DIFFUSION IN SILICATE MELTS 299 
The charge was raised quickly to the desired temperature by 
plunging it into a furnace already somewhat above that tem- 
perature. After holding it for the desired length of time the 
charge was rapidly cooled by removal from the furnace, and the 
composition of the glass at various levels in the crucible was 
determined by measuring its refractive index, the relation between 
composition and refractive index having been previously deter- 
mined on mixtures of known composition. 
It may appear that the drastic temperature differences that 
arise when the cold crucible is placed in the hot furnace would be 
bound to set up violent convection, but fortunately this is a matter 
that can easily be ascertained by running a blank test. For this 
purpose a charge was prepared in the ordinary manner and allowed 
to remain in the furnace only a few minutes, when it was removed 
and the distribution of composition determined. It was found 
that the distribution was as it should be after diffusion for a short 
period with no random variations such as would result from 
convection. 
In earlier experiments an ordinary thick-walled platinum 
crucible was used and the charge was always badly shattered in 
cooling. When not too numerous, the fragments were fitted 
together to reconstruct the original charge and the composition 
determined at various levels by removing a little powder with a 
file and determining its refractive index by the immersion method. 
The error involved in the measurement of the distance from the 
top or the bottom of the charge was large, and the results were 
only rough approximations. In later experiments the crucible 
was made of platinum foil 0.03 mm. thick. On contracting, the 
glass pulls the weak walls of the crucible with it and remains 
unshattered. The platinum foil was then peeled off and the cylin- 
der of glass was ground to a wedge whose edge was parallel to the 
axis of the cylinder. The faces of the wedge were polished and 
the refractive index of the glass at various points was determined 
on the goniometer by the method of minimum deviation. The 
exact distance of the points from the bottom of the cylinder was 
measured by means of the scale on the centering screws of the 
goniometer. 
