f 521 ) 



be negative, the quantities Q— j and (-~\ always positive. For the 



latent heat of mixing q^ and q^ can never become negative. 



So in the ideal case « z= 0, «' ==0 the meltingpoint-curve always 

 begins to descend at the highest temperature, and to ascend at the 

 Unrest temperature, so that in this case a mininnim is excluded. 

 This appears also from the fact that the condition for a minimum 



2\ — T 

 is |5'>-^- (loc. cit. p. 168), so that for /?' =r this can never 

 -^ 1 



occur, and the meltingpoint-curve will therefore gradually descend 

 from 7\ to T,. 



That a maximum cannot occur in am/ case for normal components, 

 whatever value a or «' may have, — provided a' be larger than a — 

 has been pro\'ed already in my first communication (loc. cit. p. 156). 



The equatioiis (5) and (5r^) give rise to the following discussion. 



.... rdT\ 



In the limiting case q^ = {q^ hnite) we have f - — = 



00, 



rdT 



T\ fdT\ /dT\ 



r,]^~^-'> \-r~ ] — ^^ "7~J = ^^ s<^> tliat the Uxo meltingpoint- 



\d 



curves will approach to the type A (fig.l). 



Ii^-± 



'n 



Ija^L 



{dT\ fdT\ fdT\ 



For q, = cc, -7- and - will a])proach to 0, -— to 00 (on 



account of tiio term e''), but { -,] to a /iinit, viz. -, as e~^' 



converges to 0. This gives the limiting-type B (fig.l). 



When q, = [q, unite), we have (-- 1 ^^"^1 f -— )=rO; f — ) 



