ATOMIC WEIGHTS 



41 



then reduced by electi-ic heating over finely divided metallic nickel. The 

 gain in weight of the nickel represented the amount of oxygen absorbed. 

 In some of the experiments the liberated nitrogen was condensed, at 

 the temperature of liquid air, in cocoa-nut charcoal, and its weight also 

 was determined. Two series of determinations were made, on nickel from 

 different sources, but for present purposes these may be treated as one. 

 For three of the measurements corrections are given for the nitrogen 

 occluded by the mixed nickel and nickel oxide, which corrections I have 

 applied in the following table of Gray's results : 



Weight NO. Weight 0. Weight N. 



.31384 .16729 



.64304 .34300 .30010 



.50672 .27025 



.54829 .29221 



.61862 .32981 .28885 



.62622 .33401 .29234 



.62128 .33111 



.54469 .29029 .25432 



.52001 .27715 .24270 



.62103 .33103 .28998 



From these weights the subjoined values for K are derived. 



The general mean of the three series is 



N = 14.0104, ± .0011 



The accurate calculation of molecular and atomic weights from gaseous 

 densities is really an affair of very recent times. The gases, as measured, 

 show divergencies from Avogadro's law, and the crude density ratios there- 

 fore require correction, as we have already seen in reference to the atomic 

 weight of hydrogen. It seems best, however, to assemble the actual meas- 

 urements first, and to apply the corrections to the entire mass of data 

 afterwards. 



