Practical Thermometric Standard. 545 



to 500° C, the temperatures being measured by means of a 



platinum wire extendino- along the axis of the tube. The b 



• • • i i 



term is very large for the soft glass because it is heated so 



near its softening point. It was verified, however, that the 



glass would stand a vacuum without change of volume at 



450° C. The expansion curve was not quite accurately a 



parabola, but this did not affect the results, as the value of b 



was calculated to tit the observations at 0°. 100°, and 445°, at 



which points the correction was actually required. 



The value obtained by Chappuis for the S. B P. depends on 

 two thermometers of Verre JDur and porcelain, the expansion 

 corrections of which are given in the next three lines. The 

 expansion formula for the hard glass was determined by ob- 

 serving the linear expansion of a long tube between 0° and 

 100°. This range suffices for the most accurate determination 

 of c, but is not very satisfactory for b, because the maximum 

 difference from linearity at 50°, upon which the measurement 

 of b depends, is more than 60 times less than the difference 

 due to b at 445°, the temperature to be determined by the 

 application of the correction. If, for instance, my value 

 were adopted for the coefficient b, Chappuis's result for the' 

 S. B.P. would have to be lowered to 444 o- 0, i. e. by no less 

 than 1 0, 26. It is probable that my glass was a harder 

 specimen, and it is also quite possible that my value of b may 

 be too small; but observing the close agreement in the values 

 of c for the two specimens, I feel convinced that Chappuis's 

 value of b for hard glass is too large. Chappuis employed 

 two methods for the expansion of porcelain : (1) the method 

 of the mercury weight-thermometer between 0° and 100° ; 

 (2) the Fizeau method, on a piece of tube one centimetre in 

 length, between 2° and 82°. The two methods gave the same 

 value for c. The result for b by the mercury method is 

 evidently much too large *, and the observations for b by this 

 method were rejected as being not sufficiently numerous or 

 concordant. The second method is a very delicate one, but 

 as the difference due to b over the range of the experiments 

 would amount to only a quarter of a wave-length of sodium 

 light, it would be difficult to make a very satisfactory deter- 

 mination. According to Bedford's formula, the difference 

 would be less than a tenth of a wave-length. Even admitting 

 that b could be satisfactorily determined under these condi- 

 tions, it by no means follows, with a substance like glass or 

 porcelain, that the value of b would be the same at 400° as 

 at 40°. The observations of Bedford, extending over a 

 range ten times as great (which would make the measurable 



* My own observations by this method in 1893 gave a similar result. 



