

1892.] Spectra of the Flames of Metallic Compounds. 119 



good deal of hydrogen. We have employed two methods of making 

 such a mixture. The first was by passing a stream of dry hydrogen, 

 mixed with carbonic oxide, over reduced nickel in a glass tube, and 

 burning the issuing gas in a double jet with oxygen either outside 

 or inside the burning gas. The nickel was freshly reduced at a 

 gentle heat with hydrogen, and allowed to cool in carbonic oxide. 

 When quite cold the stream of mixed hydrogen and carbonic oxide 

 was found to take up quite enough nickel at the temperature of the 

 room, and would continue to do so for some hours. After a time, 

 however, the nickel required to be again warmed in a current of 

 hydrogen, when some water was given off, and the metal recovered 

 its sensitiveness. Another plan was to pass a stream of hydrogen 

 through a (J -tube containing a little of the liquid nickel compound 

 in the bend. The result was the same in each case, but the propor- 

 tion of vapour of the compound was more easily varied (by simply 

 varying the proportion of carbonic oxide in the stream of gas) in the 

 former method. The mixed gas and vapour burnt in air with a 

 ;smoky flame, but in a full supply of oxygen with a bright yellowish- 

 green flame without visible smoke. The first jet we used was of 

 platinum, but nickel- carbonyl deposits nickel at a red heat, so that 

 the platinum soon became coated with a thick deposit of nickel, which 

 choked the orifices. This nickel adhered so closely to the platinum 

 that it could not well be removed mechanically, and had to be dis- 

 . solved off. We found it, therefore, more convenient to use a jet 

 made of a piece of porcelain tube, about 1 cm. in diameter, with a 

 narrow porcelain tube, fitted by means of a cork, in the axis of the 

 wider tube. The mixed gas and vapour were passed either through 

 the inner or through the outer tube, and oxygen through the other. 

 'The porcelain being a bad conductor, no nickel was deposited on it, 

 except close to the orifice, whence it could be easily removed 

 mechanically without disturbing the apparatus. The porcelain, of 

 course, added some lines to the spectrum, but these were easily 

 -detected. In fact, we noticed only the lines of sodium, calcium, and 

 lithium. 



The spectrum of the flame of the nickel-carbonyl thus diluted con- 

 ,sists of two parts : (1) the spectrum of the main body of the green 

 flame, (2) that of the base of the flame when the oxygen is outside, 

 and of the surface of the small inner cone when the oxygen is inside, 

 the flame. 



The spectrum of the main body of the flame consists of a series of 

 shaded bands, brightest in the green, but extending on the red side 

 beyond the red line of lithium, and on the violet side well into the 

 blue, though with rapidly diminishing distinctness. These bands have 

 their sharp bright edges on the more refrangible side, that is, they 

 are turned in the opposite direction to the bands produced by electric 



