Recycled crust controls contrasting source compositions of Mesozoic and Cenozoic basalts in the North China Craton

We explore Fe/Mn and Nb/Ta ratios of basalts as potential tracers for differentiating melts of recycled mafic crustal lithologies from peridotitic melts. Trace elements and Fe/Mn ratios of the Mesozoic and Cenozoic basalts from East China were analyzed by ICP-MS. Low Nb/Ta ratios (15.4 ± 0.3 (2σ, n = 45)), high Nb and Ta contents (60.1 and 4.01 ppm) and high Fe/Mn ratios (64.7 ± 1.5 (2σ, n = 45)) characterize the <110 Ma basalts. Mesozoic basalts with ages >110 Ma are characterized by superchondritic Nb/Ta ratios (20.1 ± 0.3 (2σ, n = 25)), low Nb and Ta contents (10.8 and 0.54 ppm) and slightly lower Fe/Mn ratios (60.0 ± 1.1 (2σ, n = 25)). Both the Mesozoic and Cenozoic basalts have Fe/Mn ratios higher than basaltic melt formed by partial melting of peridotite at the same MgO and CaO levels. Although both the Mesozoic and Cenozoic basalts are characterized by highly fractionated REE patterns, the >110 Ma basalts have island arc-type trace element patterns (i.e., depletion in Nb and Ta), whereas OIB-type trace element patterns (e.g., no depletion in Nb and Ta) are characteristic of the <110 Ma basalts. Based on D_Fe/Mn values for olivine, clinopyroxene, orthopyroxene and garnet, high Fe/Mn ratios and negative correlations of Fe/Mn with Yb (Y) of the <110 Ma basalts suggest clinopyroxene/garnet-rich mantle sources. The lower Fe/Mn ratios and positive correlations of Fe/Mn with Y and Yb in the >110 Ma basalts suggest orthopyroxene/garnet-rich mantle sources. Combining these data with Sr-Nd isotopes, we present a conceptual model to explain the Nb/Ta ratios and PM-normalized trace element patterns of the >110 and <110 Ma basalts. Preferential melting of recycled ancient lower continental crust during Mesozoic lithospheric thinning resulted in (1) peridotite-melt/fluid reaction that formed the orthopyroxene/garnet-rich mantle sources for the >110 Ma basalts, and (2) peridotite + rutile-bearing eclogite mixing that formed the clinopyroxene/garnet-rich mantle sources for the <110 Ma basalts. The choice of models may indeed be arbitrary and non-unique, but the goal is to seek relatively simple forward models that explain the characteristics of the lavas, and the differences between the >110 and <110 Ma basalts, in a relatively consistent geodynamic framework.