samples without the inserted iron atoms, as illustrated by their EXAFS study. They obtained a first contribution in the uncorrected radial distribution at about 1.5 Å,which was similar to the measured magnetite and maghemite reference materials. The corresponding Fe-O distance is typical of a ferric cation (about 1.95 Å), larger than our refined 1.84 Å value (Table 3) (Figure 2 and Table 3).