226 PROCEEDINGS OF THE AMERICAN ACADEMY. 



third fraction was dissolved in nitric acid, and, after the excess of acid 

 had been evaporated, the solutions were diluted to the same volume, 

 above live cubic centimeters. Photographs of the absorption spectra 

 were made, the solutions being contained in test tubes of the same 

 diameter, and various lengths of exposure being employed. The sama- 

 rium band A 401 was faintly visible in fraction 2923 only. This is not 

 surprising, however, since, as has been stated before, no fractions had 

 been removed from the less soluble end of the series in the course of all 

 the sixty- seven series of crystallizations from concentrated nitric acid. 

 Praseodymium could not be detected in fractions 2923 to 2932, but its 

 strongest absorption band, A 444, appeared with regularly increasing 

 intensity in fractions 2935 to 2944. 



In order to find out how much praseodymium can be detected in this 

 way, and at the same time to estimate the proportion of praseodymium 

 in the latter fractions, spectrograms were made, on one plate and with 

 the same exposure, of the spectra of most of the solutions previously 

 examined, and also of others similarly prepared from fractions 1302 

 and 1614, which had been removed at earlier stages of the fractionation. 

 To a solution of fraction 2926, in which no praseodymium could be de- 

 tected iu the first photograph, 0.1 percent of praseodymium oxide was 

 added in solution, and the original concentration was restored by evap 

 oration. The absorption spectrum of this specimen wa.s photographed 

 on the same plate as above, and again after the praseodymium content 

 had been increased to 0.2, 0.5 and 1.0 per cent respectively. In the 

 negative the first addition of 0. 1 per cent made a distinct impression. 

 From this it was concluded that fractions 2923 to 2932 contained less 

 than that proportion of praseodymium. This assumption can scarcely 

 be objected to on the ground that a considerable amount of praseodym- 

 ium in fraction 2926 might have been brought within the limits of de- 

 tection by means of a relatively small increment, because the intensity 

 of the praseodymium band appeared to be proportional to the amounts 

 of praseodymium added to the original solution. 



Since these standards and the other solutions differed little except 

 in praseodymium content, those giving praseodymium spectra of equal 

 intensity could be safely assumed to contain the same percentage of 

 this impurity. The results of the comparison of the different spectro- 

 grams are as follows : 



Since a proportion of one tenth of a per cent of praseodymium lowers 

 the atomic weight of neodymium by only four one thousandths of a 



