We report on the major element composition of basaltic glasses from the Mid-Atlantic Ridge transecting the equatorial mega-fracture zones from 7¿S to 5¿N (65 stations, 10--20 km sampling intervals, 3.5--5 km water depth range). Many of the basaltic glasses are Na2O, SiO2, and MgO rich, similar to other basalt glasses erupted along the deepest regions of the mid-ocean ridge system, suggesting melt generation by relatively low degrees of partial melting at rather shallow depth in the upper mantle. Along the ridge axis, the compositional variations show regular and systematic long-wavelength trends with a major discontinuity at the complex St. Paul transform fault, just south of St. Peter and Paul islets. A corresponding long-wavelength trend in upper mantle potential temperature, mean pressure, and degree of melting and crustal thickness variation is inferred using parameterized petrologic decompression melting models. A 600-km-long, nearly linear negative gradient in these parameters is apparent from the Charcot fracture zone (FZ) to the St. Paul FZ. Over the length of this gradient, the upper mantle potential temperature drops by about 70¿C, the mean degree of partial melting changes from 7% to 10%, and the inferred crustal thickness varies between 3 and 6 km.

The gradient along the ridge axis is unaffected by the mega-transform fault offsets, implying that a broad (approximately 2000 km wide across-axis and 600 km long along-axis) cold zone is present in the upper mantle just south of the equator. At the discontinuity across the complex St. Paul transform fault, the gradients in inferred potential temperature, mean degree of partial melting, and crustal thickness abruptly change sign, respectively increasing by 80¿C, rising from 7% to 10%, and changing from 3 to 6 km. The discontinuity is clearly related to the Sierra Leone plume affecting the Mid-Atlantic Ridge around 1.7¿N, as also evident from Pb, Nd, and Sr isotopic variations previously reported on the same glasses (Schilling et al., 1994) and the K2O variation reported here. The cause of the petrologically inferred cold zone and large gradient in the upper mantle south of St. Peter and Paul islets remains more speculative.

On the basis of a passive mantle upwelling flow model (Phipps Morgan and Forsyth, 1988) applied to the specific geometry of the equatorial Atlantic, we reject the simplest hypothesis that the cold zone is produced by the compounding cooling effect caused by the very long and densely distributed transform fault offsets in the equatorial Atlantic. The result of this test remains paradoxical in view that good correlations exist between segment length, maximum along-ridge axis relief per segment, mean segment depth, and per segment average bulk compositions of the erupted basalts, and corresponding mean degrees of melting. Other possible causes for the gradational cold zone are briefly explored. These include the evolutionary history of the region with respect to adjacent continental mantle, lithosphere age, thickness, and temperature, and tectonic mode of opening of the Atlantic, as well as large-scale convective motion associated with continental dispersion. No definite conclusions can be reached. However, we emphasize that the petrologically inferred upper mantle thermal structure in the equatorial Atlantic is quite robust and independent of the petrologic decompression melting models considered and their underlying detailed assumptions.

Large seismic S wave velocity variations are predicted over the 0--150 km depth range of the upper mantle, based on the reported correlation of Na8 with S wave velocity reported by Yan et al. (1989). Thus detailed seismic tomographic mapping could be used to test further the cold upper mantle zone hypothesis for the equatorial Atlantic. ¿ American Geophysical Union 1995