WOOD. — ANOMALOUS DISPERSION OF SODIUM VAPOR. 373 



trace of sodium vapor was formed. A different method was accordingly 

 adopted, whicli yielded excellent results. 



It is clear that if we employ light of two different wave-lengths in 

 illuminating the interferometer, the introduction of the metallic vapor 

 will shift the two fringe systems by different amounts, the fringes ap- 

 pearing and disappearing as they get into and out of step in succession. 

 If both systems are shifted in the same direction, which is the case when 

 both wave-lengths are on the same side of the absorption band, we can 

 measure the dispersion by counting the number of fringes which pass the 

 cross-hair of the telescope between two successive disappearances. If, for 

 example, the number is found to be fifteen, we know that one set of waves 

 is retarded or accelerated fifteen wave-lengths, under the same conditions 

 which cause a shift of sixteen fringes for the other waves. All uncer- 

 tainty as to the movement of the fringes during a period of minimum 

 visibility was avoided in the method employed, as will appear presently. 



To determine the dispersion close to the D lines we require highly homo- 

 geneous light of two different wave-lengths, which would yield a close 

 double line at the point at which we wish to measure the dispersion. 



This was accomplished by breaking up the radiation of a helium tube 

 into a Zeeman triplet and cutting out the middle component by means of 

 a Nicol prism placed with its plane of vibration perpendicular to the 

 lines of magnetic force. A large Rhumkorff magnet with an ammeter 

 in circuit was employed, the key for making and breaking the circuit 

 being placed close to the observing telescope of the interferometer. 

 The distance between the components of the double line obtained when 

 a current of 30 amperes was flowing through the magnet was found to 

 be 5^^ of the distance between the sodium lines. This was determined 

 very easily by counting the number of turns of the interferometer screw 

 necessary to cause the fringes to pass through a given number of phases 

 of maximum visibility, the method being the one commonly employed in 

 the laboratory in the determination of the relative wave-lengths of the 

 sodium lines. The magnetic field was now thrown off, and the temper- 

 ature of the heating coil gradually raised. As the sodium vapor formed 

 the fringes were counted as they drifted across the field. The current was 

 thrown into the magnet at frequent intervals for a fraction of a second, 

 and the fringes were seen to pass through successive phases of visibility 

 and invisibility. The momentary disappearance caused by the magnetic 

 field did not interfere with the counting, for the fringes moved slowly 

 and with the regularity of clockwork. The exact moment at which the 

 fringes disappeared could not of course be accurately determined, since 



