158 Prof. E. W. Wood. 



higher frequency than D 2 being less than one, and for all waves of 

 lower frequency than DI, greater than one. In other words, I have 

 been able to form a complete anomalous spectrum by means of sodium 

 vapour, in which all of the colours between the extreme red and 

 violet are present, with the exception of a range of perhaps twenty 

 Angstrom units at the D lines. 



Previous investigations have been made exclusively with sodium 

 flames, which have been made to assume a prismatic form. In view 

 of the great uncertainty regarding the prism angle, density of the 

 vapour, and its molecular condition, that is, whether dissociated or 

 not, it seemed advisable to employ non-luminous vapour, obtained by 

 heating metallic sodium in some inert gas or in vacuo. 



The first experiments were made with a large prism of cast iron, 

 furnished with windows of mica or thin plate glass, in which 'the 

 metal was heated in an atmosphere of hydrogen. Very beautiful 

 results were at once obtained, but certain peculiarities of the vapour's 

 action showed that the refraction was due chiefly to the action of a 

 non-homogeneous medium, the planes of constant density being hori- 

 zontal. Great trouble was had with the windows, which soon became 

 covered with a white deposit, which cut off most of the light. It 

 being apparent, however, that the oblique faces played but a very 

 small part, the effect being due almost wholly to the variable density 

 of the vapour, it seemed best to make the most of this circumstance, 

 and dispense with the trouble entirely by removing the glass plates to 

 such a distance from the heated vapour that no deposit took place. 



FIOK 1. 



The arrangement finally adopted was simply a tube of glass about 

 30 cm. long, provided with plate-glass ends cemented on with sealing- 

 wax. Hydrogen, dried by passage over calcium chloride, was con- 

 ducted into and out of the tube by means of two fine glass tubes 

 arranged as shown in fig. 1. 



The diameter of the tubes should not be over 2 mm., and they 

 should lie close against the sides of the large tube in order not to cut 

 off any of the light. The most suitable diameter for the large tube is 

 2 cm. As the experiments to be described in this paper will, I feel 



