Short Spark-Discharges, 703 



(1) The minute irregularities of the surface might oppose 

 sharp points in spite of the final polishing with soft 

 rouge, and the pressing of the surfaces together 

 before taking the zero reading. 



{'2) A particle of dust of too high a resistance to appre- 

 ciably affect the resistance of the galvanometer 

 circuit might carry the high potential discharge. 



(3) A particle of dust not at first in contact with both 



surfaces might be made to bridge the gap by the 

 electrostatic attraction while testing. 



(4) Any vibration of the building or even sound-vibrations 



might decrease the distance momentarily. 



It was therefore thought best to occasionally disregard 

 certain values which were much too low to be consistent 

 with the rest of the series. All of the observations are re- 

 corded in the data sheets, and except for two values noted in 

 the table for fig. 9, all of the observations not given a weight 

 of zero are shown on the plots. The weight of zero was 

 generally given when the discharge was so premature that 

 the zero drift had not been obtained. 



The individual curves show plainly that there is no sudden 

 change in the potential gradient within the range covered 

 by them. In every case, the potential seems to be a linear 

 function of the distance. It is to be noted that the shortest 

 distance actually determined was 3 fifi, which corresponded 

 to an apparent movement of only one-tenth of a fringe. It 

 was found also that a shorter distance — unmeasured — re- 

 quired but 0*2 of a volt to produce a discharge which, at the 

 rate found by the previous observations of this series, would 

 indicate a spark-distance of only 0*2 /x/jl. Such short dis- 

 tances could very readily have been accurately measured by 

 using a larger number of reflexions had it been thought 

 worth while to do so. 



After each discharge it was found that the surface had 

 cohered. An example of the coherence will be discussed in 

 an appendix. At each spark-distance, however, it ir> possible 

 to prevent coherence by using a sufficiently small capacity. 



It will be noted that there is a striking difference between the 

 different series of observations, a- shown by tig. 10 (p. 707). 



The potential gradient was found to vary widely from day 

 to day, although on each day a linear relation between the 

 potential and distance of spark-discharge was sufficiently 

 evident. 



It was finally found that the potential gradient could be 

 always raised by thoroughly washing the surfaces with 

 alcohol before buffing them on a soft cotton wheel with dry 



3 A 2 



