( 361 ) 



(lic ('iir\e ol' llie e<|iiilil>riiiiiis hehNcon llic lliiid phases, which is 

 necessarily adended by conli-aclioii. In Ihis [xtsilion the (y^ 7') curAe 

 ()(■ (he eqiiililti'iiiiiis l)el\veen solid and lliiid |)ha8cs need not show 

 the iiiaxiniuni and the miiuinuni of y>, and oidy th(> necessity ot'tlu^ 

 two vertical tangents remains. For still iargei- values of .i* and so 

 for ,r ^ .v,,^ the two curves, \> hich are dotted in (ig. 8 and touch 

 in R, and which also touch tiu^ cui'xe ol' the tliree-|)hase-iM|uilil)riiiuis 

 in the same [)oint, are separated, and the ( /y, T) cui-ve of the equili- 

 briums between solid and Huid [)hases surrounds the equilibriums 

 between the lluid phases altogether, so that the latter could only 

 appear in consequence of retardatioii of the appearance of the 

 solid phase. 



What precedes fully explains in a graphical maimer the way in 

 which the two [ji, T, .c) surfaces get detached, and it remains oidy 



to complete the discussion ot p. 240 on the course ot ^~ , which 



has not been fully carried, out there. For the determination of this 

 quantity we ha\e tiie e((uation : 



The course of the denominator in the second member, \i/,. !'"«/', 

 has been discussed p. '2'.VA. It has been })rove(l there that a locus 

 exists in the ( T, .'■) diagram, generally consisting of two branches, 

 outside which this (piantity is negatiNe. These two branches arc 

 further apart than the points B and D' (tig. 2 of the preceding com- 

 munication), and at least in the neighbourhood of the plaitpoint, also 

 further apart than the [)oints of the spinodal curve and even the 

 connodal cnr\e. It is possible and even probable that the two 

 l)ranches of tiiis locus meet. If namely the direction of the tangent 

 in the iidlection point of an isobar points just to the point Ps 

 of the tigiire 2, the two branches coincide. And whereas in 

 the point A' the direction of the tangent is j)arallel to the ?--axis in 

 the iidlectio]! point, in iidlection points more to the right of the 

 iïiobars the tangeid mentioned assumes nuM'e and more a position, 

 directed to /^. This locus, for w hicli r,/ = 0, is therefore a cur^■c 

 closed on the right, just as is the case with the i-onnodal cur\e 

 and the spinodal curxe and the curve of the [)oints /), for whicii 



—-=:(). Oiitsi<le this region >',/<^^K J^'"' inside r./'^O — only in 



that part of the region, howe\er, that lies outside the locus of the 



