756 TRUEBLOOD. 



0°.02 C; case (iv), by 0°.005 C. While the reL^tive magnitudes of 

 these results are hardly such as would be expected, the fact that a 

 depression of the apparent temperature was obtained in each case 

 indicates the existence of a real error from the cause under investiga- 

 tion and that the porcelain joints are of some value in diminishing its 

 magnitude. The experiments were not refined, as will be inferred 

 from the method used. They were carried no further than was 

 necessary to secure the information desired. They can hardly be 

 regarded as conclusive regarding the size of the error to be expected 

 if no thermal insulation is used, for it is difficult to calculate even 

 approximately the relative insulating effects, as regards the bulb, of a 

 porcelain joint and of the long, thin stem. Without joints, or even 

 with them, the temperature gradient in the stem must be concentrated 

 near its upper end, and the insertion of an insulating joint at this point 

 may be expected to have a much larger eifect on the effective thermal 

 conductance of the combination than would be calculated on the 

 assumption of a uniform temperature gradient longitudinally in the 

 stem and radially in the joint. However, it is at least reasonable to 

 infer from the above experiments that the error in the measured temper- 

 ature drop through the plug due to the cause we are considering would be 

 of insignificant magnitude, whether insulating joints are used or not. 

 For such an error would be that occasioned by conduction down the 

 stem of the low side thermometer, with a temperature difference 

 between the head and the bulb not exceeding 5°, while in the above 

 experiments this temperature difference was at least 140°, since it was 

 observed that naphthalin would freeze on the head of the thermometer. 

 On the basis of the above data, the omission of insulating joints could 

 then affect the measured difference of temperature by not more than 

 0.02 X ifo = 0°.0007, and if we assume that the effective thermal 

 resistance of the stem is even as much as 10 times that of the insulating 

 joints, conduction down the stem of a thermometer without joints 

 would produce an error in the temperature difference of only 0°.007, 

 which is not greater than arises accidentally from uncontrollable 

 causes. 



The thermometers used in all the work on the Joule-Thomson effect 

 described in this paper have the upper porcelain joint, but not the 

 lower one. It is difficult to get an absolutely steam-tight joint, and 

 even a small leak at the bulb would obviously be disastrous, while 

 such a leak at the upper joint would be of much less moment. Besides, 

 i f porcelain joints are to be used at all, the upper one is probably of 

 more value than the lower. 



