350 



Messrs. J. C. McConnel and D. A. Kidd. [June 21 r 



The temperature on the afternoon of the 30th was not taken, but 

 the notebook contains a statement that it was colder than the 

 morning. 



Since the box was left open all night, the temperature given by the 

 thermometer on the morning of the 31st may well have been rather 

 lower than that of the ice. Between the 1st and 2nd, an apparent 

 contraction of 0*017 mm. on one side took place without change of 

 temperature. This looks as if the indicator had slipped. Making 

 allowance for these, the mean extension from the 31st, 9 h. 15 m., to the 

 5th, 16 h., follows the temperature very fairly, considering the uncer- 

 tainty of the latter. We have arranged the table to show this. But 

 during the first six hours there was an expansion on one side of 

 0'088 mm. in actual magnitude, which we attribute to a slight yield- 

 ing at the crack. Counting the contraction as a slip, and making no 

 allowance for temperature, the mean rate during the whole 150 hours 

 was 0*00019 mm. per hour per length of 10 cm. 



If we suppose that the extension during the last five days was 

 entirely due to temperature, and that the coefficient of expansion of 

 the glass of the connecting fibre was 0*000009, we have between 



— 12*5° and —8' 5° a coefficient of linear expansion of ice of 0*000034,. 

 between -8*5° and -3*7° of 0*000060, and between —3*7° and -1*0° 

 of 0*000009. 



Into the complicated question of the expansion of ice with tem- 

 perature we do not care to enter fully. We will merely cite two- 

 investigations. The best observations on the cubical coefficient seem 

 to be those of Pettersson (" On the Properties of Water and Ice,' r 

 ' Vega Expedition,' vol. 2, Stockholm, 1883). We deduce from his 

 figures the corresponding linear coefficients, supposing ice to be 

 isotropic in this matter. With ice from ordinary distilled water he 

 obtained 0*000053 between -12° and -2°. This ice began to con- 

 tract at some point between —0*35° and —0*25°. With ice from the 

 purest water he could obtain, the coefficient rose from 0*000055 

 between -17° and —10° to 0*000057 between -4° and -3°, and 

 then decreased, till it changed sign at some point between —0*15° and 



— 0*03°. Ice containing 0*014 per cent, of chlorine, in the shape of 

 salts, began to contract at —2*5°. In these experiments the water 

 was frozen in the dilatometer, so there was no chance of the impuri- 

 ties being expelled by the process of solidification as in the case of 

 ice formed slowly on the surface of some depth of water. His purest 

 water, however, was so good as to be seriously affected by boiling 

 for a short time in a clean glass vessel. 



The coefficient of linear expansion has been determined directly by 

 Andrews ('Roy. Soc. Proc.,' June, 1886). He found 0*0000505 

 between -18° and -9°, and 0*0000735 between -9° and -0°. It 

 is possible that the difference between the determinations of these two 



