Page:Proceedings of the Royal Society of London Vol 69.djvu/117

Rh

This was ascertained (<i) by an optical method, (b) by a spher- ometer reading to O'OOl mm. In the first method the two plates A and B were placed successively in a vertical position on the prism table of the spectrometer, which was turned until one edge of the image of the reflected slit of the collimator coincided with the inter- section of the cross threads; micrometer readings were then taken. The divided circle was then rotated until the image of the same portion of the slit was seen reflected from the other face of the plate, and a second set of readings taken. The whole series of operations was repeated several times and with different parts of the circle. By this method the faces of one plate A were found to deviate from true parallelism by 12" of arc, while those of the other plate B were optically parallel.

In the second method the thickness of each plate was determined at four points, each about 2'5 cm. from the four angles along the diagonals ; the results were as under :

Plate A. Plate E.

6 -056 mm. 6 '044 mm.

049 -043

04^> -045

vt-i ,, v*-j. ,

049 -043

Mean... 6'049 Mean... 6 '044

Both of these methods show that no sensible error could arise from imperfect parallelism.

The refractive indices of twenty-four samples were determined with a refractive angle of prism of 60 a 8' 29" (mean value of numerous concordant measurements), the remaining sixteen samples with a refractive angle of 60 6' 7" (mean value also of concordant measurements) ; in every case the angle was measured by the usual method of keeping the telescope of the spectrometer in a fixed position and rotating the prism.

It is possible that the former and higher value arises from some fragment of dust or extraneous matter between one of the quartz blocks and body of prism, and the supposition is apparently borne out by the fact that though the prism has been subsequently re-built on frequent occasions, yet the latter and lower value has invariably been obtained.