418 Prof. Wood on the Magneto- Optics of Sodium Vapour 



small errors in wave-length readings will make large errors 

 in values deduced from them. The severest test of all is to 

 calculate the rotation at the centre from values of the constants 

 determined from two medium values of 8 (say 90 or 180) . 

 outside of the D lines. This has been done in each case, and 

 in almost every case the calculated value has turned out too 

 small. I have given all of the tables of data thus far ob- 

 tained, and some of them are doubtless less accurate than 

 others. By means of an improved heating coil (which is in 

 process of construction), of porcelain and strip platinum, I 

 hope to be able to hold the conditions more nearly steady 

 and obtain still more accurate results. It will also be necessary 

 to secure greater rigidity of the grating. For large rotations 

 at the centre, say above 180, we cannot be quite sure of the 

 exact value, as the best that can be done is to regulate the 

 heat so that the condition at the centre is either of maximum 

 brightness or darkness. Errors ot say 15 or 20 degrees can 

 be made easily. It is quite possible too that the formula is 

 incomplete. It seems more than likely that the molecular 

 currents play some part, and that the formula built up on the 

 hypothesis of the Hall effect is incomplete. As we shall see 

 presently, however, it represents the rotation outside of the 

 D lines with a surprising degree of accuracy, while between 

 the lines it gives in most cases a curve which is elevated 

 somewhat above the experimental curve. It is possible that 

 the refractive index plays some part, though I regard this as 

 very doubtful. Midway between the D lines the refractivity 

 is equal to unity, rising and sinking rapidly to the right 

 and left of this point. It is not worth while to experiment 

 further with the formula until still more accurate data are at 

 hand, which I am confident of obtaining in the autumn. 

 Another matter still to be determined is whether the 



variation of the ratio y results from errors, or whether it 



increases with the density of the vapour, as appears to be 

 the case. These points will be taken up again in the second 

 paper. 



For densities giving rotations of 50°, 90°, and 180°, at the 

 centre, the calculated value agrees with the observed to 

 within the probable experimental error, for, as I have said, it 

 is impossible to fix the rotation at the central band to within 

 say 15. In the case of denser vapours the discrepancies are 

 much too large to be attributed to errors of this sort. 



Outside of the D lines the calculated values agree re- 

 markably well with the observed, even in the case of the 

 densest vapour used, as will be seen by referring to the 



