Sue E. Kesson

Mineralogy and dynamics of a pyrolite lower mantle

There is a growing consensus that the Earths lower mantle possesses a bulk composition broadly similar to that of the upper mantle (known as pyrolite). But little is known about lower-mantle mineralogy and phase chemistry,, especially at depth. Here we report diamond-anvil cell experiments at pressures of 70 and 135 GPa (equivalent to depths within the Earth of about 1,500 and 2,900 km, respectively) which show that pyrolite would consist solely of magnesian-silicate perovskite (MgPv), calcium-silicate perovskite (CaPv) and magnesiowüstite (Mw). Contrary to recent speculation,, no additional phases or disproportionations were encountered and MgPv was found to be present at both pressures. Moreover, we estimate that, at ultra-high pressures where thermal expansivities are low, buoyancy forces inherent in subducted slabs because of their lithology will be of similar magnitude to those required for thermally driven upwelling. So slabs would need to be about 850 °C cooler than their surroundings if they are to sink to the base of the mantle. Furthermore, initiation of plume-like upwellings from the core–mantle boundary, long attributed to superheating, may be triggered by lithologically induced buoyancy well before thermal equilibration is attained. We estimate that ascent would commence within ∼0.5 Gyr of the slab reaching the core–mantle boundary, in which case the lowermost mantle should not be interpreted as a long-term repository for ancient slabs.

Origin of kimberlites and related magmas

Abstract Rare earth fractionation in kimberlites implies that they were produced by partial melting in the presence of residual garnet, in accordance with the widely held belief that kimberlites were formed by small degrees of partial melting of a garnet lherzolite lithology in the upper mantle. However, recent discoveries in some kimberlites of diamond xenocrysts containing syngenetic inclusions of majorite, and of xenoliths which originally contained majoritic garnet are suggestive of a deeper, transition zone origin for kimberlites. Experiments on a synthetic Group I kimberlite were carried out using an MA-8 apparatus to evaluate this possibility. At 16 GPa and 1650°C, majorite garnet (13% Al2O3) and β-M2SiO4 crystallize together on the liquidus, showing that this kimberlite magma could have been produced by a small degree of partial melting of a majorite +β-M2SiO4 (or γ-M2SiO4) assemblage in the transition zone (400–650 km). However, the first appearance of garnet well below the liquidus at 10 GPa implies that this typical kimberlite composition could not have been produced by a small degree of partial melting of garnet peridotite at depths shallower than 300 km, and casts doubt on conventional models of kimberlite petrogenesis. Isotopic, trace element and geochemical similarities imply a genetic relationship between kimberlites and ocean island basalts (OIBs). However, kimberlites were derived from a source possessing a higher Mg-number, and lower Na2O, Al2O3 and CaO contents than the OIB source. It is proposed that the ultimate source regions both of kimberlites and OIBs lie in the transition zone, in a boundary layer comprised of mixed domains of subducted former harzburgite and aesthenospheric pyrolite. The boundary layer was refertilized by partial melts derived from garnetite (former subducted oceanic crust) trapped on the 650 km discontinuity.

Phase relations, structure and crystal chemistry of some aluminous silicate perovskites

Solid solution of ∼ 25 mole% Al2O3 expands the compositional stability field of Mg,Fe silicate perovskite well beyond the limits encountered in the simple ternary system MgOFeOSiO2. Aluminous perovskites synthesised in laser-heated diamond anvil cell experiments at 55–70 GPa from starting materials on the compositional join between Mg3Al2Si3O12 and Fe3Al2Si3O12 (pyrope and almandine) can contain as much as 90 mole% of the ferrous end member. However Fe0.75 Al0.50Si0.75O3 perovskite could not be synthesised. Predictions that garnet coexists with aluminous perovskite at these pressures are unsubstantiated. These new perovskites are approximately isochemical with garnet and accommodate the full complement of Al2O3 (25 mole%) even at ∼ 70 GPa. Some contain as much as 30 mole% Al2O3, and solid solution is probably facilitated by temperature. However, there is certainly no evidence to substantiate a recent proposal that the capacity of perovskite to accommodate Al2O3 in solid solution is progressively inhibited by pressure. Magnesian silicate perovskite should therefore have no difficulty in accommodating the mantle inventory of Al2O3 in solid solution throughout the entire lower mantle pressure regime. There is no reason to expect that a new aluminous phase would be stabilised at depth within the lower mantle. Nor would exsolution of an aluminous phase at core-mantle boundary pressures be a plausible explanation for the D″ layer. Aluminous perovskites are almost always rhombohedral R3c rather than orthorhombic Pbnm, and their unit cell volumes increase by about 3% as 75 mole% of ferrous iron replaces magnesium. These new perovskites are slightly non-stoichiometric, with modest amounts of an M2(Al,Si)O5.5(M =Mg,Fe) component in solid solution. Crystal chemistry fundamentals successfully predict the site occupancy of minor and trace elements in magnesian silicate perovskite.

Partitioning of MgO, FeO, NiO, MnO and Cr2O3 between magnesian silicate perovskite and magnesiowüstite: implications for the origin of inclusions in diamond and the composition of the lower mantle

Abstract Syngenetic mineral inclusions in diamond provide valuable information about the environment in which the diamond originally crystallized. It was earlier proposed that diamonds containing a “forbidden” inclusion assemblage of magnesiowustite (Mg number ∼ 85 and NiO ∼ wt%) plus enstatite En 94–95 may have originally formed in the lower mantle. Enstatite would therefore correspond to the retrogressive transformation product of magnesian silicate perovskite. This hypothesis has been assessed by determining the partition behaviour of MgO, FeO, NiO, MnO and Cr 2 O 3 between perovskite and magnesiowustite at lower mantle pressures (30–50 GPa). Experiments were carried out starting with synthetic olivine that was heated with an infrared laser beam in a diamond anvil high pressure cell. Run products were characterized by transmission electron microscopy and X-ray microanalysis. Perovskite (Mg number ∼ 95) and magnesiowustite (Mg number ∼ 86) are produced by the disproportionation of olivine Fo 90 , whereas Fo 85 yields perovskite plus magnesiowustite with Mg number of ∼ 94 and ∼ 78, respectively. NiO is always strongly partitioned into magnesiowustite ( K NiO mw/pv (wt%) is 6.3 ± 5.1, whilst MnO and Cr 2 O 3 show moderate preferential partitioning into magnesiowustite K MnO = 2.4 ± 1.4 and K Cr 2 O 3 = 2.1 ± 0.9. The partition behaviour of all five oxide species as observed between the enstatite and magnesiowustite inclusion assemblage in diamond is entirely consistent with equilibration at lower mantle pressures. If these rare minerals did indeed form in the lower mantle, as suggested by the above experiments, then their compositions can be used to assess various classes of lower mantle model bulk compositions. A perovskititic lower mantle would be required to be highly magnesian (Mg number ∼ 95), which is unacceptable from a geophysical perspective. A lower mantle possessing significant enrichment of FeO and SiO 2 as compared to the upper mantle, would be comprised of perovskite and magnesiowustite with Mg numbers substantially lower than those of their counterparts in diamond. However, mass-balance considerations indicate that the lower mantle could well possess a bulk composition similar to that of a depleted lithology (Mg number ∼ 92) derived from “pyrolite”. There is therefore no requirement for any profound compositional differences between the upper and lower mantle.

Hexagonal Ba-ferrite: a good model for the crystal structure of a new high-pressure phase CaAl4Si2O11?

Abstract A new calcium aluminosilicate phase of composition CaAl4Si2O11 has been encountered amongst the transformation products of CaAl2Si2O8 (anorthite composition) at 14 GPa (Gautron et al., 1996). X-ray diffraction (XRD) confirms that its crystal structure is essentially the same as that of a new complex CaAl-silicate (abbreviated CAS phase) first reported by Irifune et al. (1994). The crystal structure of the CAS phase has been investigated by transmission electron microscopy (TEM). It has a hexagonal unit cell with lattice parameters a = 5.4 A and c = 12.7 A, and its space group is either P6 3 mc , P 6 2c or P6 3 mmc . It is proposed that this CAS phase has a six-layer, close-packed structure so that Z = 2 and density is 3.94 g cm−3, reasonable for a phase stable at transition-zone pressures. The most plausible model for the structure of this phase arises from published refinements of hexagonal Ba-ferrites. This postulated P6 3 mmc structure consists of octahedral layers, 3 4 occupied, separated by 12-coordinate Ca atoms, and by Al and Si in face-shared octahedra and in complex trigonal bipyramidal polyhedra, i.e. some Si would be five-fold coordinated. Observed TEM and XRD data are compared with calculated reflection intensities for this CAS model.

Immobilization of U.S. Defense Nuclear Wastes Using the Synroc Process

United States defense nuclear wastes are presently in tank storage, partly as sludges comprising Fe, Mn, Ni, U and Na oxides and hydroxides, together with 0.5 to 5 percent of fission products and actinides (exclusive of uranium). The relative proportions of Al, Fe, Mn, Ni, U and Na in the sludges from different tanks vary considerably, except that (Fe+A1+Mn) are by far the major components and Fe is more abundant than Mn. (Typical compositions of some calcined sludges from Savannah River (Garvin, personal communication) are given in Tables 1, 2, 3). In this paper we will briefly describe how the SYNROC process, utilizing straightforward technology, can be readily adapted to the problem of defense waste immobilization, yielding a dense, inert ceramic waste form, SYNROC-D.

[BaxCsy][(Ti, Al)3+2x + yTi4+8-2x-y]O16 Synroc-type hollandites I. Phase chemistry

A series of [BaxCsy][(Ti, Al)3+2x + yTi4+8-2x-y]O16 hollandites, synthesized at 1250°C and coexisting with ‘reduced ’ rutile, demonstrates complete solid solution between barium and caesium endmembers, and simultaneously between Ti3+ and Al3+. The presence or absence of rutile has only a minor effect on stoichiometry. For barium endmember hollandites (y = 0) the stoichiometry (i. e. tunnel site occupancy) ranges from 1.08 ≼ x ≼ 1.14, whilst for caesium endmember hollandites (x = 0) 1.32 ≼ y ≼ 1.51. Neither x nor y correlates with the nature and proportions of trivalent species. An appropriate stoichiometry for the aluminous barium end-member is confirmed as Ba1.14Al2.29Ti5.71O16. The composition BaO. Al2O3. 5TiO2 yields this same hoilandite, and not the supposed phase ‘BaAl2Ti5O14’. The phase ‘BaAl2Ti4O12’ does not exist, while the composition BaO. Al2O3. 4TiO2 crystallizes to an assemblage containing the hollandite mentioned above. Reinterpretation of published X-ray diffraction data substantiate these conclusions and are consistent with a 5c supercell for hollandite. Superlattice ordering in [BaxCsy][(Ti, Al)3+2x + yTi4+8-2x-y]O16 hollandites may be commensurate or incommensurate, with typical multiplicity values (m) and tunnel-site occupancies (x + y) correlating with increasing caesium content per formula unit throughout the series. Barium end-members and barium-rich hollandites, with Cs+ ≼ 0.33 and tunnel-site occupancies of 1.03‒1.15 display 4.5 ≼ m ≼ 5.0. Intermediate hollandites with 0.40 ≼ Cs+ ≼ 0.70 and tunnel-site occupancies ranging from 1.14 to 1.23 possess superstructures with 5.5 ≼ m ≼ 5.7, whereas caesium endmembers and caesium-rich hollandites have tunnel-site occupancies between 1.12 and 1.51 and 5.9 ≼ m ≼ 6.3. For barium or caesium endmembers, multiplicities fail to correlate with tunnel-site occupancies, but do increase with increasing percentages of molar Al3+/(Al3+ + Ti3+) in the structure. Superlattice periodicity is considerably more sensitive to changes in the barium‒caesium content of tunnel sites than to variation in the nature of the trivalent species. Long-range superlattice order is determined not so much by the tunnel cations as by the trivalent species. With more than about one Al3+ per formula unit, one-dimensional (uncorrelated) ordering is suppressed, and three-dimensional order occurs almost exclusively. Hollandite superstructures, and thus their stoichiometries, are determined both by mutual repulsion between large cations within individual tunnels, and intertunnel interaction between large cations. The ceramic high-level nuclear wasteform, Synroc, contains a titanate hollandite belonging to the above series. It has been suggested that the capacity of Synroc to immobilize caesium may be impaired if caesium and barium are not incorporated solely in hollandite, but are partitioned between hollandite and additional titanate phases or hollandite-related structures. No such phase has been encountered in the synthesis of the above hollandite series or in Synroc, prepared according to current specifications, because the trivalent species are present in sufficient abundance to allow the incorporation of all barium and caesium in hollandite. Consequently two-component titanates (for example Cs2Ti6O13 or Ba2Ti9O20), do not appear in the phase assemblage. Moreover, the trivalent species do not comprise Al3+ alone but also include some Ti3+, which promotes more favourable structural modifications and kinetics. Furthermore, the phase assemblage includes ‘reduced’ rutile, which effectively prohibits crystallization of two-component titanates with [Ba, Cs]/[Ti] ratios higher than that in hollandite, and also three-component [Ba, Cs] [Ti, Al]3+-titanates other than hollandite. When these three criteria are satisfied, the appearance of additional, potentially undesirable phases in the Synroc mineralogy is suppressed, and all barium, caesium (and rubidium) may be successfully immobilized in hollandite.

[BaxCsy [(Ti, Al)3+2x+y Ti4+8-2x-y] O16 Synroc-type hollandites - II. Structural chemistry

High-resolution transmission electron microscopy and selected-area electron diffraction show that all phases of the general formula [BaxCsy [(Al, Ti)3+2x + y Ti4+8-2x-y] O16, 1.08 ≼ x + y ≼ 1.51 have the hollandite-type substructure. These hollandites display commensurate and incommensurate superlattices owing to the ordered insertion of large cations (Ba2+, Cs+) into the (2, 2) tunnel interstices of the octahedral (Al, Ti) O6 framework. Multiplicity (m) of a supercell is defined as dsupercell divided by d002 for the subcell. Ordering may be one-dimensional, in which case the cation sequences between (2, 2) channels are independent, three-dimensional with lateral correlation between tunnels, or a combination of both. One-dimensional superstructures yield commensurate multiplicities of 4 in all phases except an aluminous caesium hollandite where m = 6. Three-dimensional superstructures are both incommensurate and commensurate, with 4.50 ≼ m ≼ 6.59. Multiplicities correlate directly with caesium content per formula unit, establishing a maximum in caesiumrich hollandites. Among barium (y = 0) and caesium endmembers, (x = 0) multiplicities increase modestly with increasing Al3+: (Al + Ti)3+ content. Superstructure dimensionality is largely determined by the nature and proportions of the trivalent species, rather than the tunnel cations; one-dimensional order is commonplace in hollandites rich in trivalent titanium but rare in aluminous hollandites. High-resolution electron microscopy supports the interpretation of incommensurate superstructures as fine-scale intergrowths of commensurate microdomains with m = 4, 5, 6 or 7. For aluminous hollandites, rare examples of structural modifications involving tunnels of different cross-sectional dimensions may be found, i. e. T(2, n), 1 ≼ n ≼ 3 intergrowths. As all specimens are sensitive to the electron beam, prolonged irradiation at high electron fluxes can initiate the transformation of single-crystal hollandite to single-crystal rutile. A mechanism for this transformation is proposed, whereby the hollandite crystals initially adjust their multiplicity to six. Growth fronts on {101}holl subsequently propagate through the crystals consuming hollandite and leaving rutile: the structure of the interface between the phases is believed to contain components of rutile possessing antiphase boundaries. In this reconstructive transformation, [100] of the newly formed rutile invariably lies almost parallel to [110] of the original hollandite. Less severe electron irradiation or argon ion beam milling causes crystals to twin polysynthetically. The superlattice properties of [BaxCsy [(Ti, Al)3+2x + y Ti4+8-2x-y] O16 hollandites are integrated with those of other hollandites to identify and evaluate the factors responsible for the stoichiometries and preferred superstructures of hollandites in general. These factors include electrostatic repulsion between large cations in the same tunnel, interaction between cations in neighbouring tunnels, the shielding capacity of the octahedral framework, and kinetic effects.

Reaction between magnesiowusite of lower mantle composition and core-forming Fe-Ni alloy at 1-40 GPa

Abstract One class of models for the early history of the Earth requires the present-day inventory of siderophile elements in the mantle to have been established by equilibrium partitioning between core-forming metal and mantle minerals at high pressures and temperatures deep inside the Earth. We have accordingly carried out reconnaissance experiments on the partitioning of nickel between model lower mantle magnesiowüstite (Mg′ = 85 and 1.3 wt% NiO) and a model core-forming alloy, Fe94Ni6 (~7 wt% Ni) at pressures between 1-40 GPa and temperatures ranging from 1200 °C to 2000 °C. Reversal experiments were also attempted. Our results highlight the difficulty of attaining equilibrium partitioning in this system and imply that partition coefficients derived from unreversed experiments should accordingly be viewed with reservation. Our data nevertheless imply that the concentration of NiO in lower mantle magnesiowüstite in equilibrium with core-forming metal with ~7 wt% Ni would be extremely low, e.g., about 0.2 wt% NiO. Moreover, equilibrium seems to be fairly insensitive to the effects of either pressure or temperature, and so it is unlikely that magnesiowüstite could acquire 1.3 wt% NiO simply by equilibrating with core-forming metal under special high P-T conditions early in Earth history. Alternative hypotheses for the present-day siderophile element inventory of the mantle are accordingly preferred.