84 ABSORPTION SPECTRA OF SOLUTIONS. 



position of the solvent, in order to give the bands with about half their 

 normal intensity, would be 7 or 8 per cent water and the rest alcohol. 

 In other words, we again find complete agreement between solutions of 

 neodymium chloride in the two alcohols. 



If the fact described under the last heading, that the relative inten- 

 sities of the two sets of bands depend only upon the ratio of water to 

 neodymium chloride in solution, should be found to hold even for concentra- 

 tions of one-tenth or one-hundredth of those employed here, this ought to 

 furnish a very convenient optical method of detecting rather small quan- 

 tities of water in alcohol ; for it is apparent that with a quarter normal 

 solution, 1 per cent of water gives the bands due to the aqueous solution 

 with sufficient intensity to be seen easily with a small spectroscope if a 

 layer of a centimeter or so in depth is used. Accordingly, to detect an 

 amount of water as small as 0.01 per cent, it would only be necessary to dis- 

 solve in the alcohol enough anhydrous neodymium chloride to make a 2T 

 normal solution, and fill a glass tube with the solution, so as to get a layer 

 from 50 to 100 cm. deep, when the bands due to water should easily be seen. 



NKODYMIUM CHLORIDE ANHYDROUS. (See Plate 68.) 



This plate was made in order to see whether the spectrum of the an- 

 hydrous salt is identical with that observed when the salt is dissolved in 

 pure methyl or ethyl alcohol. The anhydrous salt was in the form of a 

 very fine powder, and contained in a bottle with a tight-fitting glass stopper. 

 An image of the Nernst filament was thrown on the surface of the powder 

 in contact with the walls of the bottle, and this image was in turn focussed 

 on the slit of the spectroscope by means of the concave spectrum mirror. 

 The light falling on the grating was necessarily very faint; therefore, rather 

 long exposures were necessary; but this caused no inconvenience, since 

 the Nernst lamp burns so steadily that it needed no attention whatever. 

 In order to show as well as possible both the strong and the weak bands, 

 a series of exposures were made on the same film, the times of exposure, 

 beginning with the strip nearest the numbered scale, being 30 minutes, 

 1 hour, 1 hours, 2 hours, and 2J hours. On account of the fact that the 

 beam of light had to pass through the glass condensing lenses, as well as 

 the glass walls of the containing bottle, the spectrum ends at about A 3450 

 for the strip nearest the comparison spectrum, and at A 3600 for the one 

 nearest the scale. 



The comparison spark spectrum in this case was made by using zinc 

 terminals instead of the carbon terminals employed throughout the rest 

 of the work. Since there is usually some accidental shift between the 

 successive strips on a film, and since no light but that of the Nernst fila- 

 ment was used in making the five strips on Plate 68, it is evident that no 

 accurate wave-length measurements could be made by a comparison with 

 the spark spectrum on this plate. In fact, the position of a given absorp- 

 tion line, which appeared both on the film and on the red-sensitive plate, 

 was found to differ by as much as 10 Angstrom units as measured from the 

 two negatives. Hence it was necessary to determine the position of one 

 or more of the absorption lines by comparison with a spark spectrum which 



