An dersen — System Anorthite- For 'sterite- Silica. 441 



mixtures made up according to the course of the approximate 

 curves. The points could be determined by quenchings of 

 neighboring mixtures on each side of the curve, or by quenchings 

 on one mixture near the curve, determining the point on the 

 liquidus corresponding to the composition of that mixture (i. e., 

 the temperature where the primary phase disappeared) and 

 the temperature where the second phase appeared or disap- 

 peared. The composition of this point was then extrapolated 

 from the slope of the liquidus. Provided that this slope was 

 known with sufficient accuracy, and the mixture examined was 

 not too far removed from the curve, this method would give 

 results as exact as those inferred from quenchings on several 

 mixtures on each side of the curve. It appeared that, except- 

 ing the curve G N, the boundary curves deviated but little 

 from straight lines. Their course could therefore be deter- 

 mined by a comparatively small number of mixtures. 



The isotherms. — It would require thousands of quenching 

 experiments, carried out on hundreds of mixtures, actually to 

 determine the course of the isotherms ; that is to say, to map 

 the surface of the liquidus with such accuracy that any point 

 on this surface could be located in the diagram or space model 

 within the limits of error of experimental determinations. 

 Such work in the case of a comparatively simple system like 

 the present would appear to be of doubtful utility. The num- 

 ber of quenchings carried out, besides those determining quin- 

 tuple points andboundary curves, are therefore comparatively 

 few. Nevertheless, the general course of the isotherms (as 

 shown in fig. 10), within the part of the system where thermal 

 work has been possible, may be considered fairly well estab- 

 lished. It is obvious that, in general, points on the isotherms 

 may be determined with a sufficient degree of certainty by 

 interpolations based on determined points. 



Normal crystallization in different sections of the ternary field. 



In the following paragraphs a discussion of the course of 

 crystallization in different types of mixtures will be given. It 

 is always assumed that the cooling proceeds at such a rate that 

 perfect equilibrium obtains throughout the range of tempera- 

 ture considered. This will be referred to as the normal crys- 

 tallization. 



For the sake of convenience the system (as represented in 

 the diagrams, figs. 9 and 10) is divided into different parts by 

 lines connecting the composition points of the solid phases with 

 the quintuple points. To make the relations clearer the prin- 

 cipal types of crystallization curves are illustrated in the 

 diagram of fig. 11, which is an arbitrarily distorted representa- 

 tion of fig. 9 with this difference, that the spinel field BCK 



