REINTERPRETATION OF REACTION PROGRESS AS A RECORD OF THE GEOMETRY OF FLUID FLOW DURING METAMORPHISM: WATERVILLE LIMESTONE, SOUTH-CENTRAL MAINE AND SOME GENERALIZATIONS

Differences in progress (ξ) of the infiltration-driven reaction Mu + Ank + Qtz=Bt + Pl + Cal + CO₂ in the Waterville Limestone at »450°C and »3.5 kb were first interpreted in terms of reactive fluid flow in chemically isolated channels that correspond to individual cm-wide mica-rich (high-ξ) and mica-poor (low-ξ) layers (Ferry, 1987). Along a 90-m traverse across layering X_CO2 and δ¹⁸O_Cal are the same within error of measurement in 8 pairs of adjacent layers that differ in ξ by as much as 2.4±0.01 mol/L. Fluid composition evidently was homogenized by diffusion across layering at a scale >>1 cm, consistent with an earlier study (Bickle et al., 1997), and adjacent layers were not chemically isolated. The apparent paradox of cm-scale variations in ξ and m-scale cross-layer homogenization of fluid composition is simply explained in terms of layer-by-layer differences in the amounts and compositions of minerals prior to reaction (especially Ank and Pl). Significant differences in X_CO2 (up to 0.11) and δ¹⁸O_Cal (up to 3.0‰) occur between samples »10 m apart along the traverse. Reactive fluid flow therefore was indeed largely layer-parallel, but the minimum detectable size of channels is larger than previously recognized. Sampling at the dm-scale indicates channels were as small as »1 m wide. The channels may have been smaller, but cross-layer homogenization of fluid composition places a fundamental lower limit on the interrogation distance (ID) over which the geometry of fluid flow can be determined using geochemical tracers. For X_CO2 and δ¹⁸O, the ID varies from up to »10 m in Barrovian terrains (Vermont, Central Alps), up to »1 m in Buchan terrains (Maine), to <7 mm in very low-P contact aureoles (Beinn an Dubhaich, Skye). The duration of metamorphism and lithology appear to be the principal controls on the ID. Additionally, cross-layer H₂O-CO₂ interdiffusion greatly complicates estimation of time-integrated fluid flux from ξ in single samples when mineral reactants and/or products are solid solutions and rocks have variable modes and mineral compositions prior to reaction at a scale smaller than the ID. Because study sites in Vermont and Maine are entirely composed of marl, reactive fluids could not have been locally derived by dehydration of pelites with decarbonation driven by H₂O-CO₂ interdiffusion between carbonate and pelite.