Springer-Verlag 2002 Phys Chem Minerals (2002) 29: 633 – 641 DOI 10.1007/s00269-002-0257-3 ORIGINAL PAPER T. Yamanaka Æ T. Fukuda Æ J. Tsuchiya Bonding character of SiO2 stishovite under high pressures up to 30 Gpa Received: 7 January 2002 / Accepted: 6 May 2002 Abstract The charge density and bond character of the rutile-type structure of SiO2 (stishovite) under compression to 30 GPa were investigated by X-ray diffraction study using synchrotron radiation and AgKa rotating anode X-ray generator through a newly devised diamond-anvil cell. The valence electron density was determined by least-squares refinement including the j parameter and the electron population in the X-ray atomic scattering parameters. The oxygen j-parameter of SiO2 is 0.94 under ambient conditions and 1.11 at 29.1 GPa and the silicon valence changes from +2.12(8) at ambient pressure to +2.26(15) at 29.1 GPa. These values indicate that the electron distributions are more localized with increasing pressure. The difference Fourier map shows the deformation of the valence electron distribution and the bonding electron population in residual electron densities. The bonding electron observed from the X-ray diffraction study is interpreted by molecular orbital calculations. The deformation of SiO6 octahedra and the bonding electron density of stishovite structures are elucidated from the overlapping electron orbits. The O–O distances of shared and unshared edge of SiO6 octahedra change with the cation ionicity. The repulsive force between the two cations in the adjacent octahedron makes its shared edge shorter. The pressure changes of the apical and equatorial Si–O interatomic distances are explained by the electron density of state (DOS) of Si and electron configuration. Keywords SiO2 stishovite structure Æ Synchrotron radiation Æ High pressure up to 30 GPa Æ Valence electron population Æ Si–O bond character under compression T. Yamanaka (&) Æ T. Fukuda Æ J. Mimaki Department of Earth and Space Science, Graduate School of Science, Osaka University 1–1 Machikaneyama Toyonaka, Osaka, 560-0043 Japan e-mail: b61400@center.osaka-u.ac.jp Tel.:/Fax: +81-6-6850-5793 Introduction A rutile-type SiO2 polymorph (P42/mnm Z ¼ 2) was first synthesized by high-pressure apparatus (Stishov and Popova 1961) and stishovite was discovered in the Meteor Crater, Arizona (Chao et al. 1962). The phase diagram of SiO2 high-pressure polymorphs has aroused great geophysical interest, since it was confirmed to be one of the major substances in lower mantle. Singlecrystal structure analysis was first executed under ambient conditions by Sinclair and Ringwood (1978). The crystal-structure analyses under pressure were carried out up to 6 GPa by Sugiyama et al. (1987) and 16 GPa by Ross et al. (1990). Systematic single-crystal structure analyses of IVb-group cation dioxide and metal dioxides MO2 with rutile-type structure have been made by Baur and Khan (1971), Bolzan et al. (1997) and Yamanaka et al. (2000). The electron-density distribution in stishovite has been investigated from X-ray diffraction study (Spackman et al. 1987). We previously investigated the electron-density distribution by monopole refinement (j refinement), which was originally derived by Coppens et al. (1979), based on the variable atomic scattering factor, and which discussed the bonding nature by the complementary study of molecular orbital calculation (Yamanaka et al. 2000). Molecular orbital calculation gives a precise definition of electron density of state. Electron orbital overlapping and bonding energy cause a deformation of MO6 octahedra of rutile-type structures from the bond character (Camargo 1996; Gibbs et al. 1997, 1998; Mimaki et al. 2000). Band structure and charge distribution of rutile-type members have been often discussed (Arlinghaus 1974; Jacquemin and Bordure 1975; Simunek et al. 1993). Recently, Kirfel et al. (2001) reinvestigated the electron-density distribution by extinction-free structure refinement using high-energy synchrotron radiation, and applied multipole expansion up to 7th order. To this they compared the result obtained from band-structure calculation. 634 The concept of electron negativity has been applied for the approximation of the covalency/ionicity scale (Chelikowsky and Burdett 1986). The present investigation elucidates the change in Si–O bonding character and the valence electron distribution under high pressure by j refinement. We devised a new type of diamondanvil cell (DAC) for single-crystal X-ray diffraction study under compression (Yamanaka et al. 2001). The valence electron distribution under compression was discussed on the basis of the diffraction intensity measurement using synchrotron radiation. The pressure change of the bonding electron observed from X-ray diffraction study is interpreted by the optimized pair potential and molecular orbital calculation. We aimed to evaluate the effective charge and valence electron population of SiO2 under high pressure. Experimental arranged with topotactic orientation. Thus, the shear stress in both diamond crystals is equivalent under compression. The other side of the diamond plate is placed on the base plane of the angleadjusting steel hemisphere. The present DAC can be easily installed on a four-circle diffractometer. Detailed specification of the new DAC was described in Yamanaka et al. (2001). For diffraction studies at 5.23, 9.26 and 12.3 Gpa, SiO2 a single crystal of 40 · 60 lm wide and 40 lm thick was placed in the Re gasket hole of 200 lm in diameter. The preindented gasket keeps a thickness of about 100 lm. Pressure markers of ruby chips and pressure-transmitting media were also kept in the hole. The media at 5.23, 9.26 and 12.3 GPa was an alcohol mixture with methanol: ethanol:H2O ¼ 16:3:1. The hydrostatic condition could be guaranteed within this pressure range. Argon gas was also used as the transmitting medium in the case of experiments over 15 GPa in order to preserve the hydrostatic condition. The single crystal of 40 · 20 · 20 lm (thickness) was placed in the gasket hole at 29.1 GPa with argon-transmitting media. The new assembly greatly improves the accuracy of the structure analysis. Diamond-plate windows have the following advantages for single-crystal diffractometry: lower X-ray mass absorption, much higher-pressure generation over 50 GPa, no powder rings from the window and a wider observable 2h-angle and transparent window. Sample preparation Single crystals of the rutile-type SiO2 were prepared under high pressure of 12 GPa at 1300 C by a 6–8 type multianvil highpressure apparatus. Anhydrous amorphous silica was placed together with 10 mol% of flux material Li2WO4 in the anvil. The starting sample was kept at 1300 C for 3 h and then the temperature was gradually cooled to 900 C. The grown crystals were transparent and elongated and had a maximum size of 100 · 80 · 500 lm. Homogeneities and chemical impurities in the synthesized crystals were examined by EPMA and no trace element was found. Their crystallinity was tested by optical microscope and X-ray precession camera. Almost cube-shaped crystals with about 60 lm in edge length were selected for the X-ray diffraction experiment. High-pressure diffraction study using the new diamond-anvil cell (DAC) Single-crystal structure refinements under high pressure encountered many difficulties, such as non-hydrostaticity, a large blind region due to the limited aperture angle of the pressure cell, large X-ray absorption from the window and limitation of sample size. So far, single-crystal X-ray diffraction studies of ruby at 31 GPa by Kim-Zajonz et al. (1998) and pyrope garnet at 33 GPa by Zhang et al. (1998) have been reported. Our new system solved these difficulties and made diffraction study possible, enabling the discussion of electron density distribution under pressures up to 50 GPa. Generally, beryllium hemispheres or plate windows were used as backing plate of DAC for the diffraction intensity measurements of single crystal, due to their very high X-ray transmittance. However, many broad and strongly spotted powder rings from beryllium windows often overlap the diffraction peaks of the sample and deform the peak profiles. These obstacle rings interrupt the intensity measurement. Further, beryllium windows cause a limitation of pressurization because of its softness. We tried to find windows more efficient than beryllium (Yamanaka et al. 2001). Our new cell consists of large single-crystal diamond plates supporting the diamond anvils and the angleadjusting steel discs. Large single crystals of (100) platy diamonds of about 2 carats were prepared by cutting and polishing the grown crystals. The diamond crystals had a size of 6 · 6 mm wide and 2 mm thick. Plate windows of 2 mm thick are directly fixed on the (100) table plane of a brilliant-cut diamond anvil with culet sizes of 400 and 600 lm, as shown in Fig. 1. Both plate windows and anvils are Diffraction intensity measurement using synchrotron radiation and AgKa radiation In the present diffraction study, as shown in Table 1, intensity measurements were conducted using an AgKa rotating anode X-ray generator with radiation of k ¼ 0.5608 Å at 5.23, 9.26 and 12.3 GPa in our laboratory and using SR with k ¼ 0.40799 Å (E ¼ 30.388 eV) at a pressure of 29.1 GPa at BL02B1 in SPring-8 at Nishiharima; the synchrotron radiation (SR) source with 8 GeV and 100 mA provides a critical wavelength of 30 keV. SR has excellent characteristics for single-crystal diffraction study under high pressure using DAC. SR emitted from the bending magnet provides a source intensity about 104 times greater than that from the conventional rotating anode X-ray generator. It has the following great advantages for high-pressure diffraction study: high signal/noise ratio of the diffraction intensity, detection of the weak diffraction peaks, precise determination of diffraction intensity, very small extinction effect and short measuring period. The incident beam was monochromated by Si (111) double crystals. The beam is focused on the DAC by a Pt-coated mirror and a collimator guide pipe leads to DAC to reduce the background intensity. A divergent slit of 100 lm in diameter was adopted, because the gasket hole was 200 lm in diameter and the sample size several 10 l cross. Diffraction intensity measurement was carried out using a fourcircle diffractometer with a scintillation counter. The new DAC together with the short wavelength allowed reflections with d > 0.44150 Å (2h < 55) to be observed. The number of observed reflections is four times more than those obtained from the laboratory source. A large number of reflections with precisely observed intensities enable a more precise observation of electron density distribution at high pressures. Intensity measurement was carried out by the /-fixed and x-scan mode, scan speed 1 min)1, scan width 1 in x-rotation, step interval 0.01 in consideration of peak broadening at high pressure. Reflections up to hkl ¼ 661 with dmin ¼ 0.4679 Å at 2h ¼ 51.657 were observed. A total of 147 reflections was observed and 57 crystallographically independent reflections with Fo>3r(Fo) were used for the least-squares refinement. X-ray absorption correction was made for SiO2 stishovite samples. A linear absorption coefficient of each sample is listed in Table 2. Since all of the coefficients were negligibly small (for example l ¼ 0.334 cm)1 at 29.1 GPa) and the sample size was extremely small, the absorption correction of the sample was not considered for diffraction intensity corrections. Because the total thickness of the diamond was larger than 2 mm, the absorptions of the diamond-anvil and diamond-plate window were taken into account in spite of the very low linear absorption coefficient. The 635 Fig. 1 Diamond-anvil cell. Large single-crystal diamond plates and 1/8-carat brilliant cut diamonds were used as the windows and anvils, respectively. Both (100) plates of diamond single crystals were directly fixed on the anvil table plane in topotactic direction absorption coefficient was l ¼ 0.1772 cm)1 in the case of the diffraction study at 29.1 GPa using SR with k ¼ 0.40799 Å. Electron density analysis In the first stage, conventional least-squares refinements of the five data sets at 0.0001, 5.21, 9.20, 12.20 and 29.1 GPa were carried out with the following variable parameters: scale factor, positional parameter of oxygen atom, anisotropic thermal parameters and isotropic extinction parameter Gex based on a crystalline mosaicity (type-I model) (Becker and Coppens 1974). Higher-rank rather than second-rank thermal parameters derived from the anharmonic oscillator model are not considered, because SiO2 has a quite high Debye temperature and intensity was measured at 300 K. The full matrix least-squares refinement was carried out using the program RADY (Sasaki and Tsukimura 1987). After the conventional structure refinement, electron density analysis was executed by monopole refinement introducing the j parameter in the atomic scattering factors. According to the pseudopotential model, inner-core electrons are frozen with bonding effects; but valence electron clouds are deformed due to coordination and thermal atomic vibration, because their interactions with the core electrons are relatively weak. Accordingly, they are more sensitive to the interatomic potential being affected by the coordination of the adjacent atoms. Since the deformation electron densities of Si are supposed to be naturally very small because of spherical electron orbits, except for only a slight excitation of the d electron, a monopole refinement was applied instead of the multipole deformation density. The j parameter (Coppens et al. 1979; van der Wal and Stewart 1984) was applied in the atomic scattering factor of Si, which is an indicator of the radial distributions of the electron. The j parameter used in this analysis is expressed by: qvalence ðrÞ ¼ Pvalence j3 qvalence ðj rÞ ; ð1Þ where qvalence (r) is the ground-state density of the free atom. Pvalence indicates the valence-shell population, which is the occupancy parameter of core and valence electrons. j3 is a normalization factor. The atomic scattering factor f(s) used in the structure refinement is modified from a Hartree–Fock approximation based on the isolated atom model. Perturbed valence electron density was f ðs=2Þ ¼ X Pj;core fj;core ðs=2Þ þPj;valence fj;valence ðjj ; s=2Þ þ fj0 þ ifj00 : ð2Þ 636 Table 1 Diffraction intensity measurement conditions Pressure 1 atm Diffractometer Wave length Energy Monochrometer Gasket Pressure media Scan mode Crystal size (lm) RIGAKU AFC5 Moa (0.7107 Å) 150 kV 50 mA Graphite (002) – – x)2h 50 · 60 · 80 2h angle sinh/k 2h angle/MoKa Ref. (observed) Ref. (independent) 120 1.219 120 210 126 5.23 GPa 9.26 GPa 12.3 GPa RIGAKU AFC6R AgKa (0.5608 Å) 150 kV 50 mA Graphite (002) Spring steel M:E:W=16:3:1 /-fix x-scan 60 · 40 · 40 53 0.796 69 81 25 47 0.711 61 79 25 The valence scattering of the perturbed atom at s/2 ( ¼ sinh/2k) is given by: fM -core ðjj ; s=2Þ ¼ fj ; M -coreðfree atomÞ ðsin h=k 1=jj Þ : ð3Þ A localized electron distribution as indicated by j ¼ 1.0 implies a more ionic character in the bonding nature. The detailed formalization is discussed in Yamanaka et al. (2000). The valence charge of the cation was introduced by the population parameter. The effective charge was determined from the j parameter of the oxygen atomic scattering factor. The parameters of P and j could be refined simultaneously as variable parameters. The j parameter and population parameter were obtained by minimizing of the reliable factor R. In this refinement, the factors of R and wR are defined by: R ¼ S(||Fobs|)|Fcal||)/Sw|Fobs| and Rw = [Sw(||Fobs|)|Fcal||)2/Sw|Fobs|2]1/2, where w ¼ 1/r2(|Fobs|). The valence electrons around the atomic position and the bonding electron distribution cannot be separately evaluated by structure refinement. The covalency of the bond character is estimated from the effective charge. The difference Fourier synthesis and the population parameter show an electron deformation density. The effective charge q is obtained by the spatial integration of the difference electron density by: Z q ¼ 4p r2 DqðrÞdr : ð4Þ Then q has a correlation with the j parameter. Results 29.1 GPa HUBER (512.1) SR (0.40799 Å) 8 GeV 100 mA Si(111) double crystal Spring steel Ar /-fix x-scan 40 · 60 · 20 54 0.809 70 82 26 52 1.074 99 147 57 c=a ¼ 0:6378 þ 0:7986 103 P 9:2359 106 P 2 : ð7Þ The lattice constant of c decreases almost straight with pressure and indicates an opposite curvatures to the a edge. Therefore, the increment of c/a becomes small with increasing pressure, as shown in Fig. 2. The change of cell volume is expressed by the following equation VP ¼ 46:594ð1 3:0162 103 P þ 8:2944 106 P 2 Þ : ð8Þ The unit-cell volume at 29.1 GPa was reduced by as much as 8.6%. Isothermal bulk modulus KT (GPa) was calculated from the volume change using the Birch– Murghanan equation of state, which is shown in Table 3. The large KT value of 295(5) GPa indicates that stishovite is noticeably hard crystal and high-pressure polymorph of SiO2. The present value is a little smaller than those reported by Ross et al. (1990) and Sugiyama et al. (1987) but very similar to Andrault et al. (1998). This may be because we observed the volume change up to much higher pressure than the former two experiments. Lattice constant change The lattice constants were determined by least squares based on the d values of 15 25 reflections in the range 20 < 2h < 30. The lattice constants, axial ratios of c/a and unit-cell volumes are presented in Table 2. The lattice constant ratios of a/ao and c/co are plotted in Fig. 2, together with c/a/co/ao. The lattice constants as a function of pressure are represented by the following equation: a ¼ 4:1805ð1 1:2820 103 P þ 5:5046 106 P 2 Þ ð5Þ c ¼ 2:6674ð1 0:4878 103 P 4:1811 106 P 2 Þ ; ð6Þ and the ratio of c/a changed with pressure is Interatomic distance as a function of pressure The converged structure parameters of SiO2 stishovite are shown in Table 2. The data at ambient pressure are in good agreement with the previous experiment data (Sinclair and Ringwood 1978; Sugiyama et al. 1987; Ross et al. 1990). Interatomic distances of stishovite structure to 29.1 GPa are presented in Table 4. The SiO6 octahedron has a site symmetry of mmm, with four equatorial bonds and two apical bonds, as shown in Fig. 3. The apical Si–O bond of 1.8111(9) Å is much longer than the equatorial bond of 1.7559(9) Å under ambient conditions but the former is more compressive and becomes closer to the latter with increasing pressure, as seen in Fig. 4. The volume compression of the SiO6 octahedron is expressed by the following equation: 637 Table 2 Structure parameter. R = S(||Fobs|)|Fcal||)/Sw|Fobs| and Rw = [Sw(||Fobs|)|Fcal||)2/ Sw | F o b s | 2 ] 1 / 2 , where w=1/ r2(|Fobs|) Pressure 1 atm 5.23 GPa 9.26 GPa 12.3 GPa 29.1 GPa Radiation a(Å) c(Å) c/a V(Å3) abs. coeff (cm)1) No. ref. R(F ) wR(F ) Si (000) b11 b33 b12 O (xx0) b11 b33 b12 MoKa 4.1812(1) 2.6662(3) 0.6377 46.61 1.463 126 0.0253 0.0243 – 0.0045(1) 0.0037(5) 0.0002(2) 0.3063(1) 0.0051(2) 0.0036(3) )0.0009(3) AgKa 4.152(1) 2.6590(8) 0.6404 45.84 0.759 25 0.0440 0.0234 – 0.0126(27) 0.0261(18) 0.0004(21) 0.3063(20) 0.0075(37) 0.0031(17) 0.0005(35) AgKa 4.134(1) 2.6540(7) 0.6420 45.36 0.767 25 0.0312 0.0104 – 0.0055(21) 0.0142(12) 0.0021(14) 0.3056( 9) 0.0051(29) 0.0090(28) 0.0009(29) AgKa 4.118(2) 2.649(1) 0.6433 44.92 0.774 26 0.0345 0.0227 – 0.0088(20) 0.0103(13) 0.0009 (12) 0.3058(19) 0.0104(21) 0.0054(18) 0.0007(35) SR(30KeV) 4.044(6) 2.619(2) 0.6476 42.83 0.334 57 0.0330 0.0282 – 0.0035(11) 0.0131(13) 0.0019(15) 0.3039( 7) 0.0095(13) 0.0074(17) 0.0004(16) the octahedron found in this experiment indicates a precursor phenomenon for the post-stishovite transition. The SiO6 octahedra are linked along the c axis with the shared edge of O1–O2(sh). The edge is much shorter than the unshared edge of O1–O1(unsh), which is equivalent to the cell edge of the c axis. In spite of the shorter interatomic distance, the shared edge is more compressive than the longer unshared edge, as seen from Table 4. Hence, the ratio of O1–O2(sh)/O1–O1(unsh) decreases with increasing pressure, which is a peculiar phenomenon contrary to the Pauling sense. The greater shortening of the shared edge than the unshared edge with pressure is explained by the shielding effect due to Si–Si repulsion along the c axis (Sugiyama et al. 1987; Ross et al. 1990) and probably by tightening the Si–O bond with pressure due to the change in the electron density of state. Bonding character change with pressure Fig. 2 Axial compression. Axial ratios of a/ao and c/co are plotted as a function of pressure. ao and co are the lattice constants at ambient pressure. Deformation of the unit cell under compression is expressed by c/a/co/ao V ðSiO6 ÞP ¼ 3:6858ð1 2:7619 103 P þ 4:487 106 P 2 Þ : ð9Þ The volume decreased almost linearly with increasing pressure. The octahedron is less compressible than the unit cell in the comparison shown by Eq. (9). SiO6 has a tendency towards the structure in which both apical and equatorial bonds are at the same distance under compression. Stishovite transforms to a CaCl2 structure (Pnnm, Z ¼ 2) at about 53.2 GPa by second-order transition with oxygen displacement (Andrault et al. 1998). The CaCl2 structure is suggested as a post-stishovite phase in several experiments (Tsuchida and Yagi 1989; Kingma et al. 1995). The deformation of Structure refinements from two sets of diffraction intensities obtained at ambient pressure and 29.1 GPa provide valence electron densities in the unit cell. The great improvement of high-pressure single-crystal structure analysis using the SR source and newly devised DAC enables clarification of an electron distribution of SiO2 stishovite at 29.1 GPa. In order to estimate the valence electron distributions from the j parameter, the reliable factor R was minimized with optimization of the variable of the j parameter and population parameter. The value of the j parameter for the oxygen atom was 0.94 at ambient pressure and 1.11 at 29.1 GPa. Intensity data sets at 5.21, 9.20 and 12.20 GPa taken from AgKa radiation cannot provide a meaningful j parameter, because the observed reciprocal space is not large enough to determine the precise value. The j parameters of Si and oxygen are presented in Table 5. Fourier transform of f(s) including the population parameter (P) of the valence electrons in Eq. (2) defines electron density q(r) and the spatial integration of q(r) brings the effective charge of oxygen atoms from Eq. (4). 638 Table 3 Isothermal bulk modulus Present data Andrault et al. (1998) Ross et al. (1990)b Ross et al. (1990) Sugiyama et al. (1987) Weidner et al. (1982) KT (GPa) dKT/dP Pmax (GPa) Data Sample and remark 292(13) 291 302(5) 287(2) 313(4) 306(4) 6(fixed) 4.29 2.60(0.8) 6(fixed) 6(fixed) 29.1 53.2 16 5 17 6 Single crystal Combination dataa Single crystal 6 9 Single crystal Brillouin scattering a The cell volume data are at pressures of 0.0001 15 GPa from Ross et al (1990), at 24.6 49.4 GPa from Hemley et al. (1994) and at 48.1 53.2 GPa from Andrault et al. (1998) when dKT/dP is fixed to 6 b Table 4 Interatomic distances. Abbreviations of equatorial and apical bond are indicated by eq and ap and those of shared and unshared edge are by sh and unsh, respectively Pressure 1 atm 5.23 GPa 9.26 GPa 12.3 GPa 29.1 GPa Si–O(eq)(Å) Si–O(ap)(Å) ap/eq Vol(SiO6)(Å3) O1–O2(sh)(Å) O1–O1(unsh)(Å) O1–O3(Å) sh/unsh 1.7559(9) 1.8111(9) 1.0314 7.374 2.2906(10) 2.6662(3) 2.5226(4) 0.8591 1.750(11) 1.798(4) 1.0274 7.266 2.277(5) 2.6590(8) 2.509(10) 0.8563 1.748(8) 1.784(2) 1.0205 7.178 2.275(3) 2.6540(7) 2.498(10) 0.8572 1.742(13) 1.781(4) 1.0223 7.134 2.262(5) 2.649(1) 2.482(17) 0.8539 1.724(3) 1.738(2) 1.0081 6.806 2.242(3) 2.619(2) 2.448(2) 0.8567 Fig. 3 Stishovite structure. Although all three oxygen atoms are crystallographically equivalent, they are distinguished for the SiO6 octahedron. Hatched section, including Si, O1 and O2 atoms, indicates the same plane as the Fourier map in Fig. 6 and the deformation electron density map in Fig. 8 Fig. 4 Pressure change of apical and equatorial Si–O bonds. Open circles are from Ross et al. (1990) Since, as expressed by Eq. (1), the electron distributions are more localized with increasing pressure, a smaller j parameter implies more bonding electrons and intensifies the covalent-bond nature. Our previous study of rutile-type MO2 (M ¼ Si, Ge and Sn) indicates that the j parameter of SiO2 stishovite has a relatively strong covalent bond in comparison with the two other compounds (Yamanaka et al. 2000). After the refinement using atomic scattering factors based on the spherical electron distribution model, the deformations of electron distributions of SiO2 at ambient pressure and 29.1 GPa are disclosed by difference Fourier maps, as shown in Fig. 6a and b, respectively. The maps are the section of the coplanar atoms of Si, O1 and O2 parallel to the plane (110) (Fig. 3). The map at ambient pressure shown in Fig. 6a is very similar to that of Hill et al. (1983) and Spackman et al. (1987). A positive peak with height of 0.7 e Å)3 is found at almost mid-position on the Si–O bond. Four positive residual densities around the cation are also recognized at 0.4 Å from the metal position. A residual electron peak position found under ambient conditions is located at 0.86 Å from the Si atomic position and that at 29.1 GPa is 0.77 Å. The valence electron tends to be more localized at higher pressure. The localized density implies a more ionic character under higher compression. The non-spherical residual electron density around the Si site is probably induced by the overlapping orbital of the d electron state of Si and the p electron of oxygen, resulting in a d–p–p bond. Hence, the noticeable residual 639 Table 5 j parameter and effective charge j parameter of oxygen Residual electron peak position from Si Effective charge Ambient conditions 29.1 GPa 0.94 1.11 0.86 Å +2.12(8) 0.77 Å +2.26(15) 2.08 2.21 0.43 0.45 0.44 0.45 0.52 0.87 0.53 0.45 0.80 0.58 Mulliken population analysis Si net charge Overlap population Si–O (apical) Si–O (equatorial) 3s 3p 3d electron density on the Si–O bond indicates a bonding electron. A large negative density in the O1–O2 shared edge plays a role in preventing cation repulsion. The bonding electron distribution at 29.1 GPa is less remarkable compared with that at ambient pressure. This feature is caused by the observed j parameter and effective charge. By using Eq. (4), the effective charge of Si at ambient pressure and 29.1 GPa is +2.12(8) and +2.26(15), respectively. The dipole moment is calculated by Eq. (5) from the effective charge and interatomic distance. All data obtained from the charge density analysis indicate that SiO2 stishovite becomes more ionic with increasing pressure. The result of the apparent relative ionicity is presented in Table 5. Discussion Pressure effects on the structure can be explained by the virial theorem in statistical physics: * + N X N owij NKB T 1 X Pext ¼ rij V 3V i¼1 j>i orij * + N X N NKB T 1 X ð10Þ ¼ Fij rij V 3V i¼1 j>i where V ¼ volume, N the number of particles, wij the interatomic potential between the i and j particles, rij the distance between the i and j particles, Fij the force which comes up from the j to the i particle, T the absolute temperature and kB the Boltzman constant. Pext indicates an external pressure generated by DAC. The compression implies the equilibrium state between external pressure and internal pressure induced from the crystalline bonding force. The energy of PextV is the summation of the product of interatomic forces and interatomic distances, where V is the compressed unitcell volume. Interatomic distances can be determined directly by diffraction study as a function of pressure. The compression of these distances induces the compression of the lattice constants of bulk crystal. The term Fig. 5 Shared edge/unshared edge. O1–O2(sh) and O1–O1(unsh) edge distances of SiO6 are shown as a function of pressure. Open circles are from Ross et al. (1990) of interatomic force (F) can be obtained by lattice dynamic experiments. However, the X-ray diffraction study gives the electron-density distributions, including valence electrons and bonding electrons. The charge density analysis based on the diffraction intensities provides a view of the effective charge of ions. The apparent electric dipole moment (l) of the Si–O bond can be defined by the product of charge (q) and mean interatomic distance (r): l ¼ rðqSi qO Þ : ð11Þ The dipole moment (lobs) can be experimentally determined from the observed effective charge of Si and O. qSi and qo are the observed value obtained from the present j refinement in Eq. (2). The moment is lobs ¼ 4.51(1) · 10)29Cm ¼ 13.54D at 1 atm and lobs ¼ 4.69 (3) · 10)29Cm ¼ 14.064D at 29.1 GPa. These results indicate that the Si–O bond becomes more ionic with increasing pressure. The charge distribution reveals a significant admixture of covalency in the chemical bonds of SiO2 and the effective charge of Si is far from a formal charge. Our results are consistent with energy band calculations (Svane and Antoncik 1987). The significant d-electron population indicates some degree of non-sphericity of valence electron distribution around the cation. The difference Fourier map of SiO2 (Fig 6a, b) reveals apparently non-spherical electron distribution, which is represented by the residual electron density around the Si atom. In order to investigate the bond character of rutiletype structures of SiO2, we carried out the cluster discrete-variational Xa (DV-Xa) method molecular orbital calculation (Averill and Ellis 1973; Rosen et al. 1976). The DV-Xa method is based on a self-consistent field Hartree–Fock–Slater approximation. The cluster in the 640 Fig. 6a, b Difference Fourier map projected on (110) plane. Contours are at intervals of 0.2 e A3 and positive and negative contours are expressed by solid and broken line, respectively. Residual valence-electron density is revealed around cation and bonding electron distribution on the Si–O bond. The difference Fourier map at ambient pressure is shown in 6a and the map at 29.1 GPa is presented in b Fig. 8 Deformation electron density map at 29.1 GPa obtained from molecular orbital calculation. The section is the same projection as Fig. 6. The increment of the contours is 0.005 e in the range between )0.04 e and +0.04 e. Solid lines and broken lines are positive and negative electron distribution contours Fig. 7 Electron density of state of 3s, 3p and 3d electron of Si in SiO2 stishovite at 29.1 GPa present calculation is {SiO6Si10O38}44. Mulliken population analysis was used to analyze the local electronic properties. It is noted that absolute values from the analysis are dependent on the atomic basis set, but they are meaningful when they are calculated in the same set under different pressure conditions. The partial density of state (PDOP) of the SiO2 valence electron under ambient conditions and at 29.1 GPa are given in Fig. 7, which shows a significant decrease in Si 3s and 3p PDOS at 29.1 GPa. This indicates the accumulation of positive charge of Si under high pressure. In comparison with Si 3s and 3p electrons, little change in the PDOS of the 3d electron was found. This indicates that the 3d electrons have the role of bonding electron between the Si and O atoms. The detailed procedure of the calculation was described in our previous paper (Mimaki et al. 2000). Figure 8 shows the deformation density map of SiO2 stishovite at 29.1 GPa, which shows features very similar to the difference Fourier map shown in Fig. 6b. The bonding electrons in the map are located a little closer to oxygen compared with the residual electron observed from the difference Fourier. The overlap of the electronic orbitals causes the deformation of octahedral coordination SiO6 of the stishovite structure and the bond character of the covalency. The d electron of cations increases the degree of the d–p–p bond in Si–O. The ratio between the shared and unshared edge distance has a strong relation with 641 the interatomic repulsive force between two cations Si– Si and the degree of p bonding of Si–O. The ratio of O1– O2(sh)/O1–O1(unsh) in Table 4 decreases with increase in ionicity. Hence, the more negative electron density between two Si atoms revealed by the difference Fourier map (Figs 6a, 7b) indicates the more shielding effect with increasing pressure. The peculiar pressure changes of bonding characters found in O–O(sh)/O–O(unsh) and Si– O(ap)/Si–O(eq), as mentioned in the earlier section, can be explained by the electron density of state (DOS) of Si and the electron configuration. The electron densities of state obtained from molecular orbital calculation are in good agreement with the results from X-ray photoelectron spectroscopy (XPS) (Barr et al. 1991). References Andrault D, Fiquet G, Guyot F, Hanfland M (1998) Pressureinduced Laudau-type transition in stishovite. Science 282: 720– 724 Arlinghaus FJ (1974) Energy bands in stannic oxide (SnO2). J Phys Chem Solids 35: 931–935 Averill FW, Ellis DE (1973) An efficient numerical multicenter basis set for molecular orbital calculation: application to FeCl4. J Chem Phys 59: 6412–6418 Barr TL, Mohsenian M, Chen LM (1991) XPS valence band studies of the bonding chemistry of germanium oxides and related systems. Appl Surface Sci 51: 71–87 Baur WH, Khan AA (1971) Rutile-type compounds IV. SiO2, GeO2 and comparison with other rutile-type structures. Acta Crystallogr (B)27: 2133–213917 Becker PJ, Coppens P (1974) Extinction within the limit of validity of the Darwin transfer equations II. Refinement to extinction in spherical crystals of SrF2 and LiF. Acta Crystallogr (A)30: 148– 153 Bolzan AA, Fong C, Kennedy BJ, Howard CJ (1997) Structure studies of rutile-type metal dioxides. Acta Crystallogr (B)53: 373–380 Camargo AC, Igualada JA, Beltràn A, Lhusar R, Longo E, Andrès J (1996) An ab initio perturbed ion study of structural properties of TiO2, SnO2 and GeO2 rutile-lattice. Chem Phys 212: 381–391 Chao ECT, Fahey JJ, Littler J, Milton DJ (1962) Stishovite, SiO2, a very high-pressure new mineral from Meteor Crater, Arizona. J Geophys Res 67: 419–421 Chelikowsky JR, Burdett JK (1986) Ionicity and the structural stability of solids. Phys Rev Lett 56: 961–964 Coppens P, Guru Row TN, Leung P, Stevens ED, Becker PJ, Yang YW (1979) Net atom charges and molecular dipole moments from spherical-atom X-ray refinements, and the relation between atomic charge and shape. Acta Crystallogr (A)35: 63–72 Gibbs GV, Hill FC, Boisen Jr. MB (1997) The SiO bond and electron density distributions. Phys Chem Miner 24: 167–178 Gibbs GV, Boisen MB, Hill FC, Tamada O, Downs RT (1998) SiO and GeO bonded interactions as inferred from the bond critical point properties of electron density distributions. Phys Chem Miner 25: 574–584 Hemley RJ, Prewitt CT, Kingma KJ (1994) In: Hemley RJ, Prewitt CT, Gibbs GV (eds) Silica: physical behavior, geochemistry and materials application Mineral Society of America. Washington DC, pp 41–81 Hill RJ, Newton MD, Gibbs GV (1983) A crystal-chemical study of stishovite. J Solid State Chem 47: 185–200 Jacquemin JL, Bordure G (1975) Band structure and optical properties of intrinsic tetragonal dioxides of group-IV elements. J Phys Chem Solids 36: 1081–1087 Kim-Zajonz J, Werner S, Schulz H (1998) High-pressure singlecrystal X-ray diffraction study on ruby up to 31 GPa Z Kristallogr 214: 331–338 Kingma KJ, Cohen RE, Hemley RJ, Mao HK (1995) Transformation of stishovite to a denser phase at lower-mantle pressures. Nature 374: 243–246 Kirfel A, Krane H.-G, Blaha P, Schawarz K, Lippmann T (2001) Electron-density distribution in stishovite, SiO2:a new highenergy synchrotron-radiation study. Acta Crystallogr (A)57, 663–677 Mimaki J, Tsuchiya T, Yamanaka T (2000) The bond character of rutile type SiO2, GeO2 and SnO2 investigated by molecular orbital calculation. Z Kristallogr 215: 419–423 Rosen A, Ellis DE, Adachi H, Averill FW (1976) Calculations of molecular ionization energies using a self-consistent-charge Hartree–Fock–Slater method. J Chem Phys 65: 3629–3634 Ross NL, Shu JF, Hazen RM, Gasparik T (1990) High-pressure crystal chemistry of stishovite. Am Mineral 75: 739–747 Sasaki S, Tsukimura K (1987) Atomic positions of K-shell electrons in crystals. J Phys Soc Jpn 56: 437–440 Simunek A, Vackar J, Wiech G (1993) Local s, p and d charge distributions, and X-ray emission bands of SiO2: a-quartz and stishovite. J Phys Condens Matt 5: 867–874 Sinclair W, Ringwood AE (1978) Single-crystal analysis of the structure of stishovite. Nature 272: 714–715 Spackman MA, Hill RJ, Gibbs GV (1987) Exploration of structure and bonding in stishovite with Fourier and pseudoatom refinement methods using single-crystal and powder X-ray diffraction data. Phys Chem Miner 14: 139–150 Stishov SM, Popova SV (1961) A new modification of silica. Geokhimiya 837–839 Sugiyama M, Endo S, Koto K (1987) The crystal structure of stishovite under pressure up to 6 GPa. Mineral J 13: 455–466 Svane A, Antoncik E (1987) Electronic structure of rutile SnO2 GeO2 and TeO2. J Phys Chem Solids 48: 171–180 TsuchidaY, Yagi T (1989) A new, post-stishovite high-pressure polymorph of silica. Nature 340: 217–220 van der Wal RJ, Stewart RF (1984) Shell population and j refinements with canonical and density-localized scattering factors in analytical form. Acta Crystallogr (A)40: 587–593 Weidner DJ, Bass JD, Ringwood AE, Sinclair W (1982) The singlecrystal elastic moduli of stishovite. J. Geophys Res 87: 4740– 4746 Yamanaka T, Kurashima R, Mimaki J (2000) X-ray diffraction study of bond character of rutile-type SiO2, GeO2 and SnO2. Z Kristallogr 215: 424–428 Yamanaka T, Fukuda T, Hattori T, Sumiya H (2001) New diamond-anvil cell for single-crystal analysis. Rev Sci Instrum 72: 1458–1462 Zhang L, Ahsbahs H, Kutoglu A (1998) Hydrostatic compression and crystal structure of pyrope to 33 GPa. Phys Chem Miner 25: 301–307