of continuity, so that the angular measures were made between the directions of the return ray after reflection when the mirror moved in opposite directions. In this way the speed of the mlrror was as good as doubled, and the possible constant errors inherent in the reference to a fixed direction for the sending telescope were eliminated. The essentials of the apparatus are shown in fig. 4. The revolving mirror was a rectangular prism M of steel, 3 in. high and 1% in. on a side

i; in cross section,

/' /Z

which was driven

A by a blast of air

gf" /' WWW acging on twohfan-7"»

"' '"2l""- B w ees, not s own

eu 1i"::2 ' ' if:: ' ' ' l l "K in the fig., one at " —- if 5 < i the top, the other

'i"

"""—55' at the bottom of

lujll ' the mirror. NPO

5 i i is the object-end of

FIG 4 the fixed sending

telescope the rays

passing through it being reflected to the mirror by a prism P. The receiving telescope ABO is straight, and has its objective under O. It was attached to a frame which could turn around the same axis as the mirror. The angle through which it moved was measured by a divided arc immediately below its eye-piece, which is not shown in the figure. The position AB is that for receiving the ray during a rotation of the mirror in the anti-clockwise direction; the position A'B' that for a clockwise rotation. In these measures the observing station was at Fort Myer, on a hill above the west bank of the Potomac river. The distant reflector was first placed in the grounds of the Naval Observatory, at a distance of 2551 metres. But the definitive measures were made with the reflector at the base of the Washington monument, 3721 metres distant. The revolving mirror was of nickel-plated steel, polished on all four vertical sides. Thus four reflections of the ray were received during each turn of the mirror, which would be coincident were the form of the mirror invariable. During the preliminary series of measures it was found that two images of the return ray were sometimes formed, which would result in two different conclusions as to the velocity of light, according as one or the other was observed. The only explanation of this defect which presented itself was a torsional vibration of the revolving mirror, coinciding in period with that of revolution, but it was first thought that the effect was only occasional.

In the summer of 1881 the distant reflector was removed from the Observatory to the Monument station. Six measures made in August and September showed a systematic deviation of +67 km. per second from the result of the Observatory series. This difference led to measures for eliminating the defect from which it was supposed to arise. The pivots of the mirror were reground, and a change made in the arrangement, which would permit of the effect of the vibration being determined and eliminated. This consisted in making the relative position of the sending and receiving telescopes interchangeable. In this way, if the measured deflection was too great in one position of the telescopes, it would be too small by an equal amount in the reverse position. As a matter of fact, when the definitive measures were made, it was found that with the improved pivots the mean result was the same in the two positions. But the new result differed systematically from both the former ones. Thirteen measures were made from the Monument in the summer of 1882, the results of which will first be stated in the form of the time required by the ray to go and come. Expressed in millionths of a second this was:-

Least result of the 13 measures 24-819

Greatest result .... 24-831

Double distance between mirrors 7-44242 km. Applying a correction of +12 km. for a slight convexity in the face of the revolving mirror, this gives as the mean result for the speed of light in air, 299,778 km. per second. The mean results for the three series were:-

Observatory, 1880-188I V in air=299,627

Monument, 1881 V, , =299,694

Monument, 1882 V, , =299,778

The last result being the only one from which the effect of distortion was completely eliminated, has been adopted as definitive For reduction to a vacuum it requires a correction of +82 km. i Thus the final result was concluded to be

Velocity of light in vacuo =299,8601km per second. This result being less by 50 km. than that of Michelson the latter made another determination with improved apparatus and arrangements at the Case School of Applied Science in Cleveland. The result was

Velocity in vacuo=299,853 km. per second. So far as could be determined from the discordance of the separate measuies, tlée mean error of Newcomb's result would be less than 1

I0 m ut making allowance for the various sources of systematic error the actual probable error was estimated at =*=30 km, 625

It seems remarkable that since these determinations were made, a period during which great improvements have become possible in every part of the apparatus, no complete re determination of this fundamental physical constant has been carried out. The experimental measures thus far cited have been primarily those of the velocity of light in air, the reduction to a vacuum being derived from theory alone. The fundamental constant at the basis of the whole theory is the speed of light in a vacuum, such as the celestial spaces. The question of the relation between the velocity in vacuo, and in a transparent medium of any sort] belongs to the domain of physical optics. Referring to the preceding section for the principles at play we shall in the present part of the article confine ourselves to the experimental results. With the theory of the effect of a transparent medium is associated that of the possible differences in the speed of light of different colours.

The question whether the speed of light in vacuo varies with its wave-length seems to be settled with entire certainty by observations of variable stars. These are situated at different distances, some being so far that light must § fI“ff;'Ve be several centuries in reaching us from them. Were [eng-¢|, there any difference in the speed of light of various colours it would be shown by a change in the colour of the star as its light waxed and waned. The light of greatest speed preceding that of lesser speed would, when emanated during the rising phase, impress its own colour on that which it overtook. The slower light would predominate during the falling phase. If there were a difference of ro minutes in the time at which light from the two ends of the visible spectrum arrived, it would be shown by this test. As not the slightest effect of the kind has ever been seen, it seems certain that the difference, if any, cannot approximate to 1.1;-Gffypm part of the entire speed. The case is different when light passes through a refracting medium. It is a theoretical result of the undulatory theory of light that its velocity in such a medium is inversely proportional to the refractive index of the medium. This being different for different colours, we must expect a corresponding difference in the velocity.

Foucault and Michelson have tested these results of the undulatory theory by comparing the time required for a ray of light to pass through a tube hlled with a refracting medium, and through air. Foucault thus found, in a general way, that there actually was a retardation; but his observations took account only of the mean retardation of light of all the wavelengths, which he found to correspond with the undulatory theory. Michelson went further by determining the retardation of light of various wave-lengths in carbon bi sulphide. He made two series of experiments, one with light near the brightest part of the spectrum; the other with red and blue light. Putting V for the speed in a vacuum and V1 for that in the medium, his result was

Yellow light ...... V: V1 = I-758

Refractive index for yellow . 1-64

Difference from theory +0-12

The estimated uncertainty was only o-02, or 1+ of the difference between observation and theory.

The comparison of red and blue light was made differentially. The colours selected were of wave-length about o~62 for red and O'4Q for blue. Putting V, and Vi, for the speeds of red and blue light respectively in bi sulphide of carbon, the mean result compares with theory as follows:-

Observed value of the ratio V, -, Vp 1-0245 Theoretical value (Verdet) 1~025

This agreement may be regarded as perfect. It shows that the divergence of the speed of yellow light in the medium from theory, as found above, holds through the entirel spectrum. The excess of the retardation above that resulting from theory is probably due to a difference between “ wave-speed ” and “ group-speed ” pointed out by Rayleigh. Let ng. 5 represent a. short series of progressive undulations of constant period and wave-length. The wave-speed is that required to carry

a wave crest A to the position of the crest B in the wave time.