ARTICLE IN PRESS

Biomaterials 28 (2007) 2908–2914 www.elsevier.com/locate/biomaterials

2D mapping of texture and lattice parameters of dental enamel
Maisoon Al-Jawada,Ã, Axel Steuwerb, Susan H. Kilcoynec, Roger C. Shorea, Robert Cywinskid, David J. Wooda a Leeds Dental Institute, University of Leeds, Leeds, LS2 9LU, UK FaME38 at the ILL-ESRF, 6 rue J Horowitz, 38042 Grenoble, France c Institute for Materials Research, University of Salford, Salford, M5 4WT, UK d School of Physics and Astronomy, University of Leeds, Leeds, LS2 9JT, UK b

Received 19 December 2006; accepted 16 February 2007 Available online 25 February 2007

Abstract We have used synchrotron X-ray diffraction to study the texture and the change in lattice parameter as a function of position in a cross section of human dental enamel. Our study is the first to map changes in preferred orientation and lattice parameter as a function of position within enamel across a whole tooth section with such high resolution. Synchrotron X-ray diffraction with a micro-focused beam spot was used to collect two-dimensional (2D) diffraction images at 150 mm spatial resolution over the entire tooth crown. Contour maps of the texture and lattice parameter distribution of the hydroxyapatite phase were produced from Rietveld refinement of diffraction patterns generated by azimuthally sectioning and integrating the 2D images. The 002 Debye ring showed the largest variation in intensity. This variation is indicative of preferred orientation. Areas of high crystallite alignment on the tooth cusps match the expected biting surfaces. Additionally we found a large variation in lattice parameter when travelling from the enamel surface to the enameldentine junction. We believe this to be due to a change in the chemical composition within the tooth. The results provide a new insight on the texture and lattice parameter profiles within enamel. r 2007 Elsevier Ltd. All rights reserved.
Keywords: Enamel; Hydroxyapatite; Apatite structure; Synchrotron X-ray diffraction; Texture; Preferred orientation

1. Introduction Dental enamel is the most highly mineralised and hardest biological tissue. It is comprised of approximately 96% mineral, 3% water, and 1% organic matter (noncollagenous protein) by weight [1]. The mineral is nonstoichiometric calcium hydroxyapatite (Ca10(PO4)6OH2) with carbonate, fluoride, sodium, and magnesium ions frequently found within the structure. These hydroxyapatite (HA) crystallites are laid down as nanorods with crosssectional dimensions of 50 nm  25 nm and up to 1 mm long [2]. Clusters of these nanorods, known as prisms, contain around 1000 crystallites. They are approximately 5 mm in diameter and may be up to several millimetres long, and the majority are arranged with their long axes at approximately 901 to the enamel-dentine junction (EDJ).
ÃCorresponding author. Tel.: +441133438331.

E-mail address: m.al-jawad@leeds.ac.uk (M. Al-Jawad). 0142-9612/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.biomaterials.2007.02.019

The orientation of prisms in enamel has been studied in the past using electron microscopy. Although this is a valuable tool for finding the prism shape and size in a particular plane of enamel, it is a qualitative technique and does not give detailed, quantitative information on the degree of alignment in different parts of a tooth. Previous work using X-ray diffraction on human dental enamel has established the space group and lattice parameters as P63/m ˚ ˚ (hexagonal) and a ¼ 9.441(2) A and c ¼ 6.878(1) A respectively [3–5]. However, these values were obtained from measurements of powdered enamel collected from several teeth, and as a result any information on the spatial variation of the lattice parameters and texture relating to the growth of the HA crystallites was lost. It has been reported from grazing-incidence synchrotron X-ray diffraction experiments that there is a higher degree of crystallite alignment in surface enamel compared to enamel close to the EDJ [6]. However, only linear slices from EDJ to surface were probed in these experiments. In an earlier

ARTICLE IN PRESS
M. Al-Jawad et al. / Biomaterials 28 (2007) 2908–2914 2909

study Hirota examined the tilting of the enamel-prism orientation in a human canine using laboratory twodimensional (2D) X-ray diffraction [7], however only 12 points within the tooth were measured and therefore the information obtained about the prism orientation cannot be for the whole tooth. In this paper we aim to show for the first time how synchrotron X-ray diffraction can be used to determine the basic crystallographic parameters of the HA phase across a whole intact tooth section, allowing us to explore composition and texture on the sub-millimetre length-scale. Characterising the orientation distribution of the anisotropic apatite crystallites of dental enamel aids the fundamental understanding of the natural growth and formation of dental enamel, and provides insights into how synthetic enamel-like materials may be developed.
2. Materials and methods 2.1. Specimen preparation
The sample used in this study was a section of an adult mandibular second premolar (LR5). The tooth was collected with informed consent from a patient undergoing routine orthodontic extraction at the Leeds Dental Institute. The extracted tooth had its pulp removed and was sterilised by autoclaving prior to storage at 4 1C in a thymol solution to prevent bacterial growth. A precision diamond blade cutter was used to cut the tooth into 500 mm thick longitudinal sections perpendicular to the buccal and lingual surfaces. The sections were then polished by hand to remove any surface roughness. A photograph of the section used for this study is given in Fig. 1. The four arrows mark the tracks plotted in Fig. 10.

lattice parameters distribution maps. The 2y angle-range for this experiment was 2y ¼ 5–301—for comparison this corresponds to a 2yrange of 9–581 on a conventional lab-based X-ray diffraction apparatus ˚ with CuKa radiation of wavelength l ¼ 1.54 A. For our experimental setup, the main diffraction peak of the HA phase (2 1 1) was located in the centre of our 2y range at 16.771. Vacuum tube slits were used to focus the X-ray beam to a diameter of 150 mm on the sample. A 500 mm thick tooth section was mounted in transmission geometry onto a travelling sample platform such that the tooth could be scanned in two orthogonal directions perpendicular to the beam. A charge-coupled device (CCD) 2D detector with 2048  2048 pixel resolution was mounted behind the sample and perpendicular to the incident beam for the collection of 2D diffraction images. The cross-hairs of a telescope were positioned in line with the beam centre in order to align the beam position on the tooth. A schematic of the experimental setup with an example 2D diffraction image is shown in Fig. 2. A single diffraction image had an exposure time of 5 s and therefore a 150 mm-resolution map of the tooth on a grid of 10 mm  7 mm could be collected in approximately 8 h by moving the sample relative to the beam in an x and y direction.

2.3. Data analysis
2D diffraction images were pre-processed with the ESRF software Fit2D [9]. Each image was sectioned into 51 slices [10] and these integrated slices were then used to create Intensity versus 2y patterns for Rietveld refinement [11], i.e. 72 diffraction patterns per 2D image. Conventionally only one Bragg reflection in the diffraction pattern is used in texture analysis therefore any slight changes in sample volume as a function of position would affect the sample absorption and could affect the texture coefficient obtained. However, using the Rietveld method minimises the effect of variations in sample volume since all reflections are used to obtain the fit. In addition, in our study, we used a scaling factor in the refinement procedure and found that changes in the scale factor were small as a function of position within the tooth, indicating that variations in section thickness were negligible. A total of 1095 diffraction patterns were refined and therefore an in-house automated batching procedure was written and used to input the patterns into the GSAS Rietveld refinement software [12]. The instrument parameters such as X-ray wavelength, sample to detector distance, and peak-shape profile were determined using a LaB6 standard sample. These parameters were then kept fixed for refinements of the data. The scale parameter and background parameters (four terms) were refined first. The lattice parameters and crystallite size (Lorentzian particle broadening term) were refined next, starting from the values for HA taken from Young [3]. Finally the texture was refined using a spherical harmonics function [13] and the preferred orientation values for the 002 reflection were extracted from this. The values of preferred orientation ranged from 0 (randomly oriented) to 3.5 (strongly textured). The quality of the refinement was determined by least squares methods where the goodness of fit increased as w2 approached unity whereby: w2 ¼ R2 wp R2 e ,

2.2. Synchrotron X-ray diffraction
The measurements were taken on the XMaS beamline [8] at the European Synchrotron Radiation Facility (ESRF) using an X-ray ˚ wavelength of l ¼ 0.82 A (equivalent to X-ray energies of 15 keV) and a sample to detector distance of 163.09 mm. The 2y angle-range for our experiment was limited by our experimental setup and the sample to detector distance. We compromised on the very high-angle HA reflections in order to position the 2D detector closer to the sample and therefore greatly reduced our counting times. This allowed us to collect more diffraction patterns per tooth and obtain higher resolution texture and

where Rwp is the weighted R-factor and Re the expected R-factor.

Fig. 1. Photograph of 500 mm thick tooth section from a human adult lower right premolar. The arrows mark the tracks through the tooth plotted in Fig. 10.

Fig. 2. Experimental setup at the XMaS beamline, ESRF, with an example 2D diffraction image.

ARTICLE IN PRESS
2910 M. Al-Jawad et al. / Biomaterials 28 (2007) 2908–2914

3. Results 3.1. Preferred orientation in enamel Preferred orientation has both a magnitude and a direction. Through our analyses of the X-ray diffraction data we have been able to quantify both these parameters. In order to obtain an overview of the preferred orientation in a tooth section the 2D X-ray scans were arranged to form a composite map of CCD images of the tooth, as shown in Fig. 3. Each small square in the image is one 2D diffraction pattern. The centres of adjacent diffraction patterns are 150 mm apart. The shape of the tooth can clearly be seen from this composite image. The darker patterns in the middle of the tooth are from dentine and the lighter patterns covering these are from the enamel. At the surface the enamel is thinner, therefore there is a halo of darker images along the edge of the tooth where there is partial air scattering. This can also be seen in the fissure where there is a gap between the two cusps. Additionally, it can be seen that below the fissure there is a circular region of enamel which is darker than the surrounding patterns. It is likely that this is caused by a fissure caries lesion which has partially demineralised the enamel in that area. In Figs. 4a–d four individual diffraction scans from different parts of the tooth are shown. Figs. 4a, c and d illustrate the change in texture direction in the 002 plane at different positions within the enamel. Variations in intensity around diffraction rings are indicative of texture in the tooth enamel. The strongest texture (the most extreme variation in intensity) was found in the 002 reflection (2y ¼ 13.71)—labelled in Fig. 4a. A line through zero degree has also been marked in Fig. 4a. Fig. 4b shows the diffraction pattern from dentine. Here the peaks are much broader indicating that the crystallites are smaller. There is also much less variation in intensity around the

Fig. 4. (a), (c) and (d) Illustrate the change in texture direction at difference positions within the enamel. (b) Shows the poorly crystalline nature of dentine.

Debye rings indicating that the dentine is much less textured than the enamel. The intensity pattern around the Debye ring of the 002 reflection was used to evaluate the texture direction. The intensity was integrated over 3601 in a narrow band containing the 002 reflection and plotted versus the azimuthal angle. Fig. 5 shows a typical example of the resulting curve for one diffraction scan where there are two pronounced peaks separated by approximately 1801. By fitting these peaks to a Gaussian peak shape, the deviation angle, f, of the crystallite axis relative to vertical was determined. By applying this procedure to each of the diffraction images, a map of the local orientation of the texture in the 002 direction in the enamel is obtained. Fig. 6 shows a map of the orientation angles overlaid onto the composite image of the tooth. In Fig. 6, for clarity, only every fourth value of f has been drawn. It can be seen from this that the texture direction in the 002 reflection is approximately perpendicular to, and follows the contour of the EDJ. From our knowledge of the structure of dental enamel it would appear that the preferred orientation in the 002 direction approximately follows the direction of the enamel prism arrangement. 3.2. Rietveld refinement The extent of preferred orientation in the tooth section has been quantified using Rietveld refinement. An example of a typical 1D Intensity versus 2y diffraction pattern together with the calculated pattern is shown in Fig. 7. The

Fig. 3. Composite of 2714 diffraction images arranged spatially to show the outline of the tooth specimen.

ARTICLE IN PRESS
M. Al-Jawad et al. / Biomaterials 28 (2007) 2908–2914 2911

Fig. 5. Typical intensity versus Azimuthal angle curve for the 002 reflection showing the pronounced texture in this sample. The left hand peak has been fitted to a Gaussian.

open circles are the observed data points, and the solid line is the calculated diffraction pattern. Below the pattern is a plot of the difference (observed–calculated), and beneath are the tick marks for the 2y peak positions for the calculated diffraction pattern of HA. The difference plot shows that the agreement between observed and calculated data is generally very good with a typical value for w2 of $1.5. The final parameters obtained from this refinement are given in Table 1. Similar refinements were carried out on all diffraction patterns. A contour map showing the change in magnitude of preferred orientation in the 002 diffraction peak has been plotted in Fig. 8. The 002 preferred orientation parameters have been extracted for the ‘zero degree’ slice of each 2D diffraction image (see Fig. 4a). Areas with higher values of preferred orientation parameter are more strongly textured i.e. the crystallites are more aligned to the ‘zero degree’ direction in these areas. Areas with low texture coefficient have less well-aligned crystallites. 3.3. Change in lattice parameters Both the a- and c-lattice parameters of HA were refined in each diffraction pattern and after carrying out Rietveld refinements of 1095 data sets it was noticed that neither the a- nor c-lattice parameters were constant as a function of position. Note that the variation was not due to X-ray wavelength drift as a function of time, but was clearly dependent on the position within the tooth. Trends in the lattice-parameter changes have been plotted in Fig. 9 relative to the average lattice parameters. The relative percentage change in the a-lattice parameter (Ra) across the tooth section was calculated using:   an À a0 Ra ¼ Â 100%, a0 where an is the refined lattice parameter for diffraction pattern n, and a0 is the overall average lattice parameter. The same type of equation was used for defining the c-lattice parameter variation, Rc. Both Ra and Rc have been

Fig. 6. Texture direction of the 002 reflection of hydroxyapatite crystallites in enamel, calculated using 2D diffraction images.

Fig. 7. Typical diffraction pattern including the raw data (circles), the calculated diffraction pattern (solid line), the difference, and the tick marks for the 2y peak positions for the calculated diffraction pattern of hydroxyapatite.

ARTICLE IN PRESS
2912 M. Al-Jawad et al. / Biomaterials 28 (2007) 2908–2914 Table 1 Refined structural parameters for typical diffraction pattern of dental enamel (HA) Phase hydroxyapatite Space group ˚ a (A) ˚ c (A) a, b, g ˚ V (A3) Y(particle) 002(SH) w2 P63/m (#176) 9.4660(9) 6.9018(3) a ¼ b ¼ 901, g ¼ 1201 535.49(8) 7.83(9) 1.9(1) 1.5

Y(particle) is the coefficient for Lorentzian particle size broadening, and 002(SH) is the 002 spherical harmonic preferred orientation term.

Fig. 8. Texture distribution map generated from the calculated texture coefficient via Rietveld refinement.

plotted as contour plots in Figs. 9a and b. It can be seen from these figures that the a- and c-lattice parameters vary between À0.6% and +0.3% of their average values. In all cases the uncertainty in the lattice parameters did not exceed 1 Â 10À3, and the values for the average lattice ˚ ˚ parameters were a0 ¼ 9.5165(6) A and c0 ¼ 6.9394(2) A. These plots clearly reveal that there is a systematic variation of the lattice parameters as a function of position within the enamel. Figs. 10a–d shows plots of Ra and Rc as a function of distance from the surface for the four tracks through the enamel indicated in Fig. 1. In all regions, Ra (filled symbols) decreases with increasing distance from the tooth surface, especially around the cusps. In Figs. 10b and c there is second peak in the Ra curve at around 600 mm from the surface indicating a region of enamel which has a higher a-lattice parameter than the surrounding enamel. This region can be seen clearly in the contour plot in Fig. 9a. Along each track, the c-lattice parameter curves (open symbols) are much flatter. This indicates that Rc is less dependent on the distance from the enamel surface than Ra. In both the a- and c-lattice parameter values, there is also a difference between the lingual and the buccal sides of the tooth.

Fig. 9. (a) a-lattice parameter and (b) c-lattice parameter contour maps showing the change in lattice parameter value at different positions around the tooth.

4. Discussion It has been seen previously that there is a higher degree of crystallite alignment in surface enamel compared to enamel close to the EDJ [6]. However, in that work, only linear slices from EDJ to surface were probed. The results from our study show that the texture distribution is much more complex than previously thought. We can see

ARTICLE IN PRESS
M. Al-Jawad et al. / Biomaterials 28 (2007) 2908–2914 2913

Fig. 10. Four tracks through the tooth section going from enamel surface to EDJ showing the change in lattice parameter as a function of distance from the enamel surface. The tracks are indicated in Fig. 1.

in Fig. 8 that HA crystallites are most aligned in the cuspal regions: on both sides of the buccal cusp and on the inner side of the lingual cusp HA crystallites are highly aligned. Conversely, along the sides of the tooth away from the cusps generally the crystallites are less ordered. It is interesting to note that the areas of high crystallite alignment match the expected occlusal surfaces of a lower second premolar [14]. This may be an evolutionary development of enamel so that the regions of enamel which are exposed to the largest load are the strongest. It is interesting to note that a strong correlation between functionality and texture in human bone has been reported by Bacon [15] where he observed that the living conditions (either on a steep hillside, or on the flat) of two Neolithic tribes radically affected the HA crystallite growth and alignment on the lower front edge of the tibia. Although dental enamel cannot regenerate itself as bone can, it is likely that through evolution the degree of crystallite alignment in different regions of a tooth has been optimised for the function of the tooth. Changes in lattice parameter can be indicative of changes in enamel crystal chemistry as well as changes in the stress/ strain-state of a material. Separating these two possible effects or even using the lattice parameters to determine compositional changes in biological apatites is not straightforward as changes in crystal chemistry can be the result of several ionic substitutions, such as Na, Mg, Cl or F, as well as variations in the carbonate content and the Ca/P ratio. However, the magnitude of the changes seen across our tooth suggest that the changes in lattice parameters plotted in Figs. 9 and 10 arise predominantly from changes in the chemical composition of the enamel in different regions of the tooth. Chemical analysis of the distribution of fluoride, carbonate and magnesium in enamel has been carried out previously by Robinson et al. By dissecting tooth sections into pieces weighing

20–50 mg, the amount of fluoridated apatite was determined by etching the tooth sections and analysing the fluoride concentration in the post-etched buffer solution [16], the concentration of carbonate was determined by dissolving each piece in acid and measuring the volume of CO2 emitted [17], and concentration of magnesium was found using an atomic absorption spectrophotometer [18]. They found that fluoride concentrations decreased in going from the enamel surface to the EDJ, while carbonate and magnesium concentrations increased (from 2% to 4–6%, and from 0.2% to 0.5% respectively) across the same distance [19]. In dentine, changes in the a-lattice parameter of up to 0.5% have been reported in going from the EDJ into the centre of the dentine [20]. This trend has been explained principally as a result of the increased substitution of CO2À for PO3À associated with the more immature 3 4 dentine crystallites. Our results are of a comparable order of magnitude and we believe them to be the result of compositional change. Comparing Figs. 9a and b it can be seen that there is more variation in the a-lattice parameter than in c. This trend has also been seen in dentine [20] where the a-lattice parameter decreased by 0.5% with increasing distance from the EDJ into the dentine, whereas the c-lattice parameter only decreased by 0.1% over the same distance. In addition to a decrease in lattice parameters going from the surface enamel to the EDJ, there is a difference in the lattice parameters on the buccal and lingual sides of the tooth indicating a change in crystal chemistry on the different sides of the tooth. This could either be due to the different functions of the two sides of the tooth or due to their slightly different oral environments, or a combination of both. Cuy et al. have generated 2D distribution maps for the hardness (H) and Young’s modulus (E) of enamel using nanoindendation [21]. In the molars they investigated, they found that the cuspal regions had higher hardness and Young’s modulus. They report values ranging from H46 GPa to Ho3 GPa and E4115 GPa and Eo70 Gpa, respectively, going from the enamel surface to the EDJ. The contour maps they generated of H and E show similar features to our lattice-parameter distribution maps, indicating that the crystallographic and mechanical properties of enamel are closely linked, therefore an understanding of both is necessary in order to fully understand the function of enamel in different parts of a tooth. 5. Conclusions Using spatially resolved synchrotron X-ray diffraction we have quantified the changes in texture and lattice parameters in dental enamel as a function of position within the tooth. With this technique, in a few hours of data collection, we have generated 2D distribution maps of both texture and lattice-parameter changes in enamel with 150 mm resolution. This has given detailed quantitative information on the degree of crystallite alignment in

ARTICLE IN PRESS
2914 M. Al-Jawad et al. / Biomaterials 28 (2007) 2908–2914 [5] Wilson RM, Elliott JC, Dowker SEP, Smith RI. Rietveld structure refinement of precipitated carbonate apatite using neutron diffraction data. Biomaterials 2004;25(11):2205–13. [6] Low IM. Depth-profiling of crystal structure, texture, and microhardness in a functionally graded tooth enamel. J Am Ceram Soc 2004;87(11):2125–31. [7] Hirota F. Prism arrangement in human cusp enamel deduced by X-ray diffraction. Arch Oral Biol 1982;27(11):931–7. [8] Brown SD, Bouchenoire L, Bowyer D, Kervin J, Laundy D, Longfield MJ, et al. The XMaS beamline at ESRF: instrumental developments and high resolution diffraction studies. J Synchrotron Radiat 2001;8(6):1172–81. [9] Hammersley AP. FIT2D: An Introduction and Overview. ESRF Internal Report 1997; ESRF97HA02T. [10] Hammersley AP, Svensson SO, Hanfland M, Fitch AN, Hausermann ¨ D. Two-dimensional detector software: from real detector to idealised image or two-theta scan. High Pressure Res 1996;14(4–6): 235–48. [11] Rietveld HM. A profile refinement method for nuclear and magnetic structures. J Appl Crystallogr 1969;2(2):65–71. [12] Larson AC, Von Dreele RB. General Structure Analysis System (GSAS). Los Alamos National Laboratory Report 2004;LAUR 86-748. [13] Von Dreele RB. Quantitative texture analysis by Rietveld refinement. J Appl Crystallogr 1997;30(4):517–25. [14] Berkovitz BKB, Holland GR, Moxham BJ. Oral anatomy, embryology and histology. 3rd ed. New York: Mosby; 2002. [15] Bacon GE. The dependence of human bone texture on life-style. Philos Trans Roy Soc B 1990;240(1298):363–70. [16] Weatherell JA, Robinson C, Hallsworth AS. Changes in the fluoride concentration of the labial enamel surface with age. Caries Res 1972;6(1):312–24. [17] Robinson C, Weatherell JA, Hallsworth AS. Variation in composition of dental enamel within thin ground sections. Caries Res 1971; 5(1):44–57. [18] Robinson C, Weatherell JA, Hallsworth AS. Distribution of magnesium in mature human-enamel. Caries Res 1981;15(1):70–7. [19] Robinson C, Shore RC, Brookes SJ, Strafford S, Wood SR, Kirkham J. The chemistry of enamel caries. Crit Rev Oral Biol M 2000; 11(4):481–95. [20] Zioupos P, Rogers KD. Complementary physical and mechanical techniques to characterise tooth: a bone-like tissue. J Bionic Eng 2006;3(1):19–31. [21] Cuy JL, Mann AB, Livi KJ, Teaford MF, Weihs TP. Nanoindentation mapping of the mechanical properties of human molar tooth enamel. Arch Oral Biol 2002;47(4):281–91.

different regions of tooth enamel not previously reported. It has also shown that the lattice parameters maps, related to changes in crystal chemistry, are more complicated that previously thought indicating that understanding heterogeneities within a single tooth is as important as realising the differences between teeth. We have shown that characterising the crystallographic properties of dental enamel is crucial in order to design optimised dental restorative materials. Finally, we have shown through this work that synchrotron X-ray diffraction is a powerful technique in the study of the crystallography and microstructure of dental enamel and it could be equally successful in the study of other biological hard tissues, in the study of synthetic biomaterials, and in the study of biosynthetic complexes. Acknowledgments This work was performed on the EPSRC-funded CRG beamline (XMaS BM28) at the ESRF. We are grateful to L. Bouchenoire and J. Wright (ESRF) for their invaluable assistance and to S. Beaufoy for additional administrative support. Also, we would like to thank the FaME38 facility for providing the VAMAS approved precise sample mounting system. Thanks to C. Sullivan at Leeds Dental Institute for producing the photograph in Fig. 1. This research was funded by the UK Medical Research Council. References
[1] Deakins M, Volker IF. Amount of organic matter in enamel from several types of human teeth. J Dent Res 1941;20(2):117–21. [2] Johansen E. In: Stack MV, Fearnhead RW, editors. Tooth enamel: its composition, properties, and fundamental structure. Bristol: I Wright and Sons; 1965. p. 177–81. [3] Young RA, Mackie PE. Crystallography of human tooth enamel: Initial structure refinement. Mat Res Bull 1980;15(1):17–29. [4] Wilson RM, Elliott JC, Dowker SEP. Rietveld refinement of the crystallographic structure of human dental enamel apatites. Am Mineral 1999;84(9):1406–14.