242 PROCEEDINGS OF THE AMERICAN ACADEMY. 



tures, as shown in Plate 3, Figure 1. The retort was half filled with 

 sodium, the molten metal being poured in through one of the apertures. 

 It was then introduced into the tube and pushed down to the centre, 

 after which the plate-glass ends were cemented on, as shown in the 

 figure. This arrangement prevented the rapid diffusion of the vapor, 

 and enabled a large supply of metal to be kept at the centre of the 

 tube. The tubes used in the earlier work required re-charging after 

 two hours' continuous operation, while the retort tube could be 

 operated for several hundred hours on a single charge. 



The tube was exhausted with a Fleuss pump and heated at the 

 centre with a large burner, the ends being kept cool by jackets 

 of absorbent cotton which dipped into pails of water. 



The illuminating beam of either white or monochromatic light was 

 focussed just within one of the oval apertures of the retort, falling 

 upon the opposite wall a little to one side of the other aperture. By 

 covering the further end of the tube with a black cloth, the fluorescent 

 spot showed against the dead black background of the second oval 

 aperture, and its spectrum was therefore uncontaminated with the 

 exciting radiations. 



A large three-prism spectrograph was constructed for photograph- 

 ing the spectra. The prisms were of clear dense flint four inches in 

 height, and the focal length of the lenses thirty-six inches. 



Since only lenses such as are used for telescopes were available, 

 the spectrum lines are not so sharp as one would wish, except near the 

 axis of collimation. By adjusting things so that the centre of the 

 fluorescent spectrum fell at this point, the definition was pretty fair 

 throughout its extent, and wave-lengths could be determined with 

 an error not greater than one or two Angstrom units. 



The photographs of the complex spectrum of the fluorescence ex- 

 cited by sunlight, obtained with this instrument (Plate 4, Figure 6), 

 showed peculiarities which made it appear of the utmost importance 

 to study the spectrum under higher dispersion. The green fluorescent 

 spot had, after repeated improvements in the apparatus, attained such 

 brilliancy that I felt sure that records could be obtained with the 

 twelve-foot concave grating. An all-day exposure was found to be 

 sufficient, the resulting spectrogram, with the iron comparison lines 

 and the wave-length scale, being reproduced on Plate 1, Figure 2. 

 The scale was printed separately, and slight errors occur, due to stretch- 

 ing and shrinking of the prints. They are not greater in any case 

 than 1.5 A. E. This plate shows the minute structure of the complex 

 spectrum, and enables us to measure the wave-lengths of the bright 

 lines with far greater accuracy than could be done with the plates 



