50 



DISCOVERY 



aperture of 4 inches, and a focal length of 19 feet. The 

 telescope was not pointed directly to the sun, but was 

 for convenience supported in a fixed horizontal position 

 and fed by an 8-inch coelostat. The latter consists 

 essentially of a plane silvered mirror, which is turned 

 by clockwork about an axis parallel to the axis of the 

 earth's rotation, and at such a rate that the light from 

 the sun is always reflected in the same direction. The 

 photographic plates used measured 10 by 8 inches. 

 The focus of the instrument was carefully adjusted, 

 first by visual observations, and then finally photo- 

 graphically by a series of exposures on a bright star. 

 Eight plates were obtained with this instrument during 

 totality, each having an exposure of 20 seconds. 

 Wlien these plates had been developed, it was found 

 that, with a single exception, they all showed images 

 of seven stars, the remaining plate having been spoiled 

 by cloud. The images were round and sharply 

 defined, and suitable for accurate measurement in a 

 micrometer. 



In order definitely to establish the existence of a 

 deflection, it is of the utmost importance that the scale 

 of the photographs should be determined with the 

 highest accuracy possible — i.e., the equivalent value in 

 angular measure of, say, i mm. on the plate must be 

 known. To provide material for determining this, 

 the same region of the sky was photographed a few 

 months later with the same instrument, none of the 

 adjustments having been changed in the meanwhile, 

 and the sun by that time having moved sufficiently 

 far from the field for photographs to be obtained in the 

 early morning before sunrise. These photographs 

 were taken when the star-field was at approximately 

 the same height above the horizon as it was during 

 the eclipse : this ensures that the effects of refraction 

 on the positions of the stars are very approximately the 

 same on the two series of plates. An additional 

 photograph of the field was also obtained on another 

 plate through the glass — i.e., the back or non-sensitised 

 side of the plate was turned towards the field — so that 

 on development the star images appeared reversed : 

 if this plate, which we will call the key-plate, and 

 one of the other plates were placed film to film, the 

 corresponding star images were approximately super- 

 posed. This considerably simplified the process of 

 mecisurement, as will now be explained. 



The key-plate and one of the other plates were 

 placed films together in a measuring apparatus. The 

 latter consists essentially of a frame for carrying the 

 photographs which can be moved up and down on a 

 slide : the plates are \'iewed by a microscope which 

 is capable of movement along a slide parallel to the 

 plate, but perpendicular to that just mentioned. Any 

 part of the plate can therefore be brought into the 

 field of view of the microscope, and the two films being 



in contact, the star images on each arc in focus at one 

 and the same time. The distances apart of the two 

 images did not exceed one-third of a millimetre, and 

 this distance was measured for each star by means of 

 a micrometer in the eyepiece. By comparing first an 

 eclipse-plate and then a scale-plate successively with 

 the key-plate, the relative differences in position of the 

 stars between the eclipse- and scale-plate could be at 

 once obtained. For convenience, the components of 

 the differences in two directions were measured, viz. 

 in right ascension and declination. The latter direc- 

 tion is that joining the star to the pole ; the former the 

 direction perpendicular to this, or, in other words, the 

 direction in which the star apparently moves, owing 

 to the rotation of the earth. In this way, each eclipse- 

 plate was compared with a corresponding scale-plate, 

 and by a process which need not be elaborated here, 

 but which is not quite so straightforward as it may 

 apparently seem, the effects due to scale and deflection 

 are separated, and the displacement of each star 

 determined. 



It is important that the magnitude of the quantity 

 which is being sought should be grasped. One second 

 of arc corresponds to about one-thousandth of an inch 

 on the plates, and the star images themselves are about 

 three seconds of arc in diameter. The greatest 

 calculated displacement for any star on the plate was 

 somewhat less than one second. Yet these quantities 

 are capable of accurate measurement, and, in fact, the 

 plates were measured independently by two difierent 

 persons, who obtained substantially the same results. 



The results may be summarised in the following table: 



The first column of this table contains the reference 

 number of each star, and the following colimins 

 contain the measured and calculated displacements 

 in the two separate directions. The latter are calcu- 

 lated according to the formula i"75/'' where r is the 

 apparent distance of the star from the centre of the sun 

 measured in terms of the sun's radius as a unit. 



In view of the smallness of the quantities sought for, 

 the agreement is remarkable, the discordances being 

 well w-ithin the limits of measurement. The displace- 

 ment at the limit given by these measures would be 

 i''98, somewhat larger than the amount predicted by 



