I am aware that this conclusion is a startling one, but a little consideration will show its high probability, and a summary of all the facts proves it, I think, beyond all question.

While we have bright lines in comets, it can be shown that some of them are the remnants of flutings. Thus in Comet III of 1881, as the carbon lines died away the chief manganese fluting at 558 became conspicuously visible; it had really been recorded before then. The individual observations are being mapped in order that the exact facts may be shown. It may probably be asked how it happened that the fluting of magnesium at 500 was not also visible. Its absence, however, can be accounted for: it was masked by the brightest carbon fluting at 517, whereas the carbon fluting which under other circumstances might mask the manganese fluting at 558 is always among the last to appear very bright and the first to disappear.

In the great comet of 1882, which was most carefully mapped by Copeland, very many lines were seen, and indeed many were recorded, and it looks as if a complete study of this map will put us in possession of many of the lines recorded by Sherman in the spectrum of 7 Cassiopeia. We have then three marked species of non-revolving swarms going on all fours with three marked species of revolving ones, and in this we have an additional argument for the fact that the absence in the former of certain flutings which we should expect to find may be attributed to masking by the carbon flutings.

We have next, then, to show that there are carbon bands in the bright-line stars.

There is evidence of this. Among the bright lines recorded is the brightest carbon fluting at 517. This is associated with those lines of magnesium and manganese and iron visible at a low temperature which have been seen in comets.

But we have still more evidence of the existence of carbon. In a whole group of bright-line stars there is a bright band recorded at about 470, while, less refrangible than it, there appears a broad absorption band. I regard it as extremely probable that we have here the bright carbon band 467-474, and that the appearance of an absorption band is due to the fact that the continuous spectrum of the meteorites extends only a short distance into the blue.

If we consider such a body as Wells's comet, or the great comet of 1882, at so great a distance from us that only an integrated spectrum would reach us, in these cases the spectrum would appear to extend very far, and more or less continuously, into the blue; but this appearance would be brought about, not by the continuous spectra of the meteorites themselves, but by the addition of the hydrocarbon fluting at 431 to the other hot and cold carbon bands in that part of the spectrum.

There are other grounds which may be brought forward to suggest

that the difference between comets and the stars now under discussion is more instrumental than physical.

Supposing that the cometic nature of these bodies be conceded, laboratory work will eventually show us which flutings and lines will be added to the nebula spectrum upon each rise of temperature.

The difficulties of the stellar observations must always be borne in mind. It will also be abundantly clear that a bright fluting added to a continuous spectrum may produce the idea of a bright line at the sharpest edge to one observer, while to another the same edge will appear to be preceded by an absorption band.

III. Stars with Bright Flutings accompanied by Dark Flutings.

I also showed in the paper to which reference has been made that the so-called "stars" of Class IIIa of Vogel's classification are not masses of vapour like our sun, but really swarms of meteorites; the spectrum being a compound one, due to the radiation of vapour in the interspaces and the absorption of the light of the red- or whitehot meteorites by vapours volatilised out of them by the heat produced by collisions. The radiation is that of carbon vapour, and some of the absorption, I stated, was produced by the chief flutings of manga

nese.

These conclusions were arrived at by comparing the wave-lengths of the details of spectra recorded in my former paper with those of the bands given by Dunér in his admirable observations on these bodies.*

The discovery of the cometic nature of the bright-line stars greatly strengthens the view I then put forward, not only with regard to the presence of the bright flutings of carbon, but with regard to the actual chemical substances driven into vapour. From the planetary nebula there is an undoubted orderly sequence of phenomena through the bright-line stars to those now under consideration, if successive stages of condensation are conceded.

I shall return to these bodies at a later part of this memoir.

IV. Stars in which Absorption Phenomena predominate.

I do not suppose that there will be any difficulty in recognising, that if the nebulæ, stars with bright lines, and stars of the present Class IIIa are constituted as I state them, all the bodies more closely resembling the sun in structure, as well as those more cooled down, must find places on a temperature curve pretty much as I have placed

* "Les Étoiles à Spectres de la troisième classe."— Kongl. Svenska VetenskapsAkademiens Handlingar,' Band 21, No. 2, 1885.

them; the origin of these groups being, first still further condensation, then the condition of maximum temperature, and finally the formation of a photosphere and crust.

We shall be in a better position to discuss these later stages when the classifications hitherto suggested have been considered.

PART II.-CLASSIFICATION INTO GROUPS.

I. FORMER CLASSIFICATIONS OF STARS.

In the various classifications of the celestial bodies which have been attempted from time to time, nebulae and comets have been regarded as things apart from the stars; but from what I have stated in the first part of this paper, relating to the origin of the various groups of heavenly bodies, it is clear that it is not only unnecessary but unphilosophical to make such a distinction; and, indeed, if any such separation were needed, such a result would seem to indicate that the line of evolution is by no means so simple and clear as it really seems to be. But although it is no longer necessary to draw this distinction, it is important that I should state the various spectroscopic classifications which have been attempted in the case of the stars. With this information before us, we shall be better able to see the definite lines on which any new classification must be based to include all celestial forms.

Fraunhofer, Rutherfurd, and Secchi.

When we inquire into the various labours upon which our present knowledge of the spectra of the various orders of " stars is based, the first we come across are those of Fraunhofer, who may be said to have founded this branch of scientific inquiry in the year 1814.

Fraunhofer not only instituted the method of work which now is found to be the most effective, but his observations at that time were so excellent that he had no difficulty in finding coincidences between lines in the spectrum of the sun and of Venus.

Fraunhofer's reference in his observations runs as follows:

"I have also made several observations on some of the brightest fixed stars. As their light was much fainter than that of Venus, the brightness of their spectrum was consequently still less. I have nevertheless seen, without any illusion, in the spectrum of the light of Sirius, three large lines, which apparently have no resemblance with those of the sun's light. One of them is in the green, and two in the blue space. Lines are also seen in the spectrum of other fixed stars of the first magnitude; but these stars appear to be different from one another in relation to these lines. As the object-glass of the telescope of the theodolite has only thirteen lines of aperture, these

experiments may be repeated, with greater precision, by means of an object-glass of greater dimensions."*

He did not attempt to classify his observations on stellar spectra, but, as pointed out by Professor Dunér ("Sur les Étoiles à Spectres de la Troisième Classe," p. 3), those that he most particularly mentions are really remarkably diverse in their characteristics.

In these researches Fraunhofer was followed by Rutherfurd, who, in the year 1863, was the first to indicate that the various stellar spectra which he had then. observed were susceptible of being arranged into different groups. His paper was published in 'Silliman's Journal' (vol. 35, p. 71), and, after giving an account of the observations actually made, continues as follows:

"The star spectra present such varieties that it is difficult to point out any mode of classification. For the present, I divide them into three groups :-First, those having many lines and bands, and mostly resembling the sun, viz., Capella, ß Geminorum, a Orionis, Aldebaran, y Leonis, Arcturus, and ẞ Pegasi. These are all reddish or golden stars. The second group, of which Sirius is the type, presents spectra wholly unlike that of the sun, and are white stars. The third group, comprising a Virginis, Rigel, &c., are also white stars, but show no lines; perhaps they contain no mineral substance, or are incandescent without flame."

Soon afterwards Secchi carried on the inquiry, and began in 1865 by dividing the objects he had then observed into two types. These two types were subsequently expanded in 1867 into three ('Catalogo delle Stelle di cui si è determinato lo Spettro Luminoso,' Secchi, Parigi, 1867): first, white stars, like a Lyræ; secondly, yellow stars, like Arcturus; and thirdly, deeply coloured stars, like a Herculis and Orionis. The order of these types was not always as stated, but I have not been able to find the exact date at which the order was changed (Dunér, "Sur les Étoiles," p. 128). Secchi subsequently added a fourth type, in which the flutings were less numerous. There is little doubt that Secchi was led to these types not so much by any considerations relating to the chemical constitution of the atmospheres of these bodies, as in relation to their colours. His first classifications, in fact, simply separated the white stars from the coloured ones (see on this point Le Scopirte Spettroscopiche,' A. Secchi, Roma, 1865).

The fourth type included, therefore, stars of a deeper red colour than those of the third, and Secchi pointed out that this change of colour was accompanied by a remarkable change in the spectrum; in fact, of Secchi's four types thus established, the first and second had * "On the Refractive and Dispersive Power of Different Species of Glass, with an Account of the Lines which cross the Spectrum."-Fraunhofer, translated in 'Edinburgh Philosophical Journal,' vol. 10, October to April, 1823-24, p. 39.

VOL XLIV.

C

line spectra and the third and fourth had fluted ones. At that time the important distinction to be drawn between line- and fluted-spectra was not so well recognised as it is at present; and further the relation of spectra to temperature was not so fully considered. Secchi, as a result of laboratory work, however, at once showed an undoubted connexion between the absorption flutings in the stars of the fourth type and the bright ones seen in the spectrum or carbon under certain conditions; and although this conclusion has been denied, it has since been abundantly confirmed by Vogel and others (see Vogel, Publicationen, &c., Potsdam,' No. 14, 1884, p. 31).

Relation to Temperature.

At the time that Secchi was thus classifying the stars, the question was taken up also by Zöllner, who in 1865 first threw out the suggestion that the spectra might probably enable us to determine somewhat as to the relative ages of these bodies; and he suggested that the yellow and red light of certain stars were indications of a reduction of temperature (Zöllner, Photometrische Untersuchungen,' p. 243).

In 1868 this subject occupied the attention of Ångström with special reference to the contrasted spectra of lines and flutings. On this he wrote as follows, showing that temperature considerations might help us in the matter of variable stars (Recherches sur le Spectre solaire,' Upsala, 1868):

"D'après les observations faites par MM. Secchi et Huggins, les raies d'absorption dans les spectres stellaires sont de deux espèces : chez l'une, le spectre est rayé de lignes très-fines, comme le spectre solaire; chez l'autre, les raies constituent des groupes entiers à espaces égaux ou des bandes nuancées. Ces derniers groupes appartiennent vraisemblablement aux corps composés, et je mentionnerai, en particulier, que ceux trouvés dans le spectre de a Orionis ressemblent fort aux bandes lumineuses que donne le spectre de l'oxyde de manganèse. Supposé que ma théorie soit juste, l'apparition de ces bandes doit donc indiquer que la température de l'étoile est devenue assez basse pour que de telles combinaisons chimiques puissent se former

et se conserver.

"Entre ces deux limites de température chez les étoiles, limites que l'on peut caractériser par la présence de l'une ou de l'autre espèce des raies d'absorption, on peut s'imaginer aussi un état intermédiaire, dans lequel les gaz composés peuvent se former ou se dissocier, suivant les variations de température auxquelles ils sont assujettis par l'action chimique même. Dans cette classe doivent probablement être comprises les étoiles dont l'intensité de lumière varie plus ou moins rapidement, et avec une périodicité plus ou moins constante."

In the year 1873, I referred to this subject in my Bakerian Lecture