neighbourhood of strongly Illuminated Bodies. 233 



thousandths of a millimetre, yet a few, say an odd million or 

 so per second, might escape collision for a distance of even 

 half a millimetre, and that these might drive back the dust. 

 The difficulty here would be to see why the boundary of dust 

 and no-dust is as sharp as it is, though we tried to imagine 

 that this was caused by the convection-currents, a definite 

 amount of bombardment being necessary to drive back the 

 dust when carried towards the surface at a given rate, and 

 this definite bombardment being found at a definite distance 

 from the surface. We were now fairly landed, by the obser- 

 vations we had made, on some molecular bombardment hypo- 

 thesis or other, while before this we had been searching 

 among all kinds of mechanical, electrical, and other notions 

 of a vaguer kind. The necessity for some kind of bombard- 

 ment explanation being thrust upon us, it was not long before 

 we perceived that no considerations of free path were neces- 

 sary, but that a differential bombardment on the two sides 

 of a dust-particle would be sufficient to drive it back. In 

 other words, a dust-particle Avould move from a region of 

 greater pressure to a region of less, being urged by a force 

 depending on the size of the particle and the pressure- 

 slope. 



Consider a warm flat vertical surface of copper. Adhering 

 to the surface is a layer of condensed gases with very different 

 properties from ordinary air, which forms part and parcel of 

 the body, and constitutes a transition between genuine air on 

 the one side and unmistakable copper on the other. This layer, 

 or " Bunsen-schicht," is warmed directly by contact with the 

 solid, and by molecular diffusion it warms the next outer 

 layer, which, sending out quick-flying molecules, warms the 

 next, and so on. Ultimately a stationary state is reached, 

 and the temperature must foil continuously as we recede from 

 the surface, according to some exponential law when convec- 

 tion-currents are permitted, but otherwise linearly. 



The air-film next the solid is known to cling tightly to the 

 solid, and not to be quickly removed. Its adherence need 

 not, indeed, be one relating to individual molecules; it may 

 be only a statistical one, but its renewal is known to be such 

 a slow process that it is very unlikely that it appreciably 

 streams upward even when fairly hot. The layer outside this 

 streams upward a little, being retarded by viscosity or 

 molecular diffusion as it shears itself over the inside layer. 

 It in its turn retards the next, and so on, the velocity of up- 

 streaming increasing as we go out from the surface. But 

 this increase of velocity of successive layers only goes on to 

 a certain limit : it soon reaches a maximum, because the 



