The next table shows how for carbon dioxide with oxygen, at a tem- 
perature of about 31° C., the mean density being the critical density 
and zz, — 0.001, the difference in density depends on the mean 
2 
composition. 
4 (v, + 4) = 0.0005 A= 36"), 
0.005 WY, 
0.01 12 
0.015 6 
All these numbers relate to carbon dioxide with oxygen as admixture ; 
it is probable that these results will also be more or less applicable 
to carbon dioxide with nitrogen, hence also with air, and as in 
carbon dioxide, which had been purified with great care, KrEsom 
detected about 0.00025 mol. of air, the possibility is not excluded of 
explaining the anomalies observed with carbon dioxide, by impurities 
of air. 
The variation of the difference in density with the mean density 
reminds of a diagram concerning DE Hren’s experiments, formerly 
made by me (ef. Comm. N°. 68, Appendix p. 26; Proc. April 1901, 
p. 695); in Comm. N°. 68, Appendix p. 22 (Proc. April 1901 p. 691) 
KAMpRLINGH Onnes has derived the same diagram for the course of 
the differences in density that would result from differences of tempe- 
rature; therefore part of the deviations observed by pr HEEN are 
perhaps due to differences of temperature. 
$ 8. Survey of the experiments of Tuicuner. In the influence of 
impurities we have a complete qualitative explanation of TricHNnEr’s 
observations. The results of his second series of observations, of 
which I have used only those above the critical temperature, are 
represented in fig. 1. The positions of the floats are indicated on 
vertical lines and the points occupied by the same bulb at different 
temperatures are combined by lines. In this manner curves of equal 
densities are obtained; for each curve I have given the corresponding 
density. In this series of experiments TricHNer has first made obser- 
vations at gradually increasing, and then at decreasing temperatures; 
after each variation of temperature the observer waited till the 
temperature had become the same throughout. As abscissae I have 
not taken the temperatures themselves, but I have placed the different 
observations at equal distances, that is to say, I have taken time 
as abscissa, thus assuming that between two observations there is 
always the same interval of time, which will not probably be far 
wrong. The temperature 282°.0 C. (uncorrected). is that at which 
