204 Mr. S. P. Langley's Experimental Determination of 



E. Keeler, of this Observatory, and which seem to me favour- 

 able specimens of the accuracy attainable by this method. 



We are now prepared to test the accuracy of the various 

 formulae connecting refraction with ivave-length, though it will 

 be convenient to first prepare a table showing what this rela- 

 tion is in the visible part of the spectrum of the prism employed. 



In the following table the deviations in the visible spec- 

 trum were measured by the spectrometer, reading to 10" of 

 arc, and which has been already described ; in which for this 

 special purpose the bolometer was replaced by an achromatic 

 observing-telescope with a micrometer eyepiece, and the indices 

 of refraction were computed by the usual formula. " " in 

 the ultra-violet was measured by aid of a Soret fluorescent 

 eyepiece, and its wave-length is from Cornu. The other wave- 

 lengths are taken from Angstrom, but the unit is here the 

 micron = joVo millimetre == 10,000 times the unit of Ang- 

 strom's scale). "X " is here the symbol for the wave-length. 



The following indices in the visible spectrum, on which the 

 computations for testing the formulae are founded, are trust- 

 worthy to the fourth decimal place here given : — 



Table III. — Observed Indices in Visible Spectrum of 

 Hilger Prism. 



Line. 





d. 



n. 



A 



0-76009 

 0-65618 

 0-58890 

 0-51667 

 0-48606 

 0-39679 

 0-34400 



46 49 05 



47 15 45 



47 41 15 



48 21 05 

 48 44 15 

 50 34 05 

 52 43 00 



1-5714 

 1-5757 



1-5798 

 1-5862 

 ] -5899 

 1-6070 

 1-6266 



O 



D, 



i 



b. 



4 



F 



H, 



O 





A smooth curve drawn through points whose positions are 

 given by the above table, represents with accuracy the relation 

 between n and X in the visible part of the spectrum. This 

 method is, however, obviously inapplicable to the very ex- 

 tended invisible portion below the A line ; and accordingly 

 attempts were first made to effect the determination of corre- 

 sponding indices and wave-lengths, by extending the curve 

 derived from the above observations by means of formulae. 

 Several formulae have, it will be remembered, teen proposed 

 by physicists, expressing n as a function of X, and containing 

 constants which are to be determined by observation. But it 

 has never hitherto been possible to test these formulae far from 



