138 MESSRS. C. CUTHBERTSON AND E. PARR METCALFK 



potassium evaporates, it attacks lx)th glass and silica, so that it is not possible to 

 obtain, in vessels of these materials, a density of the vapour sufficient for a quanti- 

 tative determination of the absolute index. Eventually, however, by adopting the 

 device of heating all but a small portion of the refractometer tube to a sufficient 

 temperature, and then rapidly heating the cool part, a sufficient density of vapour 

 was attained to permit of some qualitative observations. 



Since reliable numerical results could not be looked for, no attempt was made to 

 purify the potassium beyond distilling it in vacua after prolonged heating to expel 

 occluded hydrogen. The distillate was run into capillary tubes, which were sealed 

 off in short sections, and, by a device which need not be described, the operations of 

 breaking the capillary and introducing it into the refractometer tube were performed 

 entirely in vacuo. 



Experiments were, at first, made with sodium light, but it was found that at a 

 temperature of about 220 C. the interference bands disappeared, though the light 

 was not wholly absorbed by the vapour. This effect might be accounted for by 

 supposing that the potassium contained a small quantity of sodium, or that sodium 

 was set free by the action of the potassium on the glass tube, for WOOD has shown 

 that the great dispersion of the vapour of sodium in the neighbourhood of the D 

 lines has the effect of destroying interference bands formed by light of those wave- 

 lengths, when even small quantities of vapour are present. 



The sodium flame was therefore replaced by a Bastian mercury lamp which gave 

 good bands with the wave-length 5460. Several experiments were made with this 

 light with refractometer tubes both of glass and of silica ; and it was found that 

 the evaporation of the potassium was attended by a movement of the bands in the 

 direction corresponding to a refractive index less than unity. On one occasion no less 

 than four bands were observed to pass during the heating and to return during the 

 cooling. 



These experiments at first suggested that the quantity of sodium present was 

 greater than had been suspected, and that its influence on the index dominated that 

 of the potassium. If this were so we should find that an experiment with light of 

 greater wave-length than that of the D lines would give a very high refractive index. 

 In order to test this surmise, experiments were made with the red light supplied by a 

 hydrogen vacuum tube whose H a line was so strong as to give good interference bands. 

 But in this case also the bands moved in the same direction as those formed by the 

 light of the mercury lamp. We are, therefore, driven to the conclusion that the 

 index of potassium is less than unity both for X = 5460 and for X = 6562. 



These results showed that no experiments on the indices of the alkali elements 

 were likely to repay the labour involved until the chemical difficulties had been 

 overcome and arrangements could be made for obtaining the index for very long 

 waves. It was, therefore, determined to abandon the attempt for the present, and to 

 turn to some other series. 



