Melt topology and chemistry during partial melting in contact aureoles : an experimental study and field example

This study was initiated to investigate partial melting within the high-grade metamorphic rocks beneath the Little Cottonwood contact aureole (Utah, USA), in order to understand the melt generation, melt migration, and geometry of initial melt distribution on grain scale during crustal anatexis. The emplacement of the Little Cottonwood stock produced a contact aureole in the pelitic host rocks of the Big Cottonwood formation (BC). Metamorphic isogrades in pelitic rocks range form biotite to 2nd sillimanite grade as a function of distance from the contact. Migmatites are restricted to the highest grade and resulted form partial melting of the BC formation rocks. First melt was produced by a combined muscovite/biotite dehydration reaction in the sillimanite + k-feldspar stability field. Melt extraction from the pelites resulted in restites (magnetite + cordierite + alumosilicate ± biotite) surrounded by feldspar enriched quartzite zones. This texture is the result of gradual infiltration of partial melts into the quartzite. Larger, discrete melt accumulation occurred in extensional or transpressional domains such as boudin necks, veins, and ductile shear zones. Melt composition are Si02- rich, crystallized as pegmatites, and apparently were very mobile. They were able to infiltrate the quartzite pervaisivly. These melts are similar in composition to first melts produced in the hydrothermal partial melt experiments at 2kbar between 700 - 800°C on fine grained high metamorphic rocks (andalusite-cordierited-biotite-zone) of the BC formation. The experimental melts are water rich and in disequilibrium with the melting rock. Initial melt composition is heterogeneous for short run duration, reflective a lack of chemical equilibrium between individual melt pools. Rock core scale heterogeneity decreased with time indicating partial homogenization of melt compositions. A simultaneous shift of melt composition to higher silica content with time was observed. The silica content of the melt increased due to local melt/mineral reactions. Melt textures indicate that reactive melt transport is most efficient along grain boundaries rimmed by dissimilar grains. Melt heterogeneity resulted in chemical potential gradients which are major driving forces for initial melt migration and govern melt distribution during initial melting. An additional subject of the thesis is the crystal size distributions of opaque minerals in a fine-grained, high-grade meta-pelite of the Big Cottonwood which were obtained from 3D X-ray tomography (uCT) and 2D thin section analysis. µCT delivers accurate size distributions within a restricted range (~ a factor of 20 in size in a single 3D image), while the absolute number of crystals is difficult to obtain from these sparsely distributed, small crystals on the basis of 2D images. Crystal size distributions obtained from both methods are otherwise similar. - Ce travail de recherche a été entrepris dans le but d'étudier les processus de fusion partielle dans les roches fortement métamorphiques de l'auréole de contact de Little Cottonwood (Utah, USA) et ceci afin de comprendre la génération de liquide de fusion, la migration de ces liquides et la géométrie de la distribution initiale des liquides de fusion à l'échelle du grain durant l'anatexie de la croûte. L'emplacement du petit massif intrusif de Little Cottonwood a produit une auréole de contact dans les roches pélitiques encaissantes appartenant à la Foimation du Big Cottonwood (BC). Les isogrades métamorphiques dans les roches pélitiques varient de l'isograde de la biotite à la deuxième isograde de la sillimanite en fonction de la distance par rapport au massif intrusif. Les migmatites sont restreintes aux zones montrant le plus haut degré métamorphique et résultent de la fusion partielle des roches de la Formation de BC. Le premier liquide de fusion a été produit par la réaction de déshydratation combinée de la muscovite et de la biotite dans le champ de stabilité du feldspath potassique Pt de la sillimanite. L'extraction du liquide de fusion des pélites forme des restites (magnétites + cordiérite + aluminosilicate ± biotite) entourées par des zones de quartzites enrichies en feldspath. Cette texture est le résultat de l'infiltration graduelle du liquide de fusion partielle dans les quartzites. Des accumulations distinctes et plus larges de liquide de fusion ont lieu dans des domaines d'extension ou de transpression tels que les boudins, les veines, et les zones de cisaillement ductile. La composition des liquides de fusion est similaire à celle des liquides pegmatoïdes, et ces liquides sont apparemment très mobiles et capables d'infiltrer les quartzites. Ces liquides de fusion ont la même composition que les premiers liquides produits dans les expériences hydrotheunales de fusion partielle à 2kbar et entre 700-800° C sur les roches finement grenues et hautement métamorphiques (andalousite-cordiérite-biotite zone) de la Formation de BC. Les liquides de fusion obtenus expérimentalement sont riches en eau et sont en déséquilibre avec la roche en fusion. La composition initiale des liquides de fusion est hétérogène pour les expériences de courte durée et reflète l'absence d'équilibre chimique entre les différentes zones d'accumulation des liquides de fusion. L'hétérogénéité à l'échelle du noyau s'estompe avec le temps et témoigne de l'homogénéisation de la composition des liquides de fusion. Par ailleurs, on observe parallèlement un décalage de la composition des liquides vers des compositions plus riches en silice au cours du temps. Le contenu en silice des liquides de fusion évolue vers un liquide pegmatitique en raison des réactions liquides/minéraux. Les textures des liquides de fusion indiquent que le transport des liquides est plus efficace le long des bordures de grains bordés par des grains différents. Aucun changement apparent du volume total n'est visible. L'hétérogénéité des liquides s'accompagne d'un gradient de potentiel chimique qui sert de moteur principal à la migration des liquides et à la distribution des liquides durant la fusion. Un sujet complémentaire de ce travail de thèse réside dans l'étude de la distribution de la taille des cristaux opaques dans les pélites finement grenues et fortement métamorphiques de la Formation de Big Cottonwood. Les distributions de taille ont été obtenues suite à l'analyse d'images 3D acquise par tomographie ainsi que par analyse de lames minces. La microtomographie par rayon X fournit une distribution de taille précise sur une marge restreinte (- un facteur de taille 20 dans une seule image 3D), alors que le nombre absolu de cristaux est difficile à obtenir sur la base d'image 2D en raison de la petite taille et de la faible abondance de ces cristaux. Les distributions de taille obtenues par les deux méthodes sont sinon similaire. Abstact: Chemical differentiation of the primitive Earth was due to melting and separation of melts. Today, melt generation and emplacement is still the dominant process for the growth of the crust. Most granite formation is due to partial melting of the lower crust, followed by transport of magma through the crust to the shallow crust where it is emplaced. Partial melting and melt segregation are essential steps before such a granitic magma can ascent through the crust. The chemistry and physics of partial melting and segregation is complex. Hence detailed studies, in which field observations yield critical information that can be compared to experimental observations, are crucial to the understanding of these fundamental processes that lead and are leading to the chemical stratification of the Earth. The research presented in this thesis is a combined field and experimental study of partial melting of high-grade meta-pelitic rocks of the Little Cottonwood contact aureole (Utah, USA). Contact metamorphic rocks are ideal for textural studies of melt generation, since the relatively short times of the metamorphic event prevents much of the recrystallization which plagues textural studies of lower crustal rocks. The purpose of the study is to characterize melt generation, identify melting reactions, and to constrain melt formation, segregation and migration mechanisms. In parallel an experimental study was undertaken to investigate melt in the high grade meta pelitic rocks, to confirm melt composition, and to compare textures of the partial molten rock cores in the absence of deformation. Results show that a pegmatoidal melt is produced by partial melting of the pelitic rocks. This melt is highly mobile. It is capable of pervasive infiltration of the adjacent quartzite. Infiltration results in rounded quartz grains bordered by a thin feldspar rim. Using computed micro X-ray tomography these melt networks can be imaged. The infiltrated melt leads to rheological weakening and to a decompaction of the solid quartzite. Such decompaction can explain the recent discovery of abundant xenocrysts in many magmas, since it favors the isolation of mineral grains. Pervasive infiltration is apparently strongly influenced by melt viscosity and melt-crystal wetting behavior, both of which depend on the water content of melt and the temperature. In all experiments the first melt is produced on grain boundaries, dominantly by the local minerals. Grain scale heterogeneity of a melting rock leads thus to chemical concentration gradients in the melt, which are the driving force for initial melt migration. Pervasive melt films along grain boundaries leading to an interconnected network are immediately established. The initial chemical heterogeneities in the melt diminish with time. Résumé large public: La différenciation chimique de la Terre primitive est la conséquence de la fusion des roches et de la séparation des liquides qui en résultent. Aujourd'hui, la production de liquide magmatique est toujours le mécanisme dominant pour la croissance de la croûte terrestre. Ainsi la formation de la plupart des granites est un processus qui implique la production de magma par fusion partielle de la croûte inférieure, la migration de ces magmas à travers la croûte et finalement son emplacement dans les niveaux superficielle de la croûte terrestre. Au cours de cette évolution, les processus de fusion partielle et de ségrégation sont des étapes indispensables à l'ascension des granites à travers la croûte. Les conditions physico-chimiques nécessaires à la fusion partielle et à l'extraction de ces liquides sont complexes. C'est pourquoi des études détaillées des processus de fusion partielle sont cruciales pour la compréhension de ces mécanismes fondamentaux responsables de la stratification chimique de la Terre. Parmi ces études, les observations de terrain apportent notamment des informations déterminantes qui peuvent être comparées aux données expérimentales. Le travail de recherche présenté dans ce mémoire de thèse associe études de terrain et données expérimentales sur la fusion partielle des roches pélitiques de haut degré métamorphiques provenant de l'auréole de contact de Little Cottonwood (Utah, USA). Les roches du métamorphisme de contact sont idéales pour l'étude de la folination de liquide de fusion. En effet, la durée relativement courte de ce type d'événement métamorphique prévient en grande partie la recristallisation qui perturbe les études de texture des roches dans la croûte inférieure. Le but de cette étude est de caractériser la génération des liquides de fusion, d'identifier les réactions responsables de la fusion de ces roches et de contraindre la formation de ces liquides et leur mécanisme de ségrégation et de migration. Parallèlement, des travaux expérimentaux ont été entrepris pour reproduire la fusion partielle de ces roches en laboratoire. Cette étude a été effectuée dans le but de confirmer la composition chimique des liquides, et de comparer les textures obtenues en l'absence de déformation. Les résultats montrent qu'un liquide de fusion pegmatoïde est produit par fusion partielle des roches pélitiques. La grande mobilité de ce liquide permet une infiltration pénétrative dans les quarzites. Ces infiltrations se manifestent par des grains de quartz arrondis entourés par une fine bordure de feldspath. L'utilisation de la tomography à rayons X a permis d'obtenir des images de ce réseau de liquide de fusion. L'infiltration de liquide de fusion entraîne un affaiblissement de la rhéologie de la roche ainsi qu'une décompaction des quartzites massifs. Une telle décompaction peut expliquer la découverte récente d'abondants xénocristaux dans beaucoup de magmas, puisque elle favorise l'isolation des minéraux. L'infiltration pénétrative est apparemment fortement influencée par la viscosité du fluide de fusion et le comportement de la tension superficielle entre les cristaux et le liquide, les deux étant dépendant du contenu en eau dans le liquide de fusion et de la température. Dans toutes les expériences, le premier liquide est produit sur les bordures de grains, principalement par les minéraux locaux. L'hétérogénéité à l'échelle des grains d'une roche en fusion conduit donc à un gradient de concentration chimique dans le liquide, qui sert de moteur à l'initiation de la migration du liquide. Des fines couches de liquide de fusion le long de bordures de grains formant un réseau enchevêtré s'établit immédiatement. Les hétérogénéités chimiques initiales dans le liquide s'estompent avec le temps.

Investigations into the degassing and eruption mechanisms of Nyamuragira volcano, Democratic Republic of the Congo (Africa)

One of two active volcanoes in the western branch of the East African Rift, Nyamuragira (1.408ºS, 29.20ºE; 3058 m) is located in the D.R. Congo. Nyamuragira emits large amounts of SO2 (up to ~1 Mt/day) and erupts low-silica, alkalic lavas, which achieve flow rates of up to ~20 km/hr. The source of the large SO2 emissions and pre-eruptive magma conditions were unknown prior to this study, and 1994-2010 lava volumes were only recently mapped via satellite imagery, mainly due to the region’s political instability. In this study, new olivine-hosted melt inclusion volatile (H2O, CO2, S, Cl, F) and major element data from five historic Nyamuragira eruptions (1912, 1938, 1948, 1986, 2006) are presented. Melt compositions derived from the 1986 and 2006 tephra samples best represent pre-eruptive volatile compositions because these samples contain naturally glassy inclusions that underwent less post-entrapment modification than crystallized inclusions. The total amount of SO2 released from the 1986 (0.04 Mt) and 2006 (0.06 Mt) eruptions are derived using the petrologic method, whereby S contents in melt inclusions are scaled to erupted lava volumes. These amounts are significantly less than satellite-based SO2 emissions for the same eruptions (1986 = ~1 Mt; 2006 = ~2 Mt). Potential explanations for this observation are: 1) accumulation of a vapor phase within the magmatic system that is only released during eruptions, and/or 2) syn-eruptive gas release from unerupted magma. Post-1994 Nyamuragira lava volumes were not available at the beginning of this study. These flows (along with others since 1967) are mapped with Landsat MSS, TM, and ETM+, Hyperion, and ALI satellite data and combined with published flow thicknesses to derive volumes. Satellite remote sensing data was also used to evaluate Nyamuragira SO2 emissions. These results show that the most recent Nyamuragira eruptions injected SO2 into the atmosphere between 15 km (2006 eruption) and 5 km (2010 eruption). This suggests that past effusive basaltic eruptions (e.g., Laki 1783) are capable of similar plume heights that reached the upper troposphere or tropopause, allowing SO2 and resultant aerosols to remain longer in the atmosphere, travel farther around the globe, and affect global climates.

Continuous eclogite melting and variable refertilisation in upwelling heterogeneous mantle

Large-scale tectonic processes introduce a range of crustal lithologies into the Earth's mantle. These lithologies have been implicated as sources of compositional heterogeneity in mantle-derived magmas. The model being explored here assumes the presence of widely dispersed fragments of residual eclogite (derived from recycled oceanic crust), stretched and stirred by convection in the mantle. Here we show with an experimental study that these residual eclogites continuously melt during upwelling of such heterogeneous mantle and we characterize the melting reactions and compositional changes in the residue minerals. The chemical exchange between these partial melts and more refractory peridotite leads to a variably metasomatised mantle. Re-melting of these metasomatised peridotite lithologies at given pressures and temperatures results in diverse melt compositions, which may contribute to the observed heterogeneity of oceanic basalt suites. We also show that heterogeneous upwelling mantle is subject to diverse local freezing, hybridization and carbonate-carbon-silicate redox reactions along a mantle adiabat.

Element concentrations of plagioclase and clinopyroxene in poikilitic olivine gabbros from ODP Hole 153-923A (Table 4)

Ocean Drilling Program Hole 923A, located on the western flank of the Mid-Atlantic Ridge south of the Kane Fracture Zone, recovered primitive gabbros that have mineral trace element compositions inconsistent with growth from a single parental melt. Plagioclase crystals commonly show embayed anorthitic cores overgrown by more albitic rims. Ion probe analyses of plagioclase cores and rims show consistent differences in trace element ratios, indicating variation in the trace element characteristics of their respective parental melts. This requires the existence of at least two distinct melt compositions within the crust during the generation of these gabbros. Melt compositions calculated to be parental to plagioclase cores are depleted in light rare earth elements, but enriched in yttrium, compared to basalts from this region of the Mid-Atlantic Ridge, which are normal mid-ocean ridge basalt (N-MORB). Clinopyroxene trace element compositions are similar to those predicted to be in equilibrium with N-MORB. However, primitive clinopyroxene crystals are much more magnesian than those produced in one-atmosphere experiments on N-MORB, suggesting that the major element composition of the melt was unlike N-MORB. These data require that the diverse array of melt compositions generated within the mantle beneath mid-ocean ridges are not always fully homogenised during melt extraction from the mantle and that the final stage of mixing can occur efficiently within crustal magma chambers. This has implications for the process of melt extraction from the mantle and the liquid line of descent of MORB

Trace element concentrations in zircons from ODP Holes 147-894G and 153-923A

The trace element compositions of Hadean zircons have been used in two ways to argue for the existence of Hadean continental crust. One argument is based on low crystallization temperatures of Hadean zircons that have been determined using a novel geothermometer based on the Ti content of zircons in equilibrium with rutile. The second argument is based on using the trace element abundances in zircons to calculate their parental melt compositions, especially the rare earth elements. Here we demonstrate that zircons that grow from a melt formed by basalt differentiation at modern mid-ocean ridges cannot be unambiguously distinguished from Hadean zircons on either of these grounds. Thus, we conclude that the trace element compositions of Hadean zircons are permissive of models that do not include the generation of continental crust in the Hadean.

Stable oxygen isotope ratios in zircons from oceanic crust

Lower ocean crust is primarily gabbroic, although 1-2% felsic igneous rocks that are referred to collectively as plagiogranites occur locally. Recent experimental evidence suggests that plagiogranite magmas can form by hydrous partial melting of gabbro triggered by seawater-derived fluids, and thus they may indicate early, high-temperature hydrothermal fluid circulation. To explore seawater-rock interaction prior to and during the genesis of plagiogranite and other late-stage magmas, oxygen-isotope ratios preserved in igneous zircon have been measured by ion microprobe. A total of 197 zircons from 43 plagiogranite, evolved gabbro, and hydrothermally altered fault rock samples have been analyzed. Samples originate primarily from drill core acquired during Ocean Drilling Program and Integrated Ocean Drilling Program operations near the Mid-Atlantic and Southwest Indian Ridges. With the exception of rare, distinctively luminescent rims, all zircons from ocean crust record remarkably uniform d18O with an average value of 5.2 ± 0.5 per mil (2SD). The average d18O(Zrc) would be in magmatic equilibrium with unaltered MORB [d18O(WR) ~5.6-5.7 per mil], and is consistent with the previously determined value for equilibrium with the mantle. The narrow range of measured d18O values is predicted for zircon crystallization from variable parent melt compositions and temperatures in a closed system, and provides no indication of any interactions between altered rocks or seawater and the evolved parent melts. If plagiogranite forms by hydrous partial melting, the uniform mantle-like d18O(Zrc) requires melting and zircon crystallization prior to significant amounts of water-rock interactions that alter the protolith d18O. Zircons from ocean crust have been proposed as a tectonic analog for >3.9 Ga detrital zircons from the earliest (Hadean) Earth by multiple workers. However, zircons from ocean crust are readily distinguished geochemically from zircons formed in continental crustal environments. Many of the >3.9 Ga zircons have mildly elevated d18O (6.0-7.5 per mil), but such values have not been identified in any zircons from the large sample suite examined here. The difference in d18O, in combination with newly acquired lithium concentrations and published trace element data, clearly shows that the >3.9 Ga detrital zircons did not originate by processes analogous to those in modern mid-ocean ridge settings.

Mineral composition of mid-ocean-ridge basalts from ODP Leg 187 sites

The chemical compositions of olivine, plagioclase, pyroxene, and spinel in lavas collected during Ocean Drilling Program Leg 187 in the Australian Antarctic Discordance, Southeast Indian Ridge (41°-46°S, 126°-135°E) were analyzed, and modeling of the theoretical equilibrium petrogenetic conditions between olivine and melt was conducted. The cores of larger olivine phenocrysts, particularly in the isotopic Indian-type mid-ocean-ridge basalt (MORB), are not equilibrated with melt compositions and are considered to be xenocrystic. Larger plagioclase phenocrysts with compositionally reversed zonation are also xenocrystic. The compositions of primary magma were calculated using a "maximum olivine fractionation" model for primitive MORB that should fractionate only olivine. Olivine compositions equilibrated with calculated primary magma and compositions of calculated primary magma suggest that (1) isotopic Pacific-type MORB is more fractionated than Indian-type MORB, (2) Pacific-type MORB was produced by higher degrees of partial melting than Indian-type MORB, and (3) primary magma for Indian-type MORB was segregated from mantle at 10 kbar (~30 km depth), whereas that for Pacific-type MORB was segregated at 15 kbar (~45 km depth).

Whole rock and Cr-rich spinel geochemistry of ODP Legs 152 and 163 and DSDP Hole 38-338 samples

Coring during Ocean Drilling Program and Deep Sea Drilling Project Legs 163, 152, 104, 81, and 38 recovered sequences of altered basalt from North Atlantic seaward-dipping reflector sequences (SDRS) erupted during the initial rifting of Greenland from northern Europe and likely associated with excessive mantle temperatures caused by an impacting mantle plume head. Cr-rich spinel is found abundantly as inclusions and groundmass crystals within the olivine-rich lavas of Hole 917A (Leg 152) cored into the Southeast Greenland SDRS, but only rarely as inclusions within plagioclase in the lavas of the Vøring Plateau SDRS, and it is absent from other cored SDRS lavas from the Rockall Plateau and Southeast Greenland. Eruptive melt compositions determined from inferred, thermodynamically-defined, spinel-melt exchange equilibria indicate that the most primitive melts represented by Hole 917A basalts have Mg/(Mg + Fe2+) at least as high as 0.70 and approach near-primary mantle melt compositions. In contrast, Cr-rich spinels from Hole 338 (Leg 38) lavas on the Vøring Plateau SDRS give evidence for melt with Mg/(Mg + Fe2+) only as high as 0.64. This study underlines that primitive melts similar to those from Hole 917A comprise only a small fraction of the eruptive North Atlantic SDRS melts, and that most SDRS basalts were, in fact, too evolved to have precipitated Cr-rich spinel, with true melt Mg/(Mg + Fe2+) likely below 0.60. The evolved nature of the SDRS basalts implies large amounts of fractionation at the base of the crust or deep within it, consistent with seismic results that indicate an abnormally thick Layer 3 underlying the SDRS.

Modes of occurrence and geochemistry at DSDP Holes 69-505A and 69-505B

The compositions of chrome spinels of Costa Rica Rift basalts from Deep Sea Drilling Project Site 505 vary depending on their occurrences as (1) inclusions in olivine crystals, (2) inclusions in Plagioclase crystals, and (3) isolated crystals in variolitic or glassy samples. The variations are a consequence of (1) changes of melt compositions as crystallization proceeds, and (2) contrasting behavior of olivine and Plagioclase in competition with spinels for Al and Mg. Some spinels have skeletal rims compositionally less magnesian than mineral cores; however, the cores do not appear to be xenocrysts, unlike some texturally similar spinels in Mid-Atlantic Ridge basalts.

The igneous charnockite—high-K alkali-calcic I-type granite—incipient charnockite association in Trivandrum Block, southern India

The Pan-African (640 Ma) Chengannoor granite intrudes the NW margin of the Neoproterozoic high-grade metamorphic terrain of the Trivandrum Block (TB), southern India, and is spatially associated with the Cardamom hills igneous charnockite massif (CM). Geochemical features characterize the Chengannoor granite as high-K alkali-calcic I-type granite. Within the constraints imposed by the high temperature, anhydrous, K-rich nature of the magmas, comparison with recent experimental studies on various granitold source compositions, and trace- and rare-earth-element modelling, the distinctive features of the Chengannoor granite reflect a source rock of igneous charnockitic nature. A petrogenetic model is proposed whereby there was a period of basaltic underplating; the partial melting of this basaltic lower crust formed the CM charnockites. The Chengannoor granite was produced by the partial melting of the charnoenderbites from the CM, with subsequent fractionation dominated by feldspars. In a regional context, the Chengannoor I-type granite is considered as a possible heat source for the near-UHT nature of metamorphism in the northern part of the TB. This is different from previous studies, which favoured CM charnockite as the major heat source. The Occurrence of incipient charnockites (both large scale as well as small scale) adjacent to the granite as well as pegmatites (which contain CO2, CO2-H2O, F and other volatiles), suggests that the fluids expelled from the alkaline magma upon solidification generated incipient charnockites through fluid-induced lowering of water activity. Thus the granite and associated alkaline pegmatites acted as conduits for the transfer of heat and volatiles in the Achankovil Shear Zone area, causing pervasive as well as patchy charnockite formation. The transport Of CO2 by felsic melts through the southern Indian middle crust is suggested to be part of a crustal-scale fluid system that linked mantle heat and CO2 input with upward migration of crustally derived felsic melts and incipient charnockite formation, resulting in an igneous charnockite - I-type granite - incipient charnockite association.

Trace element composition of IODP Site 304-U1309 and 305-U1309 (Table A2)

IODP Site U1309 was drilled at Atlantis Massif, an oceanic core complex, at 30°N on the Mid-Atlantic Ridge (MAR). We present the results of a bulk rock geochemical study (major and trace elements) carried out on 228 samples representative of the different lithologies sampled at this location. Over 96% of Hole U1309D is made up of gabbroic rocks. Diabases and basalts cross-cut the upper part of the section; they have depleted MORB compositions similar to basalts sampled at MAR 30°N. Relics of mantle were recovered at shallow depth. Mantle peridotites show petrographic and geochemical evidence of extensive melt-rock interactions. Gabbroic rocks comprise: olivine-rich troctolites (> 70% modal olivine) and troctolites having high Mg# (82-89), high Ni (up to 2300 ppm) and depleted trace element compositions (Yb 0.06-0.8 ppm); olivine gabbros and gabbros (including gabbronorites) with Mg# of 60-86 and low trace element contents (Yb 0.125-2.5 ppm); and oxide gabbros and leucocratic dykes with low Mg# (< 50), low Ni (~65 ppm) and high trace element contents (Yb up to 26 ppm). Troctolites and gabbros are amongst the most primitive and depleted oceanic gabbroic rocks. The main geochemical characteristics of Site U1309 gabbroic rocks are consistent with a formation as a cumulate sequence after a common parental MORB melt, although (lack of systematic) downhole variations indicate that the gabbroic series were built by multiple magma injections. In detail, textural and geochemical variations in olivine-rich troctolites and gabbronorites suggest chemical interaction (assimilation?) between the parental melt and the intruded lithosphere. Site U1309 gabbroic rocks do not represent the complementary magmatic product of 30°N volcanics, although they sample the same mantle source. The bulk trace element composition of Site U1309 gabbroic rocks is similar to primitive MORB melt compositions; this implies that there was no large scale removal of melts from this gabbro section. The occurrence of such a large magmatic sequence implies that a high magmatic activity is associated with the formation of Atlantis Massif. Our results suggest that almost all melts feeding this magmatic system stays trapped into the intruded lithosphere.