Multispecies diffusion of yttrium, rare earth elements and hafnium in garnet

We report experimental data for Y, La, Lu and Hf diffusion in garnet, in which diffusant concentrations and silica activity have been systematically varied. Experiments were conducted at 950 and 1050 °C, at 1 atm pressure and oxygen fugacity corresponding to the quartz–fayalite–magnetite buffer. At Y and REE concentrations below several hundred ppm we observe both slow and fast diffusion mechanisms, which operate simultaneously and correspond to relatively high and low concentrations, respectively. Diffusivity of Y and REE is independent of silica activity over the studied range. General formulae for REE diffusion in garnet, incorporating data from this and previous studies, are logD_REE(f) (m² s^-1) = -10·24(±0·21) - 221057±(4284)/2·303RT (K) for the ‘fast’ REE diffusion mechanism at 1 atm pressure, and log_{DREE(s)(m2 s-1)} = -_{9·28(±0·65)-} 265200(±38540)+ 10800(±2600) P (GPa) 2·303RT (K) for the ‘slow’ REE diffusion mechanism. These slow and fast diffusion mechanisms are in agreement with previous, apparently conflicting, datasets for REE diffusion in garnet. Comparison with high-pressure experiments suggests that at high pressures (>~1 GPa minimum) the fast diffusion mechanism no longer operates to a significant degree. When Y and/or REE surface concentrations are greater than several hundred ppm, complex concentration profiles develop. These profiles are consistent with a multi-site diffusion–reaction model, whereby Y and REE cations diffuse through, and exchange between, different crystallographic sites. Diffusion profiles of Hf do not exhibit any of the complexities observed for Y and REE profiles, and can be modeled using a standard (i.e. single mechanism) solution to the diffusion equation. Hafnium diffusion in garnet shows a negative dependence on silica activity, and is described by logD_Hf(m² s^-1) = 8·85(±0·38) - 299344(±15136) + 12500(±900) P (GPa) -0·52(±0·09) log₁₀aSiO₂: 2·303RT (K) In many natural garnets, diffusion of both Lu and Hf would be sufficiently slow that the Lu–Hf system can be reliably used to date garnet growth. In cases in which significant Lu diffusion does occur, preferential retention of ¹⁷⁶Hf/¹⁷⁷Hf relative to ¹⁷⁶Lu/¹⁷⁷Hf will skew isochron relationships such that their apparent ages may not correspond to anything meaningful (e.g. garnet growth, peak temperature or the closure temperature of Lu or Hf). Late-stage reheating events are capable of causing larger degrees of preferential retention of ¹⁷⁶Hf/¹⁷⁷Hf relative to ¹⁷⁶Lu/¹⁷⁷Hf and partial to full resetting of the Sm–Nd system within garnet, thus increasing the separation between garnet Lu–Hf and Sm–Nd isochron dates, owing to the fact that these systems are more significantly disturbed through diffusion as more radiogenic ¹⁷⁶Hf and ¹⁴³Nd have accumulated.