408 M. H. Becquerel's Experimental Investigations 



even the unknown rotation. It was therefore necessary to 

 compare each of the rotations with the average wave-length 

 which corresponded to it. 



It has been stated that the results obtained with different- 

 coloured rays are very nearly in the inverse ratio of the 

 square of the wave-lengths; and the departures from this 

 simple law are of the order of errors of observation. This 

 observation, then, enabled us to employ, for the determi- 

 nation of the magnetic rotation corresponding to luminous 

 rays of a given wave-length, for example the yellow rays D, 

 all the observations relative to the same gas, which in the 

 case of air and olefiant gas amounted to a hundred and fifty. 



Now the average of a hundred observations made with the 

 greatest care possible may be considered as ten times more 

 exact than a single isolated measurement ; it is clear, then, 

 that for the numbers in question, an accuracy of y 1 ^ of the 

 minute of arc may be relied on. 



It will be shown (p. 428) that the numbers deduced, as 

 has just been said, from observations relative to various 

 colours do not differ amongst themselves by more than y 1 ^ for 

 air and 3^-Q for olefiant gas. 



In the Tables that follow, the numbers are expressed in 

 minutes of arc ; and although the first decimal figure cannot 

 always be relied on, I have given the results to two decimal 

 figures, as they resulted from the calculation of the averages. 



Experimental Determinations. 



It has just been shown what was the relative importance of 

 the different corrections which affect the direct observations. 

 Among them there is one that applies to all the other correc- 

 tions ; this is the variation in the electromagnetic intensity. 



It was ascertained that this varied only within restricted 

 limits; and in order to render the numbers comparable, we 

 made this correction first of all. The tables which follow 

 contain only numbers thus corrected. The values of the other 

 corrections are given in the tables themselves, in order that 

 their relative importance may be justly estimated ; for they 

 allow of the same causes of error as the observations them- 

 selves. We now proceed to examine successively the results 

 obtained with different gases. 



Atmospheric Air. — The observations relative to atmospheric 

 air were numerous, first, because the magnetic rotations of 

 this body constituted an important physical datum, and, 

 secondly, because it was possible to effect the determina- 

 tions without interposing the tube-glasses in the path of the 

 light, and thus to verify the numbers adopted for the magnetic 



