112 THE MOTION OF A PERI'ECT LIQUID. 



then by suitable means cause liquid to flow under pressure between 

 them, the very property of viscosity, which, as before noted, is the 

 cause of the eddying motion in large bodies of water, in the present 

 case greatly limits the freedom of motion of the fluid between the two 

 sheets of glass, and thus prevents, not only eddying or whirling motion, 

 but also counteracts the effect of inertiii. Every particle is then com- 

 pelled hy the pressure behind and around it to move onward without 

 whirling motion, following the path which corresponds exactly Avith 

 the stream lines in a perfect liquid. 



If we now, by a suitable means, allow distinguishing bands of col- 

 ored liquid to take part in the general flow, we are able to imitate 

 exactly the conditions we are seeking to understand. 



[Professor Hele-Shaw here gave demonstrations of the stream lines 

 in liquids flowing under the conditions of a gradually enlarging and 

 contracting cliannel. lie proved that the condition of flow corre- 

 sponded closel}'^ with that shown in fig. 5 and not with that given in 

 fig. 4:. The method of the experiments has already been described in 

 Nature, Vol. LYIII. p. 34, though liy using glycerin instead of water 

 nmch more perfect results were obtained than in those then described.] 



But at this stage you may I'easonalfly inquire how it is that we are 

 able to state with so nuich certainty that the artificial conditions of 

 flow Avith a viscous liquid are really giving us the stream-line motion 

 of a perfect one; and this brings me to the results which mathematicians 

 have o])tained. 



The Aiew now shown represents a body of circular cross section, 

 past which a fluid of infinite extent is moving, and the lines are plotted 

 from mathematical investigation and represents the flow of particles. 

 This particular case gives us the means of most elaborate comparison. 

 Although we can not employ a fluid of infinite extent, Ave can prepare 

 the l)order of the channel to correspond Avitli any one of the particular 

 stream lines and measure the exact positions of the lines inside. 



By means of a second lantern the real floAV of a Adscous liquid for 

 this case is shoAvn upon the second screen, and you will see that it 

 agrees with the calculated flow round a similar obstacle of a perfect 

 liquid. The diagram shown on the wall is the actual figure emplo3^ed 

 for comparison and upon Avhich the experimental case was projected. 

 B}'^ this means it was proved that the two were in absolute agreement. 

 If we start the impulses as lief ore, in a row, we at once see how the 

 middle particles lag behind the outer ones, as indicated by the width 

 of the bands, showing that it is not necessarily the side stream lines 

 that moA^e more slowly. It ma}" be more interesting to you to see, in 

 addition to the foregoing case — in which, for convenience, and as quite 

 sufficient for measurement only, a semicylinder was employed — the 

 case of a complete cylinder (fig. 8). In this case two different colors 

 are used in alternate bands, and these bands are sent in, not steadily, 



