218 Mr. J. N. Lockyer. On the Classification [Jan. 10, 



of this dust in the higher regions of the air is by spectroscopic obser- 

 vations of the atmosphere containing it when it is rendered luminous 

 by electrical discharges. It becomes necessary, therefore, to make a 

 thorough investigation of the spectrum of the aurora borealis from 

 the point of view that meteoric dust, if it exists, is likely to assert 

 itself in any electrical excitation of the atmosphere. 



It is now many years since the idea was first thrown out that the 

 aurora was in some way connected with shooting stars. The connexion 

 was first suggested by Olmsted in 1833.* 



M. Zenger, in a catalogue of auroras observed from 1800 to 1877, 

 showed an apparent connexion between the brightest displays and the 

 appearance of large numbers of shooting stars, and M. Denza noted 

 the same connexion on November 27, 1872, and remarked that he 

 had noticed it before. 



In spite of these ideas, however, even after the chemical nature of 

 shooting stars was known, observers have in the main contented 

 themselves with making comparisons of the aurora spectrum with 

 the spectrum of air under different conditions of temperature and 

 pressure. 



It has never been possible, however, to reconcile the aurora 

 spectrum with any known spectrum of air. Some observers are of 

 opinion that the lines seen in the aurora coincide with air- lines, but 

 have different intensities, and they attempt to overcome this difficulty 

 by assuming that the aurora spectrum is produced under conditions 

 which we are unable to imitate in our laboratories. 



When we recognise the importance of considering the possible 

 existence of meteoric dust in the atmosphere, a comparison with the 

 spectra of uncondensed meteor-swarms is at once suggested, for the 

 more my researches advance the more does dust rather than large 

 meteoritic masses appear to be in question. 



The result of a preliminary comparison with 7-Cassiopeise and with 

 the bands in Duner's stars was communicated to the Royal Society on 

 January 9, 1888. The tables which I then gave show that there is 

 probably a very intimate relation between the spectrum of the 

 aurora and those of meteor-swarms. 



The further inquiry into the recorded observations to which I have 

 subsequently to refer, seems entirely to justify the suggestion then 

 put forward, and I now propose to show what progress has been 

 made in attacking what has always been regarded as a difficult 

 subject. I will first, however, briefly refer to the observations and 

 comparisons which have been previously made, and discuss them in 

 chronological order. 



It is necessary to state that the existing observations of aurora 

 spectra show such great differences of wave- length for what are 

 * ' Amer. Journ. Sci ' vols/35 and 36. 



