42 JOURNAL OF GEOLOGY— SUPPLEMENT 



If a liquid of composition MgSiOs is cooled, the first compound 

 which separates is Mg2Si04 (f orsterite) . There is, therefore, 

 direct evidence that the compound MgSiOj is partly dissociated 

 in the hquid state in such a manner that Mg2Si04 is one of the 

 dissociation products. A possible equation for the equihbrium 

 reaction might be written 2MgSi03^MgSi04+Si02. There may 

 be, indeed probably there is, further dissociation according to the 

 reaction Mg2Si04^2MgO+Si02, but of this there is no direct 

 evidence. There is certain evidence of the presence of Mg2Si04 

 because that compound first crystalHzes from the liquid; i.e., of 

 all the possible compounds which might exist in the liquid, Mg2Si04 

 first exceeds its solubiKty. The actual amount of the compound 

 Mg2Si04 may be quite small compared with the amount of 

 undissociated MgSiOj, but the solubility of the latter very much 

 greater. Indeed, it may be stated that, on account of the higher 

 melting-point of Mg2Si04, it may be expected to exceed its 

 solubiKty before MgSiOj, even if present in the liquid in smaller 

 amount. 



If a mixture of Mg2Si04 and anorthite, containing approxi- 

 mately twice as much anorthite as forsterite, is melted and then 

 cooled, the mineral which crystallizes first is neither anorthite 

 nor forsterite, but spinel, MgAl204.'^ There is, therefore, evidence 

 of an equihbrium reaction in the hquid of which one of the products 

 is MgAl204. It is uncertain what other products may be present, 

 but of all those present the compound MgAl204 exceeds its satura- 

 tion limit first. Magnesia was added to the mixture in the form 

 Mg2Si04 and separates from it as MgAl204, but this should not be 

 construed as meaning that magnesia has a greater affinity for 

 alumina than for sihca. The actual amount of unchanged Mg2Si04 

 in the liquid may have been many times as great as the amount of 

 MgAl204,^ but the solubility of Mg2Si04 correspondingly greater. 

 SolubiKty is the controlling factor. The soKd phases (crystals) 

 which separate from any liquid mixture of silicates cannot be 

 regarded, therefore, as giving unquaKfied evidence of the relative 



^ Olaf Andersen, op. cit., p. 436. 



2 The separation of MgAl204 even when these molecules occur in the liquid in 

 relatively small amount is connected with its high melting-point. 



