May 1 6, 1889J 



NATURE 



69 



The observations were made on February 4, 1874. There 

 was a brilliant aurora, showing a whitish light ; a direct-vision 

 spectroscope resolved this light into a brilliant line in the yellow 

 and a faint continuous spectrum. 



The "Grubb" spectroscope was directed from the window of 

 the Observatory upon the brightest part of the aurora. In the 

 first instance, an estimation by eye was made of the position of 

 the bright line by comparing it in the instrument with the spec- 

 trum of a spirit-lamp. The bright line was seen to fall on the 

 more refrangible side of the line for which Watts gives the wave- 

 length 5582 {Phil. Mag., vol. xli., 1871, p. 14), Angstrom and 

 Thalen 5583 ("Spectres des Metalloides," Nov. Act. Soc. Sci. 

 Upsal., vol. ix,, 1875, P- 29), by from one-fifth to one-fourth 

 of the distance of this line from the beginning of the band. If 

 we take one-fourth, we have A 5569*6 ; one-fifth gives A 5572"3. 

 The mean of these values gives for the 



Aurora line A 5570"9 (i) 



The cross-wires of the spectroscope were then brought upon 

 the line, and the reading 3476 showed the line to fall about 

 midway between two strong lines in the spectrum of tin, X 5564 

 and A 5587 respectively, according to my measures (" Spectra of 

 the Chemical Elements," Phil. Trans., 1864, p. 139). The 

 position of the cross was thea compared directly with those lines 

 in the spectrum of an induction spark taken between electrodes 

 of tin. The further details of this comparison are not given in 

 my note-book, but the result only, which placed the 



Aurora line at A 5571 (2) 



Consulting my map of the chemical elements, I found that 

 there was a line of tellurium very near this place — namely, at 

 A 5575 ; I therefore brought the spark from tellurium before the 

 slit, when the cross appeared on the more refrangible side of the 

 tellurium line. The measure of the distance of the cross from 

 this line came out equal to A 0003. . The place given in my 

 paper for this line of tellurium is 5575. Thalen gives for the 

 same line 5574*1 (Brit. Assoc. Rep., 1885, p. 292). If we take 

 the mean of these values and deduct 0003, we get for 



The line of the aurora A 5571*5 (3) 



There are strong lines of iron very near this position in the 

 spectrum, and I made use of these also for a further determina- 

 tion of the place of the aurora line. The cross, after having 

 been placed upon the line of the aurora, was confronted with 

 these lines in the spectrum of iron. 



The condensed account in my note-book does not give further 

 particulars of this comparison, but states only that the place of 

 the 



Aurora line came out A 5571 '5 (4) 



Summing up these determinations we have — 



(i) Eye-estimation A 5570*9 



(2) From tin 5571 'o 



(3) From tellurium 5571 '5 



(4) From iron 5571 5 



" From these values I think that we are justified in taking for 

 the aurora line, as a position very near the truth, 



A 5571 ±0*5 (5) 



Among the numerous determinations of other observers, those 

 of Prof. H. C. Vogel in 1872 [Lepzig Math. Phys. Berichtc, 

 vol. xxii. p. 285) seem to me to have great weight. A direct- 

 vision spectroscope with a set of five prisms was used. The 

 reduction of the readings of the micrometer into wave-lengths 

 was based upon the repeated measures of 100 lines of the solar 

 spectrum. 



The screw had been thoroughly examined. After each ob- 

 servation of the aurora line, readings were taken of the lines of 

 sodium or of hydroj^en. The observations extended over four 

 nights. On three nights four separate readings were obtained ; 

 on the fourth night two only. Vogel gives as the mean result 

 of the fourteen observations — 



Aurora line A 5571*3 ± 0*92 (6) 



The recent observations on the spectrum of the .-lurora by 

 Gyllenskiold, at Cap Thordsen, in 1882, de;eive special 



mention.^ With a Hoffmann spectroscope, furnished with 

 a scale, he obtained at Cap Thordsen in 1882 a mean result 

 of A 5568 ± 1*6; later, in 1884, at Upsala, with a Wrede 

 spectroscope furnished with a micrometer screw, a mean value 

 for the aurora Ime, A 5569 ± 6*2.- Gyllenskiold discusses in 

 detail nearly all the recorded observations of the spectrum of 

 the aurora from 1867 to 1882, and then brings them together in 

 a table, with such probable errors as the original statements of 

 the observers enabled him to assign to them. 



Gyllenskiold then calculates by the method of least squares 

 the mean value of all the determinations, and finds the following 

 result : — ■' 



Mean value of the 23 observations, A 5570*0 ± 0*88 . . (7) 



The recent measures by C. C. Krafft,* depart largely from 

 Gyllenskiold's mean value. Krafft found on 



1882 November 2 A 5595 



II 5586 



and measures with the same instrument made by Schroeter, on 

 November 17, gave A 5587. 



Now, though Angstrom's original value, A 5567, may not be 

 quite accurate, his observation fixed a limit towards the red 

 beyond which the aurora line cannot lie. Angstrom says : " Sa 

 lumiere etait presque monochromatique, et consistait d'une seule 

 raie brillante situee a gauche " (on the more refrangible side) 

 " du groupe connu des raies du calcium" ("Spectre Solaire," 

 Upsal, 1868, p. 42). The position of the most refrangible line 

 of this calcium group is accurately known ; according to (Brit. 

 Assoc. Rep. 1884, p. 372) 



Kirchhoff 

 Thalen .., 

 Huggins . . 



A 5580*9 

 5580*9 

 5581-0 



It is certain therefore, from Angstrom's first observation in 

 1867, alone that the aurora line lies well on the more refrangible 

 side of wave-length 5580. This limit towards the red was con- 

 firmed afterwards by Angstrom himself; he says later that the 

 yellow line falls almost midway between the second and third 

 line of the shaded carbon group (Nature, vol. x. p. 211), The 

 positions of these lines of comparison are, according to Angstrom 

 and Thalen, A 5538 and A 5583 {Acta Upsal., vol. ix. 1875, 

 p. 29). 



It follows that Kraftt's values, A 5586, A 5587, and A 5595, 

 must be from some cause inaccurate. A possible explanation 

 may be found in the small number of solar lines employed by 

 Krafft for the reduction of the measures into wave-lengths. The 

 curve was drawn through the six Fraunhofer lines, B, C, a, D, 

 E, and b. There was no control for the curve between D and 

 E, and a very small deviation of the curve from its true position 

 here would be sufficient to account for the position of less 

 refrangibility of from A 0016 to A 0024, which his measures give 

 for the aurora line. 



It should be stated that Krafft expresses regret that more at- 

 tention could not be given to the spectroscopic observations. 

 He says : — " Leider gestatteten die obligatorischen Beobacht- 

 ungen nicht, den spectroscopischen Untersuchungen die gehorige 

 Aufmerksamkeit angedeihen zu lassen. . . . Ich glaubte ausser- 

 dem diese Messungen um so mehr auslassen zu konnen, als der 

 Platz der gewohnlichen Nordlichtlinie oft und sehr genau 

 bestimmt ist," 



To sum up, we have the following values for the principal line 

 of the aurora : — 



(6) 1872, Vogel . . ; A 5571 '3 ± 0-92 



(5) 1874, Huggins 5571*0 ±0-5 



(7) Gyllenskiold's mean of twenty-three 



obser%'ers from 1867 to 1884 . . , 557o*o ± o*88 



These values agree closely, and fix within very narrow limits, 

 the position in the spectrum where we have to seek the chemical 

 origin of the line. 



Gyllenskiold, from his observations of the changes which 

 occur in the spectrum of the aurora, comes to the conclusion 



» "Observations faites an Cap Thordsen, Spitzberg. par I'Expedition 

 Surfdoise," vol. ii. part 1, "Aurores Bo.-ealss," par Carlheim-GyllenskiOld 

 (Stockholm, 1886;. 



-- Ibid. p. 166. 3 Ibid. p. i5q. _ 



4 " Beobachtungs-Ergebnisse der^ Norwegischen Polarjtation, &c. 

 A. S. Steen(Christiania, 1888). 



