Zircon (ZrSiO4) is the oldest mineral on Earth, commonly found in felsic rocks and used to date rocks. Zircon provides information about continental crust formation due to its superb stability and high content of trace elements and isotopes. Zircon megacrysts of unclear origin also occur in kimberlites, the deepest mantle-derived magmas. Recent studies found zircon in other mafic and ultramafic rocks providing evidence for zircon stability in the Earth's mantle. However, its origin remains enigmatic.

To use zircon as a geochemical tracer in mantle rocks, zircon saturation and stability in mafic mantle melts should be established. This project explores the origin and survival of mantle-derived zircon and its composition to examine processes at the lithosphere-asthenosphere boundary and their relation to kimberlite magmatism.

We investigated zircon stability in melts using synthetic kimberlites with variable carbonatitic and silicic components and natural mid-ocean ridge basalt (MORB). We used natural zircon crystals (Mud Tank, Australia). The experiments at pressures 0.5–3 GPa were conducted in a piston-cylinder apparatus (Dalhousie University, Canada), experiments at 7–15 GPa employed multi-anvil apparatus (Clermont-Auvergne University, France). Compositional profiles of the samples at the melt-zircon interface were obtained using EPMA. We determined zircon saturation: Csat(MORB) = 2.0wt% (2GPa, 1400⁰C) – 5.4wt% (0.5GPa, 1400⁰C) and Csat(kimberlite) = 3.3wt% (2GPa, 1400⁰C) – 6.1wt% (1GPa, 1300⁰C); and to calculate the Zr diffusion coefficient: DZr(MORB) = 4.69·10-9cm2/s (1GPa, 1350⁰C) – 8.3·10-8cm2/s (0.5GPa, 1400⁰C) and DZr(kimberlite) = 6.56·10-8cm2/s (2GPa, 1400⁰C) – 6.3·10-7cm2/s (2GPa, 1350⁰C). Our results show that zircon stability increases with pressure. Zircon is more stable in basalt than kimberlite but less stable in silica-rich, low-carbonatitic kimberlite. The obtained Zr diffusion coefficients align with the previous estimates for basaltic melts, whereas Zr diffusivity and solubility in kimberlites are significantly lower than predicted by existing models.