of Kirchhoff and Bunsen (Wiss. Klassiker, No. 72, p. 71) gives the following account as coming from the latter author who was then, 1895, still living:

"Bunsen had for sometime been engaged in the study of flames as colored by various salts and had used these colors as a means of analysis.

"For the purpose of distinguishing between flames having apparently the same color, he interposed colored glasses and colored solutions--and with considerable success.

"It was during a conversation on this subject that Kirchhoff suggested to him that the different colors in the flame might be more completely separated by means of a prism; later Bunsen was aided by Kirchhoff in carrying out this idea."

This led to their two joint papers, published in Pogg. Ann. in 1860 and 1861, where the question is first raised as to whether each element has a fixed and characteristic spectrum of its own, which is independent of the kind of flame in which it is brought to incandescence, and independent of the compound of which the element may be a component.

After examining the spectra of a large number of compounds in which one and the same element appears as a factor, and after employing a variety of heat-sources ranging from the hot oxyhydrogen flame to the cold flame of sulphur, they conclude that the positions, at least, of the spectral lines, for any of the alkali metals, are not affected by the remainder of the compound in which the metal appears, or by any range of temperature, or by any of the various chemical processes going on in the flame employed.

Upon this experimental foundation, they established the prism as an instrument whose delicacy, for qualitative analysis, is unsurpassed; and in their second paper applied it to the discovery of two new elements, caesium and rubidium.

The peculiar merit of Bunsen in the field of spectroscopy is then that he perfected a simple method which is capable of detecting the presence of an element by the use of quantities vastly smaller than are required by any other known method.

It opened to chemists, therefore, an entirely new field of investigation and one which, as the sequel has proved, is not confined to our own planet or even to the solar system.

Colored flames had been previously studied by Talbot, Herschel, Swan and Miller, but the work of Kirchhoff and Bunsen was needed to convince chemists that the prism was an instrument both reliable and sensitive.

In this connection, it is interesting to note that it is precisely the deviations from the general principle above mentioned, viz., constancy of position for each line, which form the most. interesting subjects of present spectroscopic investigations. Among these may be mentioned the difference between the spectra of various salts of any one metal as determined by Mitscherlich, the shift of position due to motion in line of sight, the shift due to pressure at the luminous source.

Since 1889, Bunsen had not been engaged in active teaching. As a matter of course, learned societies sought to honor themselves by heaping upon him memberships, degrees and orders. These, however, ought hardly to be mentioned on the same page with the genuine admiration which his marvelous skill of hand, his wonderful clearness of thought and his marked originality called forth from the scholars and students of two continents. And even this praise can hardly have a place beside the loving esteem which his simple-mindedness and his warmheartedness gained for him among all his personal friends.

THE WAVE-LENGTH OF THE CORONA LINE.

By C. A. YOUNG.

PROFESSOR CAMPBELL'S note in the October number of this JOURNAL seems to require a word from me, as having been the original author of the apparent misidentification of the corona line with the "1474" line (A 5317) of the cromosphere spectrum. I am obliged to say frankly, although with some natural regret, that I see no reasonable ground on which to contest the conclusion of Sir Norman Lockyer and Professor Campbell, that the wave-length of the corona line is 5303, and that the line is not identical with the conspicuous chromosphere line at 5317 as I have hitherto supposed. The spectrum photographs which Pro fessor Campbell has kindly sent me appear to be conclusive on this point; and I presume that those of the English spectroscopist are no less so, although I have not yet seen any detailed exhibition of his data such as would enable one to examine them critically.

The explanation given by Professor Campbell to account for the original mistake is doubtless correct, and if the totality had lasted a few seconds longer at the eclipses of 1870 and 1878 I should have detected the error myself. I had planned to set the micrometer wires on the corona line just before the end of totality, and to watch the reappearance of the chromosphere spectrum, followed by the reëstablisment of the ordinary dark line spectrum. Had I succeeded in so doing, the non-coincidence of the corona line with its much more brilliant chromospheric neighbor could not have possibly escaped notice. But on both occasions totality ended prematurely, while I was still searching for other coronal lines in the violet region of the spectrum.

If the weather is favorable next May the eclipse observer will need at the proper moment only a glance to verify or disprove the identity of the corona line with the old "1474."

It is, I think, worth noting that at λ 5303 there is no line of the solar spectrum, either dark or chromospheric; nor, so far as I know at present, is there at that point any line belonging to the spectrum of any terrestrial element yet investigated. The new determination of wave-length frees "Coronium" from all appearance of puzzling affinities or "entangling alliances" with other bodies less exceptional in their character.

PRINCETON, N. J.

Nov. 4, 1899.

DENSITY OF CLOSE DOUBLE STARS.

By ALEXANDER ROBERTS.

IN connection with the theory of the relation of the temperature and density of stars, it is important to put in evidence what is known of the size, mass, or density of celestial gaseous bodies.

The need for such evidence will be manifest when we consider that there is no final certification of the truth of any law, even if we admit its theoretical soundness, so long as it lacks the complete verification that follows from a comparison of theory with the facts deduced from observation.

As regards mass we, of course, can only ascertain this in the case of double stars whose parallax and orbital elements are known. And to determine the density necessitates further a knowledge of the size of the component stars of the system, and this can only be arrived at indirectly by an examination of the light variation of those binary stars whose orbits lie in the same plane as the line of sight.

In the present paper I would seek to deal with the densities. of four Algol stars, the light variations of which have been under observation at Lovedale for several years.

Before dealing with the stars severally it is necessary to

indicate the general expressions for the density of any system. Let us employ the following designations: