14 C. Barus — Viscosity of Steel and 



A few remarks on the tables are here in place. Large 

 angles cp, for the measurement of which Gauss' method is in- 

 convenient, were calculated directly from the tangents. The 

 error is everywhere within 1 per cent. In table 3 and more 

 particularly in special experiments the observations on the 

 constancy of the scale reading when the temperatures of both 

 upper and lower wires are identical, prove that errors due to 

 differences of viscosity in the cold wires are negligible. No 

 motion is perceptible until heat has been applied. The dif- 

 ficulty encountered in fixing the zero of time has already 

 been mentioned. The curves are therefore correct in their 

 vertical dimensions, but may have been shifted laterally as 

 much as is indicated in each table by the interval between 

 the parts marked "vapor on" and "vapor off." It has also 

 been stated that the results for mercury are too large because 

 of the difficulty in accurately defining the length of the 

 hot part a. Finally two sets of experiments were made 

 with soft wire (Rods, Nos. 1, 2, 11, 12) because the viscous 

 properties of soft wire are not sharply definable. The curves 

 lie sometimes above, sometimes below the mean zone for 

 annealed 450°. This vagueness of the soft state is largely 

 due to strains incidentally impressed. When a positive twist 

 immediately follows a negative twist, or vice versa, the values 

 of <p (cast, par.) are larger than for the case of an untwisted 

 wire. It is therefore to be noted that in the arrangement of 

 apparatus employed the permanent set of any prolonged torsion 

 is added to the following torsion, particularly in the case of 

 high 1. But this discrepancy is probably negligible. 



Deductions. 

 Viscosity and Temperature. — The behavior of a given steel 

 wire varies with the character and with the amount of twist it 

 has received. These variations are not insignificant. Hence 

 it is expedient to construct a diagram of the mean viscous 

 motion for each degree of hardness, and then to discuss the 

 secondary variations with reference to this diagram. In fig- 

 ures 2, 3, 4 and 5, ^>, the angular torsion in radians (p. 1) 

 is exhibited as a function of time in hours, when wires in all 

 degrees of hardness are exposed to degrees of, temperature (T) 

 of 100°, 190° and 360° respectively. The series is made com- 

 plete when the rate of twist (r) is 3°, the values of <p correspond- 

 ing to jr==100° (tables 1 to 6) where r = 6° having been divided 

 by 2 to effect this reduction. This operation is probably not 

 rigorous; but it is sufficiently correct for the present purposes. 

 In the case of J'=190 o experiments were made both for r=3° 

 (fig. 3), and for r=6° (fig. 4). All the curves consist of two 

 distinct parts, an anterior ascending branch showing the motion 

 at the high temperature 1\ and a descending branch showing 



