THIRD SERIES.— REFBANGIBILITY OF HEAT. 197 



consequent want of parallelism of the refracted rays, the scatteriug of these rays 

 in consequence of the imperfect poMsh of the surfaces, the unequal intensity of 

 the rays in different parts of the section of the cylinder, and lastly, from the 

 want of homogeneity of the rays of heat from any source, which the method would 

 serve to measure, were the other imperfections removed, just as in the com-se of 

 the total reflection of light, prismatic colours are successively presented. 



59. My first rude attempts shewed all this very clearly. As the diagonal 

 abot the lozenge (Fig. 1) shortened, total reflection obviously succeeded to par- 

 tial, and the change was not only very great, but near one point very rapid. 

 The point where the most rapid increase took place, is obviously that where the 

 greater proportion of the incident rays underwent total reflection, and might 

 therefore be taken as a mean representation of the quaUty of the heat. Still the 

 change was too gradual to enable one by mere inspection to determine this point 

 with accuracy, and I speedily resolved to take the sure but laborious method of 

 ascertaining at a number of points intermediate between total and partial reflec- 

 tion the intensities of the reflected heat, and by constructing a curve having mea- 

 sures of the diagonal of the lozenge (a function of the angle of incidence) for 

 abscissae, and intensities for ordinates, I endeavoured to discover graphically for 

 what value of the former the measure of the latter increased most rapidly, in other 

 words, where the tangent made the greatest angle with the axis, or where was the 

 point of contrary flexure of the ciUT^e. 



60. Plate XIII, Fig. 3, may represent such a curve. I have found that when 

 the diagonal of the lozenge was 14.5 inches, the reflection was in all cases nearly 

 total, or the galvanometer was little affected by any increase of the angle of inci- 

 dence. This effect, measured by the vertical line AB, was denoted by 100. 

 When the diagonal was increased to 15.0, the effect was reduced, we shall suppose, 

 to 90, expounded by the line CD, at 15.5 by EF, and so forth. An interpolating 

 curve drawn through the points so iixed, would have its greatest inclination to 

 the axis AX, when, for a given variation of the diagonal, the decrement of the 

 intensity was a maximum, in other words, at the determining angle for the predo- 

 minating part of the heat used. Such a point of contrary flexure would there- 

 fore determine the mean index of refraction of the given kind of heat by the aid 

 of the formula above investigated, whilst the form of the curve would lead to some 

 conjecture at least, respecting the distribution of heat of the more or less refran- 

 gible kinds in the given ray. Heat of low refrangibUity being the last to be to- 

 tally reflected, would cause the curve to droop fastest near the extremity B, the 

 more refrangible rays would be cut off at the other end I of the curve. 



61. I lost no time in verifying the general truth of the principle, and also of 

 the received doctrines respecting heat, by examining the quality of the heat which 

 reached the pile at different stages of total reflection. If, as M. Melloni first 

 rendered probable, heat of low temperature is least refrangible, and vice verm ; 



