726 



large when Ave give 8ucli values to m, n ai)fl p, that one or more 



of the conditions (20) are satisfied. It is apparent tiiat tliis may be 



casually only for m ^=0 or ?i ^ or p =iO or 1 — m — n — /; = 0. 



At the same time the following is apparent. When we give to 



111, n and /) such values tiiat f.i. ,/■ becomes =0, then in (11) 



d: d? dS d: 



-^ becomes infinitely large, so that-—, — and —- become infinitely 



ox ' am oil op 



large at the same time. When, however, we liave chosen the com- 



posants in such a way that «, = « then only - and — become 



■ on op 



OS 

 infinitely large, while - remains finite. 



dm 



When a liquiii has the composition : 



/. = xX + .vF + zZ + . .. 



wherein A. )' etc. represent components, then liie stability requires 

 that for all values of ilr, i/y etc. 



dx+ -d!j+ ...] yo (21) 



0.Ï oy J 



When we imagine L to be divided into 



L = hL, - (1 — x) /., 

 wherein : 



L, = (a- -f dx,) X + (y + dy,)Y+ .... 



L, = (.,■ + dx,) X + {;/ + dy,) Y+ ... 

 then must 



i;<x5, +(i-x)?, 



from which (21) is following. When we now express the composition 

 of L in coraposanis viz.: 



then if follows in the same way that 



-^dm -\- -^dn 4 >0 



Om dn J 



must be true for all values of dm, dn etc. 



{To be continued) 



Leiden. Inorq. Chem. Lab. 



(: 



