In comparing the bright lines of the electric spectra with the dark lines of the solar spectrum, I used the slit as narrow as I had employed it for the examination of the solar spectrum alone. Notwithstanding the brilliance of the electric spark most of the lines seen in the spectrum were not very bright, and in order to render these more distinct I diminished the intensity of the solar spectrum by allowing the sunlight before it reached the slit to pass through coloured glasses. Plates I. and II. give the representation of the result of these observations. I have endeavoured to reproduce the variation in the intensity of the several bright lines, as far as is possible in a drawing, by using lines of three different degrees of thickness. Most of the bright lines appeared considerably broader than they would have done if their light had in the strictest sense been homogeneous, but I have not represented this in the drawing, except in some specially prominent cases in which the lines appeared as broad bands. An example of such bands is seen in the broad green lines which occur in the zinc spectrum, one of which has the limits 1996 and 2000, and the other the limits 2014 and 2018. The horizontal line joining the lower ends of the vertical lines, which represent the bright bands, is intended to connect these bands together, and to shew that the chemical symbol Zn (zinc) applies to both the vertical lines.

The bright lines, produced by the incandescence of the atmospheric air, were scarcely visible owing to the small distance between the electrodes', and to the slight breadth of the slit. Of these atmospheric lines I have only given a representation of one group in the yellow, and one in the green, although I have observed many more. Indeed with respect to the lines of the spectra of the metals, this drawing does not make pretensions to completeness, as I have only endeavoured here to depict those lines which are the most prominent.

If we compare the spectra of the different metals with each other, several of the bright lines appear to coincide. This is especially notice

1

Compare Van der Willigen, Poggendorff's Annalen, Bd. 106, p. 615.

ON THE SPECTRA OF THE CHEMICAL ELEMENTS.

11

able in the case of an iron and magnesium line at 1655.6, and with an iron line and calcium line at 15227. It seems to me to be a question of great interest to determine, whether these and other similar coincidences are real or only apparent; whether the lines in question actually fall one upon the other, or whether they lie very close together. I believe that my method of observation does not possess the requisite accuracy for the purpose of answering this question with any degree of probability, and I think that a larger number of prisms and an increased intensity of the light will prove necessary. This might probably be accomplished by substituting the continuous current of a powerful battery for the inductionspark of a Ruhmkorff's coil.

I close this section with the following remarks. The position of the bright lines, or, to speak more precisely, the maxima of light in the spectrum of an incandescent vapour, is not dependent upon the temperature, upon the presence of other vapours, or upon any other conditions except the chemical constitution of the vapour. Of the validity of this conclusion Bunsen and I have assured ourselves by experiments made for that special object, and I have confirmed it by many observations made with the extraordinarily delicate instrument just described. The appearance of the spectrum of the same vapour may, nevertheless, be very different under different circumstances. Even the alteration of the mass of the incandescent gas is sufficient to effect a change in the character of the spectrum. If the thickness of the film of vapour, whose light is being examined, be increased, the luminous intensities of all the lines increase, but in different ratios. By virtue of a theorem which will be considered in the next section, the intensity of the bright lines increases more slowly than that of the less visible lines. The impression which a line produces on the eye depends upon its breadth as well as upon its brightness. Hence it may happen that one line being less bright, although broader, than a second, is less visible when the mass of incandescent gas is small, but becomes more distinctly seen than the second line when the thickness of the vapour is increased. Indeed, if the luminosity of the whole spectum be so lowered that only

the most striking of the lines are seen, it may happen that the spectrum appears to be totally changed, when the mass of the vapour is altered. Change of temperature appears to produce an effect similar to this alteration in the mass of the incandescent vapour. If the temperature be raised, no deviation of the maxima of light is observed, but the intensities of the lines increase so differently that those which are most plainly seen at a high temperature are not the most visible at a low temperature. This influence of the mass, and of the temperature of the incandescent gas, explains perfectly why in the spectra of many metals those lines which are the most prominent, when the metal is placed in the gas-flame, are not the most distinct when the spectrum of the induction-spark from the metal is examined. This is most clearly seen in the case of the calcium spectrum. I have found that if a wet string, or a narrow tube filled with water, be placed in the metallic circuit joining the Leyden jar which gives the spark, and if the electrodes be moistened with a solution of chloride of calcium, a spectrum is obtained which coincides precisely with that seen when a chloride of calcium bead is placed in the colourless gas-flame. Those lines appear absent which are the most distinct when an entire metallic circuit is employed. If the narrow column of water be replaced by a column of larger sectional area and of shorter length, a spectrum is produced in which both those lines seen in the flame, and those obtained by the intense spark, are equally plainly visible. In this experiment we see the mode in which the calcium spectrum, as given in the flame, changes into that produced by the bright electric spark.

13

ON THE REVERSAL OF THE SPECTRA OF COLOURED FLAMES.

In the course of the experiments already alluded to, which Foucault' instituted on the spectrum of the electric arc formed between the carbon points, this physicist observed that the bright sodium lines present were changed into dark bands in the spectrum produced by the light from one of the carbon poles, which had been allowed to pass through the luminous arc; and when he passed direct sunlight through the arc he noticed that the double D line was seen with an unusual degree of distinctness. No attempt was made to explain or to increase these observations either by Foucault or by any other physicist, and they remained unnoticed by the greatest number of experimentalists. They were unknown to me when Bunsen and I, in the year 1859, commenced our investigations on the spectra of coloured flames.

In order to test in the most direct manner possible the truth of the frequently asserted fact of the coincidence of the sodium lines with the lines D, I obtained a tolerably bright solar spectrum, and brought a flame coloured by sodium vapour in front of the slit. I then saw the dark lines D change into bright ones. The flame of a Bunsen's lamp threw the bright sodium lines upon the solar spectrum with unexpected brilliancy. In order to find out the extent to which the intensity of the solar spectrum could be increased, without impairing the distinctness of the sodium lines, I allowed the full sunlight to shine through the sodium flame upon the slit, and, to my astonishment, I saw that the dark lines D appeared with an extraordinary degree of clearness. I then exchanged the sunlight for the Drummond's or oxyhydrogen lime-light, which, like that of all incandescent

L'Institut, 1849, P. 45.

solid or liquid bodies, gives a spectrum containing no dark lines. When this light was allowed to fall through a suitable flame coloured by common salt, dark lines were seen in the spectrum in the position of the sodium lines. The same phenomenon was observed if instead of the incandescent lime a platinum wire was used, which being heated in a flame was brought to a temperature near to its melting point by passing an electric current through it.

The phenomenon in question is easily explained upon the supposition that the sodium flame absorbs rays of the same degree of refrangibility as those it emits, whilst it is perfectly transparent for all other rays. This supposition is rendered probable by the fact, which has long been known, that certain gases, as for instance, nitrous acid and iodine vapour, possess at low temperatures the property of such a selective absorption. The following considerations shew that this is the true explanation of the phenomenon. If a sodium flame be held before an incandescent platinum wire whose spectrum is being examined, the brightness of the light in the neighbourhood of the sodium lines would, according to the above supposition, not be altered; in the position of the sodium lines themselves, however, the brightness is altered, for two reasons; in the first place, the intensity of light emitted by the platinum wire is reduced to a certain fraction of its original amount by absorption in the flame, and secondly, the light of the flame itself is added to that from the wire. It is plain that if the platinum wire emits a sufficient amount of light, the loss of light occasioned by absorption in the flame must be greater than the gain of light from the luminosity of the flame; the sodium lines must then appear darker than the surrounding parts, and by contrast with the neighbouring parts they may seem to be quite black, although their degree of luminosity is necessarily greater than that which the sodium flame alone would have produced.

1

The absorptive power of sodium vapour1 becomes most apparent

By using this expression I do not intend to restrict the meaning to the vapour of metallic sodium, but I would thereby include the vapour of any sodium compound.