Stoney — Cause of Double Lines in Spectra. 607 



But this great work will lose mucli of its availability for such inquiries as the 

 present, unless it be accompanied by equally exact determinations of the refractive 

 indices of air throughout the spectrum. We cannot even verify Balmer's law 

 without these essential co-efficients. 



POSTSCRIPT. 



A good illustration of the time-relations of the motions that are concerned in the 

 production of spectral lines can be very simply made by screwing a small hook 

 into the middle of the lintel at the top of a doorway, and hanging a heavy bob 

 from it by a piece of silk or pack-thread of such a length that the middle of the bob 

 is 39 inches from the hook — about as long as an ordinary door is wide. The 

 oscillation period of this conical pendulum will be two seconds, the same as that 

 of a pendulum beating seconds. 



Place a table in the doorway, and on it some kind of pointer, such as a candle 

 or bottle, supported by a box if necessary to make it reach nearly to the bob. 



Now start the bob in a long (approximate) ellipse. We may take this to 

 represent the motion of the electron in a molecule of sodium, as it swiftly revolves 

 in that elliptic partial which produces the great yellow sodium line. The ellipse 

 of our conical pendulum will be seen to have an apsidal motion owing to the 

 resistance of the air. It is in the same dii'ection as the revolution of the bob in 

 the ellipse. This is the right direction to represent the apsidal motion which takes 

 place in the molecule, but it is probably too swift. The apsidal circuit in our 

 apparatus may perhaps be completed in some five or ten minutes, whereas, to 

 correspond with the real event in the gas, it should take l"^ 6° 8^ to get through 

 each revolution. Further, the bob parts with its energy to the surrounding medium 

 far too hastily, and will perhaps come to rest in less than an hour. It should be 

 able to persist in describing its orbit for several months, to be like the electron. 

 There is, however, no difficulty in making allowance for these defects. We should, 

 then, suppose the bob to be given a fresh impulse some eight or nine times every 

 fortnight, to represent on the time-scale that we have chosen the recurrence of the 

 encounters between the molecule and its neighbours, which from time to time 

 revive its internal motions. Finally we are to imagine our pendulum kept going 

 in this way without intermission for thirty-two years ; by which time the number 

 of the several representative events in the apparatus will have just accumulated up 

 to being the same as the number of the corresponding actual events that are going 

 on within each molecule of the vapour of sodium in every millionth of a second. 



TEAIfS. EOT. rUB. SOC, N.S. VOL. IT., PAUT XI. 4 R 



