010 



portion as the concentration (liminislie.s, i. e. the additive rule is />// 

 no means foUoioed in ini.vtures of o.ci/(/en and ni.tro</en. From a 

 somewhat more careful inspection, and tlie comparison with the 

 last row of the table, it appears further that with increasing dilution 

 the magnetisation coefjicient approaches to the values which satisfy 

 the inverse proportiojialiti/ irith the absolute temperature, starting fi-om 

 the number lately obtained bv Wfjss and Piccard lov gaseous oxygen 

 (see also Fig. 3). 



Without anticipating in any way the theoreti<'tjl interpretation of 

 these results, which will be treated in the next paragraph, we can 

 phenomenologically express iheni as follows: 



The deviations from Cirie-Langevin'.v law. shoiim bij pure oxygen 

 at low temperature, are not an immediate consequence of the change 

 of temperature, but are caused by the increase of the density or by 

 the distance between die molecules becoming smaller. 



Finally, let us examine more closely the thermal cltange for each 

 ■concentration, by plotting ^'y as a function of T (Fig. 2). We see at 

 once that the points obtained lie upon parallel straight lines; the 



change with the temperature can therefore be represented ^) within 

 the limits of accuracy of the observations, by a relation of the form 



^) As was found by Kamerli.ngh Onnes and Oosterhuis for liquid oxygen; Gomm. 

 W. 132e. 



