gas may be made to emit light of every degree of refrangibility, so that its spectrum becomes continuous. Professor Frankland has also observed a similar phenomenon, that when oxygen and hydrogen gases are inflamed under pressure, they emit white light and show an unbroken spectrum.1 From these facts there is no doubt that, when intensely heated and under certain circumstances, gaseous bodies can be made to yield continuous spectra. This, however, in no way interferes with the fixity of position of the bright lines, nor can it influence the deductions derived from this fact.

2

In the same way each of the non-metallic elements yields a characteristic spectrum, when its vapour is heated to incandescence; but in the case of some of the elements, such as silicon, the difficulty of obtaining the spectrum is very great.

The examination of the spectrum of carbon is a subject of much interest. The character of the lines which this blue flame of coal gas and air emits was first described in the year 1857 by Professor Swan. Since that time the various spectra of the carbon compounds have been carefully examined by Dr. Attfield, Dr. W. M. Watts, and others, and it has been found that the different compounds of this element, when brought into the condition of luminous gases, either by combustion or when heated up by the electric spark, give somewhat different spectra.3 Thus this beautiful purple flame of cyanogen gas exhibits a great number of very peculiar lines, which differ in position and in intensity from the lines observed in this flame of the coal gas burning mixed with air. (See fig.

1 See Appendix D.

2 See Appendix E.

3 See Appendix C. on the Spectrum of the Bessemer Flame.

of carbon spectra, Nos. 10, 11, on the chromolith. plate facing Lecture VI.)

I may mention, in connexion with these different carbon spectra, the application of spectrum analysis to the important branch of steel manufacture which has been introduced and is well known under the name of the

Bessemer process. In this process five tons of cast iron are in twenty minutes converted into cast steel. Steel dif fers from cast iron in containing less carbon, and by the Bessemer process the carbon is actually burnt out of the molten white-hot cast iron by a blast of atmospheric air.

The arrangement employed for this purpose is shown on this diagram (Fig. 31).

The molten cast iron is run into a large wrought-iron vessel termed the converter (c), lined with refractory clay. The converter is capable of being turned round on a pivot (A), through which pivot passes a tube in connexion with a powerful blowing apparatus, by means of which air can be thrown into the bottom of the vessel, through a sort of tuyère or blowhole into the molten iron. The oxygen of the air burns out the carbon and silicon which the cast iron contains, and the heated gases issue in the form of a flame (F) from the mouth of the converter during the time that the molten iron is being burned. This flame varies in appearance, and it is of the utmost importance that the operation should be stopped instantly when the proper moment has arrived. If the blast be continued for ten seconds after the proper point has been attained, or if it be discontinued ten seconds before that point is reached, the charge becomes either so viscid that it cannot be poured from the converter into the ladle (L), from which it has to be transferred to the moulds, or it contains so much carbon as to crumble up like cast iron under the hammer.

Those who are accustomed to work this process are able by the simple inspection of the flame to tell with more or less exactitude when the air has to be turned off. To those who are uninitiated in this peculiar appearance of the flame no difference at all can be detected at the point in which it is necessary to stop; but by the help of the spectroscope this point can at once be ascertained beyond shadow of doubt, and that which previously depended upon the quickness of vision of a skilled eye has become a simple matter of exact scientific observation.

The light which is given off by the Bessemer flame is most intense, indeed, a more magnificent example of combustion in oxygen cannot be imagined. A cursory examination of the flame spectrum in its various phases reveals complicated masses of dark absorption bands and bright lines, showing that a variety of substances are present in the flame in the state of glowing gas.

By a simultaneous comparison of the lines in the Bessemer spectrum with those of well-known substances I was able to detect the following substances in the Bessemer flame: sodium, potassium, lithium, iron, carbon, hydrogen, and nitrogen. At a certain stage of the operation all at once the carbon lines disappear, and we get a continuous spectrum. The workman by experience has learned that this is the moment at which the air must be shut off; but it is only by means of the spectroscope that this point can be exactly determined.

Those who are practically engaged in working this process would like spectrum analysis to do a great deal more : they would like to be told whether there is any sulphur, phosphorus, or silicon in their steel: questions which unfortunately at present spectrum analysis cannot answer, for this very good reason, that these substances do not appear at all as gases in the flame, but that they either remain unvolatilized in the molten metal, or swim on its surface in the slag of the ore; and consequently the lines of these bodies are not seen in the spectrum of the flame.

LECTURE III. APPENDIX A.

SPECTRUM REACTIONS OF THE RUBIDIUM AND CÆSIUM

COMPOUNDS.1

CASIUM and rubidium are not precipitated either by sulphuretted hydrogen or by carbonate of ammonium. Hence both metals must be placed in the group containing magnesium, lithium, potassium, and sodium. They are distinguished from magnesium, lithium, and sodium by their reaction with bichloride of platinum, which precipitates them like potassium. Neither rubidium nor cæsium can be distinguished from potassium by any of the usual reagents. All three substances are precipitated by tartaric acid as white crystalline powders; by hydrofluosilicic acid as transparent opalescent jellies; and by perchloric acid as granular crystals: all three, when not combined with a fixed acid, are easily volatilized on the platinum wire, and they all three tinge the flame violet. The violet colour appears indeed of a bluer tint in the case of potassium, whilst the flame of rubidium is of a redder shade, and that of cæsium still more red. These slight differences can, however, only be perceived when the three flames are ranged side by side, and when the salts undergoing volatilization are perfectly pure. In their reactions, then, with the common chemical tests, these new elements cannot be distinguished from potassium. The only method by means of which they can be recognised when they occur together is that of spectrum analysis.

The spectra of rubidium and cæsium are highly characteristic, and are remarkable for their great beauty (Frontispiece, Nos. 3 and 4). In examining and measuring these spectra we have

1 Extract from Professors Kirchhoff and Bunsen's second Memoir on Chemical Analysis by Spectrum Observations (Phil. Mag. vol. xxii. 1861).