MANTLE GARNETS - A CRACKING YARN

Garnets showing variation in chemical composition occur in the metasomatised peridotitic wallrocks to intrusive pyroxenitic "dikes" in mantle xenoliths from the Matsoku kimberlite pipe. They have been examined by scanning electron microscope (SEM) and imaged using both high contrast and electron channelling backscattered electron (BSE) methods. The images revealed intricate variation in the backscatter coefficient, eta, across garnets. The pattern of variation is one of very slightly diffuse "bright-white" (high-eta) lines surrounded by paler diffuse auras, grading into darker (lower-eta) areas away from the bright lines. The bright lines are usually irregular in detail and show branching and braiding; but in some cases they form regularly spaced parallel to subparallel sets which reflect crystallographic orientation. Electron and ion microprobe analyses, including a highly exhaustive 40,000 data point electron microprobe survey across one garnet, have correlated the change in the backscatter coefficient with compositional variation. The compositional changes largely involve enrichment in Fe and Ti, and decrease in Mg of the garnet forming the bright lines and pale auras, and are consistent with those of metasomatic garnets identified by previous work on an extensive suite of Matsoku xenoliths. The high Ti and Fe garnet is also enriched in Y, Zr, MREEs, and HREEs and is considered to have formed by direct crystallisation from the melt causing metasomatism, whilst areas away from the high-eta lines show compositions similar to those of garnets in unmetasomatised rock. The bright (high-eta) lines are interpreted as delineating a pattern of fractures along which the metasomatising melt was able to penetrate the garnet, and from this melt new garnet crystallised to heal the fractures and form the high Fe-Ti garnet. Diffusion of elements outward from these melt-filled and subsequently healed fractures into the main body of the garnet produced the more diffuse zonation pattern imaged by BSE methods on the SEM.