Nitrogen Thermometer from Zinc to Palladium. 95 



very persistent and inaccessible errors throughout the history 

 of gas thermometry, have therefore now become practically 

 negligible. This gas-tight furnace possessed the further advan- 

 tage that the initial pressure of the gas, and consequently the 

 sensitiveness of the instrument, could be varied within con- 

 siderable limits. A sensitiveness as great as l mm of the mano- 

 meter scale per degree was regularly employed. The iridium 

 alloy has the disadvantage that platinum thermoelements, 

 which are necessary for recording variations in the tempera- 

 ture over the surface of the bulb, and for transferring the gas 

 thermometer temperatures to standard melting points, become 

 contaminated in the presence of iridium at all temperatures 

 above 900° C. ; the higher the temperature and the longer the 

 time of exposure, the greater the degree of contamination. 



(2) It was sought to obtain a uniform temperature over the 

 surface of the bulb by winding the (pure platinum) furnace 

 coil on the inside of a refractory magnesia tube which contained 

 sufficient iron oxide and other impurities to be a fairly good 

 conductor of heat. The winding was somewhat closer at the 

 ends than in the middle. This was further supplemented by sec- 

 ondary coils of smaller wire extending a few centimeters into 

 the tube from each end. The current in the three coils could 

 be independently regulated with the help of thermoelements 

 attached to the bulb and giving its temperature at the middle 

 and upon each shoulder (positions 2, 4, and 6, fig. 1). When 

 these temperatures had been adjusted so that the differences 

 between them were smaller than 0*5°, it was assumed that the 

 temperature over the whole surface of the bulb was constant 

 within those limits. (For the oversight in this assumption, 

 see pages 99 and 102.) 



(3) The platinum capillary and connections between the 

 bulb and the manometer were much diminished in volume. 

 Compared with the total volume of the bulb (195'Y CC ) this con- 

 necting volume amounted to *0015 in their instrument, and 

 reduced the total correction for the "unheated space" to less 

 than 5° at 1100°, a correction factor not more than one-fourth 

 as large as the best previous attainment in this direction. The 

 uncertainty of the temperature distribution in the ''unheated 

 space " was perhaps 10 per cent, making the probable error 

 from this source about 0'50°. 



(4) A special bar 25 cm in length, made up from the same 

 alloy as the bulb, was provided with a scale and its length 

 measured in a special form of comparator at temperature 

 intervals of 50° up to 1000°. The expansion was found to be 

 10 6 /3=8-84 + 0-001312, with an error of about 0*5 percent. An 

 irregularity was detected both in the expansion and subsequent 

 contraction in the region below 300°, which appeared variable 

 with the rate of cooling or heating, and in character resembled 

 the hysteresis which appears in a bar which has been subjected 



