20 ARGON, A NEW CONSTITUENT OF THE ATMOSPHERE. 



tlie pipes, aiul should he relatively rich in the heavier constituents of the atmos- 

 [iliere. The flow of water from the aspirator couhl not be niaintaiued very 

 constant, but the rate of 2 per cent was never much exceeded. The necessary 4 

 litres took about 16 hours to collect. 



The air thus ol)tained was treated exactly as ordinary air had been treated in 

 determinations of the density of atniosplierie nitrotren. Oxvcren was removed by 

 red-hot cojiper, followed by cu[tric oxide, ammonia V)y sulphuric acid, moisture and 

 carbonic aidiydride by phosphoric anhydride and potash. 



The following are the results : 



Globe empty, July lo, 14 2.81789 



Globe full, Sept. 15(12 pipes) 50286 



2-31503 

 Ordinary atmospheric nitrogen 2.31016 



Difference + .00487 



Globe empty, Sept. 17 2.81345 



Globe full, Sept. 18 (12 pipes) 50191 



2-31154 

 2.31016 



Difference + .00138 



Globe empty, Sept. 21 2.82320 



Globe full, Sept. 20 (12 pipes) 5'03' 



231289 

 2.31016 



Difference + .00273 



Globe empty, Sept. 2 1, Oct. 30 2.82306 



Globe full, Sept. 22 (12 pipes) 5' '40 



2.31 166 

 2.31016 



Difference + .00150 



The mean excess of the four determinations is .00262 gramme, oi-, if we omit 

 the first, which tlei)ended upon a vacuum weighing of two months old, .00187 gramme. 



Tile gas fiom [)repai'ed air was thus in every case denser than fiom unprejjared 

 air, and to an extent much Ijeyond the possible errors of expeiiment. The excess 

 WJI.S, however, less than had l)een expected, and it was thought that the arrange- 

 ment of the pipes could be improveil. The final delivery of gas from each of the 

 groups in parallel being so small iu comparison with the whole streams coucerued, 



I 



