Divergent T–ƒO₂ paths during crystallisation of H₂O-rich and H₂O-poor magmas as recorded by Ce and U in zircon, with implications for TitaniQ and TitaniZ geothermometry

During solidification of magma chambers as systems closed to chemical exchange with environs, the residual siliceous melt may follow a trend of rising, constant, or decreasing oxidation state, relative to reference buffers such as nickel + nickel oxide (NNO) or fayalite + magnetite + quartz. Titanomagnetite–hemoilmenite thermometry and oxybarometry on quenched volcanic suites yield temperature versus oxygen fugacity arrays of varied positive and negative slopes, the validity of which has been disputed for several years. We resolve the controversy by introducing a new recorder of magmatic redox evolution employing temperature- and redox-sensitive trace-element abundances in zircon. The zircon/melt partition coefficients of cerium and uranium vary oppositely in response to variation of magma redox state, but vary in tandem as temperature varies. Plots of U/Pr versus Ce⁴⁺/Ce³⁺ in zircon provide a robust test for change in oxidation state of the melt during zircon crystallisation from cooling magma, and the plots discriminate thermally induced from redox-induced variation of Ce⁴⁺/Ce³⁺ in zircon. Temperature-dependent lattice strain causes Ce⁴⁺/Ce³⁺ in zircon to increase strongly as zircon crystallises from cooling magma at constant Ce⁴⁺/Ce³⁺ ratio in the melt. We examine 19 zircon populations from igneous complexes in varied tectonic settings. Variation of zircon Ce⁴⁺/Ce³⁺ due to minor variation in melt oxidation state during crystallisation is resolvable in 11 cases but very subordinate to temperature dependence. In many zircon populations described in published literature, there is no resolvable change in redox state of the melt during tenfold variation of Ce⁴⁺/Ce³⁺ in zircons. Varied magmatic redox trends indicated by different slopes on plots of zircon U/Pr versus Ce⁴⁺/Ce³⁺ are corroborated by Fe–Ti-oxide-based T–ƒO₂ trends of correspondingly varied slopes. Zircon and Fe–Ti-oxide compositions agree that exceptionally, H₂O-rich arc magmas tend to follow a trend of rising oxidation state of the melt during late stages of fluid-saturated magmatic differentiation at upper-crustal pressures. We suggest that H₂ and/or SO₃ and/or Fe²⁺ loss from the melt to segregating fluid is largely responsible. Conversely, zircon and Fe–Ti-oxide compositions agree in indicating that H₂O-poor magmas tend to follow a T–ƒO₂ trend of decreasing oxidation state of the melt during late stages of magmatic differentiation at upper-crustal pressures, because the precipitating mineral assemblage has higher Fe³⁺/Fe²⁺ than coexisting rhyolitic melt. We present new evidence showing that the Fe–Ti-oxide oxybarometer calibration by Ghiorso and Evans (Am J Sci 308(9):957–1039, 2008) retrieves experimentally imposed values of ƒO₂ in laboratory syntheses of Fe–Ti-oxide pairs to a precision of ± 0.2 log unit, over a large experimental temperature range, without systematic bias up to at least log ƒO₂ ≈ NNO + 4.4. Their titanomagnetite–hemoilmenite geothermometer calibration has large systematic errors in application to Ti-poor oxides that precipitate from very oxidised magmas. A key outcome is validation of Fe–Ti-oxide-based values of melt TiO₂ activity for use in Ti-in-zircon thermometry and Ti-in-quartz thermobarometry.