THE STRUCTURE OF THE NUCLEUS. 29 



Though it has not been possible to catch tliese colors with sufficient sharpness 

 to pronounce them complementary, their sti'ong and persistent contrast will at once 

 be admitted. 



10. Polarized light. — All the colors are produced equally well by polarized 

 lio-ht. They are wiped out entirely i)etween crossed nicols. The flame color is 

 sometimes seen under these conditions, probably due to diffuse light which escapes 

 throuffh the polarizer or is depolarized at the windows of the apparatus. It is not 

 possible to keep these clear during a long series of experiments, and they must be 

 assumed to pi'oduce a ground-glass effect. Strengthened axial coloi' may moreover 

 be a subjective illusion. 



11. Geometric seque7ices of coronas. — I shall first consider the distribution of 

 coronas as i-elated to the loss of nuclei by exhaustion.' Kiiev the preliminary 

 I'esults of tables 1-3, I will proceed to investigate data of a quantitative character, 

 sei'vino- to distribute the coronas in a scale of decreasing numbers of nuclei. It will 

 not be feasible to arrive at the complete factor at once ; for the nural)er of nuclei 

 must at tlie outset be supposed to vary both with the loss or drain due to the suc- 

 cessive equal exhaustions, and with causes independent of manual interference, such 

 as are involved in the motion of the nucleus and its possible decay. The absorp- 

 tion of the nucleus by the walls of the vessel, its subsidence by gravity when 

 loaded, etc., are here included. The experiments are necessarily complicated and 

 according as one places more stress on the incidental or the normal causes, different 

 conclusions are to be drawn. It is the purpose first to determine the most potent 

 and obvious cause of dissipation. 



Accordino'ly in table 4 the effect of exhaustion alone is first fully ti-eated. 

 The laro-e globe was exhausted and refilled with filtered air about 20 times in suc- 

 cession, the pressures falling off suddenly from normal to about 18 cm. less. To 

 secure efficient filtering of the air which supplied the place of that removed, very 

 slow influx through the compressed cotton and the check valve was maintained. It 

 is assumed that with each exhaustion a definitely decreasing number of nuclei are 

 removed with the air. Thus after z isothermal exhaustions from the pressui-e^^^ to 

 the pressure J?, the residue of nuclei should be {p/l\)^; Jitter z adiabatic exhaus- 

 tions between the same limits {p/p^), ''\ where y is the ratio of specific heats, ad- 

 mitting (which is by no means the case as is afterwards shown), that the whole 

 experiment is made expeditiously enough to neglect the time losses of nuclei due to 

 the normal causes mentioned. In order to give greater probability to this assump- 

 tion I selected large exhaustions, from p^ — 76 cm. to J9 = 58 cm., or a mean pres- 

 sure decrement of 18 cm. 



In two successive experiments which showed a reasonable order of agreement, 

 the color sequences of the coronas were observed from the center outwards, while 

 the correspondina: flame or axial colors were simultaneously noted ; but here there 

 is uncertainty from the small relative thickness (30.5 cm.) of the axial layer of water 



' Loss due to subsidence of fog is not considered here but will be treated in Chap. Ill, §§ 5, 

 7, II. The time during which fog is in evidence must be made as short as the observations will 

 permit. As a rule the corona is lost by evaporation, at once. 



