Absolution Spectra of Dilute Solutions. 



127 



possible from dust, ordinary distilled water was taken and distilled a 

 second time from a metal can witli block-tin condenser, the first third 

 of the distillate being rejected. This water, filled immediately into 

 the tube, was almost free from visible dust in a layer 1 m. long. It 

 was just visibly blue. I found afterwards that water filtered through 

 a " Chamberland Pasteur candle " was even more devoid of dust than 

 that which had been freshly distilled, and some measurements were 

 made with such filtered water. There was, however, no noticeable 

 difference in the numbers. 



The actual measurements of the light transmitted by a tube of 

 water lOO'l cm. long are given in the Appendix Table I. A curve 

 smoothed as little as possible was drawn from these numbers, and 

 the values of T in Table II interpolated from it. From them the 



values of the extinction coefficient for 1 cm. of water ( e = - Og r ) 



V lOO-l/ 



were calculated. 



Table II. 



The numbers obtained by Hiifner and Albrecht (' Wied. Ann.,' vol. 

 42, p. 10, 1891) are also included in the table. The agreement is satis^ 

 factory except at wave-lengths 576*5 and 551'5. Owing to this lack 

 of agreement I repeated the measurements in this part of the spec- 

 trum several times, but never found such large absorptions as Hiifner 

 and Albrecht. 



In fig. 3 the values of e are plotted against the wave-lengths. The 

 absorption diminishes very rapidly between the solar lines A and B, 



