neighbourhood of strongly Illuminated Bodies. 235 



and, we believe, explains the quick formation of a coat when 

 the light is first turned on to any surface, and also the reason 

 why such coats in stagnant air, though broad, are hazy and 

 indefinite in outline. 



But suppose the air near the body is rapidly changed (in the 

 case of the tube by blowing through it, or by convection-cur- 

 rents in ordinary cases), the stationary state never is reached, 

 the air outside the body is always getting warmed, and the bom- 

 bardment may always continue. The function of the convec- 

 tion-currents, then, is to prevent the arrival of the stationary 

 state, and so to perpetuate the dust-repelling bombardment, at 

 the same time that they sweep off some of the already cleansed 

 air and bring up dusty air at such a rate that the bombard- 

 ment has as much as it can do to keep a layer clear. The 

 coat therefore tends to become thinner when convection- 

 currents increase ; and if currents are unfairly produced (as 

 by blowing), the coat may be swept away faster than it can 

 be renewed, and so wholly disappear. But if excess of tem- 

 perature only is the cause of the currents, then the same cause 

 which strengthens them also assists the bombardment, and 

 accordingly the coat is not swept off by such currents, but 

 may even become thicker as the temperature rises. At the 

 same time it is not to be expected that small differences of 

 temperature will much affect the coat, because of the com- 

 pensation action already explained. Moreover, we have at 

 present no theoretical guarantee that a rise of temperature 

 need thicken the coat ; it might thin it in some media. All 

 we can say is, that convection-currents have a double action, 

 both helping the formation, and causing the removal, of the 

 coat. 



But now, how does the fact of an increasing upward velocity 

 of the air as we go through successive layers from the surface 

 of a body account for the differential bombardment or greater 

 internal pressure necessary to account for the driving back of 

 the dust ? Without technicalities, we answer this question as 

 follows : — Consider a vertical flat plate at a higher tempe- 

 rature than the air. We grant that the total pressure of the 

 air is the same near a warm solid as it is anywhere else ; that 

 is, the average mean square of molecular velocity is simply 

 proportional to the temperature ; but then the average mean 

 square of velocity can be resolved into three components — 

 normal to the surface, parallel to the surface and horizontal, 

 and parallel to the surface and vertical. If the air were at 

 rest, all three components would be equal ; but in any given 

 layer the air is not at rest, it is up-streaming : consequently the 

 vertical component of velocity is greater than the average, and 



