108 PROCEEDINGS OF THE AMERICAN ACADEMY. 



In order to bring the observations upon the various rods to a common 

 scale, isochroms from the readings for each rod vpere plotted. The gen- 

 eral character of these curves is shown in Figure 18, in which the isochroms 

 corresponding to the isotherms of Figure 17 are given. From the ordinate 

 at 1000° of the isochrom for .6^/, which for convenience was taken as 

 unity for the entire set, a reduction factor was obtained by means of which 

 all the curves for all the carbons were brought to the same scale. A new 

 set of isochroms was then plotted for each of the wave lengths .75^«, .70^, 

 ,65^, .GO^a, .oo^t, .50 fi, and Ao^, in the drawing of wliich all the obser- 

 vations upon the rods were used. While this method did not bring the 

 various sets of observations into perfect agreement, the results were 

 sufficiently definite to indicate with a close degree of approximation the 

 trend of these curves for temperatures up to 1400°. The result of this 

 compilation for the wave lengths just mentioned is shown graphically in 

 Figure 19. From these curves in turn, isotherms for the temperatures 

 900°, 1000°, 1100°, 1200°, 1300°, and 1400° were plotted. These curves 

 are given in Figure 20. Had the law of increasing intensities throughout 

 the spectrum with rising temperature been that anticipated at the begin- 

 ning of this investigation, the trend of the isochroms would necessarily 

 have been such as to bring all the curves together at a common point 

 corresponding to the temperature of the acetylene flame. In other words, 

 if the spectrum of the acetylene flame were identical throughout with that 

 of the carbon rod at the same temperature, the isotherm of the spectrum 

 of the rod at that temperature would be a horizontal line. It is obvious, 

 however, that if the wave lengths of the middle of the spectrum should 

 continue to increase faster than the red and the violet, a condition would 

 presently be attained in which the ordinate of the isotherm would be 

 greater in the yellow or green than at either end of the spectrum. We 

 see indications of the approach of this condition in the diagram of iso- 

 chroms (Figure 19), from which it is evident that the curves for .65^ and 

 .QOfi would cut each other and would cut the curve for JO/j at some tem- 

 perature not far above 1400° ; whereas the isochroms for the shorter 

 wave lengths would not be likely to cut the curves for the red until some 

 much higher temperature had been reached. 



The curves in Figure 20 show the nature of this unexpected development 

 of the spectrum in a somewhat different aspect. It will be seen from 

 this figure that the growth in the extreme red so far lags behind that of the 

 full red, and this in turn behind that of the orange, and this in turn 

 behind that of the wave length .6^, that at 1400° the isotherm, instead of 

 being convex to the base line throughout, actually becomes convex. I 



