HALL. — THERMAL AND ELECTRICAL EFFECTS IN SOFT IRON. 31 



Tt will be observed that such a day's work includes six complete 

 cycles, or runs, through all the pairs of corresponding couples from MP 

 to K N and back, three runs with the main current in condition F, and 

 three with this current in condition B. If we let 



A = strength of main current in amperes, 



t w = temperature of boiling water at mid depth, 



E c = net e. m. f.. in volts, of the opposed cold couples, 



E m = " " middle " , 



E h = " " hot " , 



the sign + indicating that the couple on the warmer bar, the bar in 

 which the current runs from cold to hot, prevails, and vice versa, we can 

 express the results of the observations of October 22 as follows : 



It is evident that ohservations made without reversal of the main cur- 

 rent would be useless. This fact is probably due in part to the slight 

 misplacements of the junctions which have already been suggested and 

 in part to certain accidental causes which may vary somewhat with the 

 conditions of the experiment. 



To find an approximate value for the difference of temperature gra- 

 dient which yields any one of the E's given above, we have only to 

 divide this E by the "thermo-electric height" of copper and german- 

 silver at the mean temperature of the two opposed couples which have 

 this net E. Thus we get from the observations of October 22, if A 

 stands for difference of temperature gradient in degrees per centimeter, 

 the + sign indicating that the gradient is steeper in the warmer bar, 

 and vice versa, 



Ac A m A;, 



+ 0.0352 + 0.0072 — 0.0316 



It so happens that on this day the value found for A A was numeri- 

 cally less than the value found for A c ; but the opposite condition was 



