( 215 ) 



of which, for producing the required density gradients, the temperature 



may be varied at pleasure by forcing an electric current or an air 



current through them. 



Cooling one of the tubes by an air current causes sodium vapour 



to condense on it ; so, in course of time, drops of molten metal will 



hang from the tubes and fall off again. 



s When a photograph is made, the first slit S of the 



spectroheliograph moves across the image in the direc- 

 tion of the arrow (fig. 2), and at the same time the 

 second slit moves across the photographic plate. 



Let us suppose the openings in P and Q (fig. 1) to 

 be so adjusted, that the image of the slit in P exactly 

 coincides with the slit in Q. Then all the light which 

 passes through P and traverses the vapour along- 

 straight lines, is transmitted by Q and therefore contri- 

 butes to the intensity of the image of the tube-section. 

 Fig. 2. Waves however that deviate so much in the sodium 

 vapour as to be intercepted by the screen Q, will be absent from 

 the spectrum of the transmitted light. 



If the furnace is slowly heated to 380° or 390°, the density of the 

 vapour is pretty uniform in the middle part of the wide tube and 

 falls off towards the ends; but as the direction of these density 

 gradients nearly coincides with that of the beam of sun-light, even 

 the waves subject to anomalous dispersion will hardly deviate from 

 the straight path. The Z)-lines in the spectroheliograph retain about 

 their normal appearance. If now we blow air through the tube B, 

 density gradients are produced all around it in directions perpen- 

 dicular to the axis of that tube. The ZMines no longer show 

 the same appearance throughout the Held. In the spectrum of those 

 parts of the field where perceptible gradients occur, the /Mines 

 now appear winged ; they are indeed enveloped in dispersion bands. 

 As the width of these bands depends on the magnitude of the gradient, 

 it will, in our case, vary along the lines and reach a maximum at 

 the place in the spectrum which corresponds to the plane passing 

 through the axes of the tubes A and B. And with increasing distance 

 between «S and B (tig. 2) the width of the bands will diminish. 



Let us consider the monochromatic images of the tube-section 

 produced by the spectroheliograph if the camera slit is set at different 

 distances from the /J-lines. 



With the camera slit at > 5850, outside the region of the dispersion 

 band of D, the illumination of the Held is uniform (see the Plate, «) ; 



