30 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 68 



to the wave-length under observation is lost by scattering into other 

 parts of the spectrum. Even with the slit closed the field may not 

 appear dark because of stray light in the room. General field light 

 like this is not very objectionable, for it is present both when 

 observing" the zero of energy with the shutter interposed in front 

 of the slit, and when measuring the intensity in the spectrum, and 

 is eliminated. 



In a well-designed spectroscope of moderate dispersion, produc- 

 ing a spectrum the intensity in which ranges lOO- or even i,ooo-fold, 

 scattered light may cause no trouble. But when, as in the present 

 research, the whole spectrum from the violet to 20 /x is contained in 

 a dispersion of only about i-l degrees and the intensities range 

 100,000-fold, this disturbance became exceedingly troublesome. In 

 the solar spectrum at 10 /x, for instance, the field-intensity due to 

 scattered energy may be over 100 times the true energy belonging 

 there. The safest expedient under such circumstances is to use a 

 sifting train ^ or auxiliary spectroscope forming" a spectrum on the 

 slit of the main spectroscope and so adjusted as to allow only that 

 part of the spectrum desired to pass through the main slit. The 

 scattered light of other wave-lengths is then deviated in the main 

 spectroscope to its proper place in the spectrum formed, and is 

 negligible. 



A very large proportion of the time and labor consumed in the 

 research was devoted to the elimination of errors from this stray 

 light, but it would seriously break the continuity of the exposition 

 to explain it in full here. Accordingly the subject has been rele- 

 gated to Appendix II, but the main principles of the method employed 

 to correct for stray light follow. 



Since nearly all of the radiation of a Nernst glower is of less wave- 

 length than 4 fji, it follows that nearly all of the stray light produced 

 by scattering is transmissible by quartz. Hence if the interposition 

 of a quartz plate at a point beyond 4 /x, where the ordinate of the 

 energy curve is y^, reduces the ordinate to y.^, the true ordinate proper 

 to the ray which should be found there does not exceed yi — y2- 

 But since a quartz plate reflects approximately 15 per cent of the rays 

 at wave-lengths found in the Nernst glower spectrum above 4 /x, it 

 is clear that the ordinate yo, which is due wholly to stray light trans- 

 missible by quartz, would have reached i.iSyo if the quartz had pro- 

 duced no reflection. Hence the true radiation could not have ex- 

 ceeded y^— i.i8y._,. Although rays of less wave-length than 4 fx. 



* Memoirs National Academy of Sciences, IV, p. 159, 1888. 



