164 Prof. Magnus on the Polarization of Heat at 100° C, 



observing the first deviations was retained, since, as appears from the 

 numbers given below, the values obtained on repeating the same 

 experiment agree as closely with one another as can be expected 

 in such experiments. The vessel serving as source of heat was, 

 in the first place, so turned that the radiating surface was ver- 

 tical and turned to one side (say, to the left). It was then 

 turned horizontally towards the ceiling, then vertically towards 

 the other side, and finally horizontally towards the floor. For 

 each of these positions the magnitude of the deviation was de- 

 termined by turning the reflecting mirror in succession vertical 

 towards one side, then horizontal towards the ceiling, and ver- 

 tical towards the other side. The construction of the apparatus 

 did not allow of its being turned horizontally downwards. 



For each of these positions the observation was repeated three 

 times, and the mean of these observations, taken immediately one 

 after the other, was adopted. When the radiating surface is di- 

 rected upwards or downwards, we should have expected equal 

 values for similar positions of the mirror. Nevertheless, in spite 

 of all care to avoid the introduction of errors through variations 

 in the adjustment, the values obtained were invariably higher 

 when the radiating surface was turned upwards than when it was 

 directed downwards. The cause of this is evidently the superior 

 warmth of the roof of the room in these experiments above the 

 floor. The former was, in fact, warmed by the air ascending 

 from the radiating vessel and from the neighbouring bolt-head 

 of boiling water. The roof accordingly radiated more heat than 

 the floor or the walls of the room, which was not artificially 

 heated ; and a portion of this heat, reflected by the radiating 

 surface itself, reached the pile along with that directly radiated. 

 Hence also the fact is accounted for, that the difference found 

 on turning the radiating surface upwards and downwards was 

 greatest when the radiating surface was of polished metal ; for 

 in such surfaces the radiating-power is very small, while the 

 reflection is greater than from any others. Since through the 

 wintery cold during which these experiments were conducted the 

 window-side and the back of the room had sometimes different 

 temperatures, large screens; were set up to diminish this injurious 

 influence, at half a metre on both sides of the radiating vessel. 

 In observations such as these, indeed, the greatest care is requi- 

 site. The galvanometer used was the very sensitive one con- 

 structed for former experiments, and since improved. 



If theheat radiated at 100° C. had not been polarized, the values 

 must have been the same at every position of the radiant surface 

 for each of the positions mentioned of the reflecting surface. 

 The numbers given below, however, prove that the deflections 

 of the galvanometer were smaller at every position of the radiant 

 surface when its plane of reflection coincided with that of the 



