Brucite, Mg(OH)2, was investigated by Raman spectroscopy to pressures of 36.6 GPa under nonhydrostatic conditions and to 19.7 GPa under quasi-hydrostatic conditions. Several new Raman lines are first observed at 4 GPa, demonstrating the existence of a high-pressure structural change. One of the new lines grows over a broad pressure interval, and this growth can be explained by a resonant interaction with the Eg translational mode. Raman data are compared with recent infrared spectroscopy, X-ray diffraction, and neutron diffraction studies of brucite. The structural change is likely to involve displacement or disordering of the H atoms, consistent with neutron diffraction results. The Ramanactive O-H stretching vibration in brucite decreases with pressure at the rate of -7 cm-II GPa, larger than the pressure dependence of the infrared-active stretching vibration by more than a factor of ten. The primary differences in the Raman spectra of Mg(OH)2 and Ca(OH)2 are that the O-H vibrational frequencies in Mg(OH)2 vary linearly with pressure, and the O-H stretching vibration band width increases with pressure at a rate that is an order of magnitude lower for Mg(OH)2 than for Ca(OH)2' INTRODUCI'ION cused on determining the brucite-peric1ase dehydration curve, which also constrains the thermodynamic propThe high-pressure behavior of brucite, Mg(OH)2, is of erties of H20 under conditions of the lower crust and considerable interest for studying such diverse phenommantle. Advances in multi-anvil press technology have ena as dehydration reactions at high pressure, compresrecently enabled the dehydration reaction to be studied sion-induced arnorphization, and the behavior of hyup to 15 GPa and 1500 K (Leinenweber et aI., 1993; drous minerals in the Earth's upper mantle. The discovJohnson and Walker, 1993). Detailed analyses of the reery of large numbers of magnesian silicates containing sults require accurate characterization of the thermodystructurally bound OH (Finger and Prewitt, 1989; Kannamic properties of brucite. zaki, 1991) has raised questions concerning the potential Pressure-induced amorphization has now been docurole of such phases in the Earth's interior. Because of its mented in many materials. Portlandite, Ca(OH)2, which structural and chemical simplicity, brucite serves as a is isomorphous with brucite, amorphizes when comuseful prototype for hydrous and layered minerals at high pressed above 11 GPa at room temperature (Meade and pressures. Thermodynamic properties of brucite have been Jeanloz, 1990). In addition to the loss of X-ray diffraction investigated in several recent studies. The equation of peaks, significant changes in the Raman and infrared state has been measured under shock compression (Sispectra of this material have been observed (Kruger et makov et aI., 1974; Duffy et aI., 1991) and under high al., 1989; Meade et aI., 1992; Duffy, in preparation). In static pressures (Fei and Mao, 1993; Catti et aI., 1995; contrast, brucite has been found to be stable to 78 GPa Duffy et aI., 1995). Measurements of the thermal expanby X,ray diffraction (Fei and Mao, 1993) and to 34 GPa sivity at both ambient and high pressure have also been by infrared spectroscopy (Kruger et aI., 1989). As appears reported (Redfern and Wood, 1992; Fei and Mao, 1993). to be the general case for materials undergoing these tranNeutron diffraction has been conducted at elevated pressitions, the amorphization of Ca(OH)2 has been intersures on both normal (Catti et aI., 1995) and deuterated preted as the result of a frustrated phase transition. Howsamples (Parise et aI., 1994). The structure and bonding ever, no such phase transition in brucite has been deof Mg(OH)2 has been investigated theoretically using the tected by shock compression, static compression, or inHartree-Fock approximation (Sherman, 1991; D'Arco et frared spectroscopy experiments over a broad pressure aI., 1993). and temperature range. The phase equilibria of brucite have been the subject Brucite crystallizes in the trigonal CdI2 structure (P3m1) of high-pressure experimental investigations from the (Bernal and Megaw, 1935; Petch and Megaw, 1954; Elle1940s to the present. Such studies (e.g., Bowen and Tutman and Williams, 1956; Zigan and Rothbauer, 1967). tle, 1949; Kennedy, 1956; Irving et aI., 1977) have foThis is a layered structure in which each Mg ion is sur0003-004X/95/0304-0222$02.00 222 TABLE 1. Pressure dependence of the Raman modes of brucite . iMiJP' " .. ". . Mode (em-I) (em-'/GPa) (em-I) E.,(T) 280.0 5.40 0.15P 280 359.6 0.60 + 0.04P 383.8 2.18 408.1 4.21 0.20P A,.(T) 444.7 6.93 0.15P 443 E.,(R) 727.5 725 A,.(I) 3652.0 -7.68 3652 3661.3 -5.34 DUFFY ET AL.: HIGH-P PHASE TRANSITION IN BRUCITE 223 Fig. 1. Crystal structure of brucite. The large spheres are 0 atoms, the intermediate spheres are Mg atoms, and the small spheres are H atoms. The O-H bonds are directed along the c axis. rounded by a distorted octahedron of 0 atoms (Fig. 1). The Mg ions lie in planes with the 0 ions above and below them in a sandwich arrangement. The O-H bonds are perpendicular to these planes. The brucite layers are stacked along the c direction and held together by weak interlayer forces. There is conflicting evidence regarding the nature of these forces. On the basis of the interlayer o distances, Bernal and Megaw (1935) concluded the layers were held together by weak dipole forces. X-rayemission spectra, however, support the existence of some H bonding between the layers (Haycock et aI., 1978). Ab initio Hartree-Fock studies, however, have failed to find evidence for H bonding at ambient or elevated pressure (Sherman, 1991; D'Arco et aI., 1993). Factor group analysis indicates there are six allowed lattice modes for brucite: three of these are infrared active, and three are Raman active (Mitra, 1962). In addition, there are Ramanand infrared-active internal modes. The lattice vibrations consist of translational modes that correspond to vibrations of the O-H units that are either parallel [A,.(T) and A2u(T)] or perpendicular [E.(T) and Eu(T)] to the c axis. There are also rotational vibrations of the OH ions: E.(R) and Eu(R). The internal modes are symmetric (Raman-active) and antisymmetric (infrared-active) O-H stretching vibrations [A,.(I) and A2u(I)]. Ambient-pressure polarized Raman and infrared spectra (Dawson et aI., 1973) from normal and deuterated samples have been used to make mode assignments (Table 1). -...-....------.Data from this study. .. Data from Dawson et al. (1973). EXPERIMENTAL TECHNIQUE Brucite was synthesized in a piston-cylinder apparatus at 1.5 GPa and 1073 K under H20-saturated conditions. The samples were produced in the same experiment as those used in high-pressure X-ray diffraction studies (Fei and Mao, 1993; Duffy et aI., 1995). The crystals were transparent platelets with lateral dimensions up to 50 #m and thicknesses of < 10 #m. Ambient-pressure Raman spectra and X-ray diffraction confirmed that the samples were brucite and no impurity phases were detectable. Raman experiments were carried out in a Mao-Bell diamond-anvil cell with 600-#m culet type I diamonds. The sample was loaded into a hole 300 #m in diameter and 75 #m thick in a T301 steel gasket. Brucite samples were compressed both nonhydrostatically (with no pressure medium) and under quasi-hydrostatic conditions using Ne as a pressure medium. Raman spectra were recorded with a multichannel Raman microprobe (Dilor XY) in a backscattering configuration using a charge-coupled device (CCD) detector with 1024 x 298 channels (e.g., Hemley, 1987). The excitation source was Ar+ laser operated at either 488.0 or 514.5 nm at powers of 100 m W or less. The laser light was focused onto the sample and Raman signal collected using a Leitz L25 objective. Data accumulation times ranged ITom 100 to 3000 s. Peak positions and widths were determined by fitting the spectra to Lorentzian line shapes with background subtraction. Pressures were determined from the fluorescence of small ruby chips (1-5 #m) distributed through the sample volume (Mao et aI., 1986). Pressure was measured both before and after collecting Raman spectra at several positions within the chamber. In experiments with no pressure medium, the R, and R2 ruby fluorescence peaks were not always well resolved, and pressure differences of -1 GPa could develop across the sample chamber. For the Ne medium experiments, the pressure varied by