of the Centres of Emission of D Lines of Sodium. 1021 



position of the prism can be • determined by bringing these 

 two images into exact coincidence. This method obviates the 

 50 per cent, loss o£ light which results from the use o£ NicoPs 

 prisms. As the double-image prisms had apertures consider- 

 ably less than that of the quartz block, they were placed 

 midway between the condenser lenses and the image and 

 source, as shown in the figure. With prisms of too small 

 aperture, however, placed close to the source and image, the 

 separation of the images would be insufficient. 



The source of light was a Meker burner operated by an 

 air-blast charged with the spray of a very dilute solution of 

 sodium chloride (a saturated solution diluted with 1000 parts 

 of water) obtained by means of an atomizer. The burner 

 was surrounded by an iron chimney with a vertical aperture 

 measuring 25x5 mm., these dimensions giving the best results 

 with the polarizing prisms employed. The three images of 

 the aperture, the central one of twice the brilliancy of the 

 two lateral ones, were thrown upon the sodium bulb, which 

 was mounted in the current of hot air rising from a large 

 tube of fire-clay with a Meker burner at the bottom. The 

 sodium bulb was first coated with a 

 deposit of magnesium oxide by holding- 

 it above a burning ribbon of the metal. 

 The deposit was then carefully wiped 

 off with the exception of a rectangular 

 patch and a small horizontal strip as 

 shown in fig. 2. This facilitated the 

 adjustment of the polarizing separator 

 and the spectroscope, a constant devia- 

 tion instrument designed by Broca- 

 Pellin, similar to the type now supplied 

 byHilger. 



The bulb was rotated in its wire 

 supports until the central band of light (the two super- 

 posed images) fell upon the strip of magnesia ; the collimator 

 of the spectroscope was pointed at the illuminated strip, and 

 a sharp image of the latter formed on the slit by means of a 

 small Dunoyer condenser which is free from aberration if 

 monochromatic light is employed, as is the case in the present 

 work. 



The spectrum was now examined with an eyepiece. If 

 both D lines were visible the quartz block was rotated until 

 one or the other completely disappeared. It sometimes 

 happens that only D 2 appears at the top of the image, while 

 both Dj and D 2 are seen at the bottom : this results from the 

 circumstance that, owing to the finite size of the source 



