C. Bancs — Viscosity of Solids. 183 



made with greater facility, if the solid operated on is such that 

 special instability of molecular configuration is superinduced 

 by heat, instead of electrical action. Such a solid is hard steel, 

 in which in addition to the ordinary thermal instability, what 

 may be called a carburation instability of molecular configura- 

 tion asserts itself, even at mean atmospheric temperatures, and 

 in the homogeneous metal. Inasmuch therefore as the gist of 

 Maxwell's theory is instability of configuration, it follows that 

 the evidence which can be derived with reference to it, from 

 hard steel, must be unique in. character: for despite the 

 extreme hardness and elasticity of tempered steel, instability 

 of molecular configuration demonstrably exists,* and is distri- 

 buted uniformly throughout the metal ; moreover the number 

 of unstable groups can be made to vary over an enormous 

 range, at pleasure. 



I must distinctly state, however, at the outset, that Maxwell 

 limits his considerations to configurations of molecules. The 

 responsibility of fusing Clausius's and Maxwell's theories rests 

 with me. The step is dictated by the behavior of steel, in 

 which the integrity of the molecule is certainly invaded with- 

 out producing essential differences in the character or history 

 of the viscous phenomena. § 13. I may note that the occur- 

 rence of: chemical change makes the hypothesis verifiable. 



4. Perhaps the experiments already made on the viscosity of 

 steelf are a sufficient guaranty for the deductions of this paper ; 

 but as the above remarks clearly show, that data tending to 

 throw light on the ultimate nature of viscosity are urgently 

 called for, I shall add some further experimental results. To 

 obtain these I made use of a perfected form of the apparatus 

 described elsewhere.;): 



For several purposes touched upon in the course of the present 

 work, it is necessary to indicate the theory of the said appara- 

 tus more fully than was done in the earlier paper. Given a 

 continuous straight steel wire of length Z, to which a con- 

 venient rate of twist, r, has been imparted. Consider two 

 right sections whose distance apart is the unit of length, and 

 let %(p be the amount of viscous angular motion§ of the first 

 relative to the second, during the given small time t, for the 

 fixed rate of twist r. To fix the ideas, let the wire be adjusted 

 vertically, and provided with an index to register angular 

 motion at a distance V above the lower end. Then will the 

 motion of the index due to viscous detorsion of two sections 



* B. and S. : this Journal, III, xxxii, p. 276, 188G. 



f B. and 8.: this Journal, xxxii, p. 444, 1886; xxxiii, p. 20, 1887. 



% B. : this Journal, xxxiv, p. 2, 1887. 



§ In the earlier paper (xxxiv. p. 1)1 erroneously called this quantity (p instead 

 of 2<l> ; but this inadvertancy does not conflict with the purposes of the data there 

 given. 



