We have measured the infrared spectra and x-ray diffraction of CaMg(CO3)2-dolomite to pressures of 50 GPa at 300 K. We observe both splittings and disappearances of x-ray diffraction peaks initiating near 20 GPa, as well as new bands in the infrared spectrum of dolomite. The onset of the changes in both the x-ray and infrared data appears to be gradual, and thus kinetically impeded: this is consistent with previous shock results. The infrared and x-ray data are consistent with dolomite adopting a calcite-III-like structure. Our results indicate that high-pressure polymorphism in dolomite could stabilize CaMg(CO3)2 in the deep mantle, and thus that high-pressure polymorphs of dolomite could represent the main reservoir for carbon storage within Earth's lower mantle.
 Dolomite has been investigated to a maximum static pressure of only 16 GPa using Raman spectroscopy [Kraft et al., 1991], and with x-ray diffraction to 11 GPa [Martinez et al., 1996; Fiquet et al., 1994; Ross and Reeder, 1992]. For comparison, shock data extend to 115 GPa on dolomite [Grady et al., 1976; Kalashnikov et al., 1973]. The static studies document that dolomite is stable to 16 GPa at 300 K, with the CO3 group being notably incompressible. However, shock-compressed dolomite shows the presence of a rate dependent low- to high-density phase transformation commencing near 27 GPa [Grady et al., 1976]. Accordingly, we examine dolomite at pressures to 52 GPa to determine whether a phase transition occurs in this material when statically compressed to pressure conditions similar to those of the shock experiments.
2. Samples and Experimental Techniques
 Our starting material was single-crystal dolomite samples. Zero-pressure x-ray analysis yielded lattice parameters in excellent accord with previous results. Electron probe analysis confirmed the purity of the starting material.
 Angular dispersive x-ray data were obtained for dolomite from 0 to 52 GPa at the Stanford Synchrotron Radiation Laboratory using a Si (111) monochromator tuned to a fixed energy of 17.038 keV. Dolomite was ground into 1–5 μm diameter particles and placed in Mao-Bell style diamond anvil cells equipped with 350 μm culets. A 16:3:1 mixture of methanol:ethanol:water was used as a pressure medium. Small (1–5 μm) ruby grains were placed throughout the sample for monitoring pressure and determining pressure gradients using the standard ruby fluorescence method [Piermarini et al., 1975]. Three to six separate ruby grains in the sample were measured at each pressure. Flakes of gold foil were placed along the edge of the sample for sample-to-detector distance calibration.
 Infrared spectra were obtained for dolomite from 0 to 52 GPa. Dolomite with a 1–5 μm particle size was mixed with KBr in a ratio of 90% KBr by weight to 10% CaMg(CO3)2. Such powder spectra have been demonstrated to be in outstanding accord with results from reflectance data for vibrations associated with the carbonate group [White, 1974]. High-pressure spectra were obtained with high-infrared transmission diamonds within the evacuated sample chamber of a Bruker 66 V FTIR spectrometer using a KBr beamsplitter and a liquid-N2 cooled MCT detector. Spectra were taken at ∼2 GPa intervals from 2–52 GPa at room temperature. All spectra are reported with 4 cm−1 resolution, and were collected on both compression and decompression for multiple samples.
3. X-ray Results
 The x-ray diffraction patterns correspond to those of normal dolomite up to 15 GPa. As pressure is increased to ∼20 GPa, the  peak splits (Figure 1a) and the  and  peaks (usually seen as one reflection) dramatically weaken (Figure 1b). By 20 GPa, the latter reflections are almost unresolvable, and the peak that was initially the  peak near 2.7 Å splits into two components (Figure 1a). This doublet remains stable to ∼50 GPa, but reverts to a single peak on decompression. These shifts are similar to the changes in the x-ray diffraction pattern when calcite transforms to calcite-III [Fiquet and Reynard, 1999]. The calcite-I to calcite-III transformation is characterized by the appearance of a strong doublet near the location of the  peak of the low-pressure phase (assigned as the  and [−112] peaks of the calcite-III structure [Smyth and Ahrens, 1997]). All other diffraction lines of the calcite-III pattern are weak, with intensities less than 20% that of the  peak. The high-pressure phase of dolomite appears to show a similar intensity pattern to that of calcite-III, with one doublet (assigned as the  and [−112] reflections) and an additional asymmetric and weak feature (assigned as [−313] and [−204]). No other new or unexplained peaks are observed (Figure 1), and the intensity pattern, peak locations, and number of peaks each lead us to attribute the high pressure x-ray pattern to a calcite-III-like structure.
4. Infrared Results
 The splitting of vibrations of the CO32− - group are known to be diagnostic of high-pressure transitions in initially calcite-structured materials [Adams and Williams, 1980; Fong and Nicol, 1971]. We have thus used high pressure infrared spectra to further constrain the properties of this phase transition. The ambient-pressure infrared spectra of CaMg(CO3)2-dolomite contains three types of vibrations between 500 and 1600 cm−1: (1) the sharp band of the in-plane bending vibration (υ4) of the CO32− group at 724 cm−1 at zero pressure, (2) the out-of-plane vibration (υ2) of the CO32− group at 877 cm−1, (3) and the broad band of the asymmetric stretching vibration (υ3) of the CO32− group, centered at 1448 cm−1. Representative high-pressure spectra are shown in Figures 2a and 2b. A splitting of the asymmetric stretching band (υ3) is clearly visible in the 29.3 GPa spectrum, with both high- and low-frequency shoulders appearing about the central peak. This splitting initiates as low as 20 GPa, but it becomes pronounced at pressures above ∼25 GPa. At 46 GPa, there are clearly three central peaks accompanied by lower amplitude peaks on both the low- and high-frequency sides of the broad central peak. As the sample is decompressed, these splittings are reversible. Other vibrations of the carbonate group undergo similar, although less dramatic splittings. A high-frequency shoulder of the υ4 peak is clearly visible by 40 GPa (Figure 2b). At 51.9 GPa, the υ4 and υ2 peaks (which converge under pressure) have split into at least three separate overlapping peaks, with a possible additional lower frequency shoulder near 760 cm−1. In order to determine whether the transition we observe is kinetically impeded, a sample was externally heated to a temperature of 200°C while held at 43 GPa. The heating increased the amplitude of the new peak arising from the split. Hence, we conclude that our infrared samples contain a mixture of both low- and high-pressure phases to pressures of at least 43 GPa, with the conversion being enhanced by increased temperature. As such, the transition is likely kinetically impeded at 300 K. At 40–50 GPa, five distinct components of the υ3 vibrations are clearly resolved, implying that at least two distinct carbonate ion environments, each likely with low symmetry, are present. This interpretation is supported by the similarity of the high-pressure infrared dolomite spectra with that of calcite-III (Figure 3) at comparable pressures. The splitting of the υ3 vibration is compatible with the structural differences between υ3 in calcite-I and calcite-III [Williams et al., 1992]. For comparison, both the x-ray diffraction pattern and infrared spectrum of CaCO3-aragonite are completely distinct from our results [Kraft et al., 1991]. Moreover, the calcite-II structure also has a different x-ray diffraction and infrared spectrum from our results: calcite-II has five resolvable x-ray diffraction lines with relative intensities of 20% and above [Merrill and Bassett, 1975]; furthermore, no resolvable splitting of the ν2 vibration has been resolved in the calcite-II spectra [Adams and Williams, 1980]. Therefore, both the x-ray diffraction and infrared results are consistent with a transition to a calcite-III like structure becoming pronounced at 20–25 GPa in dolomite compressed at 300 K, and we thus refer to this phase as dolomite-II. Our results are in complete accord with the observation of a sluggish low- to high-density phase transition in shock compression experiments, which show that transformation takes place at 27 GPa and moderate temperatures [Grady et al., 1976]. Finally, we note that observations of the direct transition from calcite-I to calcite-III show that considerable hysteresis is associated with this transition, implying that even within CaCO3 the I-III transition is kinetically impeded [Liu and Mernagh, 1990].
 We calculate lattice parameters for the high-pressure phase using the assumption that the new reflections can be indexed identically to the monoclinic calcite-III structure. However, the new CaMg(CO3)2-phase likely has a doubled unit cell relative to calcite-III, but this effect is not resolved in our diffraction patterns. Calcite-III crystallizes in space group C2 with a = 3.746 Å; b = 4.685 Å; c = 3.275 Å; β = 94.4°, and has a volume change relative to calcite of 4% at 4.1 GPa [Smyth and Ahrens, 1997]. We are able to resolve the [−112] and  lines of this structure in all of our patterns, and the [−204] and [−313] reflections of the calcite-III in two of our patterns. Using a bulk modulus of 94 GPa and a pressure derivative of 4 [Ross and Reeder, 1992] for dolomite, we calculate a volume of of 259.4 Å3 at 20 GPa (the pressure at which the phase transition initiates). Our best-fit modeling for the dolomite-II structure based on our diffraction patterns, assuming a β-value of 94.4° in accord with the estimated value for CaCO3-III [Smyth and Ahrens, 1997], yields a corresponding volume of 249.1 Å3. Thus, we estimate that dolomite-II has about a 4% change relative to dolomite-I, a result clearly compatible with the magnitude of volume change observed for calcite-III. Our results are not particularly sensitive to our assumed β−value: shifting the value by ±4.4° changes the inferred volume of the high-pressure phase by less than 1%. The relative equations of state of dolomite and dolomite-II relative to a mixture of aragonite plus magnesite are shown in Figure 4a. We calculate a volume change of ∼4% for breakdown of dolomite-I to aragonite + magnesite at 20 GPa. Accordingly, CaMg(CO3)2 has a comparable volume to an aragonite/magnesite assemblage at pressures within the dolomite-II stability field. CaMg(CO3)2 is expected to have a larger entropy than the two-phase assemblage: any high temperature cation disorder, as occurs at lower pressures within dolomite, should further increase the entropy [Luth, 2001]. Furthermore, we find that reactions with silicate assemblages at conditions such as the breakdown of dolomite-II + Mg-perovskite to magnesite + Ca-perovskite at the top of the lower mantle, also produce a negligible volume change (Figure 4b). Therefore, thermal effects appear likely to stabilize CaMg(CO3)2 at high pressures; the volume change that destabilizes dolomite at lower pressures disappears at the onset of the high pressure phase.
 At pressures higher than 20–25 GPa, the x-ray patterns and IR spectra of CaMg(CO3)2 change to those of a calcite-III-like structure. Due to its sluggish and kinetically impeded nature this phase may become stable at lower pressures; the pressure range of this transition is in complete accord with previous observations of shock compressed dolomite [Grady et al., 1976]. Our observations of a transition occurring in compressed dolomite have potentially important consequences for carbon sequestration in the Earth. The high pressure polymorphism we document is expected to enhance the thermodynamic stability of CaMg(CO3)2 relative to mixtures containing CaCO3-aragonite and MgCO3-magnesite indicating that this phase could be a major reservoir for carbon in the deep mantle.
 This work is supported by the NSF and the W. M. Keck Foundation. We thank H. Scott, C. D. Martin, J. Adler and L. Whitfield for helpful comments.