Dixon — A Thermo-electric Method of Cryoscofij. 



59 



in parts may continue for some time supercooled, and we may get too great a 

 deflection. Experience shows that the smallest depression is obtained if the solu- 

 tion is allowed to cool 0'1° to 0-2° below its freezing-point before inoculation. 



It will often be found that the mean of t!ie second pair of readings 

 indicates a somewhat larger deflection than that of the first pair by about 

 1 per cent. This seems to be due to the slow cooling of the support of the 

 junction in the solution. It will be found that readings after the junctions 

 have been in the freezing-chamber about 15 min. do not tend to be greater 

 than the preceding ones. In tlie natural routine the second pair of readings 

 are made about 15 min. after tlie junctions have been put in position. A 

 third pair of readings made in a similar manner will plainly show if the 

 apparatus has reached a steady state. If the observations have been 

 satisfactory, they should not diverge from one another by more than \ per 

 cent., and with care greater accuracy may be attained. 



To calibrate the apparatus sodium-chloride solutions of known strength 

 are introduced into one of the test-tubes. The deflection produced by the 

 depression of the freezing-point of each is observed. These depressions being 

 known by the work of Eaoult, Loomis, Nernst, and Abegg, we obtain the 

 value in degrees centigrade of a millimetre deflection of the light-spot. 



The following table exhibits the figures of one of these calibrations. The 

 individual readings are recorded to give some idea of the acouracy of the 

 arrangement : — 



The scale reads continuously from left to right : 250 mm. marks its middle 

 point. 



In this table are recorded the two positions of the spider-line in the spot of 



light on the scale for three successive observations of the freezing-point of 



each solution. The deflection corresponding to this freezing-point is obtained 



by subtracting the second from the first, and halving the result. It will be 



seen that deflections obtained in this way diverge only slightly from the inean, 



K 'it 



