the turned-over seam hammered until a tight joint is formed. These tubes can be heated and cooled any number of times, and can be kept in operation for an hour or two, at the end of which time the sodium is generally used up, a moss-like deposit of oxide gradually filling up the tube. These tubes can be used over and over again without deterioration, and are most satisfactory in every respect. Their only fault lies in their conductivity, the sealing-wax softening and the glass plates falling off, but this can be prevented by wrapping a strip of cloth around each end, and wetting it from time to time. I have had one tube made with water-jackets at each end, but it seems to have no especial advantage, and is more complicated. Porcelain tubes are quite satisfactory, but I prefer the iron on the whole.
By employing a tube of about 5 cm. diameter I have succeeded in projecting the anomalous spectrum, but the appearance is so very inferior to that of the phenomenon when seen subjectively that I should never think of employing the method.
No trace of any influence of the fluted absorption bands on the dis- persion has been detected.
It is a matter of some interest to know what becomes of the energy absorbed by a sodium flame, or by non-luminous sodium vapour. Most of it is undoubtedly spent in raising the temperature of the vapour ; this rise of temperature I have shown experimentally by making the dispersion tube the chamber of a radiophone, and illu- minating it with intermittent light. It was found that the radiophone sang loudly when illuminated by an intense beam of light which con- tained waves of the frequency of the sodium radiations, but became almost silent when these frequencies were removed.
Prismatic analysis was found to weaken the light too much, con- sequently a very thin film of cyanine was used which was quite opaque to the region of the D-lines while transparent to the rest of the spectrum. Sunlight concentrated by a large mirror was employed for the illumination.
It has always appeared to me thinkable that the absorbing sodium ions, when in resonant vibration, might emit a certain amount of light laterally. Kayser mentions in his new work on spectroscopy an experiment made by Miller, in an attempt to detect this phenomenon if it existed. The possibility of such a lateral radiation is discussed by Drude in his ' Lehrbuch der Optik " (page 493), and is shown to depend on a small value of the constant r in the equation of motion of an ion vibrating under the influence of a periodic electric force. For sodium vapour r is calculated to be something between 10 and 100. If it were equal to one we should expect a lateral radiation of light, which, says Drude, has never been observed.
While experimenting with the very intense beam of sunlight in the
