WOOD. — AN03IAL0US DISPERSION OF SODIUM VAPOR. 367 



forms, the glass is attacked and rendered opaque by the reduction of 

 silica. The vapor, however, has a most remarkable viscosity, which I 

 am at the present time investigating, which makes it possible to form a 

 mass of great density separated from the glass plates which close the 

 ends of the tube by a high vacuum. If we place a number of pieces of 

 clean sodium in a tube of hard Jena glass, the ends of which are closed 

 with small pieces of thin plate glass, and exhaust the tube on a mercurial 

 pump, on heating the under side strongly with small Bunsen flames, the 

 sodium vapor shows very little inclination to distil to the cold parts of the 

 tube. It condenses, to be sure, on the upper side of the tube, but is given 

 off so much more rapidly from the surface of the molten metal than it can 

 diffuse to the upper portion, that the density gradient is very steep. Ob- 

 servations on the deviation produced by the non-homogeneous cylinder 

 show that the equivalent prism has a form similar to that shown in Fig- 



FlGURE 1. 



ure 1, the density gradient being steeper near the bottom of the tube. To 

 secure good definition it is therefore necessary to place in front of the 

 tube an opaque screen perforated with a wide horizontal slit. The tube 

 thus prepared, used in the manner to be presently described, shows the 

 strong anomalous dispersion in the vicinity of the D lines with great dis- 

 tinctness. If the flames are made to play upon the upper surface of the 

 tube, after some of the vapor has condensed there, the spectrum straight- 

 ens out, and the dispersion presently occurs in the opposite direction, 

 showing that the base of the prism is always against the heated portion 

 of the tube regardless of gravity. 



In a tube 25 cms. long, 1 cms. of which are heated red hot on all 

 sides by means of an iron wire carrying a heavy current, a mass of sodium 

 vapor may form of sufficient density to give a displacement of several 

 thousand helium (D3) fringes; notwithstanding this, practically no trace 

 of the vapor can be detected beyond the heated portion. This condition 

 can be maintained for an hour or more, owing to the slowness of the dif- 



