4 C. JBarus — Viscosity of Steel and 



twist be alternately applied positively and negatively, this ex- 

 pansion error is eliminated by commutation. This safeguard 

 was invariably applied. When the twist stored amounts to 

 180° for each wire, its sign may often be reversed and readings 

 made without necessarily readjusting the mirror. For smaller 

 twists the mirror must be readjusted. Larger twists were not 

 applied because of liability to permanent set of the wire so 

 twisted. 



The wires used are Stubs' best steel, hardened and tempered 

 electrically, as described elsewhere.* 



Experimental Results. 



Residual twist. — The following experiments have an introduc- 

 tory and suggestive character. Two glass -hard wires with their 

 ends bent loop-shaped are fastened firmly together at one end 

 of each. The other ends are then twisted z° against each other 

 and also fastened. This twisted system of glass-hard wires, or 

 system in which a rate of twist of 3° or 6° has been stored, are 

 then annealed at divers temperatures during stated times as 

 shown in the tables. The amount of twist lost during anneal- 

 ing is measured by determining the angle between the planes 

 of the upper and lower loops of each wire before and after 

 annealing. A gallows arrangement by which a needle is sus- 

 pended over a divided circle by, the steel w T ire itself is used for 

 measurement. 



The tables contain the number, length (X) and diameter (2^) 

 of the glass-hard wires used for each couple ; also the tempera- 

 ture and the time (A) in hours, during which the couples are 

 annealed. They contain, furthermore, the amount of viscous 

 detorsion (J 1} J 2 ) or twist lost during annealing, for each of the 

 times specified, and each of the wires ; finally, the amount of 

 twist stored in each wire of the system. This amount is equal 

 to the twist nominally applied, minus the angle between the 

 plane of the loops. Allowance is made for the amount of per- 

 manent set produced by mere manipulatory twisting of the cold 

 wires. When this exceeds 10°, the A is discarded. The error 

 of angular measurement is probably not greater than a few 

 degrees. 



In the second part of the table the wires S and ^are soft. 



* U. S. Geol. Survey, Bull. 14, p. 29, 1885. Compare with the above method 

 the one used by Dr. Strouhal and myself in studying the relation of hardness and 

 viscosity (this Journal, xxxii, p. 7, 1886). 



