50 Prof. H. A. Rowland on a 



As the relative values are more important for spectroscopic 

 work than the absolute , I take this value without further 

 remark. It was utilized as follows : — 



1st. By the method of coincidences with the concave 

 grating, the wave-lengths of 14 more lines throughout the 

 visible spectrum were determined from this with great accuracy 

 for primary standards. 



2nd. The solar standards were measured from one end of 

 the spectrum to the other many times ; and a curve of error 

 drawn to correct to these primary standards. 



3rd. Flat gratings were also used. 



4th. Measurements of photographic plates from 10 to 19 

 inches long were made. These plates had upon them two 

 portions of the solar spectrum of different orders. Thus the 

 blue, violet, and ultra-violet spectra were compared with the 

 visible spectrum, giving many checks on the first series of 

 standards. 



5th. Measurements were made of photographic plates 

 having the solar spectrum in coincidence with metallic 

 spectra, often of three orders, thus giving the relative wave- 

 lengths of three points in the spectrum. 



Often the same line in the ultra-violet had its wave-length 

 determined by two different routes back to two different lines 

 of the visible spectrum. The agreement of these to y^- 

 division of Angstrom in nearly every case showed the accuracy 

 of the work. 



6th. Finally, the important lines had from 10 to 20 measure- 

 ments on them, connecting them with their neighbours and 

 many points in the spectrum, both visible and invisible ; and 

 the mean values bound the whole system together so intimately 

 that no changes could be made in any part without changing 

 the whole. 



This unique way of working has resulted in a table of wave- 

 lengths from 2100 to 7700 whose accuracy might be estimated 

 as follows : — 



Distribute less than y^ division of Angstrom properly 

 throughout the table as a correction, and it will become 

 perfect within the limits 2400 and 7000. 



The above is only a sketch of the methods used. The 

 complete details of the work are ready for publication, but I 

 have not yet found any journal or society willing to under- 

 take it. 



Description of the Table. 



The first column gives the name of the element whose 

 wave-length has been measured. If a letter stands at the 

 left, it is the " name " of the line in the solar spectrum. A(o) 



