AND RATIO Ol'' TIIKIH ATOMIC WKICIITS. Idf) 



To the wftter produced in tliis experiment, baiium liydroxide was added, and 

 the whole showed no trace of turijidity. It was then evaporated to the bulk of 

 five cubic centimetres ; of course caibon dioxide was not admitted. Even 

 when reduced to this bulk, there was no turbidity. It is obvious, thei'efore, that 

 the hydroo-eu used in these experiments did not contain carbon compounds. It is 

 precisely the same in quality, obtained from the same apparatus, with the same solu- 

 tion, as that used in all my experiments so far ; the only change ever made has been 

 the introduction of fresh distilled water to take the place of that removed by 

 electrolysis. 



It' now this carbon dioxide was formed ))y the oxidation of organic matter 

 adhering to the asbestos, it was supposed that each cubic centimetre of carbon 

 dioxide foi-med represented a cid^ic centimetre of oxygen withdrawn iiom the 

 globes but not used in forming water. 



In some expeiiments, nitrogen was also found. The quantity was small. 

 When the globes for oxygen contained air, and were exhausted and filled with 

 oxygen, a certain amount of nitrogen remained in the globes. If the globes were 

 exhausted again when partly filled with oxygen, the nitrogen was mostly eliminated. 

 As the amount of nitrogen found agreed fairly in amonnt with what might be 

 expected from the previous contents of the globes and the degree of the last 

 exhaustion, no hesitation was felt in counting this nitrogen as having come from 

 the globes of oxygen. 



12. — SYNTHESIS OF WATER. SOURCES OF ERROR. 



In the weio-hing of the hydrogen, no source of appi'eciable constant error has 

 yet been thought of. The magnitude of the accidental ei-rors must be learned from 

 the concordance of the extieriments. 



In the weighing of the oxygen, leakage might act in two opposite directions. 

 If leakao-e occurred before the combustion began, the amount so lost would be ex- 

 ti-acted by the Toepler pump in the exhaustion preliminary to the combustion, 

 and so would seem to have been used in the combustion. If the leakage occurred 

 in the contrary direction, when the half-exhausted globe was hung on the balance 

 for the final weighing, the amount of oxygen taken from the globe would appeal' 

 less than the truth. Now the leakage during exhaustion, if noticeable in amount, 

 would be readily detected by the behavior of the pump during the preliminary 

 exhaustion. A very slight leak can be detected when the vacuum obtained is 

 something like a ten-tliousandth of an atmosphere. Such a leak would be the 

 moi-e certainly detected, since the pump was always left at rest for fifteen minutes 



