386 PROCEEDINGS OF THE AMERICAN ACADEMY. 



that a fairly uniform cylinder of vapor could be formed in an exhausted 

 tube, the absolute value of the index for any given wave-length could 

 only be guessed at. 



The results obtained on the dispersion with the interferometer showed 

 that results correct to within 5 per cent could in all probability be obtained 

 by this method. 



The density of sodium vapor at different temperatures has been inves- 

 tigated by Jewett * for a range comprised betweed 368 and 420. It was 

 accordingly determined to take the temperature of the vapor as a measure 

 of its density. 



A thermo-couple of iron and constantan wire was made by holding 

 the ends of the wires together, both being connected to a storage battery 

 giving an E.M.F. of about twenty volts. The other wire of the battery 

 was then touched momentarily to the tips of the two wires, the small 

 arc which formed on separation fusing them together very neatly. I do 

 not remember to have seen this method of making thermo-couples de- 

 scribed, and have mentioned it as it was found very satisfactory. The 

 fused bead which united the wires was then hammered out into a 

 thin disc, and the junction mounted in the axis of the sodium tube, the 

 wires being insulated with thin glass tubes which passed through a larger 

 tube, as shown in Figure 4. 



The couple was calibrated in baths of molten lead, zinc, tin, and 

 aluminium before and after the experiment. The determinations at the 

 lower temperatures and densities were made with the light from the 

 helium tube; those at high temperatures, with the green light from a 

 mercury tube. This was necessary, since a mass of very dense sodium 

 vapor not only absorbs helium light strongly, but owing to its high dis- 

 persion makes the fringes invisible, the different wave-lengths in the D3 

 line (which is of course of finite width) being shifted by different amounts. 

 If the increase in the density with increasing temperature followed the 

 same law as in the experiments of Jewett, the optically determined 

 density curve should agree with that found by gravimetric method era- 

 ployed in his work. For example, he found that the density at 412° 

 was six times as great as at 385°, and we should consequently look for a 

 fringe shift at the higlier temperature six times as great as at the lower. 

 This, however, was by no means the case, the fringe shift being only about 

 twice as great. This is on the whole exactly what we should expect 

 under the conditions of the experiment, for we nuist remember that our 



* I'liil. Mag., i, 1902, page 551. 



