208 N. L. BOWEN 
part of the column higher than that of the lower part. With this 
end in view the crucible was suspended in most cases in a part of 
the furnace where the temperature increased upward. In one 
experiment recorded here this method of eliminating convection 
was replaced by a method involving the use of a bath of molten 
gold to obtain a uniform temperature. The platinum crucible 
containing the charge was protected from the molten gold by a 
tube of silica glass.t. The silica glass was rather soft at the tem- 
perature of the experiments and required reinforcement by a tube 
of Marquardt porcelain. The one result obtained by this method 
showed no significant difference from a result obtained by the 
method of suspending the bare crucible in the furnace. There 
seemed, therefore, to be no reason for preferring the use of the 
gold bath and it was not carried farther. 
To make assurance doubly sure in the way of minimizing 
possible convection, the charge was made small. This acts in 
three ways: to make a possible lateral difference of temperature 
small, to render difficult the initiation of convection currents, and 
to decrease the velocity of possible currents. The crucibles used 
were, therefore, 5 to 6 mm. in diameter and Io to 20 mm. deep. 
In order to simplify the conditions of diffusion the crucibles were 
right circular cylinders without flare or rounded bottoms. 
The temperature was kept constant partly by continual watch- 
ing and regulation and partly by using current from a storage 
battery. In some cases an automatic regulator was used, designed 
by W. P. White of this Laboratory. 
In each case the heavy material taken was diopside. It was 
first melted and then chilled to a firm cake of glass ( 7 2.8 54) 
in the bottom of the crucible. The lighter material put in the top 
was one of the plagioclases Ab.An,, Ab,,An,, or Ab,An, (c= 
2.483, 2.533, 2.591 respectively). The temperature was in all 
cases about 1500°, that is, it was above the melting temperature 
of both layers, so that the experiments deal with diffusion of one 
liquid silicate into another. 
« The use of silica glass in this manner was suggested by J. C. Hostetter, formerly 
of this Laboratory, who also kindly worked the glass into the desired form. 
Se 
