[BARNES & COOKE] THE DENSITY OF ICE 149 
to be less than it otherwise would have been. But this advantage is 
ofiset by the use of mercury as the liquid, owing to the great specific 
gravity of this metal. The most serious objection to the instrument, 
however, is the fact that the walls of the inner chamber give way 
under the strain and thus form crevices. 
We now come to Bunsen’s dilatometer, the construc- 
tion of which is shown in the accompanying diagram. A 
number of the defects of the dilatometers previously 
constructed were overcome in this instrument. The vary- 
ing internal pressure, caused by the alteration of the level 
of mercury in a capillary tube, was obviated by arrang- 
ing the instrument so that the mercury would be forced 
out into a beaker, which could be afterwards weighed, 
thus maintaining the internal pressure on the instru- 
ment constant. Also the difficulty about fracturing the 
sides of the vessel during the formation of ice was elim- 
inated by starting the ice formation at the end “a” of 
the instrument and gradually extending the solidifica- | 
tion down to the juncture of the water and mercury | 
at b. But, unfortunately, the uncertainty of the cork a 
c _ : ; : - Fie. 2. Pliicker 
in Kopp’s apparatus, which difficulty was avoided by ,,4 Cons 
Plücker and Geissler, was reintroduced into this instru- Dilatometer 
ment, and may be regarded as its weak point. The apparent disad- 
vantage owing to the deformation of the walls of the instrument 
does not interfere with the accuracy of the instru- 
4 ment, as this deformation is the same before and 
i | after the formation of the ice. 
u This instrument is undoubtedly the most accu- 
(EN rate of all the dilatometers used in these determin- 
ations. But there is one possible disadvantage, 
which it shares with all dilatometers caused by the 
formation of the ice in the instrument itself. 
The last general method to be considered is 
that of weighing. This is capable of several slight 
modifications, so that it would be well to consider 
the plans adopted by Brunner and Nichols separ- 
b ately. 
Brunner’s experiments were designed primarily 
with the object of ascertaining the coefficient of 
expansion of ice between 0° and —20° C. How- 
ever, the measurements made supplied the neces- 
sary data for the computation of the density at 
Fic. 3. Bunsen’s 170 The plan is indirect, as no weighings were 
Dilatouieter actually carried on at zero, for which tempera- 






