62 Scientific Proceedings, Tloyal Dublin Society. 



As has already been pointed out, the influence of the temperature of 

 the freezing-bath on the apparent freezing-point deflection is due to the 

 difference in the behaviour of water and salt solutions on freezing. In the 

 latter the crystals remain separate and the ice is finely divided. The 

 difference of density between it and the solution causes it to rise up somewhat 

 more rapidly, tends to aggregate it at the upper surface, and so permits the 

 lower layers of the solution to supercool to a small extent. In the distilled 

 water, on the other hand, the ice adheres to the walls of tlie tube, and forms 

 a lining to it, so that supercooling of the lower layers is less favoured. 



The convergence temperature and tlie constant k of Nernst and Abegg's 

 correction are dependent largely on the heat-capacity of the solution, and 

 consequently it is of importance tliat the two test-tubes should contain 

 approximately the same amount of liquid ; otherwise the rate of exchange of 

 heat between the solution and the freezing-bath and the water and the 

 freezing-bath will be different. But this is no serious difficulty within wide 

 limits. Thus the deflection due to the depression of freezing-point of a 

 solution containing 1'5 gm. NaCl in 100 gm. of water was found to be 

 115-6 mm., when the solution and the water stood at a level of 3'3 cm. in 

 similar tubes; it was reduced to 115'4 mm., when the depth of the salt 

 solution was increased to 5 cm. Of course it is easy to arrange that both 

 tubes should contain the same amount, and so have practically the same 

 heat-eapacity. 



Change of resistance of the circuit due to temperature-changes is guarded 

 against by completely immersing the eureka or nickel of the couple in the 

 freezing-chamber, while the resistance of the galvanometer is kept constant 

 by its being enclosed in tlie tliermostat. The complete immersion of the 

 connecting-piece of the couple in the freezing-chamber also secures the 

 elimination of thermo-electric effects due to want of uniformity in this wire. 



From what has been said, it appears that the thermo-electric method 

 is capable of considerable accuracy, even when only two junctions are 

 employed. Of course if it were desired to work to greater accuracy, there 

 would be no reason why the number of junctions should not be increased, 

 thus greatly increasing tlie galvanometer deflection for the same temperature 

 interval. In the work in which we were engaged, however, this would have 

 been undesirable, as a comparatively large range was required. 



But even with a pair of junctions, the hundredth of a degree could 

 be measured with certainty. With this accuracy very small quantities of 

 fluid may be dealt with. The small quantities required render the method 

 particularly suitable to physiological work. Its differential character might 

 also be applied with advantage to comparing the freezing-points of different 

 fluids ; for example, in a comparison of jugular and carotid blood. 



