Alkali Salt Vapours in Flames, 121 



which give 



2 log r 2 \r x 



The shape of the region emitting light is rather like an 

 ellipsoid of revolution with the long axis vertical and the 

 bead at the lower focus. I have fouud it convenient in 

 practice to observe the greatest horizontal diameter and the 

 corresponding value of r. To find the maximum value of 

 the diameter we differentiate 



, q iL(x-r) 



c= IrfT, e2K ' 



and put dyjdx=0 and z=0, which gives 



where r m and x m denote the values of r and x at a point where 

 the diameter is a maximum. 



A steady flame was obtained by passing the coal-gas and 

 air from an ordinary Bunsen burner through a box containing 

 baffle-plates to thoroughly mix them. The mixed gases were 

 burnt from a vertical brass tube about half an inch in 

 diameter. A brass tube about |- inch in diameter and 

 3 inches long was supported so as to be concentric with 

 the first tube, and the top of this second tube was about 

 one inch higher than the top of the other. The inner 

 cone of the flame was just below the top of the outer 

 tube. This arrangement gave a very steady flame the 

 diameter of which was nearly constant for 3 inches above 

 the burner. The outer tube prevented the wobbling of the 

 flame. A small bead of the salt to be investigated was sup- 

 ported on a platinum wire about 1 cm. above the tip of the 

 inner cone. Two fine platinum wires supported on adjustable 

 stands were adjusted so that their ends just touched the 

 surface of the region emitting visible light at opposite ends 

 of: a horizontal diameter of its widest part. The distance 

 from the bead to the end of either wire was measured, and 

 also the distance between the ends of the two wires. These 

 distances are r m and 2y m . We have then x m = y/r m 2 — y n , 2 . 



To find the coefficient of diffusion of the salt vapour we 

 require also to know the velocity of the flame gases. Small 

 luminous particles moving up the flame could be seen to 

 travel along parallel vertical straight lines for several inches 

 above the burner. The velocity was found by means of a 



