98 Composition of the Atmosphere 



Shortly prior to this, it was believed that when the mercury was raised 

 and lowered in pipette B, the amount of water adhering to the walls of 

 the lower bulb and graduated portion would vary, and that these differ- 

 ences might play a very important role in determining the percentage of 

 oxygen. Experiments were accordingly made to find the oxygen per- 

 centage in air more or less dry, but these were unsuccessful. Finally, it 

 was decided that if an excess of water was present in the pipette the same 

 relative amount of water would probably adhere to the glass walls each 

 time. 



A large number of experiments were made to determine the amount 

 of water which should be added to obtain a clear meniscus for reading and 

 to insure constancy in the amount of water in the pipette. At first minute 

 quantities of water were used, the attempt being made to secure only 

 enough to saturate the gas with moisture. It was assumed that when the 

 mercury was lowered the liquid water would adhere to the inside of the 

 upper bulb, so that only the mercury would enter the constricted portion 

 of the pipette and the lower bulb, and that no liquid water would be pres- 

 ent. Under these conditions, however, it was found very difficult to set 

 the meniscus, and the following routine was finally decided on: 



FOURTH ROUTINE, AND RESULTS OBTAINED. 



Outline of fourth routine. The nitrogen resulting from the previous 

 analysis was stored temporarily in the potassium-pyrogallate pipette, the 

 capillary tube leading to the carbon-dioxide absorption chamber C also 

 being filled with nitrogen. The stop-cock a was next removed and the 

 mercury in the pipette B raised up through the capillary to the stop-cock. 

 Water was then added and the mercury simultaneously lowered until 

 there was a layer of 17 mm. of water above it in the capillary tube. 

 The stop-cock a was again put in place, the nitrogen withdrawn from 

 the pyrogallate reservoir, and the sample taken and analyzed as usual. 



Results with fourth routine. The analyses made with this routine were 

 continued from March 28 to April 10, 1911. The results are given in 

 table 59. When these results are compared with those of earlier experi- 

 ments, it will be noted that notwithstanding the differences in tempera- 

 ture, barometer, direction of the wind, and other conditions, the oxygen 

 determinations show a striking uniformity, the variations previously 

 found practically disappearing. 



This new routine was also used for the analyses of cylinder air, which 

 provided an excellent control of the apparatus and method. The results 

 for the period between March 28 and April 14, 1911, are given in table 60. 

 Here again it is seen that variations in the oxygen percentage are rare, 

 and as used the method may be assumed to give constant results. 



