120 Mr. J. Buchanan on the 



Three rods of carbon of 1*2 centim. diameter, such as are 

 manufactured for use in arc-lamps, were taken, One of these 

 was thinned down into a conical form at the end destined to 

 go into the hot bath. This end, and that of another rod 

 intended to go into the cold bath, were electroplated with a 

 thin shell of copper. To this copper shell were soldered the 

 platinum wires, whose behaviour with the carbon I wished 

 to examine. These two rods were then placed vertically and 

 clamped in position with coppered ends dipping into their 

 respective baths. The upper ends of these carbon rods were 

 cut off square and the third rod laid horizontally, so as to 

 form with the vertical rods the third side of a vertical rect- 

 angle. 



To ensure good contact one end of this third rod was bound 

 by a number of turns of clean iron wire to the upper end of 

 the rod whose lower end dipped into the cold bath. The 

 other end of this third rod rested on the top of the rod the 

 lower end of which dipped into the hot bath. The surfaces 

 of contact were here scraped flat and kept pressed together 

 by a weight of 3 or 4 pounds. 



The platinum wires that formed the other element of the 

 thermo-couple were of ordinary commercial material, as was 

 mentioned above. 



The galvanometer used in the secondary circuit is a very 

 delicate mirror instrument of about \ ohm resistance. 



The cell used in the primary circuit as a standard was a 

 Daniell. The graduated wire (BC in fig. 1) was of German 

 silver. 



The baths, both hot and cold, were oil-baths. 



The ends of the platinum wires of the thermo-couple that 

 were exterior to the baths were joined by binding-screws to 

 copper wires. These joints were insulated from one another 

 by several layers of calico, and then tightly wrapped up to- 

 gether. Thus in the secondary circuit the only exposed con- 

 tacts of dissimilar conductors were those of the copper wires 

 joined up with the brass binding-screws of the key, of the 

 galvanometer, and of the sliding piece on the graduated wire. 

 During the experiment, care was taken to screen these exposed 

 contacts as much as possible from external radiation. More- 

 over, all the apparatus was set up the night before, so that on 

 the days of the experiments no metal joints whatever in the 

 secondary circuit were touched by hand. 



Some preliminary experiments I had made showed the 

 necessity for adopting the arrangements just described. The 

 results of these preliminary trials exhibited very curious 

 irregularities, not altogether got rid of even by those latest 



