40 uemokiIì; della società' 



of the spectrum of the third order, these two poiuts will be fonnd to bave cast 

 shadows, the former ou E^ of the 3*'^ order, and the latter near of the 3"^^ or- 

 der. By a siinple calculatiou it is therefore proved that a certain line in Hj has 

 the wave length 3930.1, and another line near the wave length 3444.6. Hence 

 I could secare in my photographs three fixed points, one at each eHd and one in 

 the middle. 



If therefore, by means of a linear dividing engine, the spectrum is provided with 

 a scale, in which the wave-leugths of the ruled divisions coincide with the wave- 

 lengths of the three fixed points , the correct vaine for any ray may be read di- 

 rectly. 



Accordiugly I bave ruled my scale in this way, but I bave fonnd, in carrying 

 the process into effect, that one or more of the three determinatious of wave-leugths 



o 



by Angstrom must be slightly incorrect, because I cannot secure exact coincidence 

 with ali the fixed points at once. The error, if it were attributed to either the wave- 

 length of D2 , or G, wonld only reqnire to be abont one ten-millionth of a millime- 

 tre , and on comparing Angstrom with Mascari, Ditscheiner ftnd van der Willigen, 

 much greater discrepancies are observed. 



Even with this error, over a largo part of the photograph the reading is correct 

 within a fractiou of a single division , and in those portions where Angstrom has 

 drawn by hand the sanie region that the photograph depicts, the coincidence of 

 the largo lines is beantifally shown, while the errors in the minor lines in the chart 

 of the great Swedish physicist are plainly exhibited. Por example, he shows no li- 

 nes between Hj and 3988, while in the photograph there are more than 20; so also 

 between 4101 and 4118 he has no lines, while the photograph has 17; from 3925 

 to 4205 his chart exhibits 118 lines, while my originai negative has at least 293. 



Again , as illustrating the difiìculty of depicting the relative intensity of lines, 

 Angstrom has drawn 3998 nearly equal to 4004, in reality it is mach fainter, and 

 ÌDstead of being single, is triple; this is very visible in the enlarged picture. If the 

 reader will examine the relative positions to the scale of wave-lengths, he will also 

 find, that while the large lines generally coincide the smaller ones in many cases 

 differ to the extent of nearly a whole division. 



But it is surprizing that Angstrom's scale should be so correct when we cousider 

 the great difficulties against which he had to conteud, for it requires a criticai in- 

 vestigation to detect any errors at ali. 



On comparing the wave-lengths that my scale attributes to the ultra-violet rays 

 L, M, N, with the wave lengths that Mascari assigns to those lines, et appears that 

 both of US give nearly the same vaine to M, but his values of L and N are less 

 than mine by 2 and 3 divisions. When however we examine the processes by which 

 the two results Avere reached, mine being a direct photograph with a machine ru- 

 led scale , and his being derived from minute photographs matched together , and 

 read by a microscope, I fiel confident that mine are not incorrect to any great de- 



