DIFFUSION IN SILICATE MELTS BON 
magnitude of diffusion in silicate melts, which was the objective 
in mind when the work was undertaken. 
VALUES OF DIFFUSIVITY 
We cannot speak of a diffusivity “‘constant” in connection 
with the present results, but we may take the amount of diopside 
which penetrates to the surface layer as an indication of the average 
diffusivity of diopside in a liquid mixture of diopside and plagio- 
clase. In this sense we find the “average diffusivity” of diopside 
in Ab,An, mixture from Experiment No. 24 (Fig. 1), =0.015 in 
cm.” per day, in Ab,An, mixture from Experiment No. 21 (Fig. 3), 
k=o.14, and in Ab,An, mixture from Experiment No. 25 (Fig. 5), 
k=o.2. We therefore observe a progressive increase in the value 
of k with increase in the amount of anorthite in the diffusion 
mixture. Since it is well known that plagioclases rich in anorthite 
afford less viscous liquids than those rich in albite, we have further 
evidence of the increase of rate of diffusion with decrease of vis- 
_cosity. If we compare the results of Experiment No. 21 (Fig. 3) 
and 37 (Fig. 4), we find again an increase in the value of the diffu- 
sivity, being k=o.14, and k=o.3, respectively.t In both these 
examples, however, the plagioclase Ab,An, was used, the difference 
being that an increasing proportion of diopside was added. Asa 
consequence, a higher value of the diffusivity is found for the 
mixture which was richer in diopside and therefore of lower vis- 
cosity. All of the results are therefore consistent with the assump- 
tion that the diffusivity varies inversely with the viscosity, which 
is in turn dependent on composition. 
By way of comparison of the diffusivities here found with 
measured values of diffusivities for other substances, it may be 
noted that for common salt diffusing in water k=1 at 15°C., for 
gold in molten lead at 492°C., k=3, for solid gold in solid lead at 
150°C., k=o.0043 in the same units as those used above. The 
tIn Figure 4 no calculated curve is shown for the reason that the calculated 
curve for k=o.3 sensibly coincides with the observed curve. Since the final result 
will be uniformity in all cases, whether the diffusivity varies with composition or not, 
a close approach of the calculated curve to the observed curve is to be expected in 
cases where diffusion is far advanced (Fig. 4). On the other hand, where the upper 
layers have not yet been affected, the greatest divergence between observed and 
calculated values is to be expected (Figs. 1 and 2). 
