612 PROCEEDINGS OF THE AMERICAN ACADEMY. 



coefficient of viscosity with more precision than that indicated by the re- 

 sults hitherto obtained. It would appear that there may be some cause for 

 the discrepancies other tlian the mere accident^ errors which are present 

 in every investigation. A further examination of the subject would, at 

 least, seem justifiable. 



Without reference to the definition of viscosity from the point of view 

 of the Theory of Gases, it may be defined as that property of a gas by 

 virtue of which resistance is offered to the motion of one portion of the 

 gas relative to the contiguous portion. The coefficient of viscosity of the 

 gas is defined as the force per unit surface which must be applied par- 

 allel to the surface of a plane placed in a mass of gas at unit distance 

 from a fixed plane, parallel to the former, so that the moving plane may 

 have unit velocity in the direction of the force. To measure this coeffi- 

 cient, which is usually denoted by //., a shearing strain must be produced 

 in the gas and the force necessary to do the shearing measured. The 

 ideal shear would be that obtained by causing one plane in a mass of gas 

 to move with a constant velocjty parallel to another fixed plane ; but the 

 planes would have to be iufinite to avoid the edge effects, so that this 

 method is not practicable. 



One of the two genei-al methods of shearing which approach this ideal 

 case is by oscillating a solid body in the gas. The resistance offered by 

 the gas to the motion of the solid is determined by observing the diminu- 

 tion of the amplitude of the oscillations. The quantity actually ob- 

 served is the arc of swing, and the decrement in the logarithm of tin's 

 arc is then calculated. In order, thsn, to determine the coefficient of 

 viscosity, there must be a mathematical expression showing the relation 

 it bears to this logarithmic decrement. The desired relation has been 

 obtained in the case of pendulums having the form of a sphere, an infi- 

 nitely long circular cylinder, and a circular plate oscillating in its own 

 plane between and parallel to two fixed disks. The cases of the sphere 

 and cylinder were first discussed by Piofessor Stokes,* while that of 

 the disk was treated by Maxwell. The other method of producing the 

 desired strain, called the transpiration method, consists in causing the 

 gas in question to flow through a tube of narrow bore, a small constant 

 difference of j)ressure being maintained between the ends of the tube. 

 O. E. Meyer f has discussed the flow of the gas in such circumstances 



* Camb. Phil. Soc. Proc, 9. Also a note to Tomlinson's paper, Phil., Trans., 

 1886. 



I I'ogg. Ami., 127, 1866. 



