and Polarization of Heat. 143 



Another series on a different day gave the following quantities 

 per cent. 91, 82, 94. Mean of the whole 86.4 : 100. 



21. Having obtained these decisive results, I proceeded to 

 operate with other sources of heat, and with different tourma- 

 lines. Anxious to avoid the interposition of glass, I had a pair 

 of tourmalines of large size cut without any support. But the 

 best kind will not bear this, and they polarized imperfectly. On- 

 ly fifteen-sixteenths (approximately) of the light in the bright po- 

 sition was stopped in the dark, whilst with the tourmalines A and 

 B every vestige of the brightest gas flame was excluded. With 

 these tourmalines (which may be called C and D) I verified the 

 general conclusions. I was unable to get sufficient effect from 

 non-luminous heat to verify the law in that case. 



22. I had two very fine tourmalines cut and mounted on ex- 

 tremely thin glass. These we may call E and F. With them 

 I was enabled to extend and verify the law of polarization even 

 to the case of non-luminous heated brass, (whose temperature 

 when warmed by alcohol, M. Mellon i estimates at 390° cent. 

 — 734° Fahr.) And it is worthy of observation that among 

 twenty-nine pairs of comparative observations, made with three 

 sets of tourmalines, and heated from the following sources, ar- 

 gand lamp, simple oil lamp, platinum rendered incandescent by 

 alcohol, and non-luminous hot brass, there was only one which did 

 not give positive indications of polarization. The effect, how- 

 ever, with non-luminous heat is extremely feeble, and the per- 

 centage very small, because it is with great difficulty that we can 

 obtain results at all with the interposition of two plates of glass, 

 and two of tourmaline (however thin), and a large portion of heat 

 which reaches the pile is derived from conduction, and therefore 

 diminishes the proportion of polarization. 



23. It is very important to observe, that in this and all simi- 

 lar cases, the effect of conduction or the secondary radiation of heat 



from screens always tends to disguise, and never to produce, the dif- 

 ferences of which we are in search ; that is, so long as the means 

 of alternate observations are taken in the way we have described.. 



