James Kelly Russell, Kai-Uwe Hess , Donald B. Dingwell

A non-Arrhenian model for the Newtonian viscosity (η) of ultramafic melts is presented. The model predicts the viscosity of ultramafic melts as a function of temperature (T), pressure (P), H2O content and for a range of melt compositions (70 < Mg# < 100). The calibration consists of 63 viscosity measurements at ambient pressure for 20 individual melt compositions and 5 high-P measurements on a single melt composition, all drawn from the literature. The data span 14 orders of magnitude of η (10-2 to 1011.8 Pa s), a T range of 880 to 2700K, pressures from 1 atm to 25 GPa, and include measurements on hydrous melts containing 0.2 to 4.4 wt% H2O. The T-dependence of viscosity is modelled with the VFT equation [log η = A + B / (T(K) − C)] whereby A is assumed to be a common, high-T limit for these melt compositions (i.e. log η= -5.4). The pressure and composition effects are parameterised in terms of 6 adjustable parameters in expanded forms of B and C. The viscosity model is continuous across T-P-composition space and can predict ancillary transport properties including glass transition temperatures (Tg) and melt fragility (m). Melt viscosity decreases markedly with increasing H2O content but increases significantly with increasing pressure and decreasing Mg# (i.e. higher Fe-content). We show strong systematic decreases in Tg and m with increasing H2O content whereas an increase in P causes a rise in Tg and decrease in m. The predictive capacity of this model for ultramafic melt viscosity makes it pertinent to the fields of volcanology, geophysics, petrology, and the material sciences. Moreover, it provides constraints on models of magma oceans on terrestrial planets and, the evolution of planetary atmospheres via magmatic degassing on exoplanets.