Several series of experiments were carried out at 1 GPa and 900¿-1100¿ in which grain boundary chemical diffusion of oxygen in dry quartzite and dunite was characterized by use of ''indicator mineral'' techniques. Most experiments involved mixing a small amount of Fe2O3 with coarse quartz or olivine powder, placing the mixture in an iron container, pressurizing in a piston-cylinder apparatus, and heating rapidly to the desired run temperature. In response to the large gradient in oxygen fugacity resulting from the juxtaposition of iron metal (i.e., the container) with Fe2O3 disseminated through the sample, oxygen transport along quartz or olvine grain boundaries took place, causing reduction of ferric iron in regions of the sample nearest the container (producing w¿stite in the dunite experiments and fayalite in those on quartz).

The flux of oxygen, and hense the diffusivity, could be calculated from the width of the observed reduction zone. Variations on the method just described included use of CuO rather than Fe2O3 as the ''indicator mineral,'' use of graphite rather than iron as the container, and placement of the Fe2O3 as a single pellet at the center of the dunite or quartzite sample rather than as a dispersed minor phase. For the dispersed-Fe2O3 experiments on dunite at 1000¿ and 1100¿C, the width of the reduced (w¿stite-bearing) zone was found to be generally proportional to the square root of run duration, in accordance with a diffusion-controlled transport/reduction process.

At 900¿, however, this relation does not hold, and the data suggest vapor phase transport of oxygen prior to closure of sample porosity. All experiments on quartzite (900¿--1100¿) were plagued by the same problem, but in this case, sufficiently long experiment durations eventually resulted in no further measurable oxygen transport, demonstrating that well-annealed quartzite is impermeable to oxygen at the run conditions. The net outcome of the experiments is a bulk grain boundary diffusivity for oxygen in dunite given approximately by D=106 exp (-355,500/RT) (D in cm2/s; R in J/deg mol). Values for quartzite are too low to measure by the techniques attempted. The information can be used to argue that heterogeneities in oxygen fugacity on the scale of centimeters to meters can be maintained indefinitely in fluid-absent, high-grade rocks of the continental crust, but similar features in a fluid-absent mantle would be quickly erased.