of Magnesium, Copper, and Manganese. 30«> 



If the substance be a metal which easily forms a carbide, 

 such as Ca, Fe, Co, Ni, <&c, and therefore adheres to the 

 surface of the graphite plate, no covering with carborundum 

 powder is required. In the case of a volatile metal, mineral, 

 or rock, &c, it is often of advantage to first crush or in some 

 other way subdivide the material, and place small quantities 

 of it all over the plate, which is then covered with a -J in. 

 thick layer of pure carborundum No. 80. But whenever 

 possible the protecting layer should be dispensed with, as 

 it always introduces lines and bands due to impurities 

 contained in the carborundum. Such metals as copper and 

 silver, which do not adhere to the graphite, may be pre- 

 vented from rolling off the plate — they are of spheroidal 

 shape in the molten state — by placing small pieces of 

 graphite round them. Particular methods employed for the 

 purpose of mounting and holding substances will be given 

 in each case. 



With the arrangement sketched in fig. 2 the spectrum of 

 the vapours given off by the substance is of course observed 

 both below and above the plate. But since the greater 

 quantity of vapour is immediately carried upwards by 

 convection, only a relatively small amount of it passes into 

 the protected space. It is no doubt for this reason (hat the 

 spectrum emission due to thermo-chemical excitation is 

 brighter and passes to a much greater distance from the 

 plate in the region above than in the protected space below. 

 Also the emission due to thermo-electrical excitation, 

 although in many cases appreciably feebler above the plate 

 than below, passes to a greater height, probably owing to 

 the upward rush round the edge of highly ionized vapours 

 from below. For these reasons it is preferable to make all 

 observations in the protected space beneath the plate, where, 

 thanks to the more tranquil state of the vapours and the 

 absence of convection currents, the two emissions are more 

 sharply differentiated than above. Only doubtful cases 

 with regard to lines caused by thermo-chemical excitation 

 are for the sake of confirmation examined in the uprising 

 vapour above the plate. 



As usual an image of the incandescent plate and pro- 

 tected space below it is sharply focussed upon the slit 

 of a spectrograph in such a way that the slit bisects the 

 plate as marked in fig. 2. The resulting spectrum shows a 

 continuous bright band caused by the edge of the incan- 

 descent graphite plate and, immediately below it, a sharply- 

 defined region of short lines and bands which constitute the 

 fringe emission (thermo-electrical excitation). Superposed 



