306 



Mr Glazebrook, On a Spectrophotometer. [Jan. 29, 



Hence 



fc=l- 



But if k represent the proportion of light lost by absorption 

 and reflexion at the faces of the vessel, we have 



r a = i a (i-k). 



tan 2 0' 

 tan 2 !' ' 



To eliminate the effects of the vessel the experiment should be 

 repeated with the vessel filled with water or some other fluid in 

 which the absorption is small ; the difference between the two 

 results will be the absorption due to the thickness used of the 

 absorbing medium. 



Of course in all cases four positions of the Nicol can be found 

 in which the two spectra will appear to have the same intensity. 

 At least two of these positions— which are not at opposite ends of 

 the same diameter — should be observed and the mean taken. In 

 this manner the index error of the pointer or circle will be 

 eliminated. A number of observations made at my request by 

 Mr F. W. Sanderson, B.A. of Christ's College, seem to shew that 

 the greatest difference between any two of a set of 6 or 8 readings 

 will be from 2° to 3°, while the mean error of twelve or fourteen 

 such sets is considerably under 1°; this result is confirmed by 

 my own observations. 



The following Table gives a series . of values of 6, & , and k 

 obtained by Mr Sanderson when the absorbing medium was a 

 small glass cell of about 1 cm. thick filled with water, to which a 

 slight blue tinge had been given by the addition of a few drops of 

 a solution of sulphate of copper. 



To consider the question a little more fully, let us, following 

 Bunsen and Koscoe, call the reciprocal of the thickness of the 

 medium which allows one-tenth of the light to pass, the extinction 

 coefficient. Let I' be the intensity of the light after passing a 

 thickness m of the medium, I being that of the incident light, and 

 E the extinction coefficient, then (Hufner, Quantitative Spectral- 

 Analyse) we have 



