Two tables are given showing in the first all the results which have been published concerning the density and coefficient of cubical expan- sion of ice, and in the second the same results tabulated separately according to the variety of ice used. The mean result for the density of natural ice at freezing point is 0-9176, while that of artificial ice is 0*9162 gramme per cubic centimetre. Only one estimation of the dilatation of natural ice is available. It is 0-0001125 for the cubical coefficient per degree C. The mean of three available results for artificial ice is 0-000160.
It was thought desirable to use a method of experiment which would yield a result both for the density and for the coefficient of cubical expansion, and in order that the work should have any value it was necessary to employ some device other than any which had been used previously.
The method consisted in weighing a quantity of water in mercury. The water was weighed both as liquid at C., and as solid at several temperatures below freezing point. If we assume values for the density of water and mercury at C., the density of ice at C. can then be calculated, if we also assume that the densities of ice and mercury are linear functions of the temperature. The coefficient of cubical expansion of ice can also be calculated from these results, but it will depend on the law assumed for the contraction of mercury, and upon the accuracy of the thermometry.
Instead of using a sinker to keep the vessel containing the water or ice under the surface of the mercury, a modification of Joly's Hydro- static Balance was employed.*
Ten values of the density of ice at different temperatures below C. were obtained in this way. The specimens of water were four in number, and the temperatures ranged from - 10'02 to - 0'37 C. The whole of the weighings taken with the final form of apparatus are included in these determinations.
In one experiment the values obtained show unmistakably that the same specimen of water may assume different densities on freezing.
The ten values of the density are set out as functions of the tempera- ture on a chart, and a graphical method is used to extrapolate for five values of the density of C. These five values have weights assigned to them proportional to the number of separate determina- tions of the density from which they are derived. The numbers thus obtained and the weights to be assigned to them are set out in the following table.
- July, ' Phil. Mag.,' September, 1888.
