132 ROYAL SOCIETY OF CANADA 
The ideal ice, so far as normal density is concerned, would be 
the ground or anchor ice, formed along the beds of small streams 
during cold, clear nights, when the surface has not frozen over. In 
this formation, the heat escapes by radiation from the bottom and the 
ice always remains at the freezing temperature, and so cannot undergo 
any strains due to unequal expansion and contraction. 
It seems very probable that ice recovers in time from any internal 
strains set up during formation, and for this reason, in cases where 
the density alters with time, it is likely that the true normal value 
is being gradually approached. The theory of the viscosity of mater- 
ials would lead one to anticipate this change in the presence of internal 
strains. 
The following facts, taken from the observations of varicus 
experimenters, seem to the writer to have a direct bearing on this 
explanation of the variation in the density of ice. 
The interior formation was used both by Plücker and Geissler, 
and by Kopp. The result of three determinations made by the former 
experimenters gave the value 0:91580 + -000008, while Kopp 
obtained 0:9078 + -0007. Kopp’s value, however, is unreliable, as 
his dilatometer was of very inferior construction, and the results 
were further vitiated by the presence of an air bubble in every speci- 
men of ice experimented with. 
The only experimenter who has used the second formation —viz., 
ice-mantles,—in his determinations, is Professor Nichols. In his paper 
he gives several abnormally high results, which are, however, unim- 
portant as far as the absolute value is concerned, as they were obtained 
by using dilatometers containing large quantities of mercury, which 
caused a considerable deformation in the walls of the vessel. But 
there are several points in the account of his experiments which seem 
to have a decided bearing on the subject in question. He says that 
certain preliminary experiments made by him “appear to indicate 
that mantles formed by the use of alcohol as a refrigerant at —5° to 
—10° are less dense than those frozen by ether and CO, at —70° by 
at least one part in a thousand. It was likewise found that one of 
the latter mantles upon standing twenty-four hours in an ice bath, 
appeared to have lost density by nearly the same amount.” This is 
precisely what one would be led to expect, as the amount of the inter- 
nal strains depends on the temperature gradient of the ice at the 
time of formation of the different layers; and also the ice would tend 
to recover from these strains very quickly when kept at 0° C., as its 
elasticity is a minimum at this temperature. Professor Nichols also 
observes that it was impossible to obtain ice-mantles of more than a 

