With this flame as a dispersing piece the photographs shown in Plate 2, fig. 8, were obtained. In comparison with some of the other photographs this example is very poor, and I have included it merely to make the set complete. An exposure of about 10 minutes was necessary, with an " Erythro " plate. Fig. 9 was taken with one of the dispersion tubes, and shows what may be called the second stage. After the exposure was over, but before the plate was removed, a small sodium flame was placed before the slit of the spectrometer, and the sodium lines impressed on the plate in their proper position, to serve as a reference. For securing photographs of the dispersion resulting from still denser vapour more light was desirable, and a small prism was accordingly substituted for the grating. This prism barely resolved the sodium lines in the negatives, but gave excellent pictures of the spectrum under strong sodium dispersion Figs. 10 and 11 were obtained in this way. In the latter the flutings in the red and green are discernible, though not very sharp.
As much of the fine detail is always lost in reproduction I have marked with arrows on fig. 11 the extreme points to which the curved branches adjacent to the absorption band can be traced in the original negative. Eye observations enable one to trace them much farther, for the tips are exceedingly faint, and the light is not very actinic.
In order to secure accurate measurements of the dispersion in different parts of the spectrum I have made use of two different methods.
The photographs were not very suitable owing to the fact that during the exposure the density of the vapour varied somewhat, pro- ducing a slight blurring of the image. Measurements of the deviation of the continuous spectrum were accordingly made with a filar micro- meter, an illuminated scale giving the wave-lengths at the points where readings were taken. Several minutes were necessary, however, to get anything like a satisfactory set of readings, and changes occurred even during that short period, as was at once apparent on repeating observations. I accordingly adopted the following method. The telescope of the spectrometer was removed and a telescope objective of nearly two metres focus put in its place.
In the focal plane of this lens a plate of glass was firmly mounted, on which a horizontal diamond scratch had been ruled. This diamond scratch was brought accurately into coincidence with the narrow continuous-spectrum image in its undeviated position. An eye-piece mounted behind the plate enabled the spectrum to be observed. When the dispersion tube was in good condition, and the deviated branches of the spectrum appeared sharp and steady, their positions on the plate were marked with a writing diamond, the line being drawn along the middle of the spectrum, which was only about a millimetre wide on the plate. The dispersion tube was then removed and various metals fed to the arc, the bright lines being marked on the plate with the
