100 REPORTS ON INVESTIGATIONS AND PROJECTS. 



In 1904 Professor Holborn, of the Reichsanstalt, increased the range of 

 this scale as far as 1600° C, the probable error of the new portion being 10°. 

 Simultaneously with this effort, work was begun at the Geophysical Labora- 

 tory in Washington with a view to increasing the accuracy of the scale, first 

 over the existing range (to 1150°), and later as much beyond this point as it 

 should prove practicable to go. 



The errors which have heretofore been present in measurements with the 

 nitrogen thermometer have been reduced by the present investigation to about 

 one-fourth their former magnitude, and the certainty of their evaluation is 

 at least proportionately increased. 



The following points have received particular attention : 



(i) To provide a uniform temperature about the bulb by a suitable ar- 

 rangement of electric-heating coils and diaphragms. 



(2) To inclose the furnace in a gas-tight bomb in which the pressure out- 

 side the bulb can be maintained equal to that within for all temperatures. 

 This offers three distinct advantages: (a) it provides against the deforma- 

 tion of the bulb through differences of pressure within and without in the 

 region of highest temperatures, where the bulb material becomes softer; (b) 

 by using the same gas within and without, there is no tendency for it to 

 diffuse through the bulb wall; (c) it enables the initial pressure to be varied 

 within considerable limits, thereby increasing both the scope and sensitive- 

 ness of the manometer. The sensitiveness in our instrument with this 

 arrangement was about three times that of the Reichsanstalt. 



(3) The expansion of the bulb material was determined with great care 

 and is probably accurate within 0.5 per cent. 



(4) The unheated space between the bulb and manometer has been re- 

 duced until the total correction in this hitherto uncertain region amounts to 

 less than 4° at 1100°. An error of 5 per cent in the determination of its 

 volume or temperature distribution is, therefore, practically negligible. 



It is probable that these changes serve to reduce the uncertainty hitherto 

 prevailing in the correction factors which require to be applied to the gas 

 thermometer in the region of 1100° to less than one-tenth of its former mag- 

 nitude. Furthermore, these improvements are equally applicable throughout 

 the region above 1100° as far as the present measurements have extended 

 (to 1550°). 



The chief source of present uncertainty is the temperature distribution 

 over the surface of the bulb in an air-bath. It would be possible to eliminate 

 this error in the lower portion of the scale by substituting a liquid bath which 

 could be stirred. In fact, this was done for temperatures below 500° in the 

 earlier work of Holborn and Day, but has not so far been tried in the present 

 investigation, because of the relatively secondary importance of the lower 

 temperatures to the ultimate purpose of the investigation (the study of sili- 

 cates). For the higher temperatures no satisfactory liquid bath has been 

 found. 



The interpretation of these measurements in terms of the melting-points 

 of readily available substances encounters certain difficulties. The melting- 

 point of pure salts is not sufficiently sharp and is somewhat difficult of inter- 

 pretation. The metals which have commonly been used for the purpose, 

 with the possible exception of nickel and cobalt, are, however, obtainable in 

 sufficiently uniform purity to guarantee an accuracy within 1° at the higher 

 temperatures. 



