172 Energy and Radiation in Incandescence Lamps. [June 19, 



preciable), that the rays of high refrangibility are more absorbed or 

 used up than those of low refrangibility, which is quite contrary to 

 all our knowledge of the laws of absorption. We are, therefore, obliged 

 to look to the filament itself for the cause of the convexity. We know 

 that in a carbon filament resistance diminishes as the temperature is 

 raised, whilst in a platinum wire the resistance increases under the 

 same circumstances. In both cases this implies a re-arrangement of 

 the molecules of the carbon or platinum and the consequent using up 

 of energy. It seems that this might be the reason of the form of the 

 curves. If it be so, then as the radiation from a filament varies as 

 the surface, and the alteration in molecular arrangement varies as 

 the mass, we ought to find that in excessively fine filaments the con- 

 vexity is much diminished, and that the radiation curve is practically 

 a straight line starting from the origin. This is found to be the case 

 in some filaments as fine as those made by Edison. It should also be 

 noticed that from the part where the radiation carve becomes straight, 

 the resistance of the filaments changes but very slowly. 



We have specially made these remarks, since in a paper read before 

 the British Association at Southport, the late Sir William Siemens 

 took exception to our regarding convection currents as being absent, 

 and stated that these currents were imparted to the air outside the 

 lamps. This objection is theoretically valid, but practically it is of 

 no moment. We propose to examine this more fully subsequently. 



A point now arises in regard to this research which is of import- 

 ance in photometric measurement. If the intensities of radiations of 

 a ray, say in the red of the spectrum of an incandescence lamp as 

 produced by two known electrical energies, are compared with the 

 corresponding radiations (say) of a candle, and if then the intensities 

 of a second ray, say in the green, are also compared, the number of 

 watts necessary to produce any temperature in which the same red 

 and green rays bear a known proportion to the same rays in a candle 

 flame may be found by a simple calculation. If the thermopile 

 be used, the same procedure may be adopted, in which case it would 

 be advisable to use rays in the invisible spectrum of low refrangi- 

 bility. For this, of course, the prism must be used with judgment, 

 and after its constants have been properly determined, which we 

 have found no difficulty ourselves in doing. With a grating the 

 deflections on the galvanometer are much smaller, and the errors of 

 observation are consequently likely to be proportionally larger than if 

 the large deflections as given when using a prism are employed. Then, 

 again, too, in using a reflection-grating, certainty must be obtained 

 that the intensity of the spectra follows the theoretical law, or at all 

 events it must be ascertained what is the deviation from it. This is a 

 point that Professor Rowland has exemplified practically, as he has 

 produced gratings in which the intensity of spectra on each side of 



