416 Messrs. Rowland and Nichols on the 



two other sources of a return charge. The surface of the 

 glass being more or less conducting, an electric charge creeps 

 over the surface from the edges of the tinfoil. In discharging 

 the jar in the usual way by a connecting wire, this surface 

 remains charged, and the electricity is gradually conducted 

 back to the coatings, and thus recharges them. If, further- 

 more, the coatings be fastened to the glass with shellac or other 

 cement, the return charge may be partly due to it; for we 

 have between the coatings not merely glass, but layers of 

 glass, cement, &c, which the theory shows to give a resi- 

 dual discharge. Besides the coatings are not planes; and 

 hence, as one of us has shown, there may be a return charge, 

 even if the glass gave none between infinite planes. If the 

 plates were merely laid on the glass without cementing, the 

 same result would follow, since the insulator would then con- 

 sist of air and glass in layers. 



In the present research these were sources of error to be 

 avoided, since the residual discharge due to the insulating- 

 plates themselves were to be compared. The condenser-plates 

 were copper disks. These were amalgamated, so that there 

 was a layer of mercury between them and the dielectric, which 

 excluded the air and conducted the electricity directly to the 

 surface of the dielectric: thus the condition of a single sub- 

 stance between the plates was fulfilled. The errors due to the 

 creeping of the charge over the surface of the dielectric and 

 that due to the plates not being infinite were avoided, the first 

 entirely and the second partially, by the use of the guard- 

 ring principle of Sir Wm. Thomson. 



Fig. 1 (PL IX.) represents this apparatus. The plate of 

 crystal, c, was placed between two amalgamated plates of cop- 

 per, a and b, over the upper one of which the guard-ring, d, 

 was carefully fitted; this ring, when down, served to charge 

 and discharge the surface around the plate, a; and so the 

 errors above referred to from the creeping of the charge along 

 the plate, and from the plate not being infinite, were avoided. 



The charging battery consisted of six large Leyden jars of 

 nearly a square foot of coated surface each, charged^o a small 

 potential. Although accurate instruments were at hand for 

 measuring the potential in absolute measure, it was consi- 

 dered sufficient to use a Harris unit-jar for giving a definite 

 charge; for very accurate measurements were not desired, 

 and the Harris unit-jar was entirely sufficient for the pur- 

 pose. The return charge was measured by a Thomson qua- 

 drant-electrometer of the original well-known form. 



The apparatus shown in fig. 1 performs all the necessary 

 operations by a half turn of the handle e. By two half turns 



