( 362 ) 



points D. It li()\vevei' we lake the iM-odiicI of ^r — and f\f wo need 



not make a diJlerenee for I lie points inside i\,f^i), and we nia^' 

 assume lliat onlside the loeus, for which /'.,/• = 0, this product is 

 negative and inside it, positive. 1 need hardiv mention, tlial Just as 



the eonnodal eur\e, the spino(Uil eur\e and the curve ^ — =^ are 



modified and (hsplaced according- to tlie temperature, also the curve 

 ?'^ = depends on tlie \alue of T. On tlie whole it will contract 

 and move towards the side of the anrhi-aqninone with increase 

 of temperature, and so follow tiie same course as the other loci 

 mentioned. 



The course of the value of tlie denominator, viz. of ^\sf, has not 

 yet been discussed. In the preceding communication 1 had thought 

 that I could leave out this discussion, first because I did not thiids; 

 t necessary at all, but also because I thought that the result of this 

 discussion could not be bi-onght under a simple form, and finally, 

 because I did JU)t wish to add another to the number of loci. 



The particularity in the course of the {[>,T) curve, for the equilibrium 

 between solid and iluid, however, to which we have had to conclude 

 in fig. 7, has proved, that the discussion is not to be evaded, at 

 least if we wish to explain fully by theoretical means, the way in 

 which the two ( i>, T,.v) surfaces get detached. And the result of the 

 discussion of the quantity Wsf has proved to be very simple — 

 and almost exactly the same as the result of the discussion concerning 

 the quantity Y^f. Just as there is a locus t"or wiiich V.fz^O, so 

 there is one for which ]r,/-=:0. Just as the curve for wdiich |^,/- = () 

 consists of two branches further from each other than the points 

 D and D' of fig. 2, which two branches meet outside the top of 

 the plait, in the same way the curve W,f=z{) consists of two 

 branches, further from each other than the points D and D' , and 

 these two branches meet also, either outside the top of the plait, or 

 inside it. And linall}' the locus, for which IT^/^O, lies entirely 

 within that for ^^ liich V,f = 0. The resemblance goes further. Outside 

 Ys/^^^ this (piaiitity is negative, and outside \Yf = () the value 

 of M^sf is negative. Inside I ",/•=() the quantity I"/' is positive and 

 increases to infinite when we reach the points D and D' , being 

 again negative inside these limits. The same applies to the quantity 

 Wsf. Inside the curve for which \r,f = 0, this quantity is positive. 

 In the curve of the |)oints I) and D' the ^■alue has increased to 

 infinite, being again negative inside the [)oints J) and D'. If \^■e 



