254 



Prof. J. Dewar. 



[Mar. 9, 



V 760(273 + T) X Y (1+^T') 

 or, with sufficient accuracy for our investigation, 



, _ 273^ V^q+^To-T) m 

 760V 273 + T \ '* 



An important error might be expected to arise from an in- 

 accurate knowledge of T the temperature of the cooled gas. An 

 alteration of a degree in the value of T would alter the value of 

 d by 



_ 273w Yp ( j 

 760Vo X 273 + T P ' 



or very approximately, a rise of a degree in the estimate of T' above 

 its true value would diminish ^ by ft d — in the present case by 4-0-J00" 

 part. 



Similarly an error of a degree in the estimate of the temperature T , 

 at which the volume of the flask was measured, would have the like 

 small effect, but in the opposite direction. 



On the other hand errors in the measurements of either Y or V 

 or of p or T would give directly proportional errors in the value of 

 d. The values of T were easily read to one-tenth of a degree, so that 

 an error in this quantity would have an effect less than 3^ on the 

 value of d. Similarly the pressure was read to J mm., leaving an 

 error of less than y-Vo on tne density. The volume V showed a 

 tendency to error in the earlier portion of each group of experiments, 

 but this was eliminated as the experiments progressed and especially 

 where the results were more important, namely, in the region of the 

 solidified gases. The greatest care was taken to insure that the whole 

 volume of the bulb was occupied by the solid. The gas was allowed 

 to enter in successive portions, each of which was liquefied and 

 solidified previous to any further admission, so as to insure the absence 

 of any vacua due to contraction. 



3. The results for oxygen seem low, the boiling-point density being 

 1*1 18, whereas former results gave 1*138. On plotting the densities 

 as ordinates to the temperatures as abscissae the observations lie very 

 closely on a straight line which (by least squares), is 



d = 1-5154-0-004420/ (2), 



t being absolute temperature. 



Such a line as (2) must in any case be only a chord of the curve of 

 liquid densities, and the nearer two observations are to the absolute 

 zero, the more nearly will the chord joining them partake of the 

 nature of a tangent to this curve at the absolute zero. Now, at so low 



