THE NEW THEORY OF SOLUTIONS. 391 



from objection, and does not rest on a firm theoretical basis 

 (see " Science Progress," iii., 229). What appears conclu- 

 sively settled is, that dissolved substances producing asso- 

 ciated solutions when pure and in the liquid state consist 

 of complex molecules ; and one interesting point, at present 

 in an unsatisfactory state, is whether the aggregation of the 

 liquid when pure is not always greater than when it is dis- 

 solved. 



Anomalous freezing-points. — It sometimes happens that, 

 as regards their freezing-points, certain solutions simulate 

 the behaviour of associated solutions. This is due to the 

 fact, first pointed out by van't Hoff, that, on freezing the 

 solution, dissolved substance solidifies along with the solvent, 

 and of course in this case the ordinary freezing-point laws 

 are no longer valid. In this way have been explained the 

 low values of the molecular lowering, and high values of 

 the molecular weight found for m-kresol in phenol, and 

 aldoxime in acetoxime, observed by Eykmann (1889), and 

 thiophen in benzene observed by Paterno (1889). It has 

 further been shown that the tendency to form mixed 

 crystals on freezing is most marked in the case of isomor- 

 phous substances, and Kuster's work (1890-91) has proved 

 that the freezing-point of an isomorphous mixture may in 

 some cases be calculated from those of its constituents by the 

 ordinarv admixture rule. In such a case, if the freezing- 

 point of the dissolved substance is higher than that of the 

 solvent, its presence will bring about a rise instead of a fall 

 in the freezing-point of the solvent. This behaviour was 

 observed by Heycock and Neville (1889) in the case of a 

 solution of antimony in tin, and by Eykmann for /3- 

 naphthol in naphthylamine. Van Bijlert (1891) in these 

 and other cases proved by actual analysis that the ice 

 separating from the solutions really contained dissolved 

 substances. 



An interesting development of this question is due to 

 Ferratini and Garelli (1893), wno discovered several anoma- 

 lous solutions of this class, and pointed out that in all 

 probability similarity in the chemical constitutions of solvent 

 and dissolved substance, rather than similarity in their 



