Measurement of the Peltier Effect. 463 



In Perrons experiments a large temperature-gradient was 

 established between the two contact surfaces, and the thermo- 

 electric power would include possible thermoelectric forces 

 arising from slight changes in the crystalline structure. In 

 the apparatus, described in this paper, the mean of the Peltier 

 forces near the ends of the crystallized rod is measured, and 

 therefore the derived value of the thermoelectric power might 

 differ from that of Perrons. Differences, arising from im- 

 purities in the bismuth rod, are also possible. 



To apply this method to the measurement of the Peltier 

 coefficient for any two metals, the copper leads n, o would be 

 replaced by one metal and the bismuth rod e by the other 

 metal. To dissipate the Joule effect and to reduce the effective 

 resistance of each junction, it is preferable to take each metal 

 in the form of a thin strip. Each strip should be approxi- 

 mately dimensioned so that the total flow of heat per unit 

 temperature-gradient is the same as for the bismuth rod. In 

 this case the resistance of the junction would be much less than 

 that of the copper-bismuth junction in this apparatus. An 

 additional lead for the current through each heating coil 

 must be soldered to each copper cylinder. The same sensi- 

 bility can be attained, in the case of small Peltier effects, by 

 increasing the number of thermo-junctions attached to each 

 cylinder, and also by reducing the size of the copper 

 cylinders. This modified apparatus would be suitable in the 

 case where only a short length of one of the metals was 

 available. The Peltier effect could also be measured at 

 various temperatures by varying the temperature of the 

 enclosure. 



An attempt was made to measure the thermoelectric power 

 of the copper-bismuth junction by passing a current through 

 one of the coils and measuring the E.M.F. of the junctions. 

 The temperature difference thus produced between the 

 junctions was small and could not be measured with sufficient 

 accuracy owing to a defective attachment of the junctions/, g. 

 One junction of the two wires projected about 1 mm. from 

 the cylinder , and although the other appeared to be satis- 

 factory the results showed that this was slightly defective. 

 Owing to the brittleness of the bismuth rod, these defects 

 were not remedied. The following results indicate the 

 existence of these defects. 



dE 

 At 22°-5 C. -^ =64-8 x lO" 6 volt per degree, 



with faulty junction at the higher temperature. 



