120 BRIDGMAN. 



axis of the central cylinder contains a linear source of heat, that is a 

 wire carrying a current, precisely as for the metals. The difference of 

 temperature between the outer surface of the inner cylinder and the 

 inner surface of the outer cylinder is measured by thermo-couples. 

 These were of the same construction as for the metals, and were 

 mounted in fine copper tubes, which were sweated into small holes 

 drilled lengthwise of the cylinders. Of course with this construction 

 the couples could not be located exactly on the surface of either 

 cylinder, but the thermal conductivity of the copper is so much higher 

 than that of the petroleum ether that practically all the temperature 

 drop takes place across the annular space of the liquid, and a rough 

 computation shows that with the dimensions used any error from this 

 cause is negligible. As a precaution against failure of perfect geometri- 



inn\unnnumn\vnHi«i»\\muuun»\H\nu»nunuu\u\u\»n 



IMwmwwuwwwwvwwwwwwwwwwwwuwwwwTOwnwmuwuwn 



Figure 12. Longitudinal section of the apparatus for measuring the 

 thermal conductivity of petroleum ether. The liquid is showTi shaded between 

 copper cylinders, with a central heating unit and two sets of thermal junctions 

 bridging the liquid. 



cal centering of the inner cylinder in the outer one, three couples 

 instead of one were used, spaced at even angular intervals around the 

 cylinders, and these were connected in parallel, so that the reading 

 obtained gave the mean of the temperature differences around the 

 cylinder, and any geometrical imperfection is eliminated. The annu- 

 lar space between the cylinders was only 1.3 mm. wide. This is so 

 narrow as to remove any error from convection in the liquid, even at 

 atmospheric pressure, and it has already been explained that such 

 error vanishes at higher pressures because of the rapidly increasing 

 viscosity. No effects were found in the measurements to suggest 

 error from such a source. 



Because of the substantial equality of temperature throughout the 

 copper cylinders, it is to be expected that errors from slight changes in 

 position of the thermo-couples, which played so large a part in the 

 measurements of the metals, would vanish. This is indeed the fact, 

 and the measurements showed a high degree of regularity, much 

 beyond that obtained for any metal. 



In making the readings, the entire interior of the apparatus was 



