OF CUBICAL EXPANSION OF ICE. 
481 
mercury used, but these errors are probably such as will not affect the result to the 
accuracy with which it is given below. Tire four values which can Ije found from 
the data available are set out in Table VI. 
Table VI. 
Experiment. 
1 
Coefficient of cubical expansion. 
•000155 
3 
4 
•000152 I 
I 
•000153 I 
•000148 J 
Mean ^000152 
Comparison of Results. 
The value '9160 for the density of ice at 0° C., is lower by two parts in 10,000 
than the mean of the results obtained by Plucker and Geissler, Bunsen, and 
Nichols. It is 1 part in 10,000 less than the mean of Nichols’s values, but is 7 
])arts in 10,000 lower than Bunsen’s value. The value '000152 for the coefficient of 
cubical expansion is 4 per cent, lower than that of Plucker and Geissler, the last 
puldished value for the directly determined cubical coefficient of artifical ice. It is 
5 per cent, lower than the mean value given in Table II. 
Conclusion. 
The results of this determination of the density and coefficient of cubical expansion 
of ice are, that Nichols’s value for the density is confirmed, and tliat Bunsen’s value 
is probably too high ; but as the same specimen of water can freeze into specimens 
of ice having different density, the use of the Bunsen ice calorimeter in absolute 
determinations must be limited to an accuracy of probably about 1 in 1,000. 
’fhe coefficient of cubical expansion seems to lie 4 or 5 per cent, less than the mean 
of previous determinations. 
The expenses of this research have been in part defrayed liy a Government grant 
from the Royal Society, and in part by the Cavendish Laboratory. I wish to thank 
Professor J. J. Thomson, F.R.S., for his kind encouragement, and my thanks are 
also due to Mr. Griffiths, F.R.S., through whom I was led to undertake the 
investigation. 
o 
VOL. CXCVIII.-A. 
