210 ANNUAL REPORT SMITHSONIAN INSTITUTION, 1925 



term of analysis, the body from which we can take nothing but 

 itself — all that is far away. 



Such a definition can not be taken literally, for then the greater 

 part of the oxides of the rare earths would be elements, since they 

 can not be reduced. One rather feels that the definition of an 

 element should not be submitted to the contingencies of analysis. 

 We should define elements by characteristics which belong to them 

 and we have carefully felt our way for a long time through the 

 uncertainties of analogies. The isolation of an element in its metal- 

 lic state is no longer necessary to characterize it as an element and 

 we must have recourse to elementary characteristics. This leads 

 to two necessary conditions: 



(1) All compounds of the same element must have in common cer- 

 tain characteristics. 



(2) Among these we can retain only those which are general. 

 The lapse of time makes possible an accurate statement of the 



problem. In the earlier times to which we have alluded, a more 

 or less confused intuition served as a guide. 



The chemical equivalent, the proportional number, the atomic 

 weight, call it what you will, is manifestly insufficient because, 

 though satisfying the two preceding criteria, cited for the case of 

 pure elements, they also hold for certain compounds, for example, 

 the rare earths, of which we can measure the molecular weights, a 

 measure scientifically precise but, in the present instance, of little 

 meaning except to a specialist. Hence the two conditions given, 

 though manifestly essential, are not sufficient. It is further neces- 

 sary that the characteristic be independent of the state of the body 

 examined. Bunsen discovered in the spectrum the first strictly ele- 

 mentary characteristic. Until then the elements had been simply 

 chemically elementary ; henceforth they will be spectroelementary. 



The method of flame spectra had an astonishing success. There 

 came at once the discovery of rubidium, caesium, and then almost 

 immediately afterwards, of thallium and indium. However, it did 

 not prove to be a wholly general method. Indeed, there are besides 

 the alkaline metals and alkaline earths very few elements suscepti- 

 ble of being revealed by the flame. Do you not recollect that when 

 Bunsen undertook with Bahr the study of yttria he found only 

 yttrium and erbium among the elements of Masander? He then 

 gave all the weight of his authority against the existence of terbium, 

 a phantom element, denied by Cleve, asserted by Delafontaine, and 

 which owed its definitive resurrection to Marignac? 



The technique of spectrum analysis, then too primitive, needed de- 

 velopment. Thalen, and then Lecoq de Boisbaudran, had recourse 

 to the electric spark, which led to a much more general method. 



