The garnet-muscovite geothermometer was refined through empirical calibration by using natural rocks metamorphosed under the physical conditions of 238–1306 MPa and 490– 700°C. Input temperatures and pressures were determined through simultaneously applying the garnet-biotite geothermometer and the garnet-biotite-plagioclase-quartz barometer, assuming that all FeO in muscovite and garnet be ferrous. Garnet was treated as the asymmetric quaternary solid solution, and muscovite as the symmetric binary solid solution. Input muscovite compositions include Fe atoms between 0.03–0.19 and Mg atoms between 0.04–0.16 on the basis of 11 oxygen atoms, and input garnet compositions include spessartine fractions between 0.01–0.289, grossular fractions between 0.028–0.273, and the Fe/Mg ratio between 3.387–18.986. The resulting garnet-muscovite geothermometer reproduces temperatures within ±50°C compared with the garnet-biotite thermometer. Total random error of ±37°C of the new thermometer may stem from the pressure uncertainty of ±200 MPa, and uncertainties of ±5% of Fe and Mg components in muscovite, and ±5% of Fe, Mg, Mn and Ca components in garnet, altogether. When there exist 10%, 20%, 30%, 40% and 50% Fe3+ in muscovite, respectively, the computed garnet-muscovite temperatures will be 1–6°C, 2–12°C, 3–16°C, 5–24°C and 7–29°C, respectively, lower than those obtained when assuming that all FeO be ferrous. The new garnet-muscovite geothermometer can efficiently reflect temperature change of typical prograde sequences and contact aureole rocks, and may be applied to low- to high-grade and low- to high-pressure metamorphic rocks.
