Conduction of Heat in Liquids. 21 



would give the ratio of their conductivities at any required 

 temperature. Christiansen seems to have seen the necessity 

 for determining the variation of conductivity in air, but made 

 no attempt to investigate it for any of the liquids. From 

 several experiments, in which both intervals were occupied by 

 air, he found, assuming 



k = Jc {l +7*}, 



where k refers to temperature 0° and k to temperature t, the 

 value of 7 to be '001504. He does not, however, apparently 

 consider the result sufficiently trustworthy to be applied to 

 the actual corrections. 



Another point requiring criticism is the extreme thinness 

 of the fluid layers. Thus, the lower was usually *1909 centim. 

 and the upper only *0214. In the experiments with air in 

 both intervals these were equal, being *0214 centim. in one set 

 and *0754 in another. The mere measurement to so great a 

 degree of accuracy would be by no means easy, and the 

 plates would require to be extremely smooth and accurately 

 plane. An error in the thickness would produce an exactly 

 proportionate error in the conductivity. Then the tem- 

 peratures of the plates separated by the thin (or air) layer 

 differed by from 10° to nearly 25°. Thus the air would be very 

 far removed from a uniform medium ; and unless the law 

 connecting the conductivity and the temperature be strictly 

 linear, a very serious error might be introduced. There 

 would also be a great deal of radiation across the fluid layers, 

 and the effects of diffusion even would require to be examined. 

 There is also the question whether there is not a species of 

 internal radiation, distinct from ordinary conduction, in the 

 case of a gas or liquid, at least when the variation in tempe- 

 rature is so extremely rapid as occurs here. 



To obtain absolute values of the conductivities from 

 Christiansen's results, the value of the conductivity in air is 

 required. The values obtained from the Kinetic Theory of 

 Gases by the formulae of Maxwell and Clausius almost agree, 

 and their mean is in C.G.S. units '0000490. The rate of 

 increase with the temperature is 7 = '00183 according to 

 Clausius, and twice as much according to Maxwell. Almost 

 identical values for the conductivity have been obtained 

 experimentally by Graetz and also by Kundt and Warburg, 

 and the former found 7=*0018. Somewhat, different values 

 have, however, been found by other observers, e. g. Winkel- 

 mann, who got a decidedly different value for y. He and 

 Graetz indulge in mutual criticisms, and neither seems to put 

 the accuracy of his results above suspicion. 



