TEMPERATURE SCALE. 



23 



W. F. Hillebrand of the Geological Survey, each showed about one per 

 cent of the sulphide still present and traces of other impurities. The 

 melting temperatures of these varied under different conditions as 

 much as 15 , and were totally unsuited to this work. 



Inasmuch as the melting points of these metals were determined 

 with thermo-elements which Professor Holborn had just calibrated 

 with the metals in use at the Reichsanstalt for this purpose, a compari- 

 son of the values obtained will show the accuracy with which one may 

 reproduce the Reichsanstalt scale entirely from local sources : 



* A single determination with one element; all others are mean values with two 

 or more elements. 



For the method of extrapolation of the scale and further informa- 

 tion regarding the use and accuracy of thermo-elements at these 

 temperatures, reference is made to the papers of Holborn and Day 

 already cited. 



For everyday use, four more elements were prepared and calibrated 

 in the same way. Of these, two are of the usual form (fig. 1 ) and two 

 are of a new design which has proved very effective in the determi- 

 nation of the melting points of non-metallic substances. It will be 

 seen from the diagram of the insulated element that the hot junction 

 is protected from the melting charge by a casing of platin-iridium 

 (0.5 mm. thick) and by a protecting tube of refractory Berlin (Mar- 

 quardt) porcelain (1.5 mm. thick) . Very early in our experiments upon 

 the mineral silicates we became aware that the conductivity of these 

 materials for heat would be much poorer than in similar charges of 

 metal. Furthermore, the charge of mineral which the furnaces could 

 carry was only one-fourth to one-third as great as the metal charges 

 used in the calibrations, because of the great difference in specific 

 gravity and the limited space which could be heated to a fairly uni- 

 form temperature. For these reasons the changes of state would be 

 less sharply marked upon the heating and cooling curves than metal 

 melting points, and it was feared that the readings of the protected 





