36 Rubens and Trowbridge — Dispersion and Absorpti 



ion 



within 0*2 per cent. Greater accuracy cannot be claimed for 

 our measurements on dispersion. 



It is evident that this arrangement is only applicable when 

 the angle of the prism is very acute and the angles of deviation 

 a are very small. In our case, the angles of the rock salt and 

 sylvine prisms were approximately 12° ; the observed devia- 

 tions, between 4J° and 6°. With angles of this size the simple 

 relation between 7, and a given above, already leads to errors 

 in the deviation 7 which may equal one minute. The error 

 becomes, however, very much smaller by reckoning from the 

 minimum deviation of the sodium line instead of from the 

 deviation ; namely, when one does not attempt the absolute 

 measurement of the deviation a but confines oneself to the 

 determination of the angle of dispersion (<x D — a^). This is, in 



fact, entirely sufficient, for after determining this angle of dis- 

 persion, it is sufficient to know the deviation for the sodium 

 line in order to calculate from this the deviations for all the 

 observed wave lengths. 



If the arrangement just described is used, then the error 

 still present in the formula y D — 7^ = const, (a^ — a^) is very small 



and may be completely ignored, if a small correction is applied 



e r) 

 to the constant -° -* 



P*. 

 The rotation of the mirror e about its vertical axis was 

 effected by means of a spectrometer on the table of which the 

 mirror was mounted. The rotation of the table could be 

 measured to within ten seconds on the graduated circle of the 

 instrument by means of a vernier. As this accuracy was not 

 sufficient for our purpose, we applied a divided micrometer 

 head to the slow motion screw of the instrument, by means of 

 which we were able to attain an accuracy of two seconds for 

 small arcs. The prism p was set at minimum deviation for 

 sodium light and kept throughout the work in \\m position. 

 Therefore the application of the formula 



* Strictly speaking j3 = y(e- p/ps ) is not correct, the correct equation being 



6a 



= are.fcj^yj 



e p 



1+ *** 



w 



noH«v4-^-VsiW 



hence 



oy cos»y+( 



\ps Q / 



If a mean value for 7, say 5°, be substituted in this equation, there appears in 



the place of the constant e s /ps " a value for ~- which differs from the constant 



by 0*1 to 0'2 per cent. 



