352 Mr. Louis Bell on the Absolute 



Combining and weighting, the mean value is 



<£ = 42 o 4'59' / -28 + 0"-2. 



The probable error is equivalent to about one part in six 

 hundred thousand in the wave-length. 



Both the glass gratings were used exclusively for the line 

 Di, which was on the whole most convenient for measure- 

 ment, D 2 being rejected by reason of the troublesome atmo- 

 spheric lines. The relative wave-lengths of a very large 

 number of lines have been so exactly determined by Prof. 

 Rowland that any one of them would have given results 

 equally valuable, and in the subsequent work with gratings 

 III. and IV., two of these standard lines were employed. 



In this second part of the investigation, the gratings as 

 before mentioned were used on the large spectrometer in 

 which the telescopes were kept at a fixed angle and the 

 grating was turned. This method is, of course, applicable 

 only to very solid instruments in which the angle can readily 

 be kept constant, and it should be further noted that it also 

 requires the use of very perfect gratings, since the grating is 

 used asymmetrically. As a result of this the spectra on the 

 two sides differ in dispersion; and if the ruling is irregular 

 either in spacing or in contour of the individual lines, may 

 differ quite widely in focal-length definition and illumination. 

 After critical examination gratings III. and IV. appeared to 

 be so nearly perfect in ruling as to be quite secure from the 

 dangers of the method. The method has moreover the dis- 

 tinct advantage of enabling the angle of deviation to be 

 varied within certain narrow limits. Hence it becomes pos- 

 sible so to arrange the apparatus as to give to some con- 

 venient line a double deflection that shall be an exact sub- 

 multiple of 360°. This once accomplished it becomes an easy 

 matter completely to eliminate the errors of the divided circle 

 and obtain a value of n<f> dependent only on the micrometer 

 constants, which in turn may be themselves almost eliminated. 

 To be sure, this method practically confines observations to 

 the spectra of a given order and limits the choice of lines for 

 measurement, but the first objection does not apply to gratings 

 of which the ruling is very nearly perfect, and since the re- 

 lative wave-lengths of a large number of lines are known with 

 very great exactness, measurements of the absolute wave-length 

 are quite comparable even if made on different lines. 



As regards the constancy of the angle between the colli- 

 mator and observing telescope, there was every reason to ex- 

 pect entire permanence throughout the experiments ; and 

 observation soon justified this expectation. The telescopes 



