(2) At the level of the base of the flame, there is an annular space, through which a stream of cold air is continually blown upwards. Thus cooled, the light is very pronounced, and the band δ, which is almost invisible in the ordinary method of examination, is plainly seen.[318] [318] Consult _Spectres Lumineux_, par M. Lecoq de Boisbaudran, Paris, 1874. See also Christofle and Beilstrom’s “Abhandlung,” in _Fresenius’ Zeitschr. f. anal. Chem._, B. 2, p. 465, and B. 3, p. 147. An apparatus (devised by Blondlot, and improved by Fresenius) for the production of the phosphine flame in medico-legal research, is represented in the following diagram:-- [Illustration] Several of the details of this apparatus may be modified at the convenience of the operator. A is a vessel containing sulphuric acid; B is partly filled with granulated zinc, and hydrogen may be developed at pleasure; _c_ contains a solution of nitrate of silver; _d_ is a tube at which the gas can be lit; _e_, a flask containing the fluid to be tested, and provided with a tube _f_, at which also the gas issuing can be ignited. The orifice should be provided with a platinum nozzle. When the hydrogen has displaced the air, both tubes are lit, and the two flames, being side by side, can be compared. Should any phosphorus come over from the zinc (a possibility which the interposed silver nitrate ought to guard against), it is detected; the last flask is now gently warmed, and if the flame is green, or, indeed, in any case, it should be examined by the spectroscope.[319] [319] F. Selmi has proposed the simple dipping of a platinum loop into a liquid containing phosphoric acid, and then inserting it into the tip of a hydrogen flame. § 297. The spectrum, when fully developed, shows one band in the orange and yellow between C and D, but very close to D, and several bands in the green. But the bands δ, γ, α, and β are the most characteristic. The band δ has its centre about the wave-length 599·4; it is easily distinguished when the slit of the spectroscope is a little wide, but may be invisible if the slit is too narrow. It is best seen by M. Salet’s second process, and, when cooled by a brisk current of air, it broadens, and may extend closer to D. The band γ has a somewhat decided border towards E, while it is nebulous towards D, and it is, therefore, very difficult to say where it begins or where it ends; its centre may, however, be put at very near 109 of Boisbaudran’s scale, corresponding to W. L. 560·5, if the flame is free. This band is more distinct than β, but with a strong current of air the reverse is the case. The middle of the important band α is nearly marked by Fraunhofer’s line E. Boisbaudran gives it as coinciding with 122 of his scale W. L. 526·3. In ordinary conditions (that is, with a free uncooled flame) this is the brightest and most marked of all the bands. The approximate middle of the band β is W. L. 510·6 (Boisbaudran’s scale 129·00).