136 Day and Allen — Isomorphism and Thermal 



(fig. 16, II) with the mixture richest in the higher melting 

 component, crystals of composition a will be in equilibrium 

 with the liquid phase b in all proportions and solidification (or 



melting) will not take place at a 

 single temperature but through a 

 range of temperature. If we now 

 plot the length of the abscissa cor- 

 responding to ab in a separate dia- 

 gram with the observed tempera- 

 ture range of solidification, adding 

 all the other possible cases which 

 will arise from the continued dis- 

 placement of the f-curves, we 

 Fig. 17. 1 . ... • ' j- 



arrive at the accompanying dia- 

 gram (fig. 17) of Roozeboom's type I. Types II and III 

 appear in the same way when the form of the (f-curves changes 

 as indicated in figs. 18 and 19. 



The physical side of the system of reasoning is readily 

 inferred from the figures. If we start with a mixture of the 

 composition indicated by ?n, (fig. 20) and temperature above 

 the melting point, crystallization will begin at a, the separating 

 crystals will have the composition b, while that of the remain- 

 ing melt approaches d. Upon cooling to e, solidification ends 

 with crystals of this composition. Melting is exactly the 

 reverse operation. Whether these first crystals of composition 

 b remain stable as such or undergo solid transformation or 

 wholly or partly redissolve, appears to remain undetermined in 

 any general way by Roozeboom's theory and may be radically 

 influenced by accompanying phenomena like viscosity and 

 undercooling : if a liquid mixture of composition a under- 

 cook to e before crystallization begins, crystals of composition 

 e will appear and no others (provided the release of heat of 

 fusion does not raise the temperature above e). Such a situa- 

 tion is certainly unavoidable in viscous mixtures like the feld- 

 spars and accounts very well for the homogeneous solidification 

 observed by us. This would classify the feldspars with type I 

 of Roozeboom's series. A comparison of our melting point 

 curve with figs. IT, 18 and 19 shows this to be the only type 

 under which it could possibly fall. There is no trace of a 

 maximum or minimum in the feldspar curve. Yogt's expecta- 

 tion that they would fall under type 3 therefore fails of 

 fulfilment from our experiments. 



That our curve so closely resembles one branch of Rooze- 

 boom's typical curve is remarkable. The difficulties of obser- 

 vation, in those portions of the curve where the viscosity 

 becomes so disturbing, are too great to enable stress to be laid 

 upon the form which our curve happens to take there, but 



