The grain-scale signature of isotopic diffusion in ice

Abstract. Diffusion limits the survival of climate signals on the water stable isotopes in ice sheets. Diffusive smoothing acts not only on annual signals near the surface, but also on long-timescale signals at depth as they shorten to decimetres or centimetres. Short-circuiting of the slow diffusion in crystal grains by fast diffusion along liquid veins can explain the “excess diffusion” found on some ice-core isotopic records. But experimental evidence is lacking as to whether this mechanism operates as theorised; theories of the short-circuiting also under-explore the role of diffusion along grain boundaries. The non-uniform patterns of isotopic deviation δ across crystal grains induced by short-circuiting offer a testable prediction of these theories. Here, we extend the modelling for grain boundaries (and veins) and calculate these patterns for different grain-boundary diffusivities and thicknesses, temperatures, and vein-water flow velocities. Two isotopic patterns are shown to prevail in ice of millimetre grain size: (i) an axisymmetric “pole” pattern with excursions in δ centred on triple junctions, in the case of thin, low-diffusivity grain boundaries, and (ii) a “spoke” pattern with excursions around triple junctions showing the impression of grain boundaries, when these are thick and highly diffusive. The excursions have widths ∼ 10 %–50 % of the grain radius and variations in δ ∼ 10−2 to 10−1 times the bulk isotopic signal for oxygen and deuterium, which set the minimum measurement capability needed to detect the patterns. We examine how the predicted patterns vary with depth through a signal wavelength to suggest an experimental procedure, based on laser ablation mapping, of testing ice-core samples for these signatures of isotopic short-circuiting. Because our model accounts for veins and grain boundaries, its predicted enhancement factor (quantifying the level of excess diffusion) characterises the bulk-ice isotopic diffusivity more comprehensively than past studies.