of the Jixed Alkalis. ,o 149 



and in passing the gas into the tube for electrization, it was 

 not easy to avoid introducing some of this moisture, which 

 must have been a saturated solution of ammonia, at the 

 same time. 



In my first trials made upon gas, passed immediately from 

 the vessel in which it had been collected into the apparatus, 

 I found the expansion of one of ammonia vary in different 

 instances from 2*8 to 2-2 measures ; but the proportions of 

 the nitrogen and hydrogen appeared uniform, as determined 

 by-detonation of the mixed gas with oxygen, and nearlv as 

 one to three in volume. 



To exclude free moisture entirely, I carefully prepared 

 ammonia in a mercurial airholder, and, after it had been 

 some hours at rest, passed a quantity of it into the tube for 

 decomposition, which had been filled with dry mercury. 

 In this case 50 parts became 103 parts by electrization, and 

 there was still reason to suspect sources of error. 



I had used iron wires not perfectly free from rust, for 

 taking the spark, and a black film from the mercury ap- 

 peared on the sides of the tube. It was probable that some 

 ammonia had been absorbed by the metallic oxides both ' 

 upon the iron and the mercury, which might again have 

 been given out in the progress of the operation. 



I now used recently distilled mercury, which did not 

 leave the slightest film on the glass tube, and wires of pla- 

 tina. The ammonia had been exposed to dry caustic potash, 

 and proved to be equally pure with that mentioned in page 

 14". 6o measures of it, each equal to a grain of water, 

 were electrized till no further expansion could be produced, 

 the gas filled a space equal to that occupied by 108 grains of 

 water. The thermometer in this experiment was at 56°, 

 and the barometer at 30' 1 inches. The wire of platina trans- 

 mitting the spark was slightly tarnished*. The 108 mea- 

 sures of gas, carefully analysed, were found to consist of 80 

 measures in volume of hydrogen, and 23 measures of ni- 

 trogen. 



* This most probably was owing to oxidation. When platina is made 

 positive in the Voltaic circuit in contact with solution of ammonia, it is ra- 

 pidly corroded. This is an analogous instance, , 



K 3 The 



