This paper addresses the question of the petrological relationships between the mantle section and the crustal section of the Trinity ophiolite. Our conclusions are based on a field survey and on petrographic and electron micro-probe study of about 200 samples. We show that the crustal section of Trinity is more developed and less chaotic than expected on the basis of previous surveys. In the Bear Creek area, we were able to describe a well preserved cumulate sequence about 1,500 m thick. The cumulate pile includes a thick (~800 m) basal part made of ultramafic cumulates (dunites, wehrlites, pyroxenites, etc…) displaying very thin (mm- to cm thick) modal layering. The most salient characterisitc of this basal section is the gradual decrease of the modal abundance of olivine from bottom to top. This paragenetic evolution is correlated with the evolution of mineral chemistry consistent with fractional crystallisation from a common parent melt. Plagioclase appears above this ultramafic sequence, in the upper half of the cumulate section, in a diffuse way at first (plagioclase pyroxenites), becoming increasingly abundant toward the top of the section. Its crystallization always coincides with that of hornblende pseudomorphs on previously crystallized pyroxenes. The layering becomes very irregular at this level and attributable essentially to textural variations. The top of the cumulate sequence is characterized by the abundance of magmatic breccias (pyroxenitic and gabbrodioritic fragments embedded in fine grained diorite). These breccias are cross cut by diabase dykes. The horizontal extent of the Bear Creek “magma chamber” is moderate (2–3 km). The lateral contact with the host peridotites and gabbros is always underlain by a screen of pegmatites reaching several hundred metres in thickness. These pegmatites are made of pyroxenites in the lowermost levels and of diorites in the upper levels. Angular xenoliths of mantle derived lherzolites are frequently observed in the layered ultramafic section, their incorporation being contemporaneous to the crystallization of the cumulates.
The field relationships and the lithological succession described above are consistent with the sudden injection of a huge batch of melt (reaching several km3) into the lithosphere (rocks at sub-solidus To) followed by fractional crystallization into the internal part of this magma body. The boniniticandesitic kindred of the parent melt is clearly revealed by the crystallization sequence. This conclusion is corroborated by the extreme depletion of pyroxenes and Cr-spinel in relatively incompatible elements (Ti, Al). The fractional crystallization trend of the Trinity cumulates is identical to the one defined by phenocrysts in present-day high-Ca boninites and is clearly distinct from that of mid-ocean ridge gabbros. The plagioclase composition is buffered around high An% values (90–95%), which is consistent with a low Na content of their parent melt and with H2O saturation at the time of crystallization of this mineral. The various so-called “gabbroic” massifs cropping out in Trinity represent individual intrusions similar to the one we have studied in detail in the Bear Creek area.
Two generations of melt migration structures are observed in the mantle section of Trinity: (1) ariegitic-gabbroic segregations in mineralogical and chemical equilibrium with the plagioclase lherzolite and whose injection is contemporaneous with high-To plastic deformation ; (2) pyroxenitic (and, less commonly, dioritic) segregations and dykes post-dating the high-To deformation and characterized by strong mineralogical and chemical disequilibrium with the host plagioclase lherzolite. The parent melts of these second generation segregations and dykes are identical to those of the crustal cumulates. The interaction between the boninitic melts, undersaturated in Al and ultra-depleted in incompatible elements, and the peridotites accounts for extreme mineralogical and geochemical variability of the Trinity mantle. Peridotites, away from reactive dykes, are, as a rule, richer in incompatible elements than the cumulates from the crustal section. The mantle peridotites of Trinity cannot be the source nor the residue of the melt that fed the crustal magma chambers. Accordingly, the mantle-crust complementarity argument that is the basis of the slow spreading mid-ocean ridge model for Trinity (Lherzolite Ophiolite Type), must be reconsidered.
A likely tectonic scenario that accounts for our data involves the evolution of a marginal, likely back-arc basin, from its opening to its closure. The ariegitic-gabbroic segregations are the witness of a low degree and shallow (~30 km depth) partial melting event experienced by the cold and relatively fertile Trinity peridotites during the first stage of opening of this basin in a transtensional regime, as suggested by the plastic flow pattern. The injection of the boninitic magma in strong disequilibrium with the lherzolite and feeding the crustal section occurred when one of the margins of the Trinity basin migrated above the zone of melting induced by dehydration of the subducting slab. This event occurred shortly before the definitive closure of the back-arc basin and of the obduction event. Paleomagnetic and geochronological data published so far are consistent with this scenario and with a life time of about 40 Ma for the Trinity basin, which is close to the life time of modern back-arc basins.