Apru 26, 1888] 



NATURE 



607 



composes peuvent se former ou se dissocier, suivant les 

 variations de temperature auxquelles ils sont assujett^s 

 par Taction chimique meme. Dans cette classe doivent 

 probablement ctre comprises les etoiles dont I'intensitd 

 de lumiere varie plus ou moins rapidement, et avec une 

 pdriodicitd plus ou moins constante." 



In the year 1873, I referred to this subject in my Baker- 

 ian Lecture {Phil. Trans, vol. clxiv. pt. 2, 1873, p. 492), in 

 which I attempted to bring to bear some results obtained 

 in solar inquiries upon the question of stellar temperatures. 



I quote the following paragraphs : — 



I. The absorption of some elementary and compound 

 gases is limited to the most refrangible part of the spec- 

 trum when the gases are rare, and creeps gradually into 

 the visible violet part, and finally to the red end of the 

 spectrum, as the pressure is increased. 



II. Both the general and selective absorption of the 

 photospheric light are greater (and therefore the tempera- 

 ture of the photosphere of the sun is higher) than has 

 been supposed. 



III. The lines of compounds of a metal and iodine, 

 bromine, &c., are observed generally in the red end of the 

 spectrum, and this holds good for absorption in the case 

 of aqueous vapour. 



Such spectra, like those of the metalloids, are separated 

 spectroscopically from those of the metallic elements by 

 their columnar or banded structure. 



IV. There are, in all probability, no compounds 

 ordinarily present in the sun's reversing (layer. 



V. When a metallic compound vapour, such as is re- 

 ferred to in III., is dissociated by the spark, the band 

 spectrum dies out, and the elemental lines come in, 

 according to the degree of temperature employed. 



Again, although our knowledge of the spectra of stars 

 is lamentably incomplete, I gather the following facts 

 from the work already accomplished with marvellous skill 

 and industry by Secchi, of Rome. 



VI. The sun, so far as the spectrum goes, may be re- 

 gar-ded as a representative of class, (/3) intermediate be- 

 tween stars (a) with much simpler spectra of the same 

 kind and stars (y) with much more complex spectra of a 

 different kind. 



VII. Sirius, as a type of a, is (i) the brightest (and 

 therefore hottest ?) star in our northern sky ; (2) the blue 

 end of its spectrum is open, — it is only certainly known to 

 contain hydrogen, the other metallic lines being exceed- 

 ingly thin, thus indicating a small proportion of metallic 

 vapours ; while (3) the hydrogen lines in this star are 

 enormously distended, showing that the chromosphere is 

 largely composed of that element. 



There are other bright stars of this class. 



VIII. As types of y the red stars may be quoted, the 

 spectra of which are composed of channelled spaces and 

 bands, and in which naturally the blue end is closed. 

 Hence the reversing layers of these stars probably contain 

 metalloids, or compounds, or both, in great quantity ; 

 and in their spectra not only is hydrogen absent, but the 

 metallic lines are reduced in thickness and intensity, 

 which in the light of V., ante, may indicate that the 

 metallic vapours are being associated. It is fair to assume 

 that these stars are of a lower temperature than our sun. 



In the same year, in a letter to M. Dumas, published in 

 the Comptes rendus^ I again pointed out that, if we con- 



^ " II semble que plus une ^toile est chaude, plus son spectre est simple, et 

 que les elements metalliques se font voir dans I'ordre de lines poids anato- 

 miques. Ainsi nous avons :— 



" (i) Des Etoiles tres brillantes, ou nous ne voyons que I'hydrogine en 

 quantitd dtiorvie, ct le magnesium. 



" (2) Des etoiles plus froides, comme notre soleil, oil nous trou\03s: — 

 H + Mg + Na. 

 H + Mg + Na + Ca . Fe, &c. ; 

 dans ces Etoiles, pas de metalloides. 



"(3) Des Etoiles plus froides encore, dans lesquelles tous les elements 

 metalliques sont associes, oil leurs lignes ne sont plus visibles, et oil nous 

 n' avons que les spectres des mtftalloides et des composes. 



" (4) Plus une etoile est agee, plus I'hydrogcne libre di^arait ; sur laterre, 

 nous ne trouvons plus I'hydrogcne en liberty." 



sider merely the scale of temperature, a celestial body with 

 flutings in it would be cooler than one which bad lines in 

 its spectrum ; and I also pointed out that, taking the con- 

 siderable development of the blue end of the spectrum in 

 white stars as contrasted with its feeble exhibition in 

 stars like our sun, we had strong presumptive evidence 

 to the effect that the stars like a Lyrae, with few lines in 

 their spectra, were hotter than those resembling our sun, 

 in which the number of lines was very much more con- 

 siderable, and I added an inference from this : "plus une 

 Etoile est chaude, plus son spectre est simple." This re- 

 lated merely, as 1 have said before, to the consideration 

 of one line of temperature. 



Vogel's Classification. 



In the year following my paper, the most considerable 

 classification which has been put forward of late years 

 was published by Dr. Vogel {Astr. Nach., No. 2000), who, 

 basing his work on the previous types of Secchi, and 

 taking into account the inference 1 drew in my letter 

 to Dumas, modified Secchi's types to a certain extent, 

 but always along one line of temperature, the leading idea 

 being, as I gather from many remarks made in Dun^r's 

 admirable memoir, to be referred to presently, that the 

 classification is based upon descending temperatures, 

 and that all the stars included in it are supposed at one 

 time or other to have had a spectrum similar to that of 

 a Lyrae.i 



This classification is as follows : — 



Class I. Spectra in which the metallic lines are ex- 

 tremely faint or entirely invisible. — The most refrangible 

 parts, blue and violet, are very vivid. The stars are 

 white. 



{a) Spectra in which the lines of hydrogen are very 

 strong. 



(h) Spectra in which the lines of hydrogen are wanting. 



{c) Spectra in which the lines of hydrogen and D3 are 

 bright. 



Class II. Spectra in which the metallic lines are 

 mimerous and very visible. — The blue and violet are 

 relatively weaker; in the red part there are sometimes 

 faint bands. The colour of the stars is clear bluish 

 white to deep reddish yellow. 



(rt) Spectra with numerous metallic lines, especially in 

 the yellow and green. The lines of hydrogen are gener- 

 ally strong, but never as strong as in the stars of Class I. 

 In some stars they are invisible, and then faint bands are 

 generally seen in the red formed by very close lines. 



{b) Spectra in which besides dark lines and isolated 

 bands there are several bright lines. 



Class III. Spectra in which besides the metallic 

 lines there are numerous dark bands in all parts of the 

 spectrwn, and the blue and violet are remarkably faint. — 

 The stars are orange or red. 



(a) The dark bands are fainter towards the red. 



{b) The bands are very wide, and the principal are 

 fainter towards the violet. 



It is pointed out that if this classification be true, there 

 must be hnks between all the classes given. Now it is 

 perfectly obvious that if this classification includes in its 

 view all the stars, and if there is a line of ascending as 

 well as descending temperatures — that is to say, if some 

 of the stars are increasing their temperatures, while others 

 are diminishing them — the classification must give way. 



It is not difficult to see. in the light of my communica- 

 tion to the Society of November 17, that it has given way 

 altogether, and principally on this wi.se. 



The idea which underlies the classification is that a 

 star of Class I. on cooling becomes a star of Class II., 

 and that a star of Class II. has as it were a choice before 



' " Car selon la thforie il faiidra que tot ou tard toutes les efoJles de \x 

 premiere classe deviennent de la seconde, et celles-ci de la troisicr. e" 

 (Duner). 



