Why do kimberlites from different provinces have similar trace element patterns?

Abstract

Analysis of the trace element contents in kimberlites from various provinces around the world, including South Africa, India, and Yakutia (Siberia, Russia), reveals remarkable similarity of the maximum abundances. In addition, we find that abundances of the rare earth elements (REE) in the South African kimberlites are highly coherent between individual elements. We suggest that the observed similarity of the trace element patterns may result from a common physicochemical process operating in the kimberlite source region, rather than from peculiar source compositions and magmatic histories. The most likely candidates for such a process are (1) partial melting at very low melting degrees and (2) porous melt flow and diffusive exchange with the host rocks. These two processes can produce the same maximum trace element abundances and similar undersaturated patterns. We argue that the porous flow, and the associated chromatographic enrichment, is preferred because it allows high saturations at relatively large melt fractions of ∼1%. Observations of enrichment of the xenolith grain rims due to an exchange with metasomatizing melts of quasi‐kimberlitic composition imply that the melt percolated beyond the source region, in agreement with basic assumptions of the percolation model. We demonstrate that the saturated REE patterns are in a good agreement with the maximum observed REE abundances in kimberlites from different provinces. The theoretical patterns are independent of the melt fraction and only weakly (if at all) depend on the source modal composition. Characteristic diverging fan‐like patterns of trace elements predicted by the percolation model are identified in kimberlites from South Africa. We propose that a high coherency of the REE patterns in the South African kimberlites results from a general dependence of all REE abundances on the calcium content. According to this interpretation, the overall depletion of the source rocks in REE with temperature (and depth) postulated by our model is a natural consequence of a decrease in the calcium content along the lherzolite trend.