r\A \ vi-n Lo ^ IY 9 l THE ASTROPHYSICAL JOURNAL AN INTERNATIONAL REVIEW OF SPECTROSCOPY AND ASTRONOMICAL. PHYSICS JANUARY NUMBER i VOLUME VII THE SYSTEM OF £ LYRAE. By G. W. Myers. Of the various theories which have been proposed to explain the light changes of variables, three have been most widely accepted : First. A body whose surface possesses different degrees of brightness^ different places, rotates upon an axis, bringing the differently illuminated portions into view at regular intervals. Second. A secondary meteoric swarm circles about a primary swarm, so as to pass at regular intervals between the observer - and the central swarm. The outlying meteors .of the two " swarms, colliding with each other at the times of periastron -"passage, produce also a periodic increase of brightness. A suit- ^able combination of these two causes of light variation will explain a large variety of the light fluctuations observed in varia- Third. The so-called satellite theory, in which two bodies whose luminous intensities may be either equal or unequal, revolve about each other in orbits whose planes pass near the solar system, and, bv the mutual eclipses of the bodies, light changes of such character are produced as to explain some sort of variability. 1 Read at the conferences held in connection with the dedication of the Yerkes Observatory, Oct. 20, 1897. P lba<;5 2 G. W. MYERS The arbitrariness involved in the fundamental assumptions of theories one and two, appears to me to be an element of weakness of so serious a nature as to compel them to yield to the third theory in all cases where the latter is applicable. By proper assumptions, indeed, regarding the extent, mode of distribution, and periodicity of the spots, or respecting the periodicity and distribution of the meteors within the swarms, as also the relative motions of the swarms themselves, any sort of light change is capable of explanation by the first two theo¬ ries ; but as has been pointed out, their well-nigh universal appli¬ cability is in itself an element of weakness, inasmuch as the theories in themselves are in no essential particular an advance on their underlying hypotheses. It is my purpose to give, in a few words, an account of a recent attempt to represent the light changes of ft Lyrae, on the basis of the so-called satellite theory. Argelander’s third curve of this star was based upon I 500 careful photometric estimates, extending over a period of nineteen years. An earlier curve by Oudemans and a later by Schoenfeld, show no discrepancies from this curve of a magnitude great enough to affect the dis¬ cussion appreciably, and, consequently, Argelander’s third curve, published in a pamphlet entitled, “ De Stella /3 Lyrae Variabili Commentatio Altera” in 1859, was taken as the basis for the discussion. This well-known curve is represented in the upper portion of the accompanying figure. INTRODUCTORY. The variability of this star was discovered by Goodericke in 1784, but aside from the recognition of the fact of its variability and the confirmation of the existence of two unequal minima by this same scientist, nothing further was done in its study until Argelander laid the foundation for a thoroughly scientific and, at the same time, an extremely practical method of studying the light fluctuations of variables. Applying his own method to Lyrae, Argelander showed the star to have two unequal minima, separated by two practically equal maxima, the entire SYSTEM OF /3 LYRAE 3 cycle of changes being completed in about I2 d 22 h ( = I2 d .gi). At the time of the I Maximum, he found the brightness of the star to rise to the 3.4 magnitude, and to fall after about three days to a secondary minimum of the 3.9 magnitude, to rise again after three days to the former brightness of the 3.4 magnitude and then, after nearly the same interval, to fall to a primary minimum of the 4.5 magnitude. Or, in other words, the bright¬ ness at the maximum corresponds to 12.33 of Argelander’s 4 G. W. MYERS grades (0.127 mag.); at the primary minimum, to 3.34 grades and at the secondary minimum, to 8.53 grades. Assuming the light changes to be due to the mutual eclipses of two revolving bodies of unequal brightness, we should have the maxima occurring when the components stand beside each other, without either being eclipsed by the other, and conse¬ quently both disks shining full-phase. At the primary mini¬ mum, the darker component lies in front of the brighter and cuts off a portion of the latter’s light, while at the secondary minimum the bright companion passes before the darker and obscures a part of its inferior brilliance. The relative positions of the components at the chief epochs are shown in the draw- ing (Fig. 2). Since the brightness of the “system ” is reduced at Minimum I (primary minimum) by 66 per cent, of its greatest brightness and at Minimum II (secondary minimum) by only 36 per cent., it is evident (i),that the disks must be assumed unequally bright or (2), that the orbital eccentricity must be assumed great or finally, that both these circumstances concur. SYSTEM OF ft LYRAE 5 ECCENTRICITY. An approximate idea of the magnitude of the eccentricity may be obtained in the following two ways : First. If we assume the motions of the components to be in conformity to Kepler’s laws we have, for the relation of the true and mean anomalies, the well-known equation : v — M- f- 2 c. sin M- f- -f e 2 sin 2 M-f- .... The lengths of the chief intervals of light change, from Arge- lander’s curve are : Min. I —Max. I —3.125 days. Max. I —Min. II —3.250 i( Min. II -— Max. II — 3-167 “ Max. II — Min. I = 3.368 “ The approximate equality of these intervals points unmis¬ takably to a small orbital eccentricity. Assuming now, the eccentricity to be so small as to render its higher powers negligi¬ ble, we may shorten the above equation into : (a) v = M-\- 2e sin M. Designating the respective values of v and M at Min. I, Max. I, Min. II and Max. II by v lf M l ; v 2 , M 2 ; v 3 , M 3 and v 4 , M 4 , and substituting in (#), there result the following four equations : (1) v x =M X -f 2 | 0.0186, and these values substituted in (1), give v i=— 3 J ° 33 '- This satisfies the intervals Max. I—Min. I, and Min. II—Min. I, and requires Min. I to lie 31 0 .5 before periastron. From (5) and (7), there result, similarly: ( M x — 209°32'4 (y) -j e = 0.0196 l V x = 209 ° 24 ' . Computing v x from Argelander’s curve, it will be found that a rigorous satisfaction of the intervals Min. I to Max. I and Min. 1 to Min. II, requires a backward displacement of Argelander’s Max. II by about 4 hours, while the intervals Min. I—Max. I and Min. I—Max. II, used in getting (*:), necessitate a backward displacement of Max. I by about 4 hours. Since now, so small a shift in these two chief epochs, corresponds to so large a SYSTEM OF LYRAE 7 change (nearly i8o°) in the position of periastron, the eccen¬ tricity of the orbit cannot be large. The mean of the two nearly equal values of the eccentricity found above, is o.0191 ±0.0033. This value is small enough to justify neglecting its second and higher powers as was done above and thereby vindicates the method of treatment. Second. It will develop later that the hypothesis of a flatten¬ ing of one or both bodies must be made. Assuming the bodies to be deformed by reciprocal tidal influence, or by whatever cause, into similar ellipsoids of revolution — a permissible assump¬ tion, since such forms are figures of equilibrium-—and denoting the ratio of the major and minor axes by “ q ,” so soon as an approximate value of q is known, a superior limit of the ratio of distance of centers, to the larger semi-major axis may be derived. Thus, assuming the bodies to be spheres, with radii equal to the respective semi-minor axes of the ellipsoids, a light curve due to two revolving spheres may be computed. A lower limit for the duration of the eclipse at Min. I, for example, may be read off from this curve. A little reflection will show that the larger q be taken, the shorter the duration of the eclipse will be. Taking then, a value of q greater than that tound later to be the approximate value and assuming that the eclipse has not begun until the light curve has fallen considerably, a value for the eclipse-duration will be obtained which is, at all events, small enough, perhaps much too small, q is later found to be 1.2, and assuming it to be 1.3, I find for inferior limit of eclipse- duration, 3 days and 4 hours, which must be at all events, small enough. But the smaller this inferior limit, the larger must be the ratio of distance of centers to the larger semi-major axis. Using the above value 3 d 4 h for eclipse-duration, a superior limit for this ratio may then be obtained, which will at any rate, be large enough. Taking now, the larger radius as unity and denoting the smaller by *, the radius vector of the true orbit by r (for this, e may be put - o), one half the distance between the nearest points of the satellite at the beginning and end of the eclipse, by we see from the figure that, 8 G. IV. MYERS CPC' ^3. 167X27°. 887 = 88° 20' and hence CPD >44°-10'. Also r V (jv T k) c s c 44 0 io' 1(X + K) I. 438 . But since x^i , I, we haver<2.87 times the smaller semi-axis of the larger ellipsoid or = 2.4 times its larger semi¬ axis. The distance between centers then, being so small com¬ pared with the dimensions of the larger body, the eccentricity could not be large or the masses would necessarily interpene¬ trate during revolution. These considerations are sufficient to warrant the assump¬ tion of a small orbital eccentricity and to justify the hypothesis, that a first approximation to the orbit may be obtained by mak¬ ing e = 0. FLATTENING. The necessity of the assumption either that the disks are flattened, or that the bodies are not yet separated, is apparent from the fact that the brightness at the maxima, does not remain constant for any considerable length of time. It has been assumed in what follows that : (1) The two bodies are distinct and separate. (2) Both are deformed into ellipsoids of revolution. (3) The periods of rotation and of revolution are equal to I2 d .9I. (Possible librations being disregarded.) SYSTEM OF /3 LYRAE 9 Taking the origin of coordinates at the center of the larger ellipsoid and assuming the brightness of the disks uniform, we may readily find from the principles of analytical geometry of space, the equation of the light curve, which may be used while the bodies stand wholly off each other. It was found to be X i + A j | / s i n 2 <£ + (R/R'y cos 2 + — = l I sin 2 <£ + (B/B'Y cos 2 (j> \ Where L denotes the ratio of the combined luminosities of both simultaneous elliptical disks to the combined brightness when the areas of these disks are greatest, i. e. y at the instant of the maxima : A. is the ratio of the intensities of the disks, k is the ratio of the corresponding semi-axes of the two ellipsoids. B ', R' and B, R denote respectively the semi-major and semi¬ minor axes of the smaller and larger ellipsoids, 6bcisthe sight line and ZOy the tangent plane to the celestial sphere at the system. o G. W. MYERS cf> is the longitude of the secondary in its orbit. As a first approximation q was put equal to R'/R = B'/ B. From the unsymmetrical character and small magnitude of the final residuals furnished by a comparison of Argelander’s curve with the computed curve, it appears later, that nothing can be gained by an attempt to improve this hypothesis by ascribing different degrees of flattening to the two ellipsoids, so long as both bodies be regarded symmetrical and their disks uniformly bright. On this hypothesis the above equation becomes i i Z= 7 0 _ . , where cf> — -—- t U being in hours). V sin 2 cf> + i / q 2 cos 2 12 P v 6 ’ An approximate value of q was found by computing light curves for various values of q , viz., i.i; 1.2; 1.25; and 1.3, through a number of symmetrically chosen points before and after the maxima, and deducting the computed values of the ordinates from those of Argelander’s curve. The correct value of q should, of course, give a curve whose ordinates, deducted from the corresponding ordinates of Argelander’s curve, would be those of a curve due to two revolving spheres, and for a time on either side of the maxima, i.e ., while the two components are wholly uncovered, such a curve must run horizontally. The value 1.2 of q gave the following nearly equal ordinates : 10.33; 10.55; 10.76; 10.78; 10.78; 10.79; 10.78; 10.78; 10.76; 10.55 ; and IO -33- This value of q was then assumed as a first approximation. An equation of a light curve applicable during the eclipses was then computed by the process suggested in the drawing, and for Min. I, the necessary equations were found to be : J‘ ' - .77“+ k- a, j * + (»-*')- P‘’ si® * [ , p sin (cb± < 90 or >> 90 , SYSTEM OF /3 LYRAE I I Putting M,= tt(i + k 2 X) (i—//). Then j- k 2 <£" — p' sin or J/ 7 — -|- * 2 — #c sin(<£" + <£). The work of computation was somewhat shortened by intro¬ ducing an additional auxiliary H h when <£">^ and by using the foregoing equations only when (/>"<-. Hj was defined by Hj= K 2 7T— 7T (i -f K 2 A.) (i-//), Whereupon Hj—k 1 <£'-*—<£—p' sin <£, or H 1 = k 2 <£'— -j- k sin (' — <£) and similarly, the equations for Min. II were found to be : ^"' =I_ ,r(i+,c* A) { H u = k 2 tt — 7T- 1 —- (i—///'), or K 2 <£"— <£-fV sin <£. A The remaining formulae hold without modification for both minima. Jj' and J If r are the ratios of the combined instantaneous brightness to the combined brightness due to the sum of the two instantaneous disks shining full-phase during the eclipses at Min. I and Min. II respectively. G, W. MYERS I 2 J r and J ir are the ratios of the same brightness to that due to the sum of the full disks, when these disks are of maximum area. The relation between J/ and / 7 on the one hand, and J n ' and J If on the other, were then derived and found to be Ji=fJ/ and Jn=fJ n ' , where f= R/ /R = B/ /B i j '"sin 2 /?-f- ^ 2 cos 2 /?’ and /3 cf) — - denotes the longitude in the orbit, counted from a point 90° ahead of the origin of (f>. (For undefined symbols see Fig. 6.) Designating by / 3 -|- ^ , the true anomaly in the real orbit, and by a that in the apparent orbit, both counted from the node, and calling p and r the radii vectores in the apparent and true orbits respectively, we have r. sin (3 = p cos a and tan a = cos i cot / 3 , wherefore p‘ 2 =r 2 . sin 2 fi-\-r 2 . cos 2 i. cos ( 3 . Differentiating the formulae for Min. I and reducing, we find 8M f SB, 8 and 8 2 k tg (f)" sin ((f)” -f- (f)) ~~ ^ 2 Ktg 1, as will be found to be the case later, the pre¬ ceding equations become Mj= (f> -j - k 2 (f> 1 — k sin (' T k sin ( 4 > — (/>') , « sin ((f) — (f)') p sin (f) 8c/> — BET, 2 k tg (f)' sin ((f)-- (f >' ) / was obtained by subtracting the ordinate of Argelander’s curve for the selected instants from the mean of the maximum ordinates, calling this difference A G, and computing J from the equation log /= 0.051 A G , which is readily derived from Pog- sjii’s scale, together with the value of Argelander’s grade. The SYSTEM OF (3 LYRAE 3 values of M Iy M in H f , H IV are computed directly from values of J obtained from the above curve, and then an approximate value of <£, interpolated from tables computed from the M's and H’s with (f) as an argument, are then corrected by the above differen¬ tial formulae. Thus it is possible by these formulae to compute a light curve from known elements, and also to solve the converse problem. To compute these tables the values of k, were required. k being unknown, an approximation of its value was obtained thus : calling b m the brightness of a star of magnitude m , and b M that of a star of magnitude M, by Pogson’s scale log = log / 0.4 A M, where £s,M i s the difference of the brightness in stellar magni¬ tudes. From this we have Brightness at Min. II Brightness at Min. I 1-8536 Brightness at Max. Brightness at Min. I The first of these gives (<0 and the second w I + K 2 A — a k 2 A I —a K 2 -j- K 2 A I —j— a * 2 + K 2 A I — a k 2 -j- k 2 A m, where a denotes the portion of the disks common to both bodies at the middle of the eclipses. By a few simple transformations of these equations, it may be readily seen that q > 1.0205, *• e - the disks must be flattened, and 0.8323 1.5241. The values of /c selected, were then 0.8323, 0.9049, 1.0000, 1.2883 an d *-5241 and a table such as was mentioned above was computed with the argument <£ or <£' according as k< i, or k> i. G. W. MYERS 14 The values of r y for various points before and after the min¬ ima, with a correct hypothesis for re and i, the inclination, must all be approximately equal, since they are computed on the hypothesis of a circular orbit, k- 1.5241 gave a fair approach to an agreement in the various values, while the other values of /c gave widely discordant results for r. For some reasons it was more convenient to have k< i, and hence, the larger radius was assumed unity and the necessary alterations in the formulae were made to suit this hypothesis. The only possible assumption which could be made for i was shown to be -, since this gave the various values of r more nearly equal than any other assump¬ tion for it. A was found to be 0.3353 from ( d ) and (^). Recapitulating then; the following values were taken as first approximations : e - -- o, i= -, r 1.8344, k 0.6 q61, A ~ 0.3353 ar| d <7=1.2. 2 By differentiating the above formulae and combining the results, the following differential equations for correcting the approximate circular elements were derived : (I )*(\ + *)JJ,= 2KK l d*+ J " Sin *' dr + r* C0S,gsin *' d (.■'») r o d' •0» where K f = A J t f 2 cos 2 —/,)—/<£— Cpf 2 cos 2 (3 sin and A — Trq\, J3 - tt ^ 1 + ^ ^ k2 — C- 2 ?3 ‘ m q must not be included in this equation since it depends on k by equation (e). The equation applies only during Min. I. For Min. II, / t T \ tt(A+k 2 ) 7r ^ 7 , 2 p sin ' , r 2 cos 2 /? sin ' , (II) —^- dJ//=2K K n dK -|- r - dr-\ - d (i ) 2 A r p K„= A J " f - x — P + tt (/—//,)— Cp/ 2 cos 2 p sin . SYSTEM OF p LYRAE 15 The coefficients of (I) and (II) were computed for 14 points selected symmetrically along the curve, with the approximate values given above, dj, and dj n were obtained by comparing corresponding ordinates of Argelander’s curve with those of a light curve computed from the preceding approximate values. 14 observation equations lead to the following values: dK — -f 0.3586, dr— -j- 0.0999, and d(i' z ) = —0.0434 i’ being here imaginary, but numerically small, it was called 0, and dK being quite large, dq was introduced in its stead, since small changes in the fundamental data do not affect q so greatly as r, and the equations again solved, gave the corrections, dq ~= — 0.0007, an d dr= —0.0889, and the corrected values were then <71.1993 and r 1.8955, and from ( d ) and ( e ), k 0.7580 and X —0.4023. The probable error was, of course, somewhat increased by dropping V , being in the latter case ±0.1 of a “brightness.” These values were regarded as the most probable on the assumption of a circular orbit. By means of these values and differential equations, which were derived for correcting a circular into an elliptical orbit of small eccentricity, a set of 48 observation equations was obtained, for as many points of the light curve, and from them the follow¬ ing elliptical elements were obtained : T 1855 Jan 13 d 6\ 35 P 12.91 d 0 =i -937 = 27°.887 i =90 0 K - = 0.7528 rj 94 0 53' from node A o -399 0=-o° (assumed) q- = 1.203 t = 6 d i5 h -4 after Min. I. Both e and 1 were included in the equation and i' was again imaginary but very small. Using the equations derived for computing the light curve for elliptical motion, the results of computation are comprised in the columns of the table given below. The columns headed A andy B contain the computed and observed ordinates respec¬ tively, and those headed A/ B _ R , the corresponding residuals G. W. MYERS 16 from Argelander’s curve. The computed curve is shown in (Fig. i rt ) drawn in a dotted line beside Argelander’s curve drawn in a full.line. The agreement is quite close. TABLE OF COMPUTED AND OBSERVED BRIGHTNESS. Minimum I Minimum II •A 5/ b-r S-'B-R —72 0.9963 O.9956 ' +0.0007 O.9918 O.9970 —0.0052 —66 0.9870 O.9852 +0.0018 O.9836 0-9945 —0.0109 —60 0.9683 0.9701 —0.0018 O.9696 0.9833 —0.0137 —54 0.9524 0-9513 +0.001 I O.949O O.9672 —0.0182 -48 0.9147 O.9270 —0.0123 O.9258 O.9482 —0.0224 —42 0.8732 O.8849 -0.0117 0.8995 O.9217 —0.0222 —36 0.8209 0.8268 —0.0059 0.8680 0.8886 —0.0206 — 3 ° 0.7296 0.7525 —0.0229 O.8306 0.8494 —0.0188 —24 0.5836 O.6CI9 —0.0183 O.7836 0.8048 —0.0212 — 18 0-4336 0.4993 —0.0657 O.7246 0.7558 —0.0312 —12 0.3661 0.4275 —0.0614 O.6674 0.7044 — 0.0370 - 6 0.3484 O.3487 —0.0003 O.6433 0.6542 —0.0109 0 0-3433 0-3433 +0 O.6365 0.6368 —0.0003 + 6 0.3499 O.3488 +0.0011 O.6472 0.6372 + 0.0100 +12 0.3988 0.4275 —0.0287 O.6762 0.6689 + 0.0073 +18 0.5306 0.5591 —0.0285 0.7340 0.7203 +O.OI 37 +24 0.6572 O.6624 +0.0052 O.7978 0.7716 +0.0262 +30 0.7644 O.7528 +0.0116 O.8498 0.8289 +O.O2O9 +36 0.8416 O.8266 +0.0150 O.893I 0.8622 +O.O3O9 +42 0.8857 O.8845 +0.0012 0.9255 0.8997 +0.0258 +48 0.9234 O.9268 —0.0034 0.9524 0.9307 +0.0217 +54 0-9537 O.9508 +0.0029 O.9627 0-9545 + 0.0082 +60 0.9732 O.9694 +0.0038 O.9881 0.9732 +O.OI49 + 66 0.9870 O.9847 +0.0023 0.9977 0.9877 +0.0100 +72 0.9940 0.9952 —0.0012 I.OO49 0.9970 +0.0079 With the help of an eccentricity, therefore, the residuals are somewhat reduced, and, considering the errors necessarily attach¬ ing to photometric estimates, the curve of Argelander may be regarded as sufficiently well represented. The eccentricity results again almost the same as before, so that at the epoch 1855, the eccentricity did not differ materially from 0.02. Since the periastron lies near Min II, the systematic devia¬ tion of the computed from Argelander’s curve, the former lying above the latter, before, and below it, after Min II, it is highly probable, that while the secondary rounds periastron, very con- SYSTEM OF /3 LYRAE '7 siderable augmentation of brightness due to deformations of the disks, to internal friction, etc., occurs. Because of inertia, these effects could not immediately show themselves, so that the real curve would lie below the computed mean before Min II and above it after Min II, as is represented in the figure. Feeling somewhat suspicious that any one of several sets of elements lying near those just given might represent observa¬ tions equally well, it seemed worth while to test whether this same set of elements would result, if one of the circular elements upon which the elliptical ones were based, were given an arbi¬ trary change and then, by repeated applications of the method of least squares, the elements were again corrected by the obser¬ vations. An arbitrary change of —0.01 was given to k. and of — 0.07 to r and after four adjustments giving ever smaller proba¬ ble errors, the former values were again obtained. It may therefore be inferred that the above elements are the most probable. An interesting fact arising during the latter process of adjustment, was that one set of elements giving a probable error of nearly the same magnitude as the final set, gave the distance between the centers =fe§ 1.80 and the sum of the radii equal to 1.82, i. e. } the components are not yet separated. This fact in connection with the low mean density of the system points to the nebulous condition of the star. The indications are then, either that the companions are not yet separate, but in the act of separation, or that if separate, their separation has taken place comparatively recently. In either case we seem to have here the first concrete example of a world in the act of being born. SPECTROSCOPIC CONSIDERATIONS. Treating by the method of Rambaut, published in Mon. Not. 51, 316, the observations of Belopolsky made between Sept. 23 and Nov. 26, 1892, and published in Melanges math, et astrono- miques , t. VII, l. 3, I find, V = — 0.8 kilometers per second T— 26.93 September 1892 G. IV. MYERS I 8 e = 0.108 o> —79 0 17' from node a sin i — 15836000 kilometers P = i2 d .9i from light period. Lockyer observes the relative displacements of three lines and finds velocities as follows : Hy — 155.0 miles 178 = 154-0 miles ^4025= i5 8 -° mdes. The mean of these values is 155.7 miles and from a private letter of Professor Lockyer, 1 find it belongs to the epoch, August 24.46, 1893. Correcting for the motion of the Earth and reducing to kilometers, there results for the relative velocity in the line of sight 259.8 kilometers. Belopolsky’s measures give for the diameter of the absolute orbit of one of the components 31672000 kilometers. From the absolute orbit furnished by Belopolsky’s observations, the velocity for the above date was found to be of the above relative velocity. The semi-major axis of the relative orbit is then A = 3.168/2X 31672000 — 50175000 kilometers (calling 7 T i =~ - ) and for the ratio of the masses 2 a ' A = m M), or m /M = -— • Assuming that the F-line observed by Belopolsky was pro¬ duced by the smaller component, we have A /a = (Mm ) /m = 1 -|- 8 ' k 3 , (S' —ratio of densities of components), and assuming his line to be produced by the larger, there results, A /a = (M- f- m) /m 1 -j- 1 S'k 3 . Using the values of a/A and k given above, we find 8 '= 5.083, or 1.081. This furnishes a means of deciding which of the components produced Belopolsky’s F-line. Since it is quite improbable that one of the two bodies, so near to each other as is the case here, should have a density 5.083 times that of the other, the latter SYSTEM OF ft LYRAE 19 value is assumed to be the correct one, and hence, that Belopol¬ sky’s F-line was due to the larger component. Using now the well known relation M+m = A* /F* and substituting the values of A in astronomical units and of P in years, there result, M- f- m — 30.56 solar masses and since m/M= 1 / 2.168, M= 20.91 solar masses and m — 9.65 solar masses. Calling 5 the solar mass, H the solar radius, R' and R, the semi-major and semi-minor axes of the minor ellipsoid, the density of the larger companion in terms of the solar density and S 2 of the smaller, Sj = (M/S) q* (H /R')z, (q~R' R as before) and substituting former values, = 0.00058. But 8 2 = 1.08 Sj (8' = 1.08) , hence 8 2 — 0.00063. The mean density of the system is then somewhat less than the density of air, i. e ., comparable to nebular density. It appears then that (3 Lyrae furnishes us a concrete illustration of the actual existence in space of a Poincare figure of equilibrium. Using the chief epochs of Lindemann’s curve constructed from Plassmann’s observations, the following equations result: cos (M x + 49° 57') = —0.0563/^ = — 3*204 (with Argelander’s e) cos (M t 4 - 93 0 5 2 ') = — 0.0338/V = — 1.920 cos ( M 1 -)- 136° 23'.5) = — 0.01771 /e — — 1.010. The impossibility of these relations, on the hypothesis that e is not greater than it was in 1855, requires us to infer that since Argelander’s time the eccentricity of the system has grown larger. Comparing the spectroscopic with the photometric 20 G. IV. MYERS observations, it is also seen that a large motion of the line of apsides has occurred since 1855, though its precise amount can scarcely be ascertained with any considerable degree of certainty. The following results, therefore, seem to be quite clearly indicated by the preceding discussion : 1. The photometric estimates of /3 Lyrae’s variability may be explained easily within the limits of the errors of these esti¬ mates by the aid of the satellite theory. 2. The bodies may be regarded as similar ellipsoids of revo¬ lution. 3. The orbit of the secondary body is nearly circular and its plane passes almost exactly through the Sun. 4. The common flattening of the ellipsoids differs but little from 0.17 and, aside from librations, the periods of rotation and revolution are equal. 5. The larger body is about 0.4 as bright as the smaller. 6. The distance of centers is extremely small, about \ l/% of the semi-major axis of the larger ellipsoid. 7. From Lindemann’s chief epochs for 1892 the orbital eccentricity of the system must have increased from 1855. Belopolsky’s spectroscopic observations indicate the same. 8. The motion of the center of gravity of the system with respect to the Sun is very small. 9. The semi-major axis of the orbit of the companion is about 50,000,000 kilos. 10. The mass of the larger body is 21 times, and of the smaller 9.5 times the solar mass. 11. The densities of the companions are nearly the same. 12. The mean density of the system is comparable with atmospheric density, or the “system” (for such, I think, it must now be called), is in a nebulous condition. 13. In conclusion, it may be said that the spectroscopic and photometric observations, which were available to me for the foregoing discussion, so far from being widely discordant, as some have thought, agree with each other remarkably closely. SYSTEM OF LYRAE 21 The strong absorption lines in the spectrum of this star point to the presence of a powerfully absorbing atmospheric envelope about the nuclei of the masses. From the dynamical theory of gases we know that such an atmospheric layer would arrange itself about the combined mass of the system so that portions of equal density would be in equipotential surfaces. These surfaces would, in the immediate vicinity of the surfaces of the masses, conform somewhat closely to the surface of the bodies, but they would rapidly lose the abrupt curvatures at the surface and become more and more nearly spherical. A rough attempt to represent this is shown in the subjoined figure. The equipotential surfaces, shown by dotted lines, would dispose themselves svmmetrically about the center of gravity of the system as a center, and if the density of the larger com¬ panion were very materially greater than that of the smaller, the center of gravity would lie far within the larger compan¬ ion. It might even happen that the center of gravity of the two bodies should lie beyond the geometrical center of the larger ellipsoid. Where it would lie would depend wholly upon the density of the various parts of the mass of the two bodies. It is then readily seen that the atmosphere might be, and prob¬ ably is, so arranged as to permit the remote end of the smaller 22 G. W. MYERS to shine through a shallow layer of it and thereby to permit the smaller to appear brighter, even though it might be intrin¬ sically darker than the larger body. No violence is done to theory, at all events, by such an assumption. This distribution of the atmosphere would also explain the absorption bands, which are seen in the spectrum of this star more distinctly at Min I than at Min II. The continuous spectrum, which would be produced most distinctly by the smaller body, must at the same time appear fainter. This accords with spectroscopic observations also. Although some of the ideas given above may seem a little venturesome, let it be remembered that the peculiar character of the observations of this star leads one to expect an explana¬ tion of a somewhat, unusual nature. That the ellipsoids are similar is, of course, an arbitrary assumption. In conclusion, let it be observed that an attempt at a formal representation of the condition of things prevailing in the sys¬ tem of ft Lyrae, leads to the assumption of a single body (such as Poincare’s or Darwin’s figures of equilibrium). The above has, of course, only a formal significance, but on account of the poverty of observational material at my disposal an attempt to push the discussion farther on a mathematical basis could not have proved profitable. It is believed, however, that the discus¬ sion may help us to orient our views with regard to this wonder¬ fully interesting star. Fig. i b represents the most probable relative dimensions of the bodies and orbit of the system, as based on Argelander’s photometric estimates up to 1859. Pro¬ fessor Pickering has kindly offered to place all the earlier esti¬ mates of /3 Lyrae’s brightness at the writer’s disposal, and it is the latter’s intention to make a full investigation and discussion of them at the earliest possible date.