IN QUARTZ OF LIGHT IN THE VISIBLE REGION OF THE SPECTRUH. 
271 
producing a steady lithium flame would be a very great convenience. Tlie readings 
were taken with a dense direct-vision prism as the only dispersive system and a 
half-shadow angle of 5°. 
( 3 ) Sodium. 5896‘16 and 5890'19. 
The lines of the sodium doublet were by far the most difficult to read of the series 
now described. They could only be separated by using the highest available dispersion 
on the constant-deviation spectroscope, when the actual separation of the centres of 
the lines amounted to mm. Under these conditions the flame spectrum, even with 
the help of oxygen, was quite inadequate as a source of light. The actual readings 
were taken from a carbon arc with a thread of glass in the hollow core of the upper 
carbon. The main trouble was the reversal of the lines, which frequently appeared 
as a yellow band crossed by two very narrow black Fraunhofer lines. These reversed 
lines were admirably adapted for setting the spectroscope, the best possible adjustment 
being obtained when a black Fraunhofer line could be seen down the centre of the slit 
of the polarimeter, this slit being narrowed to about ^ mm. in order to cut out the other 
line of the doublet; they were, however, quite useless for taking the actual readings. 
The device finally used to secure trustworthy observations consisted in levering open 
the jaws of the slit to see the character of the lines and releasing the jaws in order to 
take readings when the lines were seen to be narrow and luminous (without reversal). 
This condition could sometimes be maintained during several minutes but was never 
quite to be relied on ; the reversal of the line could usually be detected even without 
opening the jaws of the slit, by the loss of sharpness of the extinction and a sudden 
change of perhaps 2° in the readings as the neighljouring line broadened and spread 
into the field of view; this effect was particularly troublesome in the case of the less 
refrangible line. The sodium lines were read without any prism on the eye-piece, as 
the half-shadow angle could be reduced from 10° to 5° by removing this prism; this 
exceptional course was justified (a) by the small amount of tlie stray light in the 
grating spectrum {h) by the fact that the stray light would be practically a 
continuous spectrum crossed by some scores of extinction bands and would not be 
likely to deflect the extinction more in one direction than in the other ; the eye-piece 
would not in any case have been adequate to throw out stray light from the other 
sodium line. 
A very promising method of producing a bright sodium spectrum without reversal 
consisted in adding sodium salts to a silver arc. But owing to the relative feebleness 
of the light and the extremely trying character of the work of taking readings, the 
attempt to utilise this method was finally abandoned. 
( 4 ) Mercury. 5790-49, 5769-45, 5460-97, 4358-58. 
The mercury arcs used in these experiments included— 
{a) A Bastian lam%). The arc is here formed in a zigzag glass tube 11 inches long 
