490 PROCEEDINGS OF THE AMERICAN ACADEMY. 



plained by the experimental fact that it is impossible to superheat II 

 "with respect to III. In the course of these experiments it has never 

 been possible to carry II the slightest distance into the region of III, 

 the reaction running as inevitably as the reaction from a solid to a 

 liquid when the solid is heated to the melting point. Furthermore 

 this reaction, when it does run, runs with great velocity, since the tem- 

 peratures are sufficiently high. The change is furthermore accompa- 

 nied by a comparatively small change of volume. This evidently 

 accounts for the curve having been overlooked by Tammann. The 

 supposed crossing of the two curves I-II and I-III found by Tammann 

 is to be explained by slight errors in the pressure measurements. The 

 greatest difference between the curves is 30 kgm. This is the second 

 bit of evidence alluded to on p. 482, that Tammann's absolute pressure 

 measurements may be in error. The turning point of the curve found 

 by Tammann is at the same temperature as that found here, —35°. 

 The impossibility of superheating II also explains Tammann's inability 

 to find points on the II-L curve. Apparently it is impossible to reach 

 this. II appears to be surrounded on all sides by other solid phases. 



The behavior of III with respect to II is not the reverse of II 

 with respect to III. It is possible to subcool III greatly with re- 

 spect to II. The amount of subcooling possible depends on a va- 

 riety of factors. One of the most important of these is the interval 

 of time which has elapsed since the water had previously existed in 

 the form 11. It is a fact verified repeatedly on all the other curves as 

 well as on this particular one, that it is very much easier to obtain a 

 phase after it has once been present in the apparatus. Thus it is pos- 

 sible to subcool III at least as far as —70° with respect to II, provided 

 that II has never been formed. But if II has been formed previously, 

 the reaction from III to II is very likely to run on passing from 

 the III to the II region. This explains why it is possible to move up 

 along the II-I curve, pass to the I-III curve, and then on decreasing 

 temperature pass again to the I-III curve. 



The points obtained experimentally have been classified with these 

 facts in view. The points of March 25, 29, and April 5 obtained at 

 temperatures above —35° after coming from lower temperatures on 

 the I-II curve have been listed on the I-III curve in the tables. 



The points are plotted in Figure 9. These points are the mean of 

 the points found with increasing and decreasing pressures at tempera- 

 tures very close to each other. "Within the limits of error the points 

 lie on a straight line. The two lowest points are irregular because of 

 the extreme slowness of the reaction. It is interesting to notice that 

 the line as drawn is heading straight for the absolute zero, the equilib- 



