
TO A THERMO-ELECTRIC DIAGRAM. 129 
employ two thermo-electric circuits, whose hot and whose cold junctions are 
immersed in the same vessels; and to plot the curve whose abscisse and ordi- 
nates are simultaneous readings of the electro-motive forces in the two circuits. 
In every case I have tried, the curve thus obtained is almost accurately a para- 
bola, most of the few deviations yet observed being in the case of silver and other 
metals at temperatures not very much below their melting points—under cir- 
cumstances, in fact, in which we should naturally expect that the law would no 
longer hold. There are, also, cases in which the whole electro-motive force is 
so small, even for very large differences of temperature, that very much more 
delicate apparatus would be required for their proper investigation. And there 
are cases in which the neutral point is so far off that for moderate ranges of 
temperature the curves obtained are sensibly straight limes. J imtend to examine 
these cases with care—the former by using more delicate galvanometers, the 
latter by employing metals which are practically infusible. The difficulty of 
obtaining wires of such metals has been the chief one I have had to face. 
“Tf we assume the experimental curve to be a parabola, then it is easily 
seen (Proc. May 29, 1871) that in each circuit the electro-motive force must be 
a parabolic function of some function of the absolute temperatures of the junc- 
tions. And, as in the iron-silver, iron-zinc, iron-copper, iron-cadmium, &c., 
circuits, this function has been proved to be simply the absolute temperature 
itself (at least, within the range of mercury thermometers), it is probable that 
such is the general law, at least for ranges of temperature short of those which 
materially alter the molecular structure of the metals employed. 
“The second method consisted in employing two pairs of circuits, all four 
hot junctions being in the same heated substance, and all four cold junctions 
kept at a common temperature. The members of each pair acted on a differ- 
ential galvanometer (as explained in Proc. Dec. 19, 1870) in such a way as to 
eliminate the term containing the square of the absolute temperature. In this 
case the readings of the galvanometers should be simply proportional to one 
another, and likewise to the differences of absolute temperature of the junctions. 
The method is exact in theory, but by no means easy in practice, especially 
with the very limited number of metals capable of resisting a high temperature 
which I could manage to obtain. That a very exact and useful thermometric 
arrangement can be made on this principle admits of no doubt, when we 
examine the results of the experiments. 
“The third method consisted in assuming the parabolic law, and the follow- 
ing consequence of it which follows directly by the use of THomson’s general 
formule. These may easily be reproduced as follows :—Suppose a sliding ring 
or clip to be passed round the wires, so as to press together points of the wires 
which are at the same temperature, ¢. Its effects are known by experiment to 
be nil, whatever be its material. Let it be slid along so that the temperature 
VOL. XXVII. PART I, 21 

