employed an improved form of apparatus (Fig. 32), which in every respect is much to be preferred to that described in our first memoir. In addition to the advantages of being more manageable and producing more distinct and clearer images, it is so arranged that the spectra of two sources of light can be examined at the same time, and thus, with the greatest degree of precision, compared both with one another and with the numbers on a divided scale.

In order to obtain representations of the spectra of cæsium and rubidium corresponding to those of the other metals which we have given in our former paper, we have adopted the following course.

We have placed the tube g (Fig. 32) in such a position that a certain division of the scale, viz. No. 100, coincided with Fraunhofer's line D in the solar spectrum, and then observed the position of the dark solar lines A, B, C, D, E, F, G, H, on the scale: these several readings we called A, B, C, &c. An interpolation scale was then calculated and drawn, in which each division corresponded to a division on the scale of the instrument, and in which the points corresponding to the observations A, B, C, &c. were placed at the same distances apart as the same lines on our first drawings of the spectrum. By help of this scale, curves of the new spectra were drawn (Fig. 24, p. 59), in which the ordinates express the degrees of luminosity at the various points on the

scale, as judged of by the eye. The lithographer then made the designs represented in the Frontispiece from these curves.

As in our first memoir, so here we have represented only those lines which, in respect to position, definition, and intensity, serve as the best means of recognition. We feel it necessary to repeat this statement, because it has not unfrequently happened that the presence of lines which are not represented in our drawings has been considered as indicative of the existence of new bodies.

We have likewise added a representation of the potassium spectrum to those of the new metals for the sake of comparison, so that the close analogy which the spectra of the new alkaline metals bear to the potassium spectrum may be at once seen. All three possess spectra which are continuous in the centre, and decreasing at each end in luminosity. In the case of potassium this continuous portion is most intense, in that of rubidium less intense, and in the cæsium spectrum the luminosity is least. In all three we observe the most intense and characteristic lines towards both the red and blue ends of the spectrum.

Amongst the rubidium lines, those splendid ones named Rb a and Rbẞ are extremely brilliant, and hence are most suited for the recognition of the metal. Less brilliant, but still very characteristic, are the lines Rb S and Rby. From their position they are in a high degree remarkable, as they both fall beyond Fraunhofer's line A; and the outer one of them lies in an ultrared portion of the solar spectrum, which can only be rendered visible by some special arrangement. The other lines, which are found on the continuous part of the spectrum, cannot so well be used as a means of detection, because they only appear when the substance is very pure, and when the luminosity is very great. Nitrate of rubidium, and the chloride, chlorate, and perchlorate of rubidium, on account of their easy volatility, show these lines most distinctly. Sulphate of rubidium and similar salts also give very beautiful spectra. Even silicate and phosphate of rubidium yield spectra in which all the details are plainly seen.

The spectrum of cæsium is especially characterised by the two

I

blue lines C's a and Cs B: these lines are situated close to the blue strontia line Sr S, and are remarkable for their wonderful brilliancy and sharp definition. The line Cs 8, which cannot be so conveniently used, must also be mentioned. The yellow and green lines represented on the figure, which first appear when the luminosity is great, cannot so well be employed for the purpose of detecting small quantities of the cæsium compounds; but they may be made use of with advantage as a test of the purity of the cæsium salt under examination. They appear much more distinctly than do the yellow and green lines in the potassium spectrum, which, for this reason, we have not represented.

As regards distinctness of the reaction, the cæsium compounds resemble in every respect the corresponding rubidium salts: the chlorate, phosphate, and silicate gave the lines perfectly clearly. The delicacy of the reaction, however, in the case of the cæsium compounds is somewhat greater than in that of the corresponding compounds of rubidium. In a drop of water weighing four milligrammes, and containing only 0·0002 milligramme of chloride of rubidium, the lines Rba and Rb ß can only just be distinguished; whilst 000005 milligramme of the chloride of cæsium can, under similar circumstances, easily be. recognised by means of the lines Cs a and Cs B.

If other members of the group of alkaline metals occur together with cæsium and rubidium, the delicacy of the reaction is of course materially impaired, as is seen from the following experiments, in which the mixed chlorides contained in a drop of water, weighing about four milligrammes, were brought into the flame on a platinum wire.

When 0·003 milligramme of chloride of cæsium was mixed with from 300 to 400 times its weight of the chlorides of potassium or sodium, it could be easily detected. Chloride of rubidium, on the other hand, could be detected with difficulty when the quantity of chloride of potassium or chloride of sodium amounted to from 100 to 150 times the weight of the chloride of rubidium employed.

0.001 milligramme of chloride of cesium was easily recognised when it was mixed with 1,500 times its weight of chloride

of lithium; whilst 0.001 milligramme of chloride of rubidium could not be recognised when the quantity of chloride of lithium added exceeded 600 times the weight of the rubidium salt.

APPENDIX B.

CONTRIBUTIONS TOWARDS THE HISTORY OF SPECTRUM ANALYSIS. BY G. KIRCHHOFF.1

In my "Researches on the Solar Spectrum and the Spectra of the Chemical Elements "2 I made a few short historical remarks concerning earlier investigations upon the same subject. In these remarks I have passed over certain publications in silence

in some cases because I was unacquainted with them, in others because they appeared to me to possess no special interest in relation to the history of the discoveries in question. Having become aware of the existence of the former class, and seeing that more weight has been considered to attach to the latter class of publications by others than by myself, I will now endeavour to complete the historical survey.

1. Amongst those who have devoted themselves to the observation of the spectra of coloured flames, I must, in the first place, mention Herschel and Talbot. Their names need special notice, as they pointed out with distinctness the service which this mode of observation is capable of rendering to the chemist. For a knowledge of their researches I am mainly indebted to Professor W. Allen Miller, who gave an extract from them in a lecture republished in the number of the Chemical News for 19th April, 1862. It is there stated that in the volume of the Transactions of the Royal Society of Edinburgh for 1822, at page 455, Herschel shortly describes the spectra of chloride of strontium, chloride of potassium, chloride of copper, nitrate of 1 Communicated to the Phil. Mag. Fourth Series, vol. xxv. p. 250, by Professor Roscoe.

2 Published by Macmillan and Co. Cambridge and London, 1862.

copper, and boracic acid. The same observer says, in his article on Light in the "Encyclopædia Metropolitana," 1827, page 438: "Salts of soda give a copious and purely homogeneous yellow; of potash, a beautiful pale violet." He then describes the colours given by the salts of lime, strontia, lithia, baryta, copper, and iron, and continues: "Of all salts, the muriates succeed the best, from their volatility. The same colours are exhibited also when any of the salts in question are put (in powder) into the wick of a spirit lamp. The colours thus communicated by the different bases to flame afford in many cases a ready and neat way of detecting extremely minute quantities of them. The pure earths, when violently heated, as has recently been practised by Lieut. Drummond, by directing on small spheres of them the flames of several spirit-lamps, urged by oxygen gas, yield from their surfaces lights of extraordinary splendour, which, when examined by prismatic analysis, are found to possess the peculiar definite rays in excess which characterise the tints of flames coloured by them; so that there can be no doubt that these tints arise from the molecules of the colouring matter, reduced to vapour and in a state of violent ignition."

Talbot says:1 "The flame of sulphur and nitre contains a red ray which appears to me of a remarkable nature. This red ray appears to possess a definite refrangibility, and to be characteristic of the salts of potash, as the yellow ray is of the salts of soda, although, from its feeble illuminating power, it is only to be detected with a prisın. If this should be admitted, I would further suggest that whenever the prism shows a homogeneous ray of any colour to exist in a flame, this ray indicates the formation or the presence of a definite chemical compound." Somewhat further on, in speaking of the spectrum of red fire and of the frequent occurrence of the yellow line, he says: "The other lines may be attributed to the antimony, strontia, &c. which enter into this composition. For instance, the orange ray may be the effect of the strontia, since Mr. Herschel found in the flame of muriate of strontia a ray of that colour. If this opinion should be correct, and applicable to the other definite rays, a glance at 1 Brewster's Journal of Science, vol. v. 1826; Chemical News, April 27, 1861.