Excesses of Th-230 over U-238 in mid-ocean ridge basalts (MORB) require tha
t the mantle source region preferentially retains U over Th during partial
melting. Based on existing mineral-melt partitioning data, Th-230 excesses
are widely cited as evidence that partial melting beneath ridges begins wit
hin the garnet stability field, at pressures over 2.8 GPa. However, recent
experimental and theoretical studies of U-Th partitioning show that melting
in the presence of aluminous mantle clinopyroxene may also generate rh-exc
ess. In order to try to distinguish between these models we sought basalts
with independent constraints on their depth and extent of partial melting.
We report data from alkali basalts from the Antarctic Peninsula whose tecto
nic setting indicates that they formed by < 6% partial melting at pressures
of < 2 GPa, well within the spinel stability field. Their major and trace
element chemistry is best modelled by similar to 4% partial melting at pres
sures of 1-2 GPa, in excellent agreement with that inferred from the plate
structure. However, these rocks preserve large (6-26%) 230 Th-excesses, whi
ch would conventionally be ascribed to the involvement of garnet. Instead w
e show that the trace element signature and isotopic data can be reconciled
with partial melting involving residual aluminous-clinopyroxene within the
spinel stability field. These Antarctic Peninsula basalts provide the firs
t observational evidence that significant 230 rh-excesses can be produced b
y partial melting of spinel Iherzolite and challenge the perceived importan
ce of garnet in MORE petrogenesis. (C) 2000 Elsevier Science B.V. All right
s reserved.