BLACKWELL. — DISPERSION IN ELECTRIC DOUBLE REFRACTION. 655 



of glass were previously used, but their insulating power seemed in time 

 to fail. The bars are finally firmly wedged together in glass rings just 

 large enough to slip over them (Figure 3). Thus assembled, they are 

 shown in the Plate, Figure B. 



Tube. — The construction of the tube to hold these bars immersed in 

 the liquid to be observed was a difficult problem. The tube must admit 

 insulated leads to the bars from a source of high voltage, must be closed 

 liquid tight by glass end-plates, must be connected by glass tubes with a 

 supply of pure, dry carbon bisulphide, and must be chemically inert and 

 capable of being made chemically clean. It must be three inches in 

 diameter and nearly four feet long (7 X 110 cm.). 



Glass tubes were tried first. Four in succession cracked spontaneously. 

 Apparently such large tubes are under great internal strain, and when 

 this is released in grinding the ends, the glass gradually yields. Some- 

 times a night and sometimes a week elapsed before the crack came, but 

 fortunately it came before any tube was in place and filled with carbon 

 bisulphide. 



Glass is too dangerous to rely on, and special enamelled iron tubes 

 were tried. On opposite sides, as near as possible to each end, insulating 

 " lava " bushings about 2.5 cm. in diameter were screwed in before enam- 

 elling. The ends of the tubes were flanged, and had to be machined to 

 a smoothly convex surface in order to retain a good continuous coating 

 of enamel. Four attempts were made before a satisfactory tube was 

 secured. 



Closing Tube. — The necessity of having enamel on the ends of the 

 tube is due to the method of closing it. In order to avoid the use of 

 washers, or cement, which might be more or less soluble in carbon 

 bisulphide, a fine fit alone is relied upon. Glass discs for the ends were 

 cut out of "crystal plate," about 1.5 mm. thick, only those being used 

 which when finished seemed free from double refraction. " Crystal 

 plate " is ground and polished in manufacture, and consequently free 

 from small irregularities in the surface. This was shown with an opti- 

 cally flat test-plate by the smoothness of the Newton's fringes in the air 

 film between plate and disc. The enamelled ends of the tube were 

 ground, at the Clarks', optically flat and nearly parallel. The tube was 

 then closed by simply pressing the glass plates against the flat ends 

 (Figure 4). Thin rubber rings were inserted between the glass and the 

 rigid cast-iron rings as a precaution against breakage. 



When all is in place, a continuous rubber ring, or gasket, previously 

 slipped over the tube, is wired down on tube and ring, and mercury is 



