( 833 ) 

 So Hie sign of (^f\ depends on MBT q -- {Vg ~ bqY h J. or on 



\dv J g bg Vg* Vg 



d'a Xg(l — x q ) 1 — y q 2a Ay f \—y q 



d? " ~T~~ {ï+y~y ~ V 9 (T+^T Ï+//-/ 



4?/ * 

 or on (a, -f a, — 2a ls ) & (1 — x g ) — a ~f- . 



For the discussion of this last quantity we first put the firsl 

 mentioned limiting ease, in which a, and a la may he neglected with 



respect to a^, and a = a^.v 1 may be put; the value of y g being = - 



according to the table of p. 829. With' this value this quantity becomes : 



a„x \\ — x x 



2 f 3 



so positive. 



For the other limiting case for which y„ = 0, it is also positive. 



But for the intermediate cases, specially those for which a x -j- a, - a lt 



is small with respect to a, and l> x and b 3 are not equal, it will he 



negative, and shortly before its disappearance the locus ■— = 



dhp 



will lie in the region in which — is negative, and (lie existence of 



dv 1 



this locus will have little influence on the course of the spinodal 



curve, and accordingly it will not give rise to a complex plait *) 



or rather to a spinodal curve which diverges greatly from the curve 



<Ptp 



— - = 0. 



dv* 



Let us now also examine where the point in which — =0 



r/,r 2 



dhy /dn\ 



disappears, lies with respect to the curve = or to ( — =0. 



dado \dxj r 



Let us substitute the value of MRTg, x g and v g in the expression for 



— ). If this expression becomes positive, the point lies outside 

 dxj 



the curve or rather at smaller volumes than those of the curve 



( — ) = and the other way about. Then we find for : 

 \dxj v 



!) I need hardly state expressly, that in this communication I no longer attribute 

 the cause of the complexity of the plaits exclusively to the abnormal behaviour 

 of the components, to which at one time I thought I had to ascribe it, in com 

 pliance with Lehfeld. On the other hand it would be going too fur to deny the 

 abnormality of the components any influence. 



57* 



