of the tube. When, however, a slow current of the undried gas was allowed to impinge on the sodium vapour, by partly opening the stopcock A, the absorption-line vanished and an intensely brilliant but fine line appeared in its place in the centre of the broad but relatively faint emission-line. This sharply-defined narrow line resembled the ordinary D line seen in flames. It was brighter than the continuous spectrum of the glowing sides of the tube, on which it appeared to be superposed.

IV. Atmospheric nitrogen was substituted for coal-gas in the gas-holders, and was freed from traces of oxygen and dried before entering the heating-tube by passing over heated phosphorus and through a tube of CaCl2. The phenomena observed on volatilizing the sodium in this gas were in every way the same as in coal-gas. It was subsequently found that the same results could be obtained with unpurified nitrogen, or even with common air, the sodium itself effecting the purification almost immediately on vaporizing, producing at the same time a brilliant flash-in the spectroscope a brilliant but sharply-defined and narrow D line-and clouds of oxide; afterwards showing the broad hazy emission-line and the central black absorption-line, when all the oxygen in the tube had been consumed and the oxide had subsided.

These results appear to me to show that impurities in the neutral gases used are not concerned in the production of the broad hazy emission-line, for when traces of these, particularly oxygen and moisture, are known to be present and are allowed to impinge on the sodium vapour, a line is seen which is fine and sharp, showing that the region of chemical action is only a surface-layer of no great density, whilst the fainter but broad and diffuse D line, always seen when the vapour is undisturbed, evidently originates at a great depth where the vapour-density is considerable and in a region protected from chemical action by the outer relatively cool layers giving the absorption-line.

While this central region of the vapour may be considered to be well protected by the outer layers from impurities in the neutral gas employed, there still remains the possibility that the porcelain tube itself reacts with the sodium throughout its heated part, thus furnishing a continual supply of chemical energy; and some support is given to this view of the case from the fact that the bright line cannot be maintained as a wide line indefinitely without a continual addition of fresh sodium, also the tube becomes much corroded, black silicon being deposited inside: thus proving a reaction between the silicates of the porcelain and the sodium.

In the experiments which follow, the effect of such reactions between the tube and the sodium is eliminated by the use of

iron tubes in place of porcelain. The first trials were made with a short piece of iron tube about 6 cm. long and 8 mm. bore, bevelled at the ends and fitted between two hard glass tubes of the same diameter, the joints being ground to fit. This made good joints when the iron became red-hot and the glass in contact with it was softened and pressed up tight; but although satisfactory results were obtained in two or three experiments, constant trouble was experienced in the cracking of the glass while cooling. Finally the glass was discarded and a long iron tube prepared (a piece of ordinary 4-inch hydraulic tube). In order to diminish as far as possible the loss of heat by conduction, so as to maintain a high temperature for about 8 cm. in the central part, a number of deep necks were cut in the metal, as shown in fig. 2, and around these necks a thick ring of asbestos-packing was wound and a fire-clay arch placed over all. Thus the Bunsen flame could be concentrated entirely on the central piece of the tube and the temperature could be maintained inside the tube above the fusing-point of fluor-spar and aluminium, but not reaching that of silver. The rubber connexions between the ends of the iron tube and the two T-pieces gave trouble at first, but subsequently it was found that when buried in a thick layer of plaster of Paris they were completely protected from destruction by heat, the large surface afforded by the plaster of Paris forming an effectual radiator and preventing the ends from becoming too hot.

Experiment V.-With the iron tube at a bright red heat, the sodium was tipped out from the steel spoon as before, in an atmosphere of carefully dried coal-gas. Now, if in the previous experiments the D radiation was due to chemical action taking place between the sodium and the oxygen compounds of the porcelain, one ought in this experiment to find, if not an entire suppression of the bright line, at least a striking difference in the radiation. No such difference was, however, to be observed, the intensity remaining precisely the same as before and the sequence of phenomena closely resembling that shown in fig. 3. The various phases there shown were, however, prolonged almost indefinitely in time, as the sodium never became used up as before, and the distillation into cooler parts of the tube proceeded so slowly that it was necessary to allow a gentle current of gas to drive away the denser vapour giving a continuous spectrum before the D line itself could be studied. Also throughout the experiment the dark absorption-line was more intense than in porcelain, and it could easily be observed after six hours of continuous heating, when even the emission-line had become relatively narrow. This naturally follows from the

consideration that in the iron the temperature-gradient on each side of the central red-hot part of the tube is much less steep than is the case when porcelain or glass is used, and consequently there is a much greater thickness of relatively cool vapour through which the emission-line is seen.

Under the conditions of these experiments, therefore, the bright D line appears to be quite uninfluenced either by the nature of the neutral gases used and the impurities they may contain, or by the material of which the heating-tube is composed: iron giving exactly the same results as porcelain. It would hardly be safe, however, at this stage of the inquiry to infer that chemical reactions are not concerned in the production of the light; for it would be argued that, as iron becomes slightly porous at a red heat, oxygen, or at any rate some of the gaseous constituents of the Bunsen flame, might find their way into the tube by diffusion from outside, and in this way maintain a continual reaction with the sodium vapour.

In order to diminish the possibility of this diffusion inwards affecting the results, a constant pressure of a few millimetres above atmospheric pressure is maintained within the tube, and if gases diffuse in at all it must be in opposition to the outward diffusing hydrogen. It has been pointed out to me, however, by Prof. Smithells that in dealing with the D line we are dealing with a reaction that is sensitive to 180,000,000 of a grain of sodium. It is only necessary to suppose, therefore, an equivalent amount of oxygen or other reacting body to be continually present in the tube to determine the Ď radiation.

1

While I am not prepared to deny the possibility of such minute traces of oxygen or other bodies constantly finding their way into the middle of the sodium vapour, I consider that any reactions so caused could under no circumstances produce the broad ill-defined line actually observed. At the most a fine double D line would be seen similar to that of a flame tinted with a salt of sodium, where the density of the reacting molecules is not great. Moreover, if reacting bodies were diffusing in from outside the tube-the absorbing layer of the sodium vapour itself forming an effectual barrier in other directions-the action would be greatest in an annular region in contact with the sides of the tube where the incoming molecules first encountered the sodium. This should cause a brightening or widening of the D line at each end *. But there is no such inequality seen the line is quite uniform in width and brightness throughout its length; showing that if chemical reactions are producing the light, the reacting

* The D line with the optical arrangement employed represents a section of the space inside the hot part of the tube.

molecules must be uniformly distributed throughout the mass of vapour.

The most telling argument, however, and one which, taken alone, appears to me to prove beyond a doubt that the D radiation under these conditions is the direct result of the heating, is that derived from a comparison between the emission and absorption spectra.

To effect this comparison, the lamp and reflecting prism previously described are brought into operation, and a beam of white light is made to traverse one side of the heated tube, so that one portion of the slit of the spectroscope is illuminated by transmitted white light, whilst another contiguous portion is illuminated by the D radiation alone. Under these circumstances the absorption and emission spectra appear side by side in the field of view, and may be readily compared.

Figs. 3 and 4 show the corresponding phases of these spectra (denoted by letters of the alphabet). It will be noticed that the emission-line or band in c, d, and e is represented as of the same width as the absorption-band. Careful observation under various conditions as to density shows that, excepting for the dark line in the centre, the bright D line is in every respect the exact counterpart of the absorption-line, whether the broad hazy band of the dense vapour is studied, or the relatively narrow line seen with more attenuated vapour. Assuming, then, that the width of both absorption and emission lines is determined by the molecular density of the absorbing and emitting vapour, it follows that in the densest region every molecule that is concerned in the absorption is also concerned in the radiation. In other words, practically every molecule in the hot part of the tube contributes its share to the radiation. But it is surely impossible to suppose that every molecule, or even a large proportion of the molecules, is continually undergoing chemical change. The supply of oxygen or other reacting bodies-supposing they do gain access to the sodium-will never be equal to even a small fraction of the demand; also, if oxidation is proceeding, one would expect to find traces of oxide forming after several hours. But the heating may be continued for six hours at the least without touching the apparatus, and at the end of this time the D line, with its central absorption-line, is seen as clearly as at the beginning, there being no trace of any opaque clouds of oxide such as are always seen when traces of oxygen are known to be present in the tube.

There seems no possible alternative, therefore, to the obvious and simple explanation which ascribes the radiation to heat alone. If it were assumed that there exists diffused throughout the sodium vapour some substance capable of setting up

chemical reactions, it might indeed be imagined that a kind of cyclical process of alternate combination and dissociation takes place, the energy supplying the radiation being derived indirectly from the heated walls of the tube: these alternate changes being determined by differences of temperature in the reacting molecules, the cooler combining and the hotter dissociating, or vice versa. But it is practically certain that under the conditions of these experiments there must always be a large excess of sodium molecules over any others likely to produce such reactions, unless, indeed, hydrogen or nitrogen were to behave in this way towards sodium, and the spectroscopic evidence, as just explained, implies that all the free sodium molecules are concerned in the radiation.

Or perhaps it may be further argued in support of the "chemical" origin of the radiation, that the sodium molecule is itself undergoing alternate dissociation and recombination. The reasons already given against this view in the case of iodine apply even more forcibly in this instance. Thus, if the radiation were due to such action, the intensity should follow the curve representing the change of relative vapour-density with temperature, rising to a maximum at the turning-point in this curve-indicating the greatest interaction between the atoms and the molecules-but falling away to zero when the point of complete dissociation is reached. Now the most reliable recent determinations of the vapour-density of sodium indicate that at about the temperature of melting cast-iron the vapour is entirely monatomic: therefore, if we assume that at the lowest temperature of my experiments the atoms are more or less aggregated, the relative intensity of the D line, compared with the continuous spectrum of the glowing sides of the tube, should change as the temperature is increased, it should get brighter or fainter according as the actual temperature of dissociation is above or below the initial temperature of my experiments; and it should cease altogether if the temperature be raised to the point where the vapour becomes entirely monatomic.

But as a matter of fact there is no such change of relative intensity. As the temperature is increased from the point where the radiation begins to be seen, the D line follows strictly the continuous spectrum of the glowing tube: from the lowest to the highest temperature (a range of some 300 C. degrees), the gaseous radiation increases in intensity exactly in correspondence with the radiation from the solid, always keeping the same intensity (so far as the eye can judge) as the spectrum of the glowing tube.

* A. Scott, Proc. Roy. Soc. Edinb. xiv. p. 410,