340 Profs. Live big and Dewar, On the use of [April 30, 



the angle through which the grating is turned, that is to say the 

 circle must be well divided. Readings on the two sides may still 

 be taken and will help to eliminate errors of reading and errors 

 due to inequalities of temperature and so on. For a given angular 

 deviation the dispersion is double what it is in the old method 

 which facilitates accurate reading ; and it often happens that a 

 spectrum can be observed of an order which involves a deviation 

 greater than any which can be observed in the old method. Thus 

 with a grating which when fixed perpendicular to the collimator 

 will only shew the D lines in the first and second orders, it may 

 be possible to observe them with a collimating eye-piece in the 

 5th order. Moreover gratings are usually very unequal on the 

 two sides. The Rowland's grating we have used gives very feeble 

 spectra in orders above the second on one side, bright spectra of 

 those orders on the other side, and it is a great advantage to be 

 able to use the bright side alone. By way of example of the 

 accuracy easily obtainable by the use of the collimating eye-piece 

 we may cite one determination. Readings were taken of the b group 

 of magnesium lines in the 3rd, 4th, and 5th orders on both sides, 

 that is six observations of each line, the circle being divided in 

 10' and the vernier reading half-minutes. The mean of the two 

 readings of the same line in the same order on the two sides did 

 not in any case differ from the single reading by half a minute. 

 The wave-lengths deduced from the whole set were 



K K b 4 



5183-14 517220 516662 



Angstrom's Nos. 5183-10 5172-16 5166*88 



In comparing our numbers with those of Angstrom it was of 

 course necessary to determine the distance between the lines of 

 the grating according to the standard of length employed by 

 Angstrom. This was done by observations of the deviations of 

 the D and E lines in the spectra of a sodium flame and of 

 a spark between iron points respectively, as well as in the solar 

 spectrum. For spectroscopic purposes it appears to us to be 

 far the best plan to adopt Angstrom's standard of length, -whether 

 his millimetre be the most accurate attainable or not, and when 

 we get a more exact determination of absolute wave-length, 

 all numbers on Angstrom's scale can be reduced to the new 

 scale by multiplication into a constant factor. For the purpose 

 of reducing the grating space to Angstrom's scale Mendenhall 

 has proposed the use of a rather feeble iron line below E, but 

 it seems to us far better to use the mean of the two E lines, 

 because this was Angstrom's standard line on the measurement 

 of which he bestowed the greatest pains ; besides E„ is a strong and 



