THE SPECTRUM OF HYDROGEN 245 



the spectrum, if any, are unknown. It is now usually assumed 

 that molecules emit band spectra, while series spectra are 

 characteristic of atoms. Hg is the simplest possible molecule, 

 and its spectrum must be the prototype of band spectra, as the 

 Balmer series is of series spectra. The spectrum emitted by an 

 atom or molecule is at present the best weapon we possess for 

 the attack upon its constitution. It is, therefore, justifiable to 

 assume that the secondary hydrogen spectrum is destined for 

 the same part in the elucidation of molecular mechanics that 

 the Balmer series has already played in the investigation of the 

 atom. Several attempts have already been made to calculate 

 the wave-lengths of the secondary lines from various models of 

 the hydrogen molecule, but in all published instances the cal- 

 culated and observed values have disagreed completely. The 

 reverse procedure of devising a model of the molecule from a 

 study of the spectrum has so far been impossible, because too 

 little has been known of the structure of the spectrum itself. It 

 was on account of its probable theoretical importance that Prof. 

 Merton and the writer have recently {Phil. Trans., 1922) carried 

 out an extended investigation of this spectrum, in the hope of 

 preparing the way for its discussion by the mathematical 

 physicist. The work has included a study of the wave-lengths 

 and other characteristics of the lines, and a proof that the 

 spectrum is actually due to the hydrogen molecule, about which 

 there had previously been some doubt. 



It has already been stated that the secondary spectrum is 

 very rich in lines. There is in fact an average interval of only 

 two or three Angstrom units between the lines which are 

 included in the range X\ 6,000-4,000 A. It is necessary to 

 have accurate wave-length measurements in such a spectrum, 

 before searching for mathematical relations among the lines, 

 for the ease of finding spurious regularities increases rapidly 

 with the probable errors in the wave-length tables and the 

 number of lines in the spectrum. The wave-lengths of as 

 many lines as possible were accurately determined to diminish 

 ambiguities of this type. It is, perhaps, worth while describing 

 the procedure adopted in such measurements. Modern wave- 

 length measurements are always made from spectrum photo- 

 graphs, and not visually. The light is dispersed by a Rowland 

 concave grating, and the spectrum photographed on a plate 

 suitably sensitised over the required spectrum range. The 

 spectrum of the iron arc is next photographed on the same 

 plate, which must not be moved in the interim. The unknown 

 wave-lengths can then be deduced from the positions of the lines 

 relative to certain iron lines, the wave-lengths of which have 

 been determined accurately once and for all, and are accepted as 

 international standards. The positions of the lines on the 



