ti$$ Prof. Wilmer Duff on Poiseuilles Law 



float and the fixed support. Hence the lever must be very 

 light, or its centre of gravity must be in the plane of support. 

 Both precautions were taken. Each lever, including mirror, 

 weighed about 0*9 gm., and it was made of such a shape that 

 the centre of gravity was in the plane of support. The latter 

 adjustment was made by hanging the lever up by two legs 

 on an inclined piece of paper, and adjusting the mirror until 

 the end of the third leg was in the vertical plane through the 

 other two, as tested by the vertical cross-hair of a reading- 

 telescope. The length of each lever was about 1 cm., and a 

 calculation showed that the residual error in the determination 

 of the difference of level in the two vessels could not exceed 

 1/50 per cent. — (3) Surface rigidity of impure surfaces. The 

 quasi-rigidity of an impure surface (sometimes called sur- 

 face viscosity) would prevent free movement of the float, 

 and cause a tilt that might greatly affect the readings. With 

 clean vessels and distilled water it was found sufficient to 

 avoid taking readings when the water had been more than 

 24 hours at rest, or, better still, to remove the float and 

 thoroughly stir the surface before beginning readings, flow 

 between the vessels being meanwhile prevented by a 

 cap placed on one end of the tube. (I expect to record 

 some observations on surface rigidity in another article.^ 

 (•1) Inequalities of the bore of the tube. The tube with which 

 the measurements were made was carefully selected from a 

 large number that were roughly calibrated, and was 139*5 cm. 

 long and had a mean radius of 0*1675 cm. Calibrated by a 

 column of mercury in 18 steps from end to end, the radius 

 showed a maximum deviation from the mean of 1*4 per cent., and 

 a mean deviation from the mean of 0*7 per cent. The difference 

 between the greatest and least diameters of the end sections 

 did not exceed 1 per cent. It is very improbable that these 

 inequalities were sufficient to produce errors equal to those of 

 observation. — (5) Correction, for the ends of a capillary tube. 

 If the usual (empirical) correction for the ends were applied, 

 namely, three times the diameter to be added to the length 

 of the tube *, it would diminish the values obtained for the 

 viscosity by less than 1 per cent. ; but, since the velocity of 

 flow in the following experiments was exceedingly small 

 compared with those in which the correction has been found 

 necessary, and therefore the continuation of the stream-lines 

 in the direction of the tube much less marked, it would seem 

 probable that the necessary correction is much smaller. The 

 correction for the kinetic energy of the outflowing liquid * is 

 wholly negligible. 



* Thorpe and llodger, Phil. Trans, vol. clxxxv. (1891). 



