132 DRS. J. A. HARKER AND P. CHAPPUIS ON A 



APPENDIX II. [Added December 1, 1899.] 



We think we are justified in adding, in the form of an appendix, some further 

 considerations on the question of the sulphur boiling-point, the results of which we 

 obtained since the date of handing in the paper. On p. 99 of the text are given the 

 observations of REGNAULT on the variation of the boiling-point of sulphur with 

 pressure near 760 millims. The formula used by REGNAULT himself to express the 

 results of his observations over the whole range was of a logarithmic kind, and gave 

 for the pressure 760 millims. the value 448 0- 38. If, however, we disregard the 

 extreme portions of the range and find a formula to represent only those observations 

 near the normal pressure, we find for this point a value nearly a degree lower. 



Taking the four observations quoted in the text, and representing them by a 



formula 



T = a + bp + off, 



of which the constants a, b, and c are determined by least squares, we find for 

 p = 760 the value 447 '51, with residuals very much smaller than those given by 

 the logarithmic formula. We are aware that to represent four observations by a 

 formula with three constants is not giving very much latitude for probable errors, 

 but we think, nevertheless, that 447'5 gives much more nearly the true result to be 

 deduced from REGNAULT'S experiments than his own much higher figure. Accepting 

 this method of treating his observations, we further find that instead of the value 

 for dt/dp 0'082 per millim. as given by the logarithmic formula, we get 0'088, a 

 value very appreciably higher. 



If our determinations of pt, had all been made at 760 millims. pressure, or if this 

 had been the mean pressure of each different series, the value to be taken for dt/dp 

 would have been of no great consequence, but as in each case the mean pressure fell 

 appreciably below this, we thought it desirable to see how much the assumption of 

 the higher value might influence the results of our experiments. 



We gave in the text the results of some calculations on the series of sulphur 

 points taken with the Kew platinum thermometers K.I and K.3, made with the 

 object of arriving at an independent value for dt/dp. Dr. CHREE has recently com- 

 pleted for publication an investigation into the behaviour of the Kew platinum 

 thermometers, and their permanence over a considerable period, and finds that, when 

 one or two sources of uncertainty are eliminated, the values we gave for dt/dp for 

 K.I and K.3 are both somewhat too small. He has courteously permitted us to 

 state that the most probable value for this number deducible from the different series 

 of determinations of the sulphur point, which he has worked ut>, is much more nearly 



