REQUIRED TO PRODUCE DISCHARGE THROUGH VARIOUS GASES. ; i 



For this purpose stopcocks d and e are opened, and b is set so as to open communi- 

 cation between a and c. The mercury reservoir (not shown) attached to the india- 

 rublxjr hose q is raised. Mercury rises in a, driving out the air through c, and on 

 lowering the mercury, fresh air bubbles in at r, just as in the ordinary Toepler pump. 

 When the apparatus has been exhausted, stopcocks d and e are shut, b is turned so 

 as to open communication between a and r, and gas is admitted through p to any 

 desired pressure. The pressure can then be varied at pleasure by raising or lowering 

 the mercury in a. The mercury reservoir attached to the hose q can be fixed at any 

 ^iven height so as to make the mercury stand at a corresponding height in a. The 

 object of shutting off the part between the stopcocks d and e is to make the volume 

 in connection with a as small as ]>ossible, thus making it possible to obtain a greater 

 range of pressure without admitting or removing gas permanently. Another 

 advantage of this contrivance was that, when helium was to be used, less of it would 

 be required. The volume required to fill the apparatus to atmospheric pressure was 

 only about 40 cc., when a was entirely filled with mercury. 



We now come to the arrangements for producing and measuring the striking 

 potential. A large Wimshurst machine was used. It was driven at a constant 

 speed by means of an electric motor. The potential difference between its terminals 

 was measured by means of one of Lord KELVIN'S multicellular electrostatic voltmeters. 

 This instrument has been checked by Mr. CAPSTICK. He compared it with a quadrant 

 electrometer standardised by Clark cells, and found it practically correct. I have 

 also verified it to some extent myself, by observing the additional deflection produced 

 by 50 volts, this latter voltage being determined by a Weston voltmeter, believed to 

 be trustworthy. No measurable discrepancy was detected. 



Some previous experimenters have used a Wimshurst machine in direct connection 

 with the spark terminal, and have turned it till the spark passed. This method 

 succeeds fairly well when large spark lengths and, consequently, large spark 

 potentials are to be measured. But when the spark potential is small it is scarcely 

 possible to raise the potential slowly enough to prevent the needle of the instrument 

 swinging violently ; and if this occurs, it is ot course impossible to make any accurate 

 observations. PEACE (loc. cit.) failed to get consistent measurements with the 

 Wimshurst machine. He was driven to the use of a battery of storage cells. The 

 use of these, however, very much adds to the trouble of the investigation, and I was 

 able to avoid it in the following manner. The machine was shunted by means of a 

 fluid high resistance column of variable length. When running at a constant speed 

 the difference of potential at the terminals of the resistance column was constant, 

 and depended of course on the length of the resistance column in use. By gradually 

 increasing the amount of this resistance the potential difference could be raised to any 

 desired value and adjusted with the utmost nicety. 



In order further to improve the electrical steadiness of the arrangement a large 



