Analysis of experimental data on divalent cation diffusion kinetics in aluminosilicate garnets with application to timescales of peak Barrovian metamorphism, Scotland

• Published in: Contributions to mineralogy and petrology 2015-08, Vol.170 (2), p.1-27, Article 25
• Main Authors: Chu, Xu, Ague, Jay J.
• Format: Article
• Language: English
• Subjects: Cations; Diffusion coefficient; Earth and Environmental Science; Earth Sciences; Experimental data; Geochemistry; Geology; High temperature; Kinetics; Low temperature; Markov chains; Markov processes; Metamorphism; Metamorphism (Geology); Mineral Resources; Mineralogy; Monte Carlo method; Original Paper; Petrology; Rocks, Metamorphic; Statistical analysis; Water content
• ISSN: 0010-7999; 1432-0967
• DOI: 10.1007/s00410-015-1175-y

We present a new statistical framework to analyze the diffusion data for divalent cations (Fe, Mg, Mn, and Ca) in aluminosilicate garnet using published experimental data and an Arrhenius relationship that accounts for dependence on temperature, pressure, garnet unit-cell dimension, and oxygen fugacity. The regression is based on Bayesian statistics and is implemented by the Markov chain Monte Carlo approach. All reported experimental uncertainties are incorporated, and the data are weighted by the precision of the experimental conditions. We also include a new term, the inter-experiment bias, to compensate for possible inconsistencies among experiments and to represent any experimental variability not explicitly presented in the Arrhenius relationship (e.g., water content, defect density). At high temperatures where most experiments were conducted, the diffusion coefficients calculated with the new parameters agree well with previous diffusion models (e.g., Chakraborty and Ganguly in Contrib Mineral Petrol, 111:74–86, 1992 ; Carlson in Am Mineral, 91:1–11, 2006 ). However, the down-temperature extrapolation leads to notable differences at lower temperatures for common petrological applications. For example, at 600 °C, the diffusion coefficients of Fe and Mn are one half-to-one order of magnitude faster and the diffusion coefficient of Ca is about one order of magnitude slower than calculated with parameters in Carlson (2006) . Our statistical analysis also provides well-defined uncertainty bounds for timescale estimates based on garnet diffusion profiles. Application of the newly derived coefficients indicates that the timescale of the thermal peak of Barrovian metamorphism (Dalradian belt of Scotland) is about four to seven times longer than that estimated using previously published diffusion coefficients. The peak is still geologically brief, however—of the order of 10 6 years (0.75 Myr +0.70/−0.36 Myr; ±1σ). This brevity requires pulsed advective heat input, as provided by syn-orogenic mafic magmatism in these rocks.