540 Mr. G. A. Shakespear on an Interference- Method for 



though if required there would be little difficulty in reading to 

 •0000003 cm. 



In measuring by interference the relative displacement of 

 two points it is necessary that these points should be connected 

 each with a plane surface, and that during the displacement 

 these two surfaces should remain parallel to each other and 

 sensibly perpendicular to the direction of displacement. Inter- 

 ference-bands are produced between these two surfaces, and 

 any relative tilting would alter the nature of the bands ; 

 besides, in case of such a tilt the displacement would clearly 

 be different for different parts of the surfaces. 



In the earlier experiments the author used a single rod of 

 the substance (glass in the first place, iron and bronze later), 

 and the extension was measured between one point near the 

 upper end of the rod and another near the lower end, the rod 

 being suspended vertically. To the upper point a clip was 

 attached from which depended a rod of the same material as 

 that under investigation, which supported at its lower extremity 

 a right-angled glass prism with one of the equal faces hori- 

 zontal and parallel to a plane glass plate, which latter was 

 borne on levelling-screws on a table carried by a second clip 

 at the lower end of the stretched rod. The points of suspension 

 of the rod and of the stretching-weight were arranged so as 

 to be in the axis of the rod (see fig. 1); but in spite of all 

 precautions (though rods were made specially, and jv. l. 

 as nearly true as possible) it was found that on the => 

 application of the stretching-force a relative tilt pj 

 was produced between prism and mirror, and the 

 centre of the interference-rings was thrown out of 

 the field. 



The reason seemed to be the existence of a slight 

 curvature in the rod which was changed by stretching. 



The following form of apparatus w r as then de- 

 vised : — 



Three wires were used, and these were suspended 

 from a horizontally fixed disk of gun-metal, and 

 their lower ends supported a similar disk. Each 

 wire had a T-piece fixed to each end, and these 

 T-pieces moved each in a radial slot in one or 

 other of the disks, the three slots in each disk being 

 at angles of 120° with each other, and by means of radial 

 adjusting-screws these T-pieces could be adjusted at any 

 desired distance from the centres of the disks. Suppose, 

 now (fig. 2), the wires are of equal length and the disks 

 parallel; if on application of the stretching-force they are 

 displaced from parallelism, it is clear that one of the wires 



