426 PROCEEDINGS OF THE AMERICAN ACADEMY. 



CO2 snow and ether. In this was suspended the steel shell with the 

 mercury, which had been previously frozen. The temperature was then 

 raised to about —42°, and readings made at intervals of 0°.5. The 

 spontaneous rise of temperature of the bath during a weighing was so 

 small as to introduce no error. During warming, the weight increases 

 linearly, due to the larger thermal expansion of the CS2 than the mer- 

 cury. Melting is indicated by a rapid drop in the weight of about 

 0.6 gm. This drop is not instantaneous, but may extend over 1°.5, 

 showing temperature lag between the mercury and the bath while the 

 heat of transformation is being absorbed. Completion of melting is 

 indicated by resumption of the linear relation between weight and tem- 

 perature. To find the discontinuity in the weight, it is merely neces- 

 sary to extrapolate this straight line back to the point at which melting 

 began. 



There are only a few special precautions to observe. The most 

 likely source of error is the formation of minute cracks in the mercury 

 when freezing. To obviate this, the mercury was poured into the steel 

 shell already filled with CS2, so as to ensure filling of all the corners, 

 and then frozen by placing merely the very bottom of the shell in the 

 freezing mixture. Freezing took place very slowly and uniformly from 

 the bottom, the surface of the finally fi-ozen mercury being as perfectly 

 convex as that of the liquid. Pains had to be taken to scrape away 

 from time to time the minute quantities of ice condensing on the 

 suspending wire. This wire was so fine that no correction whatever 

 had to be applied for the slightly varying depths of immersion during 

 the course of an experiment. Finally, it was necessary to determine 

 the density of the CS2 at the freezing temperature separately for each 

 experiment. This arose from the fact that during the preliminary 

 work a slight quantity of ether had been accidentally spilled into the 

 CS2, enough to decrease the density by 1 per cent. This gradually 

 evaporated, producing a slow rise in the density of the CS2 to a final 

 steady value. The density was given directly by the loss of weight of 

 the mercury and steel in the CS2 at the freezing temperature. There 

 seems no question whatever but that the density of the liquid mercury 

 at the freezing point is known with the accuracy demanded here, 1 

 part in 2000, although there may be reasonable question as to whether 

 its accuracy is 60 times greater, as it must be when used to give the 

 change of volume by a subtraction. The steel shell was only a small 

 part of the total weight, and its thermal dilatation is so well known and 

 so small that there is no danger at all in calculating the density of the 

 steel at —38°. 85 from the experimentally measured density at room 

 temperature. The accuracy of this procedure is vouched for by the 



