WOOD. — ANOMALOUS DISPERSION OP SODIUM VAPOR. 379 



These values are of course only approximations, the errors becoming 

 greater as we approach D2. 



The same experiment was repeated, using a filar micrometer in the 

 eye-piece in place of the reticulated mesh, and values obtained which 

 agreed fairly well with those given above. This method, however, did 

 not admit of such a near approach to D2 as the other. 



From these results it is possible to calculate the values of the refractive 

 index within this range for the very dense vapor used in the absolute 

 determinations. These will be called observed values, and they will bo 

 compared with the values calculated from the dispersion formula for the 

 corresponding wave-lengths. 



I have already spoken of the enormous dispersion at the helium line, 

 a dispersion great enough to separate a Zeeman doublet by an amount 

 equal to nearly the width of the entire visible spectrum produced by a 

 glass prism of sixty degrees, provided we were able to construct a prism 

 of sodium vapor giving the same mean deviation, and get the lii^lit in 

 question through it. Great as is this dispersion, it is as nothing when 

 compared with that within say an Angstrom unit of one of the D lines, 

 as an inspection of the above table indicates. 



The dispersion within the range between Do and D3 is shown graphi- 

 cally in Curve 2, Plate II. 



If this curve were completed up to wave-length 5889.6, at which the 

 largest recorded deviation occurs, the point in question would be at the 

 thirty-eighth square below the unity line. To include this point it would 

 have been necessary to take the ordinates of the curve on about I of the 

 scale employed. 



Visual Observations of the Dispersion in the Visible 

 Spectrum by the Method of Crossed Prisms. 



As a check on the results obtained with the interferometer, a series of 

 measurements of the dispersion was made by the method of crossed 

 prisms. The method was identical with that described in my previous 

 paper, except that the metal was heated in a highly exhausted tube of 

 hard Jena glass instead of in a tube filled with hydrogen. These tubes 

 can be very easily prepared, and give no trouble at all. They are, how- 

 ever, apt to crack if reheated, after having been allowed to cool. It 

 is not necessary to have the sodium in separate small lumps, as the 

 action is not the formation of a number of prisms of sodium vapor, but 

 a cylinder of vapor, very deuse along the heated floor of the tube, and 

 decreasing to very nearly zero along the top. If it is desired to prepare 



