Many new observations of seismic anisotropy in the upper and lower boundary layers of the mantle are presented. These observations (from shear-wave splitting on SKS, receiver corrected S, and receiver/source corrected ScS phases) constrain the mineralogy and geodynamics in these regions, which are crucially important to mantle convection. The main target of this thesis is the lower boundary layer: the lowermost mantle; where the new (ScS) dataset greatly expands on the pre-existing coverage to reveal a global dominance of SH fast anisotropy with strength ~1.4 %. Interesting deviations are detected inside the Pacific large low shear velocity province where SV is fast; and in the 'slab graveyard' region beneath Eurasia where the angle of fast wave polarisation shows coherent regional variations in dip suggesting the presence of large-scale (~1000 km) structural features. A model of flow in the lowermost mantle is tested with the hypothesis that the anisotropy is caused by the lattice preferred orientation of MgSi03 post-perovskite. It is demonstrated that the accuracy of ray theory is inadequate to test general models of anisotropy in the lowermost mantle and that a full waveform finite frequency method is required. Finite frequency waveform results do not match the observations for three candidate post-perovskite plastic deformation models (dislocation glide on (001), (010), and (100)/{11O}). Therefore either the flow model is wrong or anisotropy in the lowermost mantle is not caused by dislocation glide deformation in post-perovskite. Anisotropy in the upper mantle causes on average 0.8 s of splitting beneath seismic stations (determined from the SKS dataset). The fast wave tends to be polarised in line with the direction of absolute plate motion in regions disturbed by orogeny in the last 540 Ma. This suggests that orogenies deform the mantle. Beneath subduction zones, anisotropy causes an average 1.3 s of splitting in events shallower than 300 km (from the S dataset). The fast direction tends to be aligned with the strike of the slab (trench parallel splitting); notable exceptions are identified in South America, the Izu-Bonin arc, a segment of the Sunda arc, and at the Hokkaido corner. Interestingly, in each of these regions the trench is migrating forward in the direction of the subducting plate's motion. With the exception of the Mariana arc, subduction zones experiencing trench roll-back all display trench parallel shear wave splitting. This suggests that trench parallel splitting is caused by mantle deformation associated with trench roll-back. Events deeper than 300 km split by an average 0.9 s; this remains true of events deeper than 520 km. The lack of depth dependence on splitting beyond 300 km hints that anisotropy is confined to the slab or is located in a region deeper than ~660 km.

No description available.

No description available.

Whether silicic volcanoes erupt explosively or effusively is controlled by the ability of magma to lose gas prior to eruption. However the mechanisms behind permeability development and obsidian generation remain an enigma for volcanologists. Torsion experiments were performed at high temperature and pressure on three-phase rhyolitic magma analogues simulating magma deforming in shallow conduit conditions. Significant loss of porosity and dissolved water occurred due to brittle fracturing of the melt. Outgassing efficiency increases with strain and is enhanced by crystal content. Fabric development as a result of strain localisation impacts sample rheology as much as crystal, bubble or water contents. Experiments suggest multiple fractures must have formed and healed during shear to account for observed water loss. Viscous suspensions of particles and bubbles were sheared to observe, in real time, the development of deformation fabrics. Bubble deformation and crystal rearrangement are enhanced by increasing crystal content and strain-rate, which magnify heterogeneities in distributions of particles and the gas phase. Development of bands of high particle content and low strain adjacent to melt-rich, high strain areas illustrate the first stage in the formation of permeable outgassing networks and reveals the governing role of textures on magma rheology and outgassing. Multiple methods were used to study outgassing and emplacement mechanisms of the Rocche Rosse obsidian flow. Strain localisation and highly localised variations in melt water content are responsible for extreme textural heterogeneity observed on micron to m scale in the flow. The formation of repeated flow bands may result from high sensitivity of obsidian degassing to water content. Complex surface deformation as opposed to conduit processes governs water and textural distribution and prolongs flow. A model is proposed consisting of multiple cycles of shear-induced outgassing, densification and re-vesiculation during ascent and emplacement to result in observed textures and water contents.

The study of carbonatites and their genesis is an active area of research, with significant implications for carbon in the mantle and the global carbon cycle. In this thesis, possible theories of carbonatite petrogenesis are investigated using two main approaches. Firstly, natural examples of extrusive calciocarbonatites in the Calatrava Volcanic Province (CVP), Spain, are explored. Secondly, experiments at mantle conditions in piston cylinder apparatus are used to reproduce characteristic textures and compositions observed at the CVP, and examine the melting behaviour of potassic carbonated lherzolite in the mantle. The extrusive pyroclastic carbonatite deposits at the CVP can potentially provide· insights into deep mantle melts. Carbonate-rich volcanism is prevalent, and closely associated with silica-undersaturated magmas such as melilitites, nephelinites and leucitites. Rounded calcium carbonate globules within silicate melt fragments in carbonate tuffs are described and analysed in detail in this study. A variety of methods are used to characterise the textures and geochemistry of the carbonate globules, to investigate whether they represent quenched immiscible carbonatite liquids or solid calcite crystals with a rounded shape. Results from trace element partitioning studies of the natural samples, and experiments based on the Calatrava compositions, suggest that the carbonate globules are rounded carbonate crystals, and do not support an immiscible origin in their formation. In an effort to understand carbonatite melting at a more general scale, experimental results in the simple synthetic system, CaO-MgO-Ah03-Si02-K20-C02, at 30 kbar are presented. A low variance approach is utilised to maximise the proportion of melt to aid in identification and analysis of small-degree partial melts. Potassium-rich carbonatite melt is found to be in equilibrium with phases typical of a carbonated lherzolite assemblage over a large temperature interval. Melt composition changes very little over the temperature range, corresponding to a region of highly potassic (rv20 wt.% K20) carbonatite metasomatism at mantle depths.

Water and alkalis such as K and Na can be delivered to the lower mantle in subducted slabs and the fate of these components in lower mantle are investigated here. A series of experiments on the MgO-Si02-H20 (MSH) and nepheline-spinel-kalsilite (NSK) ternaries have been completed using a Laser-Heated Diamond Anvil Cell and analysed using X-ray diffraction (XRD), FEO (Field Emission Gun)-probe and EDS-SEM (Energy Dispersive Spectroscopy-Secondary Electron Microscope) techniques. Phase relations in the MSH ternary appear complex. Complete melting of experiments at mantle adiabatic temperatures indicates that hydrous lithologies would melt at ambient lower mantle conditions. The stability of phase D has been confirmed at lower mantle conditions equivalent to subducted geotherms, suggesting that phase D remains stable in cold subducted slabs only. The [101] peak for this phase varies systematically as a function of pressure, thought to be related to variations in composition. Phase D was not found in high pressure experiments (>~45 GPa). A possible new phase, phase H, previously predicted to exist by Tsuchiya (2013) was observed in some experiments. This new phase may be important in storing and delivering water to the lower mantle. The NSK system, was systematically investigated at lower mantle conditions. At pressures <50 GPa and mantle adiabatic temperatures, melting occurs. Above 50 GPa, CF (Calcium-ferrite type phase) was stable in all compositions. NAL (New Aluminous phase), was found in potassium rich compositions at this pressure and is thought to be stabilised to higher pressures by potassium. This work implies that CF and NAL are both stable in sediments and MORB, below 45 GPa, but as pressure increases to >50 GPa CF alone is stable. Differentiating between NAL and CF found in diamond inclusions using potassium content may not be valid, since CF is able to contain potassium in its structure at high pressure.

Primitive melts produced beneath island arc volcanoes are rarely erupted at the surface in their original form, instead charting a huge variety of evolved compositions and testifying to the influence of intracrustal processing during magmatic ascent. The study of cumulates (coarse-grained igneous rocks) that sample directly from magma storage regions offers a chance to glimpse a 'snapshot' of this magmatic evolution. This thesis combines major and trace element analysis of mineral compositions in plutonic xenoliths with experimental petrology to explore the differentiation of mantle-derived magmas beneath volcanic island of Grenada, Lesser Antilles. The observed diversity in cumulate assemblage and texture is caused by variability in parental melt composition and post-cumulus interaction with hydrous evolved melts. The whole plutonic suite is produced in a narrow pressure window (P = 0.2 - O.S GPa) at f'V 8S0 - lOS0°C, tracing a shallow (depth :SlSkm) section of a vertically extensive volcanic system. All but the most primitive Grenadan lavas contain a significant proportion of antecryst material derived from unconsolidated cumulate piles, and hornblende only appears as a true phenocryst in the most evolved andesites. Experimentally determined phase relations of a Grenadan picrite demonstrate that the geochemically and petrologically distinct M- and C-series lavas can be generated from a common pieri tic source; the observed dichotomy results from differentiation in the uppermost mantlellower crust (f'V 1.4 - 1.8 GPa) and upper crust (f'V 0.2 GPa) respectively and the effect of pressure on the temperature of olivine and clinopyroxene saturation. Numerical modelling shows that hydrous M-series picrites are in equilibrium with a garnet lherzolite at a depth of f'V 70 km (2.3 GPa). These results are consistent with aspects of primitive melt composition that indicate the site of magma generation beneath Grenada is significantly deeper than its northerly neighbouring island of St Vincent.

Subduction is a key process in the formation of continental crust. However, the interaction of the mantle with the subducting slab is not fully understood and varies between subduction zones. In this thesis I use seismological techniques to investigate subduction beneath the Lesser Antilles Arc (LAA) and eastern Greater Antilles. I first investigate the earthquake-magnitude distribution along the LAA, which shows "bulls-eyes" of elevated b-values, implying a relative increase in smaller magnitude earthquakes. They coincide with subducted fracture zones, suggesting they serve as conduits for increased amounts of water, which acts as a lubricant. I then investigate the LAA crustal structure at 10 islands using receiver functions (RF). Synthetic modelling shows that increased amounts of crustal water content can weaken the Moho (at about 30km depth) and a mid-crustal discontinuity (about 20km) can dominate the RF result. ID velocity-depth profiles, constrained from petrological studies, reveal a dichotomy in Moho strength; it is weaker south of Martinique due to higher crustal water content. On average, both discontinuities are deeper in the north. Finally, I use shear-wave splitting analyses of local and teleseismic phases to investigate anisotropy. Local event delay times (0.21 ± 0.12s) do not increase with depth, indicating a crustal origin and an isotropic mantle wedge. Teleseismic delay times are larger (1.34 ± 0.47 s), indicating sub-slab anisotropy. The results suggest trench-parallel mantle flow, with the exception of trench-perpendicular alignment in narrow regions east of Puerto Rico and south of Martinique, suggesting mantle flow through gaps in the slab. The RF and anisotropy results show a change in arc properties between Martinique and St. Vincent, possibly marking the North-South American plate boundary. Dissimilarity in the RF study and b-value patterns indicates that crustal water does not necessarily correspond with water released at the subducted plate, but may correlate with sediment dewatering at shallow levels. Cumulatively, the three investigations offer insights into the dynamic behaviour of a region of slow subduction.

Volcanic arc magmas are usually thought to receive components from three different sources: the subarc mantle, fluids derived from the subducted oceanic crust and lithospheric mantle, and melts derived from subducted sediments. This thesis re-evaluates this model based on isotopic data for magmas from the Izu, New Britain and Mariana arcs. High (230Th/232Th) ratios of up to 2.5 and U excess over Th in samples from the Izu and New Britain arcs are shown to be inconsistent with a derivation from subducted sediments or the sub-arc mantle and are instead argued to reflect a contribution of a melt of the mafic oceanic crust. In addition to Th, the magmas require a number of incompatible elements to be added to the sub-arc mantle. Many of these elements have traditionally been interpreted to be derived from a melt of subducted sediments. Yet, radiogenic Nd and Hf isotopes in these samples do not indicate the presence of a sediment melt. In order to explain this discrepancy, a model is presented in which the component previously identified as sediment melt is a mixture of a sediment melt and a melt of the mafic oceanic crust. Mo isotopes are suggested as a novel tracer for subduction components in volcanic arc magmas and a potential tracer for deep recycling of material transported into the mantle beyond subduction zones. Mo isotope ratios are presented for samples from oceanic sediments, the altered top part of the mafic oceanic crust, and volcanic arc magmas from the Mariana arc, thus defining input and output parameters for a subduction zone. Mo is shown to be preferentially transported in an isotopically distinct fluid derived from the subducted slab. Fluid addition leads to values of 0'98Mo in the Mariana arc lavas ~0.1-0.3 %0 higher than in MORB.

Arc volcanism at subduction zones is likely regulated by the mantle wedge's flow regime and thermal structure and, hence, numerous studies have attempted to quantify the princi- pal controls on mantle wedge conditions. In this thesis, we build on these previous studies by undertaking the first systematic 2-D and 3-D numerical investigations, across a wide parameter-space, into how hydration, thermal buoyancy and toroidal flow around the slab edge influence the wedge's flow regime and associated thermal structure. We find that small- scale convection (SSC), resulting from Rayleigh-Taylor instabilities, or drips, off the base of the overriding lithosphere, is a typical occurrence, if: (i) viscosities are < 5 x 1018 Pa s; and (ii) hydrous weakening of wedge rheology extends at least 100-150 km from the forearc cor- ner. In 2-D models, instabilities generally take the form of 'drips'. In 3-D, two separate, but interacting, longitudinal Richter roll systems form (with their axes aligned perpendicular to the trench), the first below the arc region and the second below the back-arc region. These instabilities result in transient and spatial temperature fluctuations of 100-150K, which are sufficient to influence melting, the stability of hydrous minerals and the dehydration of crustal material. Furthermore, they are efficient at eroding the overriding lithosphere, par- ticularly in 3-D and, thus, provide a means to explain observations of high heat flow and thin back-arc lithosphere at many subduction zones. A preliminary study into the effects of a finite-width slab on the wedge's flow regime, which allows for toroidal flow around the slab edge, highlights that the toroidal cell can locally increase or decrease temperatures suf- ficiently to either enhance or shut down wet melting, while a hydrated wedge corner may channel trench-parallel flow. The dynamic complexities of wedge flow revealed by our mod- els may help explain the diversity in geophysical and geochemical subduction signatures.