468 W. P. White — Melting Point Determination. 



immersion. This method applied to 2*5 gram silicate charges 

 indicates that there was no perceptible error of this sort with 

 the element immersed 6 to 8 mm , thus confirming the results 

 obtained by varying the size of wire. 



VII. The use of bare thermoelements. — From the fore- 

 going it is clear that in several ways a very great advantage 

 results if the thermoelement can be inserted directly in the 

 charge. But an error may then be introduced, due to the 

 electrical conductivity of the charge itself, which can not be 

 altogether neglected. At high temperatures, practically every- 

 thing conducts electricity to some extent. 



The slightly conducting charge acts like a battery of very 

 high resistance connected in shunt across the junction.* The 

 effect on the galvanometer reading of such a battery is easily 



T 



shown to be practically equal to E s — , where E s is an E.M.F., 



T, the (very low) resistance of the shunted portion of the 

 thermoelement, S, the (very high) resistance of the shunt. 

 It follows, first, that the leakage error increases with the depth 

 of immersion of the thermoelement and the fineness of its 

 wires, and second, that by dipping into a charge portions of a 

 thermoelement far from the junction the effect can be mag- 

 nified, and so a very delicate test made for it. Indeed, for 

 this test there need be no junction at all ; if there is one, 

 either it can be kept at room temperature, or its E.M.F. can 

 be measured by making it also the junction of two other wires 

 which do not clip in the charge. In these ways it has been 

 shown that the error due to this effect, under the conditions 

 of our work, is not over one or two-tenths of a degree, either in 

 sodium chloride near 801° or in diopside near 1400°. It 

 was also shown, in the case of sodium chloride melted in a 

 crucible over a Bunsen burner, that the E.M.F. is thermo- 

 electric and is directed (in this case) from the electrolyte to the 

 colder of the two wires. Hence the error increases .with the 

 temperature differences present in the charge. The convection 

 currents of a large charge frequently reveal their presence 

 through the slight unsteadiness of the galvanometer, and in one 

 case the leakage disturbance from melt conductivity w T as 

 increased a hundred-fold by bringing a charge too near the 

 furnace top.f 



The conditions making for small leakage error are therefore : 

 Uniformity of external temperature, smallness of charge, and 

 thickness and shortness of the immersed portion of the thermo- 



* The effect would of course be the same if it were connected across any 

 other portion of the circuit. 



f It is therefore clear that the results just quoted as to the very small error 

 from conductivity of the melt should be applied with great caution in 

 extreme cases. 



