Abstract:
A survey of global ocean island basalt geochemistry shows a remarkably strong linear correlation (average rmr/rm2 = 0·95) between Zr and Nb concentrations. Fractional crystallization explains neither the range in Zr and Nb concentrations nor the fact that regression lines have intercepts significantly different from zero. A substantial, linearly correlated (average rmr/rm2 = 0·92) variability in Zr and Nb concentrations remains in the datasets after correction for fractional crystallization. Linear correlation between Zr and Nb concentrations is not expected from variable degrees of melting because the concentration of Nb is significantly more sensitive to the degree of melting than that of Zr. Our main conclusion is that the Zr/Nb systematics of ocean island basalts requires binary mixing, and more specifically mixing of two separate melts. We show that a mid-ocean ridge basalt melt is the depleted endmember, common to all examined ocean island basalts. The enriched endmember has a very variable Zr/Nb ratio. We speculate that, rather than implying the existence of several distinct enriched reservoirs, the highly variable Zr/Nb ratio could reflect intra-mantle processes such as fertilization by small-degree melts or carbonate–silicate melt immiscibility.