lUniVIMAX. — THERMODYNAMIC PUOPP:UTIKS OK WATF.U. 821 



because the pressure changes hrouj^ht ahont by changes of temperature 

 (luring the determinations of the thermal dilatation are comparatively 

 slight, so that any change of the pressure coefficient of the coil brought 

 about by the change of temperature appears in the result greatly 

 magnified. Thus for the sake of example, we will suppose that a 

 change of temperature of 20° produces a change of pressure of 400 

 kgm. at 10,000 kgm. total pressure. This figure is a fair average of 

 the results to be met with in practice. If now the pressure coefficient 

 of the coil is changed by 1% by this same rise of temperature, the 

 pressure will therel)y appear to have risen 500 kgm. instead of the 

 actual 400, introducing an error of 25% for a change in the constant 

 of the coil of only 1%. In addition to the effect of the temperature 

 coefficient of the coil, there is an effect due to the change of the zero 

 of the coil with temperature, but this change can be determined by 

 observations of the temperature coefficient of the coil at atmospheric 

 pressure and is easy to measure with the requisite accurac3\ 



The change in the pressure coefficient of the coil with temperature 

 is more difficult to determine with the desired accuracy. It would 

 not be possible to determine this by a direct calibration against the 

 absolute gauge with which the mean value of the coefficient has been 

 determined, for the reason that the absolute gauge itself is not accu- 

 rate to l)etter than 1/10%, and this would still leave a possible error in 

 the thermal dilatation of 2.5%. To affect the desired calibration, 

 some standard of pressure must be used which can be relied on to 

 remain absolutely constant. Such a standard pressure is evidently 

 afforded by the transition point of the liquid to the solid form of any 

 convenient substance at some fixed temperature. In previous work 

 the transition points of both water and mercury have been determined 

 at various temperatures with an accuracy in the absolute pressure of 

 1 / 10%. To make the calibration it is only necessary to keep the pres- 

 sure constant automatically at this known value by placing in com- 

 munication with the chamber in which is the manganin coil to be 

 calibrated another chamber in which are the liquid and solid forms 

 of the substance whose transition temperature and pressure are 

 known. This second chamber is to be kept at constant temperature 

 accurately enough so that slight changes in this temperature will not 

 produce changes of more than the allowed amount in the transition 

 pressure. For this purpose the most convenient fixed temperature 

 seems to be that of melting ice at atmospheric pressure, and the most 

 convenient substance to use mercury, because of the sharpness of the 

 freezing, and the ease with which it can be obtained pure. 



