﻿450 JBarus and Strouhal — Viscosity of Steel. 



of experiment such that the torsional couple may alone be 

 effective. The following are extremely unfavorable phases of 

 the best available values, in question : 



Z=Z'<0-2 cm ; Z>25 cm ; ra<2500*; <p<0'2°; 



p=0'041 cm ; ^=2X10 12 (ml-H- 2 ). 



In view of the small values of <p it is sufficient to compare the 

 coefficients only. We obtain nearly 



M b : Mr: M f = 1000: 280000: 20. 



Hence it appears that in the case of the chosen dimension the 

 most unfavorable effect of the bifilar and flexural couples com- 

 bined, is less than 0*4 per cent of the torsional couple. More- 

 over the torsional couple becomes more predominant as the 

 angle of deviation, cp, increases so that no serious inaccuracy 

 from the discrepancies here enumerated need be apprehended. 



2. When we operate with glass-hard steel, we encounter so 

 many mechanical difficulties, that the conditions postulated in 

 the foregoing analysis can not be immediately assumed. We 

 can not, for instance, adjust the tensions of the two wires pre- 

 cisely to equality. Nor is it easy so to store the permanent 

 torsion that the parallelism of the lines joining the upper and 

 lower points of suspension remains intact. Under these circum- 

 stances flexural and bifilar couples may have very different 

 values from the ones accepted and the moments, as a whole, 

 will form a complex aggregate. Although it is improbable 

 that the flexural and bifilar discrepancies will exceed the limits 

 of error investigated, it is none the less desirable to determine 

 them by direct experiment. This can be done with conven- 

 ience and great accuracy. 



Let the water be removed from the jar ILK and the system 

 put in vibration around a vertical and in such a way that the 

 angle <p remains small. Then the circumstances of motion are 

 given bv 



±tt*K W 8M f 



where K is the moment of inertia of the bifilar body, 1 the 

 time of a complete vibration. If we replace the steel wires by 

 threads for which r and dM f /d<p are zero, we may find the 

 value of the bifilar coefficient in terms of the other two. We 

 cite the following typical experiment: . 



Steel wires: 2 7 ,=0-025 sec; Z=26 cm ; Z<0-l cffi ; 



Brass capillary wire: T b =0'35 sec. ; Z=S0 cm ; /=0'2 cra ; l'=0'3 cm . 



If we reduce the results for capillary brass wire to the L and I 

 which obtain for steel, we deduce 7J>40 sec.=0* Hence the 



