PHYSICAL CONSTANTS OF SATURATED SOLUTIONS. 201 



fully adjusting the temperature of the oil bath, two densities could be obtained, one 

 while the temperature of the solution was close to the boiling-point, but slowly rising, 

 and the other when it was above the boiling-point, but close to it and slowly falling. 

 The former gave the density of a slightly unsaturated solution, and the latter that of 

 a slightly supersaturated solution when referred to a solution saturated at its boiling- 

 point. The mean of these two observations was considered to be the density of the 

 solution saturated at the boiling-point and similarly with the resulting solubilities. 



The results obtained at the boiling-point were found not to be as concordant as 

 those at the other temperatures ; doubtless the greater part of this is due to the 

 exceptional difficulties of the experiments. The following are some of the sources of 

 error. It was impossible to prevent the condensation of steam in the stem of the 

 pyknometer, and it was therefore necessary to estimate the length of each drop, and 

 add this length to the reading of the level in the stem. There was also an error 

 introduced by the fact that, for the purpose of reading its level, the solution had to 

 be sucked into the cold part of the stem which projected through the indiarubber 

 stopper ; on reaching this colder part, salt immediately crystallises out ; the total 

 volume thus changes, and the observed level is not that which the solution would 

 otherwise have attained. The maximum error possible from this cause was calculated 

 for the case of rubidium nitrate, and was found to be such as to give an error in the 

 density of O'l per cent. 



Another and a much more important source of error was that the reduction of 

 pressure on the surface of the solution in the pyknometer, unavoidable when filling 

 by suction, very often caused steam bubbles to form. As it was essential to fill whilst 

 a large excess of undissolved salt was still being stirred by the steam, the solution 

 surrounding the pyknometer was often semi-opaque, and consequently it might 

 happen that part of the space inside the pyknometer was occupied by an unseen steam 

 bubble, and might thus be an unobserved source of error. Numerous fillings had to 

 be rejected on this account, the steam bubbles showing themselves on the withdrawal 

 of the pyknometer from the solution. The numbers tabulated below are those 

 derived from fillings in which there were no observed steam bubbles. 



Determination of the Temperature. 



As mentioned above, the boiling-point of the saturated solutkm was considered to 

 be the constant temperature which the solution and salt reached when steam 

 was rapidly bubbled through them ; this temperature was indicated by mercury 

 thermometers, and is given in the table of results. They are, however, unconnected 

 for emergent column, because it was found to be practically impossible to apply a 

 satisfactory correction. It is hoped that, later on, when determining the osmotic 

 pressures,* these boiling-points will be accurately ascertained by means of platinum 



* These experiments are in progress, but not complete. 

 VOL. CCIII. A. 2 D 



