94 A. L. Day and i?. B. Sosman — 



ing a study of the correction factors, led the authors to believe 

 that the upper end of the existing gas scale (melting point of 

 pure copper), which has been vacillating in a somewhat irregu- 

 lar way in various hands for three-quarters of a century, had 

 been finally confined to the limits ± 0*5°, or within 1°. 

 Although this ideal had been affectionately cherished for a 

 good many years, its triumph has been shortlived. The present 

 investigation has discovered a source of error which appears to 

 have passsed unnoticed before, which operates to raise the 

 temperature scale at the copper point about 1-4°. This kind 

 of history has repeated itself with remarkable persistence all 

 through the record of high temperature research, and may, of 

 course, do so again, but the limits of uncertainty are continu- 

 ally becoming narrower, and it appears to the authors unlikely 

 that further investigation will again reveal errors aggregat- 

 ingl°. ^ . 



On the other hand, the detailed study of temperature distri- 

 bution about the bulb (page 102) in which the present error 

 was discovered, cannot but convince an experienced observer 

 that the limit of refinement in an electrically heated air bath 

 has been practically attained, and that higher accuracy in gas 

 thermometry must be sought in a liquid bath which can be 

 stirred. 



Since the publication of the Reichsanstalt scale* - in 1900, it 

 has remained the standard for all temperature measurements 

 between 400° and 1100°. Its limit of accuracy as an absolute 

 scale was estimated to.be about 3° at 1000°. 



The work of Day and Clement was mainly directed to the 

 following essential features of the problem of absolute meas- 

 urement with a constant volume gas thermometer: (1) An 

 absolutely gas-tight bulb of definite volume ; (2) uniform dis- 

 tribution of temperature over the bulb surface during the 

 measurements ; (3) the reduction of the error due to the 

 unheated capillary tube connecting the bulb with the mano- 

 meter ; (4) a more accurate determination of the expansion 

 coefficient of the bulb itself. 



The results accomplished by them in these directions may 

 be outlined seriatim as follows : (1) The bulb chosen (90 parts 

 platinum, 10 parts iridium) is quite rugged enough for meas- 

 urements as high as 1500°, and no difficulty was experi- 

 enced in maintaining a nitrogen atmosphere in it without loss 

 by diffusion or leakage. At high temperatures the material 

 becomes considerably softer, but with the help of a gas-tight 

 furnace in which nitrogen could be maintained at the same 

 pressure outside the bulb as within, neither diffusion through 

 the bulb walls nor mechanical strain was encountered. Varia- 

 tions in the zero point of the thermometer, which have been 



*L. Holbornand A. L. Day, Ann. d. Phys. (4), ii, 505, 1900 ; this Journal, 

 (4), x, 171, 1900. 



