296 ON THE WIDENING OF SPECTRUM LINES }i fr, li who observed the limit of interference when tubes containing helium, n and krypton were cooled in liquid air. Under these conditions ba which had already disappeared at room temperature again became disti and the ratios of maximum retardations in the two cases (1'66, 1'60, 1 were not much less than the theoretical 173 calculated on the supposition 1 the temperature of the gas is that of the tube. The highest value of X/S their notation N, hitherto observed is 950,000, obtained from kryptoi liquid air. With all three gases the agreement at room temperature betw the observed and calculated values of N is extremely good, but as alre remarked their theoretical numbers are a little lower than mine (14). may say not only that the observed effects are accoimted for almost comple by iDoppler's principle and the theory of gases, but that the temperatur the emitting gas is not much higher than that of the containing tube. As regards m, no question arises for the inert monatomic gases. In case of hydrogen Buisson and Fabry follow Schonrock in taking the a rather than the molecule as the moving source, so that m = 1; and furl they find that this value suits not only the lines of the first spectrun hydrogen but equally those of the second spectrum whose origin has so times been attributed to impurities or aggregations. In the case of sodium, employed in a vacuum-tube, Schonrock found a agreement with the observations of Michelson, on the assumption that atom is in question. It may be worth while to make an estimate for th lines from soda in a Bunsen flame. Here m = 23, and we may perhaps i T at 2500. These data give in (14) as the maximum number of bands X/A = 137,000. The number of bands actually seen is very dependent upon the amount of £ present. By reducing this Fizeau was able to count 50,000 bands, an would seem that this number cannot be much increased*, so that observa falls very distinctly behind calculationf. With a large supply of soda number of bands may drop to two or three thousand, or even further. The second of the possible causes of loss of homogeneity enumerated ab viz. rotation of the emitting centres, was briefly discussed many years ag a letter to Michelson J, where it appeared that according to the views t * "Interference Bauds and their Applications," Nature, Vol. XLVIII. p. 212 (1893); Scie Papers, Vol. iv. p. 59. The parallel plate was a layer of water superposed upon mercury, enhanced illumination may be obtained by substituting nitro-benzol for water, and the reflex from the mercury and oil may be balanced by staining the latter with aniline blue. But a layer of nitro-benzol takes a surprisingly long time to become level. f Smithells (Phil. Mag. Vol. xxxvii. p. 245, 1894) argues with much force that the act operative parts of the flame may be at a much higher temperature (if the word may be admi than is usually supposed, but it would need an almost impossible allowance to meet the crepancy. The chemical questions involved are very obscure. The coloration with soda ap to require the presence of oxygen (Mitcherlich, Smithells). t Phil. Mag. Vol. xxxiv. p. 407 (1892); Scientific Papers, Vol. iv. p. 15.