the Marfnetic Effect of Electric Convection. 451) 



measure at least, we left the metallic tube bare, but interposed 

 between it and the moving sectors a large piece ot* paraffined 

 ebonite, which touched neither the tube nor the sectors. 



In these conditions the passage o£ the charged sectors 

 under the astatic system produced deflexions qualitatively in 

 accord with the effects expected from the theory of convec- 

 tion and i)roperly varying with the density of the charge 

 and the velocity. As to a quantitative agreement, that was 

 not very satisfactory, but in this form of experiment the 

 distribution of the charge on the sectors is too indefinite to 

 allow of even a rough approximation. Again, as the sectors 

 rotate in the open air, it is impossible to eliminate completely 

 the effects of the air-currents, which render the system too 

 unsteady for accurate readings to be possible. 



There now- remained for us to find the cause of the re- 

 peatedly negatiA-e results of Cremieu^s experiments. 



Essential Difference betioeen the Negative and Positive 

 Experiments. — An analysis of the details of the negative 

 experiments led us to see that they all differed from the 

 positive experiments in this, that in the negative experiments 

 the moving charged surfaces and the condensing-plates, when 

 there were any, were always covered with a thin layer of 

 some solid dielectric, usually caoutchouc. In his experiments 

 Cremieu found it desirable to realize as high a surface- 

 density of the charge as possible. To do this, he placed the 

 turning disk and the fixed condensing-plates as near each 

 other as possible ; the caoutchouc coA-erings were for the 

 purpose of preventing sparks between plates and disk. 

 A priori, one could see no inconvenience resulting from the 

 presence of the caoutchouc ; in fact, one would naturally 

 suppose that it would prevent leakage and assure a more 

 perfect entrainement of the charge, for, as the charge is 

 carried on the surface of the dielectric in contact with the 

 conductor, any slipping which might occur w^ould be rendered 

 less probable. 



To test the truth of these considerations, we tried the 

 effect of covering the moving sectors in the above experiment 

 with a layer of caoutchouc. There resulted a considerable 

 diminution of the magnetic effects, without, however, the 

 intensity of the convection -current, as measured by the 

 quantity of charge going on or leaving the sectors (measured 

 as above described), showing a corresponding diminution. 

 Moreover, the nature of the deflexions of the astatic system 

 changed. At first quite distinct for the two signs of the 

 charge, they rapidly became smaller, and at the end of several 

 reversals l)pcame scarcely perceptible, especially when the 

 moving sectors were charged positively. For the negative 



