489 490 491 492 493 494 495 496 497 498 499 500 501 502  
is adjusted on an infinitely distant object, every scaledivision corresponds to l″·308. The glass grating prepared by Nobert is particularly well constructed. In a space 9·0155 Par. lines broad, there are 4501 lines drawn by a diamond. Errors of division, as tested by Nobert with a microscope which magnified 800 times, lie below 0·00002 of a Par. line. The breadth, as given by Nobert, was obtained by comparison with a standard prepared by the mechanician Baumann of Berlin, and which was a copy of the one made by the same artist for Bessel. As a proof of the excellence of this glass grating, I may state that Fraunhofer’s lines can be seen therewith in the third and fourth spectrum, and that in distinctness and richness of detail these lines far exceed those which are obtained by the refraction of light through a flintglass prism. During the observations the grating was always placed perpendicularly to the incident rays. This was accomplished, first, by always giving to the unscratched side of the grating a position such that the image of the heliostataperture reflected by it coincided with the aperture itself; secondly, by adjusting on the heliostataperture the moveable telescope used in the observations; and thirdly, by fixing the axis of the second telescope so as to coincide with the prolongation of the optic axis of the first. The scratched side of the glass grating was always turned from the incident light and towards the moveable telescope, being placed in the middle over the rotationaxis of the instrument. The observations were calculated according to the known formula e sin Θ = mλ, where e, or the distance between two scratches on the grating, had, according to the above remark, the value e=0·000166954 of a Par. inch, λ denotes the required wavelength, Θ the observed angle, and m the order of the spectrum. As the values of λ thus obtained have reference to air, they must be dependent upon its temperature and barometric pressure; I have consequently always noted these two elements, although under ordinary circumstances their influence on the measurements was found to be inappreciable. The changes in the temperature of the grating itself exercise a somewhat more important action; nevertheless since, at the time the observations were made (September and commencement of October), the temperature of the room only oscillated between 13° and 18° C., I have likewise omitted this correction.  That no appreciable errors can have thereby arisen in the mean values thus obtained—values which may be regarded as true for 15° C. and the mean barometric pressure—is readily seen on calculating the magnitudes of these corrections. Assuming the refractioncoefficient of air to be n=1·000294, to be a constant magnitude, independent of temperature and pressure, and the value of e, moreover, to hold for 15° C., w. obtain the following corrected value:— whence we conclude that the correction for log λ amounts to + 0·45 (t°–15°)–0·14 (H–0^{m}·76), expressed in units of the fifth decimal place. Accordingly a change of 2 degrees in temperature produces a change of 2″ in the value of the angle Θ, if Θ be assumed equal to 25°; this error is comparable with the error of adjustment itself. For smaller values of Θ the error will of course be smaller. The angle Θ is also subject to a correction dependent upon the absolute motion of the instrument in the direction of the path of the incident ray; this correction, however, is almost inappreciable for the observations upon which the numerical values in the following Table are founded. The wavelengths are, like those of Fraunhofer, expressed in units whose magnitude is equal to 0·00000001 of a Par. inch. Table I.—Wavelengths, inths of a Paris inch.
