of some Mocks at High Temperatures. 61 



The experiment was unfortunately brought to an untimely 

 conclusion by the breaking of the multiple-couple inside the 

 furnace during an attempt to obtain the value of K at 650° C. 

 The rock on removal from the furnace was found to have 

 been considerably altered, being much whiter and less hard 

 than before. The change is probably due to the formation 

 of very minute cracks which would explain the permanent 

 lowering of the conductivity. 



Sources of Error. 



The chief errors likely to affect the results are those due 

 to end effects, which are very difficult to allow for. These 

 errors seem likely to occur in three ways, though their effects 

 are probably small when the conditions of the experiment 

 are suitable. 



As already stated, when the furnace is heated and no heat 

 is generated in the central wire, the axial temperature is, in 

 the case of the granite at least, slightly lower than that of 

 the outer parts of the rock. If only a small quantity of heat 

 is generated in the central wire, this error will make A too 

 small and K apparently too large. This error will increase 

 with increase in 6 and with decrease in W. 



On the other hand, the ends of the furnace are filled with 

 asbestos, which is certainly a worse conductor than the rock, 

 so that the radial temperature gradient due to the central 

 wire will be much greater in it than in the rock, i. e. when 

 a large amount of heat is developed in the central wire the 

 axial temperature at the ends of the rock cylinder will be 

 too high. That the axial temperature is actually higher in 

 the asbestos ends than in the rock for large central wire 

 currents is shown by the fact that on one occasion the central 

 wire fused in the asbestos when the mean temperature of 

 the part in the rock was far below the melting-point of 

 platinum, as obtained from the readings of the ammeter and 

 voltmeter just before the fusing occurred. This high axial 

 temperature at the ends of the rock cylinder affects the 

 apparent value of K in two opposite ways. Firstly, by con- 

 duction along the axis the centre radial temperature gradient 

 is made too large, and consequently K too small. Secondly, 

 as the axial temperature is higher near the ends the resistance 

 per centimetre of the central wire, and consequently the heat 

 evolution per centimetre, is greater near the ends than in the 

 middle, so that the rate of evolution of heat per centimetre 

 in the middle, where the gradient is measured, is less 



