High Temperatures. 205 



The platinum wire became very brittle in the vicinity of the 

 junction, broke in several places and was melted together in 

 the oxy-hydrogen flame. Afterward when used for measure- 

 ment at about 1000° in the long oven (unchanged temperature 

 distribution) its reading was some 90 microvolts too low, but 

 for the melting point of copper (different distribution) it 

 gave only 9840 MV, i. e. some 400 MV too low. Sixteen 

 centimeters were at once cut off at the hot junction, first from 

 the platinum and afterward from the platiu-rhodium wire, 

 when the melting temperature of copper was given at 10,290 

 and 10,150 MY respectively. Further shortening did not 

 serve to change these figures and the pieces cut out were again 

 inserted. In this condition the element was glowed as in the 

 previous case for some hours at white heat, when the melting 

 point of copper was read at 10,216 MV — almost exactly the 

 normal value. 



The glowing also restores the bright surface to wires 

 (platinum particularly) which have become dull after exposure 

 to these gases. 



It is probable that the standard element which represents 

 the earlier temperature scale of Holborn and Wien and which 

 has been used many times for comparisons up to 1600° has 

 undergone similar changes. 



Its indications at 500° are now 10°, and at 1000°, 18° too 

 high, while the new determinations of the melting point of 

 gold differ only 8° from the old. This melting point in par- 

 ticular was well determined at that time, the greatest variations 

 from the mean in twenty-five determinations under varied con- 

 ditions being ± 4° while the corresponding variations for 

 silver and copper amounted to =b 10°. 



At the lower temperatures no fixed points were then deter- 

 mined so that no further comparison is possible. 



"We draw the conclusion from this that the temperature scale 

 once established can be maintained with certainty only with 

 the help of fixed temperatures such as the melting points given 

 above. It is further advisable to divide a set of thermo-ele- 

 ments into two groups, one to be used only in porcelain tubes 

 up to say 1200° and the other under other conditions or in 

 higher temperatures. 



For the latter, up to the present at least, the same accuracy 

 is neither to be obtained nor expected until observations with 

 the gas thermometer have been made and the technical 

 resources much extended. 



Especial attention ought again to be called to the fact that 

 in the measurement of the fall in temperature in an extended 

 space such as is necessary for the calculation of its mean tem- 

 perature, the utmost care should be taken that the elements be 



