and an Optical Method of Calibration. 425 



observation reduces the intensity too much to allow of accurate 

 settings in many experiments. The same may be said of the 

 variable rotating sector in direct observations. The loss of 

 light is at least one-half by the last two methods, and generally 

 much more. As Lummer and Murphy have shown, the direct 

 reading of the width of the slit cannot be taken as a measure 

 of the relative intensities, errors of several per cent, arising 

 in different parts of the spectrum, depending on the slope of 

 the luminosity curve. This fact makes the method of Vierordt 

 with a double slit defective. It is only with very narrow slits 

 and large dispersion that the direct readings can be taken : 

 and even in this case the true zero of the slit cannot be accu- 

 rately determined, and is subject to variation by the dete- 

 rioration of the edges. 



However, either of the above methods — varying the distance 

 of the radiant or using Nicol prisms, or placing a rotating 

 sector between the source and the slit — may serve to obtain 

 the true optical value of the slit for different widths and 

 colours. This method of optical calibration eliminates all the 

 errors of the screw and irregularities and lack of parallelism 

 of the edges of the slit, and corrects the variations in the slope 

 of the luminosity curve. With this calibration, comparisons 

 can be made far more rapidly, and as accurately as with the 

 direct use of the rotating sector, which 

 Lummer and Brodhun have shown to Fig. 4. 



be the most reliable method. Usually 

 with each setting this sector must be 

 stopped and read, a tedious but accu- 

 rate process. In order that such a 

 sector may be capable of variation 

 and adjustment while running, to ob- 

 tain a match, elaborate and careful 

 construction is required. A simple 

 disk, however, divided into several 

 sectors, say six to eight, may be 



used to determine the optical values of the slit — which 

 should be bilateral — in corresponding ratios ; and the inter- 

 mediate values may be found by interpolation, and thus 

 eliminate any further use of the disk, the screw- readings 

 being used thereafter directly, with the corrections of the 

 calibration. 



A simple cardboard disk, mounted on a whirling-table or 

 motor, with its circumference divided into any convenient 

 number of parts — eight for example — and slotted out to 

 different depths between different radii, as shown in fig. 4 

 and then slid forward between the source and the slit, so that 



