On The Relative Velocity Of Blue And Yellow Light

ASTRONOMY: H. SHAPLEY 386 PROC. N. A. S.' ON THE RELATIVE VELOCITY OF BLUE AND YELLOW LIGHT By HARioW SHAPLEY HARvARD COLLIGM OBSICRVATORY, CAMBRIDGES, MASSACHUSETTS Communicated, September 24, 1923 The following discussion of certain astronomical observations appears to show conclusively that the velocities of blue and yellow light through interstellar space differ by less than one part in twenty thousand million. The test is differential and is quite independent of the actual velocity of light, which is as yet affected by errors probably as great as one part in twenty thousand. -80 +80 FIGURE I Composite photographic (full line) and photovisual magnitude curves for cluster type variables in Messier 5. Ordinates are magnitudes in hundredths, and abscissae are phases in thousandths of a day. For the accurate measurement of the relative velocity of light of different wave-lengths, there are three principal requisites-a source that emits vani-colored light signals at controlled or predictable intervals, a fairly accurately measured base line of great extent, and a device for recording the signals sent over this base line in light of different colors. With the development of photometric methods of measuring great stellar distances, these requirements are now all fulfilled, I believe, in Voi. 9, 1923 ASTRONOMY: H. SHAPLEY 387 the study of variable stars in globular clusters. The Cepheid variables of the cluster type provide the signals; the recently determined distance of the globular cluster Messier 5 gives a base line that is 2.3 X 1017 miles in length; and the ordinary photographic plates, used in a.large telescope and with appropriate light filters, are satisfactory recorders of the transmitted light. The Signaling Apparatus.-The periodic cluster type variable stars, of the subclass which shows very asymmetrical light curves and a total variation in brightness of a magnitude or more, are the most satisfactory for this investigation. The changing brightness can be timed with considerable certainty as it passes through any magnitude on the rising branch of the light curve. The maxima are usually sharp, and they too can be timed with some success; but the rapid rise is less dependent than the maximum on the exact form of light curve adopted. Observations with polarizing photometers, by Wendell at Harvard on the star RR Lyrae, and by the writer at Princeton on the similar variable SW Andromedae, showed, more than ten years ago, that the time of median magnitude can be determined within two or three minutes. Many of the variables of the cluster type rise from minimum to maximum, more than doubling the light emission of minimum, at the rate (as they go through median brightness) of one magnitude in thirty minutes. In Messier 5 there are nearly one hundred cluster type variables, with periods ranging from six to twenty hours, and an average period of thirteen hours. In addition, this globular cluster contains a few long period Cepheids, which are of value in determining the parallax. The variability of nearly all these stars was discovered by Professor Bailey. He also investigated the periods and light curves.' For a few stars slight irregularity or variability in the period was found or suspected, but the changes are relatively small, and for the great majority it can be claimed that the periods are known within a fraction of a second. The time of a maximum, or of the passing through the median light on the ascending branch, can be predicted some years in advance with an uncertainty of only a few minutes. The explosive character of the light variation of the cluster type variables is now generally interpreted as one result of pulsations in giant stars. But the use of these stars in the present investigation depends in no way upon the assumed theory of Cepheid var