Vaterite Crystals Contain Two Interspersed Crystal Structures

See allHide authors and affiliations

Science  26 Apr 2013:
Vol. 340, Issue 6131, pp. 454-457
DOI: 10.1126/science.1232139

Double Vision

Vaterite is the least stable form of anhydrous crystalline calcium carbonate. While rarely found in geological contexts, it is an important biological precursor and occurs as a minor component in the shells of some organisms. The crystal structure of vaterite has long been debated with no model able to explain all the experimentally observed diffraction spots. Kabalah-Amitai et al. (p. 454) show that vaterite contains two coexisting crystallographic structures that form a pseudo-single crystal.


Calcite, aragonite, and vaterite are the three anhydrous polymorphs of calcium carbonate, in order of decreasing thermodynamic stability. Although vaterite is not commonly found in geological settings, it is an important precursor in several carbonate-forming systems and can be found in biological settings. Because of difficulties in obtaining large, pure, single crystals, the crystal structure of vaterite has been elusive for almost a century. Using aberration-corrected high-resolution transmission electron microscopy, we found that vaterite is actually composed of at least two different crystallographic structures that coexist within a pseudo–single crystal. The major structure exhibits hexagonal symmetry; the minor structure, existing as nanodomains within the major matrix, is still unknown.

By far the most abundant biogenic minerals are calcium carbonates, which exist in different polymorphs: calcite, aragonite, vaterite, and amorphous CaCO3, listed in order of decreasing thermodynamic stability under normal conditions. There are numerous examples of biogenic calcite and aragonite, as well as a growing list of discoveries of biogenic CaCO3 formation via transient amorphous calcium carbonate (13). Vaterite is much rarer as a geologic mineral or biomineral, possibly because of its instability. Vaterite is an important constituent of Portland cement (4) and has been found during oil field drilling (4). It has also been detected in gallstones (5) and even as part of a meteorite (6). Synthetic vaterite is common in various scenarios, such as under template-directed (7, 8) or additive-assisted (9) conditions.

In biominerals, vaterite occurs as a minority component of a larger formed structure or as the result of a defective biological process. Some examples can be found in green turtle eggshells (10), in coho salmon otoliths (11), in freshwater lackluster pearls (12), and in the aberrant growth of mollusk shells repaired after an injury (13). A rare example of deposition of vaterite as the only mineral component of an organism’s endoskeleton can be found in the body and tunic spicules of the solitary stolidobranch ascidian Herdmania momus (14) (Fig. 1).

Fig. 1 SEM images of the Herdmania momus vaterite spicules.

(A) A body spicule (long) and a tunic spicule (short). (B and C) Two different observation directions of two tunic spicules and their crowns of thorns (acceleration voltage 4 kV, in-lens detector).

Gibson (15) showed that the x-ray diffraction pattern of vaterite is distinct from that of calcite or aragonite. The first symmetry that was related to vaterite was the hexagonal symmetry described by Olshausen in 1925 (1618). More than 30 years later, Meyer (19) provided the first full crystallographic description (atomic positions) of vaterite, describing its unit cell as having pseudo–hexagonal-orthorhombic symmetry. The next model for vaterite, which was proposed in 1963 by Kamhi (17) and has become the most widely accredited structure, posits 100% occupancy by the calcium and one-third occupancy by the carbonate groups. Notably, Kamhi also reported eight weak reflections that could not be indexed, and attributed them to a superstructure, rotated 30° about the c axis from the main one (Kamhi’s hexagonal structure), with lattice parameters a = 3a and c = 2c′. Meyer (20) suggested another structure in which the basic unit cell was essentially in agreement with Kamhi’s model but the site symmetry of the carbonate groups was different. Meyer also emphasized that the observed additional spectral features (diffuse streaks along the c axis, and satellite reflections) were attributable to stacking faults perpendicular to the c axis, leading to doubling or tripling of the c lattice parameter.

The principal debate has centered on whether the vaterite symmetry is hexagonal or orthorhombic, although the Raman spectrum of vaterite did not agree with either of the proposed structures (21). Wang and Becker (22), using density functional theory (DFT), proposed a new unit cell with the P6522 space group. Recently, Le Bail (23) suggested a modified orthorhombic unit cell based on a microtwinning hypothesis. Demichelis et al. (24), also using DFT, took energy considerations into account, and this led to new vaterite structures that have greater stability in simulations. Interestingly, Demichelis et al. reported several theoretically stable structures that required only room-temperature thermal energy for their interconversions. They suggested that these frequent interconversions provide an additional source of disorder within the vaterite structure. Mugnaioli et al. (25), using automated diffraction tomography (ADT) and precession electron diffraction on a transmission electron microscope, suggested two structures with very low symmetries: a monoclinic unit cell and a triclinic one. To account for all the diffraction data obtained in the studies of Mugnaioli et al., almost the lowest symmetries have been hypothesized (25).

We used high-resolution synchrotron powder diffraction and aberration-corrected high-resolution transmission electron microscopy (HRTEM) to examine the structure of vaterite, using the vaterite spicules of H. momus ascidians as samples. Relative to the synthetic vaterite used in previous studies, the single crystals in these spicules are larger and have fewer lattice defects. The spectrum obtained from high-resolution synchrotron powder diffraction on beamline ID31 of the European Synchrotron Radiation Facility (ESRF), combined with Rietveld refinement, could not be fitted by any of the structures proposed for vaterite. The hexagonal structure proposed by Kamhi (17) gave the best fit to the data but did not explain all the experimentally observed diffraction peaks (fig. S1).

We then used HRTEM investigations of H. momus spicules to study very small volumes of vaterite, so as to ensure that data would not come from different structures. An additional advantage of using these spicules is that they are perfectly and reproducibly oriented. When indexed in the hexagonal setting, each thorn on the tunic spicule comprises a single crystal of vaterite, which grows precisely along the c axis (Fig. 2A). This allowed us to prepare TEM specimens by cutting with a focused ion beam each single crystal precisely along known crystallographic directions. By cutting the thorns parallel, perpendicular, or at other known angles to the main axis, we could study the vaterite structure along different zone axes.

Analysis of all the electron diffraction patterns collected at different zone axes (Fig. 2) shows that each thorn diffracted as a single crystal of vaterite. Furthermore, the corresponding phase-contrast images demonstrated perfect crystalline order. In Fig. 2C, an array of strong diffraction spots, as well as weaker ones, can be observed. When we used the JEMS package (26) to simulate electron diffraction patterns and high-resolution images, it was clear that the strong diffraction spots in Fig. 2C could not be fully explained by a single structural model (fig. S2 and table S1). Although both the Kamhi and Meyer models combined were needed to explain the strong diffraction spots, neither model could explain any of the weak diffraction spots. Conversely, along the other zone axis (Fig. 2A), all diffraction spots in Fig. 2E could be explained by the Kamhi model, whereas the Meyer model could explain only some of them. Using JEMS, we simulated each of the proposed models for vaterite structure, and found that none of them could explain all the HRTEM diffraction spots.

Fig. 2 HRTEM images (acceleration voltage 300 kV) and electron diffractions (acceleration voltage 200 kV) in two different zone axes.

(A) Schematic illustration of the focused ion beam cutting direction on vaterite crystals from a tunic spicule of H. momus. The electron beam in TEM is perpendicular to the cutting direction. (B to E) Different phase-contrast images [(B) and (D)] and their respective corresponding electron diffractions [(C) and (E)]. The images in (B) and (C) were produced from a sample cut at ~70° to the long axis of a single thorn; those in (D) and (E) were produced from a sample cut parallel to the long axis of a single thorn.

In seeking a solution to this conundrum, we examined additional zone axes of vaterite with the thought that these might reveal additional information. A zone axis perpendicular to the c axis (in the hexagonal setting) is shown in fig. S3. In this third zone axis, both the diffraction and the phase-contrast images revealed, perpendicular to the c axis, planar defects that were predominantly stacking faults. These observations were in agreement with Kamhi’s report of diffuse scattering perpendicular to the c axis (17). However, when looking at different areas of this phase-contrast image, we observed distinct areas with variable atomic ordering (Fig. 3, A and B), indicating the possibility that the sample was not in fact a single crystal. Fast Fourier transform (FFT) analysis of the part of the phase-contrast image that includes the planar defects yielded a pattern that could be fitted to the electron diffraction (Fig. 3, C and E) of the same zone axis. This FFT-generated diffraction-like pattern also showed a good fit to the Kamhi model. However, when FFT analysis was performed only on the area containing the distinctly different atomic structure from the selected area observed in Fig. 3B (free of planar defects), we found that it did not match the FFT pattern generated from the other area depicted in Fig. 3A (see Fig. 3F) and showed no correlation with the Kamhi model or with any other model we examined.

Fig. 3

HRTEM phase contrast image (acceleration voltage 300 kV) produced from a sample cut parallel to the long axis of a single thorn. A few distinct areas and their corresponding electron diffractions and FFT patterns can be seen. (A) Stacking faults. Despite the stacking faults, the diffraction pattern is in good agreement with the Kamhi structure. (B) Distinct atomic structure with no planar defects. (C) Electron diffraction corresponding to zone axis [210] taken from (A). (D) Microdiffraction taken from (B), showing superposition of two electron diffraction patterns from the main (Kamhi) structure (zone axis remains unaltered [210]) and an additional, distinct one (which seems not to be in the zone axis). Gamma correction was performed, but the quantitative analysis was not based on spot intensity. A few diffraction spots from the minor distinct structure are circled. (E) FFT corresponding to (A) [providing an equivalent pattern to that of electron diffraction shown in (C)]. (F) FFT corresponding to (B).

For additional verification, we acquired electron diffraction patterns using an electron beam with a diameter of only 3 to 4 nm. Changes in diffraction patterns from different positions within a single thorn were observed. In diffraction images obtained from areas that could be seen to be distinct in the phase image, such as in Fig. 3B, we observed that the strong diffraction pattern that fit the Kamhi structure was now considerably weakened and that new diffraction points from a different structure were visible (compare Fig. 3D and Fig. 3C). The latter observation demonstrates that within a “single crystal” of vaterite there are at least two distinct coexisting structures. Even the diffraction pattern observed in Fig. 3C, taken from the selective area in Fig. 3A, is a superposition of both structures, but because of the large proportion of the Kamhi structure relative to the minor structure, the diffraction spots from the latter are not observed.

This observation was also consistent over several different samples cut from the same spicule, and from different tunic spicules. Upon examining the same [210] zone axis in each of these samples, we observed that the main structure, despite the planar defects, was the Kamhi structure; the minor but clearly distinct structure is unknown. We did not observe an ordered array of these two structures, but rather an apparently random distribution.

The possibility that the observed phenomenon was induced by the focused ion beam can be ruled out. All samples prepared from the same “single crystal” showed what appeared to be a perfect single crystal along the c axis; in addition, the multiple-structure phenomenon was observed only along the specific zone axis perpendicular to the c axis. This was further verified when we prepared TEM samples with the same zone axis by means of bombardment with argon ions and observed the same structures (fig. S3).

Because the two distinct structures were observed only when imaging a sample tilted to a zone axis perpendicular to the main long axis of a single thorn (the c axis in the Kamhi model), we tried to image the structure by tilting to a zone axis parallel to the long axis of a single thorn, [001]. Although hardly any beam damage was observed at other orientations, samples tilted to the [001] zone axis were always sensitive to the electron beam, and we therefore had to work with the much lower acceleration voltage of 80 keV. We are not sure why this zone axis was the most vulnerable to the electron beam, but it seemed that intracrystalline organic macromolecules were situated mainly on planes perpendicular to the c axis (fig. S3). When the image obtained at low acceleration voltage was stable, it could be seen that although the structure diffracted as a single crystal, it was actually made up of smaller domains.

When two coexisting domains were observed, such as those seen in Fig. 4, it was evident that not only were they rotated by 30° from one another, but they also had distinct d-spacings, intercorrelated by exactly 3. The d-spacing of the first domain showed a good fit to that of the Kamhi {110} d-spacing (2.07 ± 0.05 Å), whereas the d-spacing of the minor structure was 3.63 ± 0.05 Å. In his original paper on vaterite Kamhi observed, by means of single-crystal diffraction, eight weak diffraction points that he was not able to index, but he was able to report that the observed image resembled a supercell with a lattice rotated by 30° around the c axis and with lattice parameters of a′ = 3a and c′ = 2c. Here, we instead observed two distinct crystalline structures that were not parts of a supercell but were completely different structures coexisting within a “single crystal” of vaterite. Clearly, the predominant structure was the Kamhi structure.

Fig. 4 HRTEM image (acceleration voltage 80 kV) with two lattices.

The lower lattice has d-spacing = 2.07 ± 0.05 Å and the upper has d′-spacing = 3.63 ± 0.05 Å, and the two lattices are rotated by 30°.

It is intriguing that both electron diffraction and phase contrast revealed what appeared to be perfect single crystals, and that the coexisting two structures were revealed only when the samples were tilted to the two zone axes ([210] and [001]). We stress that although all the various models suggested for vaterite have different unit cells, they all share common features: The carbonate ions are parallel to the c axis, and the calcium ions are organized in a hexagonal structure with similar inter-ion distances. Because TEM is much more sensitive to heavier atoms (calcium in vaterite) than to lighter atoms (oxygen and carbon), most of the contrast under standard conditions is derived from the calcium ions. If the calcium ions in the main and minor distinct structures we observed had matching organizations, it is unlikely that these two structures would be distinguishable in most crystallographic directions. It is therefore likely that the local calcium ion arrangement is conserved across the main and the minor structures, and that only the carbonates differ. This would produce a pseudo-epitaxial effect.

The finding that two distinct atomic structures coexist in a “single crystal” of vaterite may well explain the discrepancies and disputes generated by the experimental results and theoretical calculations published over the past 60 years (for data regarding the d-spacings derived from the FFT in Fig. 3F, and for comparisons with calcite and aragonite, see table S2). Inconsistencies among the published results are found not only in diffraction data but also in spectroscopic data such as Raman spectra and even solid-state nuclear magnetic resonance (27). In Raman spectra, for example, symmetric stretching vibration (ν1) of the carbonates reveals a triplet in some studies and a doublet in others. The Kamhi structure predicts a doublet for this peak (21). In the seminal Raman study of vaterite by Wehrmeister et al. (21), it was concluded that the vaterite, owing to stacking of the carbonate groups, might possess a lower symmetry than that proposed by Kamhi or Meyer. Lower symmetry was also recently proposed by Mugnaioli et al. (25) who, in an elegant study based on ADT and other techniques, proposed two low-symmetry structures, one of which has triclinic symmetry (space group P1), which is almost the lowest symmetry possible. The electron beam diameter in that study was 50 nm. In our study we saw domains on a much smaller scale, only a few nanometers in diameter, in two different structures. We believe these two structures were not found by other researchers because they did not have the resolution we were able to achieve with the aberration-corrected HRTEM, and so they reported an average over the two structures. Even in the work of Mugnaioli et al. (25), where a 50-nm electron beam was used, what the authors observed was a superposition of the two structures.

Our findings indicate that the vaterite crystal structure is not a single entity; rather, vaterite consists predominantly of a hexagonal structure with at least one other coexisting crystallographic structure. The two structures are rotated by 30° with respect to one another and have lattice parameter relationships: a′ = 3a. The challenge ahead is to fully characterize and solve the crystallographic structure of the minor structure and to understand how these two structures are organized relative to one another in space.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Tables S1 and S2

References (2830)

References and Notes

  1. Acknowledgments: We thank E. Zolotoyabko and W. D. Kaplan for helpful discussions, and T. Haim-Cohen and M. Kalina for help in preparing the TEM samples. Supported primarily by U.S.-Israel Binational Science Foundation grant 2010065 (B.P. and P.U.P.A.G.). Also supported by Israel Science Foundation grant 98/10 and German-Israeli Foundation for Scientific Research and Development grant 2278/2010 (B.P.), U.S. Department of Energy grant DE-FG02-07ER15899, and NSF grant DMR-1105167 (P.U.P.A.G.). High-resolution powder diffraction measurements described in this paper were carried out at beamline ID31 of ESRF (Grenoble, France).
View Abstract

Navigate This Article