Successful synthesis of fluorapatite has been achieved through ion-exchange between NIST hydroxlyapatite SRM 2910a and optical-grade fluorite. Additional intermediate F-OH apatite compositions were made through ion-exchange between the newly synthesized fluorapatite and the original hydroxylapatite. Based on solution calorimetric data collected on seven fluorapatite–hydroxlyapatite crystalline solutions at 50 °C in 20.0 wt% HCl under isoperibilic conditions, fluorine-rich series members display ideal thermodynamic behavior, whereas hydroxyl-rich compositions show negative enthalpies of F-OH mixing. Unit-cell volumes for the series are linear with composition. Relative to enthalpy and volume, therefore, there are no energy barriers to complete solid solution between the F and OH end-members.
Apatite, A5(XO4)3Z, is a common mineral in igneous and metamorphic rocks with Ca in the A site and P in the X site (Hughes and Rakovan 2002). It exhibits primary continuous solid solution through substitution of F, Cl, and OH in the Z site. It can further contain various trace elements, in the A site, making it of particular use in petrologic trace element studies and in materials design. However, the importance of the Z-site constituents, F, Cl, OH, to petrology has also been recognized (e.g., Hovis and Harlov 2010). The recent discovery that the lunar interior was not as dry as has been the long-held presumption was based in large part on OH abundance in apatite (e.g., McCubbin et al. 2010a). In spite of its obvious importance, however, the stability of the solid solution remains poorly known. A fundamental step in understanding ternary apatite stability is evaluation of the thermodynamic mixing properties.
An initial attempt at characterizing the thermodynamic mixing properties in this system addressed the thermodynamic effects of F-Cl substitution (Hovis and Harlov 2010; Schettler et al. 2011). These studies, along with structural studies (e.g., McCubbin et al. 2008), demonstrated the difficulty in both synthesizing specific apatite stoichiometries and chemical quantification of Z-site occupancy. For example, the synthesis methods employed in the work of Hovis and Harlov (2010) introduced an apparent oxyapatite chemical component into Cl-rich samples. Without knowing the quantitative effect of oxyapatite substitution on enthalpies of the samples, it was impossible to choose between two possible interpretations of the solution calorimetric data: (1) ideal thermodynamic F-Cl mixing based on non-oxyapatite-containing samples only or (2) substantial positive F-Cl enthalpies of mixing based on data for all specimens. Additional compositional uncertainty extends to all Cl-bearing samples in that there is the strong potential for OH substitution, even with efforts taken to avoid it. Other than via SIMS, which requires well-characterized apatite samples for standards, direct spot analysis for OH is not possible. For compositionally controlled synthetic apatite, it is possible to determine OH by difference in the ternary system, but this involves assuming no oxy-component and that reliable F analyses can be obtained. McCubbin et al. (2010) have discussed the F analytical problems at length.
Ion-exchange between hydroxlyapatite and fluorite
The use of ion exchange (e.g., Hovis 1988; Hovis and Roux 1993, 2008) with exchange medium CaF2, as a mechanism of converting well-characterized hydroxlyapatite into fluorapatite was explored to minimize the likelihood of incorporating significant oxycomponent in fluorapatite during synthesis. The hydroxlyapatite used was standard reference material SRM 2910a, which according to the NIST certificate of analysis is material synthesized “for use in evaluating the physical and chemical properties of calcium apatites of biological, geological, and synthetic origin.” The certificate also gives an analyzed calcium content of 38.89 ± 0.85 wt%, which is 1.00 wt% less than the ideal 39.89 wt%, but only slightly outside the uncertainty of the NIST measurement. The phosphorous content given, 18.029 ± 0.071 wt%, is less than the ideal 18.50 wt%. However, the Ca/P molar ratio of 1.667 ± 0.037 determined from the NIST analyses equates to the ideal Ca/P ratio.
The SRM 2910a prepared by NIST utilized solution reaction of calcium hydroxide and phosphoric acid at relatively low temperatures (near the boiling point of water), as described in the certificate of analysis (synthesis procedure of McDowell et al. 1977). Interestingly, comparison of measured unit-cell dimensions of the NIST hydroxlyapatite sample as received and after annealing at 750 °C for 3 days showed clear changes beyond the standard errors of the determinations (results for SRM 2910a included in Table 3). Although such changes cannot be attributed to a particular cause, decarbonation of the original sample is one possibility. Although NIST took precautions to avoid incorporation of CO2 into its sample, no reported measurements tested for the presence of carbonate in the sample. Dehydration of structural water, which has been observed in apatite (Yoder et al. 2012), is another possible cause of the observed change.
To have a perfectly flat surface for the ion exchange trial and to utilize chemically pure CaF2, polished optical grade fluorite disks (e.g., 50 mm diameter × 6 mm thick) were obtained from International Crystal Laboratories (Garfield, New Jersey). A thin layer (<<1 mm) of hydroxlyapatite SRM 2910a (previously annealed at 750 °C for 72 h) fine powder was spread across the surface of a CaF2 disk and tamped down with a spatula. The disk+powder assemblage was placed in a furnace atop a sheet of gold foil and heated in air to 750 °C. As experience with this procedure increased, three days became the typical duration of an experiment. After cooling, the assemblage was removed and the powder carefully scraped off and collected. The product then was reacted a second time under the same conditions, but using a fresh fluorite surface. Typically, the original clear fluorite surface became cloudy during the course of each synthesis reaction. Successful F-exchange in the experiments was gauged through X-ray powder diffraction analysis, which showed obvious changes in peak positions, and also unit-cell dimensions, from hydroxlyapatite to fluorapatite, as demonstrated by the data in Table 3.
Several intermediate F-OH apatite compositions were synthesized by combining carefully weighed (and dried) hydroxlyapatite and newly synthesized fluorapatite powders. Each mixture was stirred by hand with a spatula on weighing paper, then tamped down in the bottom of a small cylindrical Pt crucible and annealed at 750 °C for 72 h. These samples were remixed once each day, then tamped down again and put back into the furnace for further annealing. Although the samples were stored in a desiccator, following the first round of calorimetric experiments, the samples were reheated at 750 °C for 20.7 h in an attempt to drive off any adsorbed H2O or CO2 prior to the second round of calorimetric work. Synthesis conditions for each sample are summarized in Table 1.
Direct synthesis of end-member fluorapatite
The concerns with inadvertent incorporation of possible oxy- and hydroxylapatite component in fluorapatite during synthesis caused us to explore an additional method for fluorapatite synthesis using CaF2 and β tri-calcium phosphate (TCP), both of which were purchased “off-the-shelf.” To ensure proper calcium:phosphate ratios in the synthetic apatite, a representative aliquot of the batch of commercial TCP used was analyzed by X-ray diffraction. Through Rietveld analysis, it was found to contain 7 mol% calcium pyrophosphate (Ca2P2O7). This Ca-deficiency was corrected by adding CaCO3 and then decarbonating. X-ray analysis of the corrected material showed 100% TCP, and there was appropriate weight loss for 100% decarbonation.
Care was taken to avoid OH incorporation by careful drying of the starting materials. CaF2 and corrected TCP were individually dried at 1100 °C for 1 day. Appropriate aliquots were weighed and combined in an agate mortar, mixed under alcohol for 1 h, and dried by heatlamp. The powdered mixture was added to a long Pt capsule that had been welded shut at the bottom; the top was then gently crimped closed. The Pt tube was loaded into a silica glass tube, the silica glass tube was melted above the capsule using a torch, and a capillary was pulled. The capillary and attached silica glass ampoule containing the capsule, was attached to a vacuum pump and evacuated. The powdered sample was further dried by heating at ~800 °C while still attached to the vacuum pump. The evacuated silica tube was then sealed by melting the capillary with a torch, and the ampoule was placed in a horizontal furnace at 900 °C for 23 days and then in a Deltech furnace at 1190 °C for an additional 11 days.
H2O analysis of synthetic F-OH apatite powders by continuous flow mass spectrometry
Continuous flow mass spectrometry (CFMS) was used to quantify water contents in six of the synthetic apatites used in the present study. Measurements were made in the continuous flow mass spectrometry lab in the Department of Earth and Planetary Sciences at the University of New Mexico, using the technique and apparatuses previously described in detail by Sharp et al. (2001). Briefly, the technique involves reduction of structural OH components in solid mineral samples by reaction with glassy carbon at high temperatures. H2 is produced by reaction of the mineral sample with carbon at 1450 °C in a helium carrier gas. Product gases are separated in a gas chromatograph and analyzed in a mass spectrometer configured to make hydrogen isotope analyses in continuous flow mode. In the present study, the sample powders were dried in a desiccated drying oven at about 100 °C and stored at these conditions until the synthetic apatite powders were wrapped in silver foil and weighed. Next, the samples were placed back into the drying oven until analysis. Each sample was weighed again before analysis to confirm the absence of either weight gain or loss. At the time of analysis, each sample was dropped into the furnace using a commercially available autosampler (e.g., Carlo Erba AS 200-LS) mounted directly over the reduction tube. Sample sizes of 3 to 9 mg provided a satisfactory amount of H2 that was well above the detection limit and did not saturate the spectrum. McCubbin et al. (2012) demonstrated that the blank associated with sample introduction or from the silver itself was approximately 300 ppm H2O (0.0001 μLH2O/mg powder), which was attributed to adsorbed surface water. All analyses were well above this detection limit. Throughout the analysis routine, reproducibility of the column was checked using several H2O standards including Water Canyon biotite (USGS sample number 3149-11), BUD biotite (Bindeman and Serebryakov 2011), Banco Bonito biotite, and San Antonio biotite. For each synthetic apatite sample, we analyzed 3 to 5 separate aliquots of sample to test for homogeneity among aliquots of apatite powder. The 2σ uncertainty reported in Table 2 refers to the variability of the water contents among all of the analyses of a given sample, which was larger than the analytical uncertainty. Total time of analysis was less than two minutes for a single hydrogen analysis. Water contents for the synthetic apatite powders analyzed are presented in Table 2.
Electron probe microanalysis (EPMA) of synthetic F-OH apatite powders
Six of the synthetic apatite powders were analyzed using the JEOL 8200 electron microprobe in the Institute of Meteoritics at the University of New Mexico using Probe for EPMA (PFE) software. An accelerating voltage of 15 kV and a nominal probe current of 25 nA were used during each analysis. We conducted energy-dispersive spectroscopic measurements of the apatite powders to determine what elements needed to be included in the quantitative routine, and only peaks for Ca, P, and F were identified, indicating that possible contaminants introduced during synthesis could not be detected. F was analyzed using a light-element LDE1 detector crystal and standardized using strontium fluoride. Ca and P were standardized using a Durango apatite standard from the UNM EPMA laboratory. A 5 μm spot was used for standardization and analysis of all apatite samples. The sample powders were so fine-grained that it was not possible to conduct individual analyses on single grains; instead, the powders were mounted in epoxy and the epoxy-impregnated powders analyzed as per the techniques discussed for powder analysis by EPMA in Reed (2005). This resulted in consistently low EPMA totals, as only Ca, P, and F were analyzed and major elements in epoxy (primarily C, O, H) ignored. To both assess the quality of the data set and obtain useable compositions, approximately 30 spots of each powder/epoxy mixture were analyzed and the average structural formula of each sample calculated. The resulting structural formula for each sample was typically within ~1% of the expected stoichiometry for P and 1.5% for Ca, giving confidence that the resulting atomic ratios were reliable.
Several authors have documented that fluorine X-ray count rates change with time as a function of crystallographic orientation during electron microprobe analysis of apatite (Stormer et al. 1993; Goldoff et al. 2012; McCubbin et al. 2010b, 2011). Given the lack of control over the orientation of the powders and likelihood of equal representation of all grain orientations, F analyses of the apatite/epoxy mixtures were monitored for time-dependent count rates; it was discovered that fluorine count rates were not always constant during the course of an analysis. For the purposes of the present study, any analysis that displayed a variable fluorine X-ray count rate was excluded. Although F is not a major constituent of the epoxy used, its presence has been confirmed by SIMS techniques (McCubbin et al. 2010b). To blank-correct for this F in the epoxy, the epoxy/F-free hydroxylapatite mixture (sample 1138) was analyzed for F and found to contain 0.07 ± 0.03 wt% F. This value was used to blank-correct all of the other analyses. A summary of all EPMA analyses is provided in Table 2.
Hydroxyl cannot be measured directly by the EPMA technique; however, a missing structural component in the anion site of the apatite can be calculated on the basis of stoichiometry to compare with the CFMS results. If the other anions in the site are analyzed with sufficient accuracy (in our case F−), this missing component can be attributed to some combination of the anions Cl−, OH−, O2−, CO32−, S2−, Br−, and I− and/or structural vacancies (Pan and Fleet 2002) and/or structural H2O (Yoder et al. 2012; Mason et al. 2009). The most likely culprit for this missing component in our synthetic system is OH− or structural vacancies, due to compositional limitations imposed by the synthesis technique utilized.
Chemistry of additional synthetic and natural specimens
The most F-rich sample obtained by ion-exchange between fluorite and hydroxlyapatite could not be analyzed by the techniques above due to both the fine-grain size of the sample, and because of anticipated low levels of OH that were thought to be well below the detection limit for H2O by continuous flow mass spectrometry. To determine the composition of F-rich samples 1117 and 1127, anticipated compositions of intermediate F-OH members (based on sample weights and an assumed pure fluorapatite end-member) were plotted against the actual (measured) compositions determined by the techniques above. The resulting well-defined linear trend was utilized to yield an F:OH ratio of 0.85:0.15 for samples 1117 and 1127 (which have identical unit-cell dimensions within a single standard deviation). On subsequent plots of unit-cell dimension data it is seen that this composition fits well with data for other samples, which supports the determined composition. Although no pure-F end-member was synthesized by the F-OH ion-exchange experiments, data for the fluorapatite end-member were obtained using the samples synthesized at high temperatures (SB fluorapatite 4-13-11, described above, and fluorapatite APS-25 from Hovis and Harlov 2010).
A natural apatite specimen was obtained from the National Museum of Natural History to compare data for the newly synthesized samples with those for a naturally occurring apatite sample. The chemical analysis for Durango (Mexico) fluorapatite NMNH 144954-3 is included in Table 2 (compare with Young et al. 1969). Of course, natural samples present problems of their own, including the propensity for cationic and anionic substitutions that remove them from the simple F-OH binary system, as well as the possibility for chemical heterogeneity.
Characterization by X-ray powder diffraction
The unit-cell dimensions of all F-OH apatite samples were determined using data acquired with a Scintag PAD V X-ray powder diffractometer. Scans using Cu radiation typically were run from 15 to 72° 2Θ at a rate of 0.25°/min. NIST SRM 640a silicon, with a stated unit-cell dimension of 5.430825 Å, was used as an internal standard. Indexing of hydroxlyapatite X-ray diffraction maxima was aided by XRD data given in the NIST certificate of analysis. There also are additional sources of XRD data for fluorapatite and hydroxylapatite, such as the ICDD database. Additionally, peaks were tracked from the OH to the F end of the ion-exchange series, and vice versa; this was especially helpful for mid-compositions for which XRD data are sparse or nonexistent. Calculations of unit-cell dimensions were made using the unit-cell program of Holland and Redfern (1997).
It should be noted that X-ray diffraction maxima for the original SRM-2910a were not as sharp as those of typical well-crystallized mineral samples. Almost certainly this is related to the relatively low-temperature and rapid synthesis conditions utilized by NIST, which likely produced a sample that was less crystalline than natural geological samples. Nevertheless, XRD data for the specimens synthesized by ion exchange show well-defined peaks having 2Θ positions that are easily measured.
The calorimetric system used to measure enthalpies of solution has been described in Hovis and Roux (1993) and Hovis et al. (1998). The latter paper demonstrates that this system enables the measurement of precise enthalpies of solution for samples as small as 10 mg. In the present study apatite sample weights ranged from 48.7 to 52.2 mg.
Because of the production of CaF2 as a precipitate (fluorite) or gel, it is not possible to conduct dissolutions of F-bearing apatite in the standard solvent (20.1 wt% hydrofluoric acid) employed by the Lafayette College solution calorimetry laboratory (see more extensive discussion in Hovis and Harlov 2010). For the present study, therefore, each apatite dissolution was performed in 20.0 wt% hydrochloric acid (910.1 g, or ~1 L). All experiments took place at 50 °C under isoperibolic conditions (i.e., temperature of the water bath surrounding the calorimeter held constant) utilizing an internal sample container (Waldbaum and Robie 1970). In all cases, only one dissolution was performed in each liter of acid so as to minimize the amount of fluorine in solution.
Dissolutions for the calorimetric experiments on the ion-exchanged samples were performed on powders having the relatively fine grain sizes inherited from the original SRM 2910a hydroxylapatite. It is known that heat effects can arise from extremely fine grains (Nitkiewicz et al. 1983). If these arose, they would have occurred for all ion-exchange series samples, as all such samples had essentially the same grain sizes; this suggests that compositional differences should be robust. The additional deleterious effect of fine grains comes during weighing of samples for calorimetry as they tend to be hygroscopic. Adsorbed water can cause additional weighing uncertainties, which in turn, affect both calorimetric accuracy and precision. On the other hand, fine grains also can increase calorimetric precision by promoting rapid dissolution, resulting in less heat exchange between the calorimeter and its environment, and in turn a smaller ΔT correction for the sample dissolution period. Overall, however, calorimetric precision for the present samples is somewhat less than that experienced in previous work on silicate mineral systems, probably due to the H2O-weight-gain issue. Importantly, grain-size effects would not have been an issue either for the relatively coarse high-temperature synthetic fluorapatite sample SB fluoraparite 4-13-11 nor for the two natural apatite samples. The fact that data for these samples are compatible with those from samples of the ion-exchange series suggests that small grain size was not a major source of calorimetric uncertainty.
Unit-cell dimensions and volumes
The unit-cell dimensions and volumes calculated from X-ray powder diffraction data are given in Table 3 and plotted in Figures 1 to 3. Figure 1 demonstrates the linear behavior of the a unit-cell dimension with composition. Extension of the fitted line intersects the data point for fluorapatite SB 4-13-11 and comes within a few tenths of an angstrom of intersecting the APS-25 fluorapatite data point of Hovis and Harlov (2010).
The c unit-cell dimension (Fig. 2) changes little, if at all, with composition, as all c values are within 0.01 Å of one another. The projected data point at XF = 1.0 in Figure 2 fits well with the datum for APS-25 (ibid). On the other hand, the c-value for fluorapatite SB 4-13-11 falls somewhat below the others. The reason for this is unknown, but in all likelihood represents slight differences in chemical composition (e.g., hydroxyl or oxy component abundances) between the pure fluorapatite samples.
The unit-cell volume (Fig. 3) behaves linearly with composition. It can be described by
and fits well with data for APS-25 (ibid). The lower volume for fluorapatite SB 4-13-11 is related to the lower value for c. Linear behavior of volume with regard to composition was seen as well for F-Cl apatite crystalline solutions (Schettler et al. 2011; Hovis et al. 2011). This linearity is important for the thermodynamic implications discussed below.
Enthalpies of solution and F-OH mixing
Solution calorimetric data for the newly synthesized F-OH apatites are given in Table 4. As noted in previous papers, it is advantageous to plot negative values for these data; this allows concave-down relationships to be associated with positive heats of mixing and concave-up curves to be related to negative ones. Accordingly, negative values for heats of solution are plotted against mole fraction fluorine in the Z-site (Fig. 4). A third-order polynomial fits the data well.
The magnitude of excess enthalpy, which relates to the energetics of non-ideal mixing of the fluor- and hydroxlyapatite components, depends upon the ideal mixing line chosen. Under optimal conditions, this is determined from the enthalpy of solution of the end-members in identical structural states. The differences in enthalpy of solution of the fluorapatite samples and the extrapolated ion-exchanged fluorapatite composition, makes the choice of an ideal mixing line more difficult. On the other hand, it is reasonable to consider the calorimetric results for the F-OH ion-exchange series independently of other data due to the “connectedness” of the samples via the synthesis methods used to make them. With that in mind, a line can be drawn from the hydroxlyapatite end-member through the most F-rich member of the F-OH series at XF = 0.85. Such a line is described by the equation
Coincidentally, this line passes through data for the F-rich sample at XF = 0.720 and comes close as well to the data at XF = 0.594. Extension of the line to XF = 1.0 nearly intersects fluorapatite end-member data for both SB 4-13-11 and Durango NMNH 144954-3 after data for the latter have been corrected for Cl content (6.77 kJ/mol), based on the ideal enthalpy model of Hovis and Harlov (2010, see their Fig. 2) for the F-Cl apatite system. If the choice of end-member enthalpies of solution and hence, this ideal mixing line, is reasonable, F-rich members of the series would be interpreted as behaving ideally in a thermodynamic sense, while OH-rich samples would exhibit negative enthalpies of mixing, reflected by the concave-up fitted curve for OH-rich samples in Figure 4. The excess enthalpies of mixing (that is, those not accounted for by ideal mixing) may be expressed as
for compositions between XF = 0 and 0.679 (where Hex = 0 for XF > 0.679), and reach a maximum magnitude of −4.8 kJ/mol at XF = 0.24 (Fig. 5). Although the magnitude of this excess enthalpy of mixing depends upon the choice of ideal mixing line, and therefore on the enthalpies of solution of the end-member apatites, the use of samples with similar mode of origin maximizes the likelihood that this is a reflection of true mixing characteristics.
There is a difference of ~12 kJ/mol among the three pure-fluorapatite data points (i.e., the extrapolated F-end-member of the ion-exchanged F-OH apatite series, SB 4-13-11, and APS-25). This probably represents real differences among the samples, but it is not obvious what these might be. Possibilities include differences in crystallinity (particularly for the F-OH ion exchange series), non-stoichiometry in the samples, or sample impurities of various kinds that can occur in both synthetic and natural samples. It is clear from the unit-cell dimensions that, while the properties of “pure” fluorapatite specimens from the present study may be similar, they are not identical. While choice of SB 4-13-11 as the fluorapatite end-member would only marginally increase the negative excess mixing magnitudes, use of APS-25 as the pure-F end-member would produce nearly ideal behavior for OH-rich specimens and positive excess enthalpies in the F-rich part of the series.
The thermodynamic requirements for immiscibility within a solid-solution series is that positive excess Gibbs free energies (Gex) of sufficient magnitude exist across the F-OH apatite join under some set of pressure-temperature conditions. Gex, in turn is a function of excess enthalpy (Hex), volume (Vex), and entropy (Sex) related to F-OH mixing, where Gex = Hex + PVex – TSex (T = absolute temperature, P = pressure). Immiscibility in turn is associated with positive values for Hex and Vex, but negative ones for Sex.
The observed unit-cell volumes for the F-OH apatite system behave linearly with F:OH ratio, so Vex is zero across the entire series. The only enthalpies of F-OH mixing in the system, based on analysis of the solution calorimetric data presented here, are negative. Barring negative entropies of F-OH mixing for this system, there is no thermodynamic barrier to complete dissolution of hydroxyl ions in fluorapatite, nor fluorine ions in hydroxylapatite, at any temperature or pressure.
If apatite is to be utilized as a petrogenetic indicator, it is essential that this chemically complex system be characterized thermodynamically. The present study of F-OH apatite solid solutions makes an essential contribution to such characterization, adding to previous data for the F-Cl apatite binary (Hovis and Harlov 2010). We hope to be able to make future contributions to additional apatite binaries (e.g., Cl-OH), as well as more complex apatite systems, as new sample synthesis methods are developed. We hope as well that the solid-solid ion-exchange synthesis techniques used to make F-OH apatites for this study, and also the techniques utilized to produce end-member synthetic fluorapatites, might encourage others to develop new synthesis methods to make possible the study of additional apatite compositions.
G.L.H. and H.N. thank the Earth Sciences Division of the US National Science Foundation for support of this research through grants EAR-1019809 and EAR-809283, respectively. F.M.M. acknowledges support for this research from the NASA Cosmochemistry Program through grant NNX11AG76G. We thank Jeffrey Post and Michael Wise of the National Museum of Natural History and George Harlow of the American Museum of Natural History for providing natural samples for this and other apatite work.