Second Series. — Polarization by Refraction. 357 



of heat communicated from A to B, in the positions of Fig. 2. 

 and Fig. 3. To any one who had tried any experiment of the 

 kind, the enormous difference in the effect observed (rising to 

 40 per cent, even in my first experiments), and the instantaneous 

 nature of the change, would have been a sufficient answer. But 

 I farther shewed that the effect was wholly independent of the 

 interval between A and B, and that, therefore, the objection must 

 fall to the ground. 



19. To place the matter beyond all cavil, however, I after- 

 wards devised this experiment : I had a tin canister A, made, 

 Fig. 4, having a surface a, similar in size, figure, and position to 

 the mica plate A of Fig. 2, whose absorbed heat was supposed to 

 affect the pile differently, by being placed in a rectangular posi- 

 tion. The second mica plate B, was interposed as before, — that 

 is, between the pile P and the heated body A, the canister being 

 now filled with boiling water. The effect on the pile, of the sur- 

 face a, was now very great, exceeding, perhaps some hundred 

 times, that of the mica surface with its absorbed heat, which it 

 replaced. Yet, upon turning the canister A into successive rect- 

 angular positions, no very decided difference of effect upon the 

 pile could be observed (Dec. 21. 1835.) If any, the effect indicated 

 a, greater supply of heat in the crossed than in the parallel positions, 

 or was opposed to that which would have indicated polarization. 



20. The polarizing effect takes place only at the su? faces of 

 plates, — the absorptive effect depends upon their thickness. 

 Hence to polarize heat effectively, a minute subdivision of mica 

 into thin laminae is essential. This I formerly effected by a pen- 

 knife. I have since, however, discovered a method much more 

 effectual. A piece of mica, thrown into a brisk red fire, is split 

 up, by the expansion of the air between its films, into a multi- 

 tude of pellicles, which reflect light with almost metallic bril- 

 liancy, and polarize it intensely by transmission. Such mica 

 plates I have used, one pair being marked G and H ; the other 

 I and K. 



21. By experiments with both these pairs of plates I have 



