We have used a variety of different experiments to document isotopic fractionation by diffusion in liquid silicates. The initial set of experiments involved measuring the mass dependence of isotope mobility (or equivalently the self diffusion coefficient) in isothermal, isochemical diffusion couples (see Richter et al., 1999). The couples were made by juxtaposing glasses of the same chemical composition that had been doped isotopically to produce a large concentration step in the two isotopes

more »

... ose mobility we wanted to compare (e.g. *Oca vs %a) but with the same isotopic ratio (e.g.'8Ca/"oCa) in each half of the couple. The couples were heated to temperatures sufficient to melt the glasses and held at high temperature for sufficient time for diffusion on a measurable length scale to take place. Given the initial conditions, any significantly greater mobility of *Oca relative to **Ca would manifest itself as low (high) %daCa in that part of the couple that *Oca and 48Ca diffused into (out of). As shown in Figure 2 , this is exactly what we found. Figure 2. Ca isotope fractionation by diffusion in a Ca0-AI2O3-SiO2 glass recovered from an experiment run for 1.5 hours at 1500'C and 1.OGpa (from Richter et al., 1993) Initial conditions were a large step in both 'Oca and aCa (high concentrations xO). The locally low isotopic ratios ("CaPCa-0.250) centered on x=1000pm are explained by 'Oca diffusing faster than 48Ca by an amount D&,p(48/40)@ with B= 0.1. Error bars for each data point are la standard deviations based on 20 measurements of the &C@Ca ratio. Given results such as those shown in Figure 2 it became quite clear that chemical diffusion can indeed fractionate isotopes of elements such as Ca by measurable amounts We also carried out experiments using germanium isotopes in molten GeO, (a lower