Modification of the Ph«i E.i'penment. 519 



certain about the Josired corrections, outside the range which 

 they covered. The gas scale is now in use over so vastly 

 wider an interval than forty years ago, that it is very desirable 

 indeed that the plug experiment should be repeated over the 

 original range and extended as tar as possible towards both 

 low and high temperatures. At the same time, the greatest 

 attainable accuracy should be sought. The weak point of 

 the original method seems to me to lie in the fact that the 

 passage through the plug was not strictly adiabatic. By a 

 slight modification, suggested by the ease and accuracy with 

 which electrical measurements may now be carried out, this 

 defect of the original method may be avoided. 



Instead of making the passage through the plug adiabatic, 

 let us make it isothermal. Instead of the fall of temperature, 

 we then have to measure the energy which must be supplied 

 to keep the temperature from falling. Let a platinum 

 heating-coil be embedded in the plug, and by adjusting the 

 current in this coil let us establish a steady state such that 

 there is no fall of temperature as the gas passes through the 

 plug.^ 



This method has several advantages. It reduces the 

 measurement of temperature to a null method — we place two 

 platinum thermometer coils in the stream of gas, one before 

 and one after the passage through the plug ; we balance 

 these as tw^o arms of a Wheatstone's brido-e, and we reo-ulate 

 the heating current so as to preserve the balance. Beside 

 thus avoidino- the correction for thermal leakao-e into the 

 plug, we furthermore avoid all differences of temperature 

 which may, in the original form of the experiment, cause a 

 thermal leakage against which it is difficult to guard. 



The quantities to be measured, in addition to the fall of 

 pressure, the rate of flow of gas, and the uniform temperature 

 of the whole ap]3aratus, are now merely the heating current 

 needed for the steady state and the fall of potential in the 

 heating coil. The originally difficult part of the measurement 

 is thus performed electrically, and such measurements are 

 easy and accurate if our standards are really correct. As 

 regards the accuracy of the absolute values obtainable, we 

 substitute two electrical units for a specific heat and Joule's 

 equivalent ; and this is not, I think, to be regarded as a 

 disadvantage, in the present state of our knowledge. 



As a further precaution, we may adopt the guard-ring 

 principle and use two independent concentric plugs ; if, in 

 the steady state, both plugs are giving similar results, we are 

 reasonably sure that the inner plug is not subject to any 

 thermal leakage from outside. It is easy to distribute 



