552 HENRY A. BOWL AND 



When the flat side was parallel to the camera plate, a strip of the 

 spectrum three-sixteenths of an inch wide fell on the plate. When 

 turned ninety degrees, the plate shielded this portion and exposed the 

 rest. Using absorbents., it was thus possible to photograph a strip of 

 say the 4th spectrum between two strips of the 5th. This arrangement 

 is better than having only two edges come together. To correct any 

 movement of the apparatus during the time of exposure, I expose on 

 one spectrum, then on the other, and back again on the first. 

 Placing the negatives so obtained on a dividing engine with a micro- 

 scope of very low power and a tightly stretched cross-hair, the coin- 

 cidence of the two spectra can be measured. Owing to the large scale 

 of the photographs, about that of Angstrom, an ordinary dividing 

 engine having errors not greater than y^Vo" i ncn can be used, but the 

 negatives should be gone over at least twice, reversing them end for 

 end. Two screws were used in the engine and finally another com- 

 plete machine was constructed, giving wave-lengths direct with only a 

 slight correction. For determining the wave-length of metallic lines, 

 the same process can be used with wonderful accuracy. 



The results are given in the columns marked PL with the number 

 of the plates. The accuracy is very remarkable, and I think the figures 

 establish the assertion that the coincidence of solar and metallic lines 

 can be determined with a probable error of one part in 500,000 by only 

 one observation. 



This process not only gave me measures of the ultra violet, but also 

 new observations of the visible spectrum. So far in my work on these 

 coincidences, I have only used erythrosin plates going a little below D; 

 but cyanine plates might be used to B, or even in the ultra red, as Trow- 

 bridge has recently shown. One plate, No. 20, however, connects wave- 

 lengths 6400 and 3200. 



Thus I have constructed a table of about one thousand lines, more 

 or less, which are intertwined with each other in an immense number 

 of ways. They have been tested in every way I can think of during 

 eight or nine years, and have stood all the tests; and I think I can 

 present the results to the world with confidence that the results of the 

 relative measures will never be altered very much. I believe that no 

 systematic error in the relative wave-lengths of more than about '01 

 exists anywhere except in the red end as we approach A. Possibly 

 -03, or even less, might cover that region. 



The relative measures having thus been obtained, we have means in 

 the concave grating of obtaining the wave-lengths of the lines of metals 



