180 Dr. Wright and Messrs. Thompson and Leon. [Feb. 12, 



with Ag t Zn 5 ) in the same way as the abscissa and ordinate curve 

 shown in fig. 5, Part II, p. 35. It is noteworthy, however, that 

 hi 1st the direction of the slope of the ties indicates that throughout 

 the lighter alloy contains more silver than the heavier one, the 

 triangular graphical representation does not clearly indicate that the 

 difference in silver percentage between the lighter and heavier alloys 

 rises to a maximum and then diminishes again, as is distinctly shown 

 In i lie ordinary method with abscissae and ordi nates, as depicted in 

 tig. 4, Part II, p. 35.* Precisely the same remarks apply if the 

 analogous results obtained with bismuth-zinc-silver alloys described 

 in Part III are similarly plotted. 



In addition, however, to the employment of this improved method 

 of graphical representation, Sir G. G. Stokes deduces from a priori 

 considerations an important general principle, viz., that when 

 sufficient amount of intermixture of the constituent metals has taker 

 place a state of equilibrium is arrived at (the temperature being 

 Constant throughout), such that the presence of one ternary alloy in 

 no way affects the composition of the other ; so that the addition or 

 subtraction of a further quantity of either alloy, or of any mixture 

 of the two, does not affect the compositions, but only the relative 

 quantities present, of the two alloys; whence, if any given weights 

 of the two fused alloys be intermixed, the same weights of the same 

 alloys will separate again from one another by gravitation on standii 

 If. therefore, two given alloys, A and B, be thus related (truly 

 jugate), and in any particular experiment carried out until equilibrium 

 is reached one of these alloys, A, be formed, the other alloy, B, must 

 necessarily be also produced ; and this must be the case no matter 

 what may have been the relative proportions subsisting between 

 three metals in the mixture originally employed. 



It appeared to us of considerable interest to examine from the 

 experimental point of view whether this general principle can be 



* [Sir G. G. Stokes has pointed out to me that the diagram, fig. 1, shows at one* 

 that, inasmuch as the difference between the percentages of the solvent in two con- 

 jugate alloys vanishes for the pair, a, b, being nil for each, and again for the 

 pair which merge into one, represented by the point L, it must necessarily be a 

 maximum for some intermediate pair ; and also that, in order to preserve the con- 

 tinuity of conditions, we must, in crossing L, pass from the upper alloy to the 

 lower, and rice versa. Hence, if the entire system of ties could be determined, so 

 as to obtain every possible pair of conjugate points lying, one on one side, the 

 other on the other side, of L, and if these values were plotted on the abscissa and 

 ordinate system, the curve representing the difference between the percentages of 

 the solvent, after having ascended and attained a maximum elevation, must 

 descend again to the base line at a point corresponding with L. If we wish to con- 

 tinue the curve beyond that point, we must now take the ordinates negative instead 

 of positive, the same in magnitude as before, and the curve having crossed the base 

 line, and attained a minimum elevation, will ultimately ascend again to the final 

 l>oint on the base line. C. R. A. W., February 25, 1891.] 



