LYMAN. — THE SPECTRUM OF A CARBON COMPOUND. 317 



of the change will be understood by consulting the illustration on page 

 90 of volume 27 of The Astrophysical Journal. The brass collar A is 

 no longer provided with a screw thread as shown in the illustration, 

 but it is now made to fit into the cup B air tight by means of a cone 

 joint 2.8 cm. long. The discharge tube itself is no longer cast into the 

 collar A with Khotinski cement, but is blown on a platinum tube 

 3.5 cm. long by 1.5 cm. in diameter. This tube is soldered into the 

 collar A. By this arrangement the gas does not come in contact with 

 grease in the joints, and the danger of leak is considerably reduced. 



Measurements in the region between A 1850 and A 1675 where no 

 fiducial lines exist were made by the two slit method. 7 In the region 

 from A 1675 to A 1300 direct comparison was made with the spectrum 

 of hydrogen. 



The values of A refer to the heads of bands, and they are accurate to 

 0.3 of an Angstrom unit. In the class of the double bands marked 

 " d " the wave-length given is for the stronger component. The inten- 

 sities are represented on a scale of ten. The absorption of fluorite, 

 which begins to make itself felt near the end of the spectrum, renders 

 the relative intensities of the most refrangible bands rather uncertain. 

 As usual, the wave-lengths and frequencies are in vacuum. 



In addition to their value as standards of wave-length, the results 

 are of some theoretical importance. Deslandres in the paper just 

 quoted 8 has used his measurements of the carbon spectrum to test his 

 Laws. As the spectrum under discussion seems to form a continuation 

 of that described by Deslandres, it is interesting to see if its bands also 

 show the numerical relations described by the earlier investigator. 

 In making the comparison, however, it will be necessary to confine the 

 attention to those relations which deal with the heads of the bands, 

 for the dispersion employed does not permit of the study of the lines 

 of which each band is composed. It must also be remembered that 

 the region of high frequencies is not perfectly adapted to such a test, 

 since a small error in the wave-length is magnified in relations which 

 deal with frequencies. 



The laws under discussion are two in number: first, that a group of 

 bands may be broken up into sets of series such that the differences in 

 frequency of the heads of the bands in any one series form an arith- 

 metical progression ; second, that all the series are similarly constructed. 

 The first rule may obviously be stated in another way, — the second 

 differences of the frequencies of the heads of the bands in any one 

 series are constant. 



7 See note 2. 



8 Loc. cit. 



