A. L. Day and J. K. Clement — Gas Thermometer. 463 



reasonably probable that the gas scale can be extended to 1500° 

 or 1600° with a proportionately small error in its absolute value. 

 The immediate future of the present investigation will be to 

 undertake this extension. 



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 interpretation. The metals 

 which have commonly been used for the purpose are not obtain- 

 able commercially in sufficiently uniform purity to guarantee 

 an accuracy within 1° at the higher temperatures. This is too 

 large an error for the interpretation of the gas thermometer 

 scale in its present refinement. No effort has been made to 

 prepare metals in our own laboratory of exceptional purity for 

 the reason that such metals would not be available for general 

 use and would therefore be of little service. 



We have accordingly adopted metals which are carried per- 

 manently in stock by dealers (whose names are given in con- 

 nection with each) from whom the same metal in a nominal 

 quality equal to that which we used can readily be obtained. 

 We have analyzed these with extreme care to show the exact 

 content of the sample supplied to us. We have duplicated the 

 purchases ourselves, and have found no errors greater than 1° 

 in their melting point determinations. 



Another difficulty arises from the fact that the melting points 

 of the purest metals available for use as constants in reproduc- 

 ing a high temperature scale (zinc, silver, gold and copper) are 

 distributed in such a way that, although they may be located 

 upon the gas thermometer scale with a probable error not 

 greater than 0*5°, the calculation of a similar curve passing 

 through these points does not suffice to reproduce the scale with 

 this accuracy. In the region midway between zinc(418'9°) and 

 silver (958*5°) the error of interpolation may amount to 2° even 

 with metals of exceptional purity. Extrapolation is even more 

 uncertain. This can be avoided by locating intermediate 

 points which are equally trustworthy, if such can be found. 

 We have not been fortunate enough to find points which fulfil 

 these conditions satisfactorily but hope that we may yet be 

 able to do so. 



As the matter now stands therefore we have succeeded in 

 perfecting the constant volume gas thermometer until the 

 aggregate error affecting the measurements between 300° 

 and 1150° appears not to be greater than 0*5°, but we are 

 not yet able to offer adequate assurance that our scale can be 

 reproduced by another with this accuracy. This matter will re- 

 ceive further attention in a later paper. 



Geophysical Laboratory, Carnegie Institution 

 of Washington, August, 1908. 



