382 E xperimental Determination o/ Poisson's Ratio. [Apr. 



Professor Kennedy's Needle Gear (E). Two frames were attache 

 to either end of the test piece, and each one carried a long arm 

 such a position that the two were close alongside each other, but 

 touching. The index pointer was attached to a small brass 

 from which two strong needle points, about one-tenth of an inch aj 

 projected ; these rested in fine cross grooves which were cut on 

 arms, and any relative motion was magnified about a hundredfc 

 This instrument gives the average reading for two sides of a test pic 



One of the objects of these researches was, to ascertain wl 

 Poisson's ratio, as determined by these experiments, agreed witl 

 values as found by a comparison of tension and torsion tests, and 

 order to obtain reliable angular measurements of the twist, the aut 

 constructed an instrument (F), which consisted of two mirrors, wl 

 were attached to either end of a torsion test piece, in such a posit 

 that the doubly-reflected image of a scale, which was placed al 

 60 ft. away, coincided with the image as seen direct. A slight 

 of the test piece produces a displacement of the two scales, anc 

 is the measure of the torsion angle. The instrument is very sei 

 and reliable for small angles. 



Only a few of the samples were tested for torsion, but Me 

 Platt and Hargraves (Minutes of the Inst. of Civil Engineers, voL: 

 p. 387) have made experiments on 11 samples with the instr 

 C, and F, but as there is internal evidence that the results cannot 

 relied upon in all cases they have not been reproduced here. 



Before discussing the results it will be necessary to consider 

 far the experiments are reliable. The instruments have already ' 

 discussed, but the methods also play an important part. 



1st Method. Tensile test, measurement of elongation e and 

 contraction c. The value of l//t is c/e, and an error of I per ce 

 in either determination will affect !//* by an equal amount. 



2nd Method. Tensile test and measurement of elongation, 

 torsion test and measurement of shearing angle, a. In this case 

 = 2e 1, and when this value is about 0'2, an error either in e 

 produces a sixfold greater one in !/,. A 5 per cent, error in 

 which is not unlikely, if it is only determined for one side, won 

 absolutely spoil the conclusions. In most cases, 1 \f. found in 

 way is smaller than by the 1st Method, but, as will be seen (Ts 

 Nos. 16 and 58), it sometimes is even greater than 0'500. 



The conclusions to be drawn from the experiments with these nil 

 teen samples are : 



1. That Poisson's ratio is not a constant value for all materials. 



2. That mechanical treatment : cold rolling (No. 52) and annealii 

 (No. 67) of the metal alter it. 



3. That Poisson's ratio is sometimes a function of the stress (No 

 12,17. 23, 28, 53,61, and 68). 



