White — Specific Heats of Silicates and Platinum. 335 



temperature throughout the working chamber of the furnace, 

 has thus far remained the chief difficulty in the present work. 

 The working chamber of an electric resistance furnace is 

 cylindrical. The heat is produced in the sides and escapes 

 more or less through the ends, which are thus at a considerably 

 lower temperature. Any body within will necessarily have an 

 uneven temperature whose exact distribution depends on the 

 body, as well as upon the furnace. The correction of this 

 uneven temperature is somewhat troublesome, but will of 

 course be essential as high temperature measurements become 

 more exact.* Indeed, the comparative indifference to the 

 question of furnace temperature distribution shown by the 

 authors of much work already done is rather surprising. In 

 the present case the furnace first used, whose working chamber 

 was 16 cm high by 6 cm wide, showed by actual measurement 

 differences of about 20° between different parts of the charge. 

 The corresponding error, that is, the difference between the 

 average temperature of the charge and that of the ther- 

 moelement which enters and measures it, would be under 

 10°, or one per cent, at 1000°. It very soon appeared that 

 the precision attained in the other measurements was such as 

 to justify an attempt to diminish this furnace irregularity 

 and a new furnace was constructed 20 cm by 4'5 cm in internal 

 dimensions, in which partitions above and below the charge 

 partially shut off the cooling effect of the ends. These 

 partitions were of fire-clay 5 n,m thick covered with platinum 

 on the side toward the charge to give a reflecting surface. 

 Thermoelectric measurements outside the crucible contain- 

 ing the charge now showed a maximum difference of 15° 

 at 500° and of 6° at 1500° with a maximum systematic error 

 at 500° of perhaps 0*5 per cent between the average tem- 

 perature of the charge and that of the thermoelement used to 

 measure it. The other sources of systematic error are small, 

 and the greatest accidental errors in the final results ( with two 

 or three exceptions, which cannot be definitely accounted for) 

 ranged from 0*5 per cent at the extreme to 0*2 per cent at the 

 intermediate temperatures. These accidental errors are less 

 for platinum and for a charge of glass which was melted into 

 a single cake than for the other substances, which consisted of 

 separate lumps. From this it would appear that the accidental 

 errors are mainly connected with failure to hold the furnace 

 temperature steady for a sufficient time, and are not to be taken 

 as indicating a large systematic error. The present systematic 

 error then may fairly be inferred from the temperature dis- 



*See, e. g., Some New Measurements with the Gas Thermometer, by 

 Arthur L. Day and J. K. Clement, this Journal, xxvi, 412, 1908. Several 

 other schemes for obtaining a very uniform high temperature are now 

 under way. 



