White — Melting Point Methods at High Temperatures. 481 



sufficient to cut off 2 or 3 cm , merely enough to bring fresh wire 

 down into the gradient region, which lies near the top of the 

 furnace. The time to do this is evidently when the contamina- 

 tion has become so great that the correction for it has more 

 than the allowable uncertainty. Practically, a lowering of 

 one degree in the reading is about the limit allowed in most 

 of the work of this laboratory. 



The gas-thermometer scale is transferred to the thermoele- 

 ment by means of metal melting-points, which are therefore 

 the practical temperature standards in our work. A thorough 

 treatment of this question will accompany the forthcoming 

 paper on the gas-thermometer. Here it is sufficient to say that 

 it is useless to calibrate contaminated elements directly by the 

 metal melting-point method, since the introduction of a large 

 crucible of metal into our furnaces seriously alters the tem- 

 perature gradients which previously prevailed. Extrapolation 

 is then particularly treacherous. But with wires of the homo- 

 geneity now attained in the Heraeus elements an element may 

 be relied upon to return to its original reading when properly 

 amputated. It is hardly necessary to say that the average of 

 several calibrations is more to be relied upon than a single one, 

 and this means that in case an element has been calibrated 

 several times with concordant results, and is then cut off to 

 remove contamination, for the new element thus formed the 

 average of the preceding calibrations is more to be relied 

 upon than a single new one. The same is true of a calibration 

 established by comparison of one element with another whose 

 calibration is known to be good. The opportunity for error in 

 the comparison, if properly conducted, is less than that in a 

 curve from a single set of melting-points made with equal care. 



The calibration of contaminated elements, which requires a 

 particular temperature gradient, can be performed either with 

 charges of silicates or other salts, which will go in almost any 

 furnace, or by direct comparison with standards. The stand- 

 ards can also be used to calibrate new and perfect elements. 

 Such comparisons are also far less laborious than melting-point 

 determinations. 



For very high accuracy in direct comparison it is absolutely 

 necessary that the two junctions be at the same temperature. 

 This is secured at high temperatures, either by thrusting the 

 two elements, without electrical contact, side by side into one 

 of the 8 mm porcelain tubes and maintaining a uniform hori- 

 zontal temperature distribution around this tube (by a bath of 

 silver, a block of iron, or simply by putting the tube carefully 

 in the center of the furnace and surrounding it with a ballast 

 tube 4 cm in diameter) ; or the two elements are first inserted 

 and read, then removed, transposed and inserted again, so that 



