1896 LIBRARY UNIVERSITY OF CALIFORNIA. 01 FT OR Received Accession No. Class No. THE ELEMENTS OF THE FOUB INNER PLANETS AND THE FUNDAMENTAL CONSTANTS OF ASTRONOMY BY SIMON NEWCOMB Supplement to the American Ephemeris and Nautical Almanac for 1897 WASHINGTON GOVERNMENT PRINTING OFFICE 1895 PREFACE. THE diversity in the adopted values of the elements and constants of astronomy is productive of inconvenience to all who are engaged in investigations based upon these quanti- ties, and injurious to the precision and symmetry of much of our astronomical work. If any cases exist in which uniform and consistent values of all these quantities are embodied in an extended series of astronomical results, whether in the form of ephemerides or results of observations, they are the exception rather than the rule. The longer this diversity continues the greater the difficulties which astronomers of the future will meet in utilizing the work of our time. On taking charge of the work of preparing the American Ephemeris in 1877 the writer was so strongly impressed with the inconvenience arising from this source that he deemed it advisable to devote all the force which he could spare to the work of deriving improved values of the fundamental elements and embodying them in new tables of the celestial motions. It was expected that the work could all be done in ten years. But a number of circumstances, not necessary to describe at present, prevented the fulfillment of this hope. Only now is the work complete so far as regards the fundamental constants and the elements of the planets from Mercury to Jupiter inclu- sive. The construction of tables of the four inner planets is now in progress, those of Jupiter and Saturn having already been completed by Mr. HILL. All these tables will be pub- lished as soon as possible, and the investigations on which they are based are intended, so far as it is practicable to con- dense them, to appear in subsequent volumes of the Astro- nomical Papers of the American Ephemeris. As it will take several years to bring out these volumes, it has been deemed advisable to publish in advance the present brief summary of the work. HI IV PREFACE. The author feels that critical examination of this monograph may show in many points a want of consistency and conti- nuity. The ground covered is so extensive, the material so diverse as well as voluminous, and the relations to be investi- gated so numerous, that no conclusion could be reached on one point which was not liable to be modified by subsequent decisions upon other points. The author trusts that the diffi- culties growing out of these features of the work, as well as those incident to the administration of an office not especially organized for the work, will afford a sufficient apology for any defects that may be noticed. NAUTICAL ALMANAC OFFICE, U. 8. Naval Observatory, January 7, 1895. I^IVWITT CONTENTS. CHAPTER I. GENERAL OUTLINE OF THE WORK OF COMPARING THE OBSERVATIONS WITH THEORY. Page. 1. Reduction to the standard system of Right Ascensions and Declinations 2. Observations used 3. Semidiameters of Mercury and Venus. Table for defective illumination of Mercury in Right Ascension 3 4. Tabular places from LEVERRIER'S tables. Reduction for masses used by LEVERRIER 6 5. Comparisons of observations and tables 8 $ 6. Equations of condition. Method of formation 8 7. Method of determining the secular variations and the masses of Venus and Mercury independently 10 8. Method of introducing the results of observations on transits of Venus and Mercury ; separate solutions, A from meridian observations without transits ; B, including both meridian observations and transits 13 CHAPTER II. DISCUSSION AND RESULTS OF OBSERVATIONS OF THE SUN. $ 9. Method of treating observed Right Ascensions of the Sun. Expression of errors of observed Right Ascension as error of longitude 15 10. Treatment of observed Declinations of the Sun. Formation of equations of condition for the corrections to the obliquity and to the Sun's absolute longitude 16 $ 11. Formation of equations from observed Right Ascensions of Sun 17 12. Solution of equations from Right Ascensions of the Sun. Tabular exhibit of results of observations of the Sun's Right Ascensions at various observatories during different periods 20 13. Mass of Venus, derived from observations of the Sun's Right Ascension 24 $ 14. Discussion of corrections to the Right Ascensions of the Sun relative to that of the stars 25 v VI CONTENTS. Page. 15. Discussion of corrections to the eccentricity and perihelion of the Earth's orbit 27 16. Results of observed Declinations of the Sun. Exhibit of individual corrections to the absolute longitude and the obliquity of the ecliptic at the different observatories during different periods ^ 29 $ 17. Discussion of the observed corrections to the Sun's absolute longitude 32 18. Discussion of the observed corrections to the obliquity of the ecliptic 33 $ 19. Effect of refraction on the obliquity ; special investigation of the secular change of obliquity as derived from observa- tions of the Sun 35 $ 20. Concluded results for the obliquity, and its secular varia- tion 39 21. Summary of results for the corrections to the elements of the Earth's orbit and their secular variations as derived from observations of the Sun alone . 41 CHAPTER III. RESULTS OF OBSERVATIONS OF THE PLANETS MERCURY, VENUS, AND MARS. 22. Elements adopted for correction 43 $ 23. Introduction .of the corrections to the masses of Venus and Mercury 45 24. Introduction of the errors of absolute Right Ascension and Declinations of the standard stars 46 25. Introduction of the corrections to the secular variations. Method of forming the normal equations by periods so as to include the correction to the secular variation 49 $ 26. Dates and weights of the equations for the various periods. 52 27. Unknown quantities of the equations. Factors for changing corrections of the unknown quantities into corrections of the elements 55 28. Table of the values of the principal coefficients of the normal equations '..'...' 56 29. Order of elimination 57 30. Treatment of meridian observations of Mercury. Effect of want of approximation in the coefficients of the equations of condition 58 31. Introduction of the equations derived from observed tran- sits of Mercury 61 $ 32. Solution of the equations for Mercury 65 $ 33. Systematic discordances among the observed Right Ascen- sions of Mercury in different points of its relative orbit.. 66 CONTENTS. VII Page. 34. Comparison of the results derived from meridian observa- tions of Mercury with those derived from transits over the Sun'sdisk 69 35. Treatment of meridian observations of Venus 70 36. Results of observed transits of Venus 70 37. Equations derived from observed transits of Venus 75 38. Solutions of the equations from Venus 76 39. Comparison of the results of meridian observations of Venus with those of transits 76 40. Solution of the equations for Mars. Inequality of long period in the mean longitude and perihelion, indicated by. observations 77 41. Reduction from the equator to the ecliptic 79 CHAPTER IV. COMBINATION OF THE PRECEDING RESULTS TO OBTAIN THE MOST PROBABLE VALUES OF THE ELEMENTS AND OF THEIR SECULAR VARIATIONS FROM OBSERVA- TIONS ALONE. 42. Modifications of the canons of least squares 81 43. Relative precision of the two methods of determining the elements of the Earth's orbit 86 $ 44. Concluded secular variations of the solar elements, as derived from observations alone 87 $ 45. Common error of the standard declinations 89 $ 46. Definitive secular variations of all the elements from obser- vations alone. Matrices of the normal equations for the secular variations. Tabular statement of results 90 $ 47. Definitive corrections to the solar elements for 1850. . 95 CHAPTER V. MASSES OF THE PLANETS DERIVED BY METHODS INDEPENDENT OF THE SECULAR VARIATIONS, WITH THE RESULTING COMPUTED SECULAR VARIATIONS. $ 48. Plan of discussion 97 49. Mass of Jupiter ; general combination of results 97 50. Mass of Mars. Prof. HALL'S value adopted 99 51. Mass of the Earth, derived from the preliminary value of the solar parallax 99 52. Mass of Venus, derived from periodic perturbations 101 53. Mass of Mercury, from various sources 102 54. Theoretical values of the secular variations for 1850... 106 VIII CONTENTS. Page. CHAPTER VI. EXAMINATION OF HYPOTHESES AND DETERMINA- TION OF THE MASSES BY WHICH THE DEVIATIONS OF THE SECULAR VARIATIONS FROM THEIR THEORETICAL VALUES MAY BE EXPLAINED. 55. Comparison of the observed and theoretical secular varia- tions 109 $ 56. Hypothesis of nonsphericity of the equipotential surfaces of the Sun Ill 57. Hypothesis of an intraniercurial ring 112 58. Hypothesis of an extended mass of diffused matter, like that Which reflects the zodiacal light 115 $ 59. Hypothesis of a ring of planets outside the orbit of Mer- cury. Elements of such a ring. This hypothesis the only one which represents the observations, but too improbable to be accepted 116 $ 60. Examination of the question whether the excess of motion of the perihelion of Mars may be due to the action of the zone of minor planets 116 61. Hypothesis that gravitation toward the Sun is not exactly as the inverse square of the distance 118 62. Degree of precision with which the theory' of the inverse square is established 119 63. Determination of the masses which will best represent the observed secular variations of the eccentricities, nodes, and inclinations 121 $ 64. Preliminary adjustment of the two sets of 1 masses. Result- ing value of the solar parallax 122 CHAPTER VII. VALUES OF THE PRINCIPAL CONSTANTS WHICH DEPEND UPON THE MOTION OF THE EARTH. 65. The processional constant 124 66. The constant of nutation, derived from observations 129 67. Relations between the constants of precession and nutation and the quantities on which they depend 131 $ 68. The mass of the Moon from the observed constant of nuta- tion --- 132 69. The constant of aberration 133 70. The values of this constant, derived from observations 135 71. The lunar inequality in the Earth's motion 139 72. The solar parallax derived from the lunar inequality 142 $ 73. Values of the solar parallax derived from measurements of Venus on the face of the Sun during the transits of 1874 and 1882, with the heliometer and photoheliograph 143 $ 74. The solar parallax from observed contacts during transits of Venus.. 145 CONTENTS. IX 75. Solar parallax from the observed constant of aberration and measured velocity of light 147 76. Solar parallax from the parallactic inequality of the Moon. 148 i 77. Solar parallax from observations of the minor planets with the heliometer 152 78. Remarks on determinations of the parallax which are not used in the present discussion. Errors arising from dif- ferences of color 154 CHAPTER VIII. DISCUSSION OF RESULTS FOR THE SOLAR PARAL- LAX AND THE MASSES OF THE THREE INNER PLANETS. 79. Separate values of the solar parallax, and their general mean 156 80. Rediscussion of the motion of the node of Venus 159 81. Possible systematic errors in determinations of the parallax. 164 82. Revised list of determinations 166 83. Definitive adjustment of the masses of the three inner planets 168 84. Possible causes of the observed discordances 173 85. Adopted values of the doubtful quantities 173 86. Bearing of future determinations on the question. 175 CHAPTER IX. DERIVATION OF RESULTS. 87. Ulterior corrections to the motions of the perihelion and mean longitude of Mercury 178 88. Definitive elements of the four inner planets for the epoch 1850, as inferred from all the data of observation 179 $ 89. Definitive values of the secular variations 182 90. Secular acceleration of the mean motions 186 91. The measure of time 188 92. The constant of aberration 188 93. The mass of the Moon 189 94. The parallactic inequality of the Moon 190 95. The centimeter-second system of astronomical units 190 \S 96. Masses of the Earth and Moon in centimeter-second units.. 191 ^S 97. Parallax of the Moon 193 98. Mass and parallax of the Sun 194 99. Constant of nutation, and mechanical ellipticity of the Earth 195 100. Precession 196 $ 101 . Obliquity of the ecliptic 196 102. Relative positions of the equator and the ecliptic at differ- ent epochs for reduction of places of stars and planets . . 197 ELEMENTTlfFCONSTANTS. OHAPTEK 1. GENERAL OUTLINE OF THE WORK OF COMPARING THE OBSERVATIONS WITH THEORY. 1. In logical order, the first step in the work consists in the reduction of observed positions of the Sun and planets to a uniform equinox and system of declinations. The adopted standard of Eight Ascensions was that origi- nally worked out in my paper on the Eight Ascensions of the fundamental stars, found in an appendix to the Washington Observations for 1870, and extended to a fundamental system of time stars in the catalogue published in Yol. 1 of the Astro- nomical Papers of the American Ephemeris. This system coincides closely with that of the Astronomische Gesellschaft and the Berliner Jahrbuch, about the epoch 1870, but the cen- tennial proper motion is greater' by about 8 .08. In Declinations, the adopted standard was that of Boss, which has been used in the American Ephemeris since 1881, and on which is based the catalogue of zodiacal stars just referred to. But as Declinations generally are not immediately referred to fundamental stars, the method of reducing obser- vations to this system in Declination was not entirely uniform. Observations used. 2. The following is a general statement of the observations used, and the extent to which they were corrected, or re-re- duced. Greenwich. Dr. AUWERS courteously supplied me with the results of his re-reduction of BRADLEY'S observations both of the Sun and planets. From the beginning of MASKYLENE'S work until 1835, the Greenwich observations were completely re-reduced, utilizing, so far as possible, AIRY'S reductions. The 5690 N ALM 1 i GENERAL OUTLINE. [2 data necessary for these observations were discussed in Prof. SAFFORD'S paper, Vol. n, pt. n, which paper was prepared for this purpose. In the case of the Greenwich observations from 1835 onward, it was deemed sufficient to apply constant corrections to the Eight Ascensions, determined from time to time by comparisons of the adopted Eight Ascensions with the standard ones. In the case of the Declinations, Boss's special tables were used, but in the later years it was judged sufficient to apply the constant correction necessary for reduc- tion to Boss's standard. Palermo. PIAZZT'S observations of the Sun and Planets were completely re-reduced, the zero point of his instrument being determined from the observed Declinations. Paris. LEVERRIER'S reduction of the Paris observations from 1801 onward was made use of, applying the correction necessary to reduce the results to the adopted standard. Konigsberg. BESSEL'S clock corrections were individually corrected by the new positions of the fundamental stars, so that practically the Eight Ascensions may be considered as completely re-reduced. In the case of the other observatories, it was deemed suffi- cient to determine, by a comparison of the adopted or of the concluded Eight Ascensions and Declinations of the funda- mental stars with the standard catalogue, what common cor- rections were necessary for reduction to the standard. When, however, the period was covered by Boss's tables, the correc- tion which he gives as varying with the Declination was ap- plied. After more mature consideration, I am inclined to think it would have been better to apply a constant correction to the Declinations in every case, except those where the change with the Declination was quite large. Although these processes were somewhat heterogeneous, it is believed that the main object of referring the Declinations to a system of which the error would be a uniformly varying quantity was fairly well attained. The subsequent determi- nation of this error both in Eight Ascension and Declination is a necessary part of the work. 3] OBSERVATIONS USED. 3 The following is a list of the observatories whose observa- tions of the Sun and Planets were included in the work: Greenwich 1750-1892 Palermo 1791-1813 Paris __-- 1801-1889 Konigsberg 1814-1845 Dorpat 1823-1838 Cambridge . 1828-1844 Berlin 1838-1842 Oxford, Radcliffe 1840-1887 Pulkowa 1842-1875 Washington _ 1846-1891 Leiden 1863-1871 Strassburg 1884-1887 Cape of Good Hope 1884-1890 The number of the meridian observations of the Sun, and of the planets Mercury, Venus, and Mars, actually included in the work is approximately as follows: The Sun 40,176 Mercury _' 54 21 Venus 12, 319 Mars 4, 114 Total 62,030 Semidiameters of Mercury and Venus. 3. The reduction of the semidiarneter of the planets was a point to which special attention was given. In the case of Mercury, the adopted semidiameter at distance unity was 3".34. The values adopted by the various observatories in reducing their observations varied so little from this that in cases where the original reductions were accepted no correction was applied for the difference. So, also, when the observers applied a cor- rection for reducing the observed center of light to the actual center of the planet, no revision of this reduction was made. Such was supposed to be the case with the Paris observations. When the published Eight Ascension was that of the center of light simply, a reduction to the true center was computed by the empirical formula used in the Washington observations. If we put i for the angle between the Earth and Sun as seen from the planet, then 1 -f- cos i will represent the fraction of 4 GENERAL OUTLINE. [3 the apparent transverse diameter of the planet that is illu- minated by the Sun. It was assumed that when the illumina- tion was such that the thickness of the crescent approached zero, the point observed would be two-thirds of the way from the center of the planet to the limb, and that when the planet was dichotomized the center of observation would be five- twelfths of the way from the center to the limb. These con- ditions, with the added one that when the planet was fully illuminated the correction should vanish, suggested the em- ployment of the formula Correction = seinidiameter x (1-cos ^(5-f cos i) This correction was to be multiplied by the sine or cosine of the angle which the line of cusps made with the meridian to reduce it to Right Ascension and Declination respectively. The correction being practically the same whenever the Earth and planet return to the same positions in anomaly, it is possible to embody it in a table of two arguments, one depending on the longitude of the Earth, the other on that of the planet. Actually, however, the table was arranged in a more convenient form, in which one argument is the date at which Mercury last passed perihelion, and the other, its mean anomaly. Owing to the importance which this correction may assume, a partial transcript of the table actually employed for the reduction in Right Ascension is given on the next page. Read horizontally, the numbers show the corrections of the argument through one revolution of the planet. Vertically, they may be regarded as giving the successive corrections corre- sponding to any one position of the planet, while the Earth goes through a complete revolution. The table as actually used extended to every 10, but the values for every 00 of mean anomaly will suffice to show the general magnitude of the correction. The correction to the Declination was embodied in a similar table, which it is not deemed necessary to print at present. In the case of Venus, it seems scarcely possible to decide upoiT a value of the semidiauieter, or a law of its apparent change, which should apply to all parts of the orbit. After a 3] SEMIDIAMETERS OF MERCURY AND VENUS. careful examination of the data, it was decided to reduce all the observations with the semidiameter 8 -^- 5 +0".20 when made with modern instruments, and to use a value 0".3 greater in earlier observations. The actual reductions of all Correction for defective illumination of Mercury in R. A. Arguments: Date of perihelion passage at side, and mean anomaly "g" at top. g= * 60 120 1 80 240 300 360 s s s s s J s Jan. o __ +.19 .16 -.07 03 . OI . oo +.03 10 __ .16 -.18 -.09 -.04 .01 . oo . 02 20 .. .14 . 21 . II 05 . 02 . oo . 02 3-- . 12 .19 13 .06 -3 .00 + .01 Feb. 9__ . 10 17 .15 .08 .04 .01 . oo 19 ._ .08 . 18 . 10 .05 .01 .00 Mar. i .. .06 +. 16 . 21 . 12 .06 .02 .00 ii __ 05 . 16 .24 -.15 .08 03 .00 21 _. .04 *5 .26 .18 -^. 10 .04 .00 31 __ .03 . 14 . 20 . 12 .06 .01 Apr. 10 _. .02 . 12 +.23 . 22 .15 97 .01 20 __ .02 . IO . 20 . 18 . 09 .01 3-- + .01 .08 .18 +.24 . 21 X . II .02 May 10 __ .00 .06 15 . 22 17 -.13 03 20 __ .00 05 . 12 . 20 .12 . 16 .04 30 __ .00 4 . IO . 17 . 18 .05 June 9 __ .00 3 .09 .14 +.18 .20 .06 I9~ .00 .02 .07 . 12 . 16 . 20 -.07 29-- .01 .01 05 .09 15 . 20 .09 Tulv o . OI .OI . 04 . O7 . 1*2 . 1 1 J J I9-- . OI + .01 .03 / 05 o . II +.16 . 12 2 9-- . 02 .00 .02 .04 .09 .16 . 14 Aug. 8 03 .00 .01 -03 .07 . 16 . 16 18. .04 . oo .OI 3 .06 . 14 .18 28.. -.05 . oo + .01 . 02 05 13 Sept. 7 . . .06 .00 .00 .02 .04 . II *7 -- .07 . OI .00 + .01 .02 .09 27 .. .09 . OI .00 .00 .02 .07 + .20 Oct. 7__ . II . 02 .00 .00 + .01 05 .18 17-. . 12 .02 .00 .00 .00 .04 . 16 27 __ . 14 -3 .01 . oo . oo 3 r 3 Nov. 6 . . .16 .04 .01 . oo . oo . 02 . II i6__ .18 .06 .01 . oo . oo + .01 .09 26 .08 . 02 . oo . oo . oo .07 Dec 6 . IO . 03 --. OI . oo . oo / .06 16 . 12 J . o<; . OI . OI . oo . o T o* * - 9-5 21 o. 96 . 03 9-5 3-2 6. i 7-9 May i 0.99 . OI 8.8 4.8 8.4 5-4 ii I. 02 0.98 7-8 6.2 9-7 2. 2 21 1.05 0.95 6.6 7-5 9.9 I. 2 31 1.07 0.93 + 5.3 + 8.5 - 8.9 4-5 June 10 1.09 o. 91 3-7 9-3 6.9 7.2 20 I. 10 o. 91 2. I 9.8 4.1 9.1 30- 1.09 0.91 + 0.4 IO. O - 0.7 10. July 10 1. 08 0.93 9.9 + 2.7 9.6 20 1.05 0.95 3.0 + 9-5 + 5.8 - 8.2 30 1.03 0.07 4.6 8.9 8,2 5-7 Aug. 9 I. OO . OO 6. i 8.0 9-6 + 2.7 19 0.97 03 7-3 6.8 10. 0.8 29 -95 05 8.4 5-4 9.1 4. i Sept. 8 0-93 .07 9.2 + 39 + 7.2 -6.9 18 0.92 .08 9-7 2-3 4.5 8.9 28.... o. 92 .08 10. + 0.6 4- 1.2 9-9 Oct. 8... 0-93 .07 9.9 1. 1 2. 2 9-7 18 o-9S 05 9.6 2.8 5-4 8.4 28.... 0.97 I. 02 9.0 4.4 7.9 6.1 Nov. 7 I. 00 0.99 8.1 5.9 9-5 - 3-1 17 1.03 o. 96 7.o 7.2 IO..O + 0.3 Dec. 7~~" .06 .08 0.94 o. 92 5.6 4- i 8.3 9.1 9-3 7-5 3.7 6.6 17 .09 o. 91 - 2. 5 9-7 4-9 + 8.7 27.... .09 o. 91 -0.8 IO. O 1.6 + 9-9 20 OBSERVATIONS OF THE SUN. [12 Finally, throughout the work the equations of condition were expressed only in entire numbers, the decimals being neglected. To lessen the number of equations of condition, the residuals were divided into groups generally covering from ten to fifteen days, the length of the group being determined by the condition that the perturbations of Venus must not change much during the period. While the formation and solution of the equations of condi- tion on this system were going on, it was found that the intro- duction of the assumed coefficients x 1 and y' was a refinement productive of little or no good result. In fact, the observa- tions of the Sun proved to be much freer from annual sources of error than I had supposed, as will be seen by the tables of their results soon to be given. This is shown by the general consistency of the corrections to the eccentricity and perigee given by the work at the same or different observatories dur- ing different periods. In marked contrast to this is the discordance among values of the correction c to the relative Eight Ascensions of the Sun and Stars. This quantity it is that is affected by personal error and possibly by the effect of the Sun on the instrument. Under a perfect system of discussion it would be advisable to determine it separately for each observer. This however was practically impossible. Solution of the equations. 12. For the purposes of forming and solving the normal equations, the equations of condition were divided into groups of generally from four to eight years, the exact lengths of which will be seen from the following exhibit of results. The equations for each period were solved on the supposition that the corrections were constant during the period. Thus every separate result is independent of every other, except so far as they may depend on the same instrument or the same observer at different times. The first column shows the years through which the obser- vations extend. The second one shows to the nearest year the value of T that is, the fraction of the century after 1850. 12] SOLUTION OF THE EQUATIONS. 21 The third column shows the value of //, or that factor which, being multiplied by the adopted mass of Venus, is to be applied as a correction to that mass, to obtain the value given by the observations. All systematic errors arising from the instrument and the observer are so completely eliminated from the separate de- terminations of X that they may be regarded as absolutely independent of each other, that is as not affected by any common systematic error. We have next the relative weight assigned to each value of //, which is determined in the usual way from the solu- tion, and is, therefore, on a different scale for different ob- servatories. Next is given the value of c, or the apparent correction to the Eight Ascension of the Sun, relative to the assumed Eight Ascensions of the Stars, as given by observations during the several periods and expressed in seconds of arc, followed by the weights assigned to the separate results. The next two columns, the corrections to the solar eccen- tricity and to the longitude of the perigee, require no further explanation. Eespecting the weights ultimately assigned to these quanti- ties, and to GJ it is to be remarked that they are the result of judgment more than of computation. It is only possible to enumerate in a general way with some examples the consider ations on which they are based. In assigning the weight of c the number of observers en- gaged is an important factor in determining it. Other factors are the steadiness of the atmosphere and the adaptation of the instrument to this particular work. General consistency is an important factor in the assignment. In this respect the Cambridge observations are quite remarkable ; if their excel- lence corresponds to their consistency they must be the best ones made. It will be seen that PIAZZI'S results are thrown out en- tirely. The wide range of his values of c led to the inquiry whether more consistent results would be obtained by taking shorter periods, but it was found that the values of c varied from time to time in such an irregular way that his instrument 22 OBSERVATIONS OF THE SUN. [12 must have been affected by some extraordinary cause of error, unless some mistake has been made in interpreting or treating the observations. The Oxford values of c are unusually discordant. The pre- sumption that this discordance arises mainly from the special personal equation in observations of the Sun, described on page 17, derives additional weight from the greater relative consistency of the values of 6e" and e"dn". I have therefore allowed the values of these quantities to receive a fair weight. The value of c for Paris, 1866-'70, has received a much re- duced weight, solely on account of its excessive value. It seems that the work of one observer who made many observa- tions during this period was affected by an unusual system- atic error. Results of observations of the Sun's Right Ascension. GREENWICH. Years. T P' w c W 6e" *" o 4 i/y 1 U J i SOA ' 1 1 - JJ 41 1 I 7O o O $2 4-O. 42 O. 52 O. A 1 0^4 i j CAMBRIDGE. 1877 '78 T A O "I 2 O. 77 4-o. So O. 54 i 10 OJ O 1870 '44 .08 -[-o. 31 2 O. 2O +o. 29 o. 41 i 1847 '57 oo 4-O. 21 2 4-o. 7,1 O. }2 -fo. 10 i i854-'s8 +.06 o. 15 2 4-o. 74 o. 42 -4-o. 13 i WASHINGTON. i840-'49 . 02 0.28 4 o. 73 o. 47 o. 81 2 i86i-'66 +.14 O. II 4 o. 43 0.45 o. 25 2 i867-'72 + . 20 4/-O. 74 4 o. 39 +0.28 o. 51 2 i8y3-'78 + . 26 o. 58 4 o. 32 -(-o. 10 o. 45 2 1 879-' 84 + 32 o. 31 4 o. 60 0.35 o. 72 2 i88s-'9i -f-38 o. 02 4 0.05 o. 20 o. 18 2 KONIGSBERG. i8i I O7 1 820-' 23 28 o. 14 2 O. 22 - 59 O. 47 I . 24 -(-0.65 2 4~- 49 o. 60 -f-o. 24 I i828-'3i . 20 41.08 2 +o. 09 o. 64 o. 1 6 I 1 832-' 34 . 17 o. 72 2 o. 15 I. 72 o. 40 I 1 837-' 44 . 09 0.66 2 o. 62 2. 24 o. 87 I OXFORD. 1 840-' 45 1 846-' 5 1 i86i-'66 .07 . 01 + H +0-79 +0.35 -j-o. 36 2 2 2 +0.42 4-0-40 o. 81 +0.67 4-0.89 4-o. 10 4-0. 22 4-0. 20 I. OI O. 2 0. 2 O 2 1 867-' 7 2 +.20 o. 1 6 2 o. 24 -j-O. 2Q O. 44 O 2 1873-76 i88o-'83 + .25 + 32 0.38 -43 2 2 0-33 -f-O. 12 +0.29 o. 17 o 53 o 08 0. 2 O. 2 i88 4 -'87 -f-36 o. 24 2 -fo. 23 o. 19 4-o. 07 O. 2 OBSERVATIONS OF THE SUN. [16, 17 Results of observations of the Sun's Declination Continued. PULKOWA. Years. 61" W 6 e \\ W 1842^45 .06 -fO. 82 2 0.35 0.01 0.35 I .02 o. 10 2 0.48 +0.07 0.48 .1 i86i-'65 +.13 0.53 2 0.48 0.30 0.48 i866-'7o -f. 18 +0.27 2 0.31 0.38 0.31 DORPAT. i823-'28 .24 +0.99 2 1.26 +0.59 1.41 i i829-'32 .19 -f- 99 2 o. 7 6 -f I -34 0.91 I l8 33~'3 8 .14 4- 1 - 00 2 0.63 -fi-34 0.78 i CAPE OF GOOD HOPE. i884-'87 +.36 0.51 4 +0.05 +o. ii 0.07 2 i888-'9O -}- 39 0.84 4 -{-0.09 -f - 1 9 0.21 2 . STRASBURG. i884-'88 +.36 0.57 4 0.05 0.77 4-0.12 2 LEIDEN. i864-'69 -)-. 17 +0.14 4 o. 01 -(-0.27 0.24 2 i87o-'76 -(-.23 0.23 4 0.06 0.04 0.29 2 Correction to the Sun's absolute longitude 17. So far as mere instrumental measurement is concerned, the correction d s should be determined with greater precision than dl" in the ratio 5:2, because the errors in decimation have to be divided by the factor sin s = 0.40, in order to form dl". Allowing for this large increase in the source of error, the values of 6 1" are more accordant than those of 6 8. This is what we should expect. The values of the former quantity depend mainly upon the comparison of observations made 17, 18] OBLIQUITY OF ECLIPTIC. 33 near the opposite equinoxes, when the snn has the same decli- nation, and when the season is not greatly different. Indeed, if the season changed exactly with the sun's declination, all effects of annual change of temperature would be completely eliminated from 61", as would also in any case any constant error which is a function simply of the Sun's Declination. It is therefore to be expected that the actual probable error of this quantity will conform more nearly to that determined from the residuals than in the case of the other. For these reasons the value of dl" does not give rise to much discussion. The general result from all the observa- tories is, for dl", when developed in the form x -f- y T. x = + 0".05 y = 0".97. Obliquity of the ecliptic. 18. The determination of the obliquity rests upon an essen- tially different basis from that of the absolute longitude, in that it depends upon actual differences of measured Declina- tions, which differences are still further complicated by the fact that they are necessarily made at opposite seasons. A more detailed discussion of them is therefore necessary, and some modification may have to be made in the separate results as adopted. The following special circumstances affecting the observations are to be taken into consideration : The BRADLEY Greenwich results for 1753-^2, are derived from a manuscript communicated by Dr. AUWERS, containing the results of his very careful reduction of BRADLEY'S ob- served Declinations of the Sun, which were compared with HANSEN'S tables. The corrections were reduced to those of LEVERRIER'S tables by being computed at intervals suffi- ciently short to permit of the reduction being interpolated with all necessary precision. No reduction was applied either on account of the constant error of the Declinations determined by Dr. AUWERS himself, nor for reduction to the Boss system of standard Declinations. Hence arises the large value of Ad given by these Declinations. Consequently the value of df is 5690 N ALM 3 34 OBSERVATIONS OF THE SUN. [18 that given immediately by the instrument, on the system of reduction adopted by Dr. AUWERS, in which I have supposed that the Pulkowa refractions were used. From 17G5 to 1816 the Greenwich observations were made with the imperfect quadrant, the Declinations of which are subjected to an error which is not constant. The neces- sary corrections are derived by S AFFORD in Vol. n of the Astronomical Papers. The corrections are those necessary to reduce to Boss's system, and they vary with the Declination. Hence the arc on which the obliquity depends is not that measured with the instrument itself, but that so corrected as to reproduce as nearly as may be the standard Declinations. From 1812 onward the two mural circles were used. Up to 1830 no correction except the constant one derived by SAF- FORD was applied to the Declinations as measured with these instruments. Hence the arc of obliquity is that measured with the instrument itself without being corrected by the standard stars. After 1830 the Declinations were corrected by the tables for Greenwich given in Boss's paper. These corrections vary somewhat with the Declination, and they are different also for different periods. Hence we have here a period during which the instrumental differences of Declination were cor- rected to reduce them to the standard star- system. If the standard system were subject to no farther error than a constant one, common to all Declinations within the zodiac, which common correction would be subject to a uniform change with the time, this system would doubtless be the best one to adopt in order to obtain the secular variation in the obliquity of the ecliptic. But, as a matter of fact, the standard Decli- nations are simply the mean results of Declinations measured with different instruments. It is, therefore, a question whether we shall get any better results by applying reductions to a standard system than we should get by simply taking the mean of the instrumental results, because the system is itself only a mean of such results. It is true that the standard sys- tem depends on more instruments than the obliquity, though not on better ones; but it is also to be considered that the reductions in the case of the Sun may be different from those 18, 19] OBLIQUITY OF ECLIPTIC. 35 in the case of the stars, owing to the very different conditions in which the observations are made. Another troublesome point arises from the refraction used in the reductions. The effect of refraction is always to make the measured obliquity less than the actual one; the correc- tion to the obliquity on account of refraction is therefore a positive quantity, which is a minimum for an observatory at the equator and increase equally towards each pole. Some values of the obliquity were derived from BESSEL'S refractions of the Tabulae Regiomontance, and others from the Pulkowa tables. Since the secular variation of the obliquity is more important than the absolute value of the quantity, it is essen- tial that the standard to which all determinations of the ob- liquity are reduced should be as nearly as possible the same, and therefore that the same refraction should be used. But in reductions to standard star places we meet with the addi- tional complication that the differences in the constant of refraction might be wholly or partially eliminated by the reductions to a standard system. It would therefore be a dif- ficult question how far we should modify the values of 6s on account of the use of different tables of refraction. To avoid all these difficulties I have judged it best to make the obliquity depend mainly upon absolute measures, the reductions being made with the Pulkowa refractions. Effect of refraction on the obliquity. 19. The determination of the average or most probable effect on the obliquity produced by using the Pulkowa refractions, instead of those of the Tabulce Regiomontanw, is easily deter- mined. We divide the ecliptic into a number of equal arcs throughout the year, and by equations of condition express differences of refraction in terms of differences of Declination, and hence differences of obliquity. We thus find that at certain latitudes where observations were made, and where BESSEL'S refractions were used in the reduction, the follow- ing corrections are necessary to reduce the obliquity to the ones given by the Pulkowa refractions: Pulkowa; y = 59.S; Jf = 0".325 Greenwich;

= 3S.9; 4e = - 0".125 36 OBSERVATIONS OF THE SUN. [19 Hence I conclude that for . Dorpat; As 0".29 Konigsberg; Js = /7 .26 Cambridge; Je = - 0".21 Cape Town; At = - 0".12 The corrections to the obliquity thus derived, depending mainly on direct instrumental measurement, and reduced to the Pulkowa refractions, are designated as 6'f . The results for this quantity are given in the last column of the several tables. In the case of BRADLEY'S Greenwich results, I have taken as 6'e Dr. AUWERS'S results unchanged, assuming in the absence of any specific statement that he has used the Pul- towa refraction tables. In the case of MASKYLENE'S observations, I have, by excep- tion, used them as reduced to the standard star-system, because we have no other results at these times, and the en or of his instrument is so strongly shown that it would not do to use the results unchanged. It will be seen, however, that small weights are assigned, and that the weights diminish towards the end of the 'series. In the case of the Greenwich observations from 1812 to about 1834, no change has to be made, as the results are gen- erally or always purely instrumental, and Pulkowa refractions are used in SAFFORD'S work. From 1835 onward I have depended mainly on certain cor- rected Greenwich reductions. First, for tf 7 , I have used the results given by Mr. CHRISTIE in his very valuable paper on the Greenwich Declinations, in M. E. A. S., Vol. XLV, where the Declinations from 1836 to 1879 are reduced on a uniform system. Later, I have adopted the corrected results given in Appendix III to the Greenwich observations for 1887. In each case the result has been reduced to the Pulkowa refrac- tions. The Paris results rest on a different basis from the others, in that the zero point of the instrument depends wholly upon LEVERRIER'S Declinations of the stars, and I fear it was not always accurately determined. Observations near the winter solstice are mostly referred to one set of stars; those near the 19J OBLIQUITY OF ECLIPTIC. 37 summer to another set, the error of which may be systemat- ically different. Certain it is that the results during the early years were very discordant. The weights as given in the table are those assigned a priori, without sufficient reference to the discordance of the older results. I have felt constrained to evade a decision as to their treatment by entirely omitting their results in the final discussion. Iii the case of some other observatories it was difficult to determine exactly what refractions had been used in each special case and what reductions should be made. I have, how- ever, determined the corrections in the best way I was able. A precise determination of the secular change in the ob- liquity is of more importance for our present object than a precise determination of its amount. Hence a series of obser- vations extending through a long period of time, and made on a uniform system, has an advantage over a number of isolated values, in that any constant error with which it may be affected will be eliminated from the secular variation. Possi- ble constant differences between the determinations of the various observatories at different epochs will vitiate the sec- ular variation, but the probable amount of this error may be diminished by using a number of separate determinations, such as are presented in the preceding table. In the Green- wich transit circle we have a very uniform series, extending over a period of forty years, but giving results systematically different from other determinations. This series gives for the correction to the obliquity : Transit Circle, 1847->91 : d'e = - 0".ll i 0".06 + (0".21 i 0".46) T . . . (a) Here, in view of the uniformity of method and reduction, we may regard the mean error of the centennial variation from the discordance alone as a fair approximation to the probable mean error. It will be seen that I have here included four years (1847-'50) of the Mural Circle results. Continuing the Greenwich series backward, the question arises whether we can regard the results of the mural circle from 1812 to 1850 as comparable with those of the transit circle. 38 OBSERVATIONS OF THE SUN. [19 There is certainly nothing in the table to indicate any system- atic difference. From the combination of the two we have M. C. andT. 0., 1812->50: u u rf i <* ^/ _. ^ ^ r ro ^"^ ^ ^ da .d \dadr u it se dtx d f 4- da dr Wde + W fo In the second members of the equations a is regarded as a function of the seven quantities (a), as is also #, for which a similar equation is to be formed. The corrections of the solar eccentricity, perihelion, and mean longitude were also introduced by putting in (1) tfL = dl" + de" + dn" de" dn" (6) ^R = *L de" + ^ dn" de" dn" Introduction of the masses of Venus and Mercury. 23. The correction to the mass of Venus was introduced by taking the tabular perturbation produced by Venus on the geocentric place of the planet at the mean date of each equation as the coefficient of the unknown quantity to be determined. In computing these perturbations regard was 46 MERCURY, VENUS, AND MARS. [23, 24 had to the action of Venus on the Earth as well as ou the planet. On this system the unknown quantity finally found would be the factor by which the adopted mass of the planet must be multiplied in order to give the correction of that mass. It has already been remarked that the mass of a planet can not be determined free from systematic error by observations made upon the planet itself. Hence, the mass of Venus can be determined only from observations of Mercury and Mars, and that of Mercury only from observations of Venus and Mars. But the mass of Mercury is so minute that it would be useless to attempt to determine it from observations either of the Sun or Mars. It was therefore determined solely from the periodic perturbations of Venus. It has happened that the mass of Venus could not be deter- mined in a reliable way from observations of Mars, owing to a defect in the theory of the latter planet, which I shall men- tion hereafter, and have not yet had time to correct. Practi- cally, therefore, the mass of Venus is determined only from observations of the Sun and of Mercury, and that of Mercury from observations of Venus. Correction of equinox and equator. 24. t Could all the observations be directly referred to a visible equinox and equator, the corrections above enumerated would have been the only ones which it was necessary to include in the equations of condition. But, as a matter of fact, the observations were all referred to an assumed system of Right Ascensions and Decimations of standard stars my own system in Eight Ascension and Boss's in Declination. We must therefore introduce two additional unknowns into the equations, which I have represented in the following way: <*, the common error of the adopted Right Ascensions. #, the common error of Boss's Declinations. The first quantity will appear only in the equations derived 'from observed Right Ascensions and the second only in the equations derived from Declinations, the coefficient being unity in each case. 24] CORRECTION OF EQUINOX AND EQUATOR. 47 That the value of 6 found in this way should be regarded as a correction to the Declinations of the equatorial stars will appear by the following considerations. The mean heliocen- tric orbit of a planet as projected on the celestial sphere is undoubtedly a great circle. On the other hand, in view of the systematic discordance always found to exist in measures of absolute Declinations near the equator, and of the fact that these absolute Declinations depend upon assumed constants and laws of refraction, which are necessarily affected with greater or less uncertainty, and are otherwise subject to systematic errors, instrumental or personal, of an obscure character, but strongly shown by a comparison ot.the Declina- tions derived from the work of different observatories, it can not be assumed that these Declinations are free from sys- tematic error. JSow, m one circle ot Decimation, say the equator, we may expect that the error will be nearly constant around the sphere, since the causes of error will generally be nearly constant for any one Declination. This conclusion is confirmed by a comparison of the best star catalogues. Moreover, between the zodiacal limits, the error in each par- ticular case is not likely to differ very greatly from the error at the equator. Even if the difference should be considerable the various values of the error of the different Decimations must have a certain mean value, so that in the case of each particular star, or each region of the heavens, we may conceive the actual error to be divided into two parts one the mean value in question, and the other the deviation from this mean. The latter is probably smaller than the former, and in any case can not very well be determined from observations of the planets. But the condition that the planet moves on a great circle of the sphere admits of the mean value being deter- mined with great precision. It should, therefore, be included in the equations of condition. The value of <*, the common error of all the Eight Ascen- sions, can obviously not be determined from the equations in .Eight Ascension alone, because the only result that such observations can give us would be the values of the Eight Ascensions referred to some assumed equinox. The coefficient of a would therefore completely disappear from the equations 48 ' MERCURY, VENUS, AND MARS. [24 of condition in Eight Ascension. But since the same unknown quantities are introduced into the equations of condition in Eight Ascension and in Declination, the requirement that the two sets of equations shall give common values of these quantities does away with this indetermiriatiou and enables determinate values to be found. In fact, this method does not differ in principle from that usually adopted, in deriving the Eight Ascensions of stars from observations of the Sun. The latter consists in deriving the Sun's absolute longitude from observations of its Declination and absolute Eight Ascensions of the stars by comparing them with the Sun. In the same way we may. consider that, in observations of the planet, the Sun's absolute longitude is derived from observations of Decli- nations of the planet, and then a comes out from the observa- tions in Eight Ascension. I have deemed it absolutely necessary that all the equations of condition should be solved by the method of least squares. By this method alone can the results of the observations as regards separate values of the elements and constants be prop- erly brought out. But the work of constructing and solving a system of nine thousand equations of condition, each involv- ing twenty unknown quantities, would be extremely laborious, and might even require a century for its completion, if done in the usual way. It was therefore necessary to adopt every device by which the labor could be reduced to a minimum. One device was the dropping of all superfluous decimals in the coefficients of the equations. Since tbe errors thus produced would be purely accidental, it follows that if the sum of the products obtained by multiplying the value of each unknown quantity by the error of its coefficient in the equation of con- dition is but a small fraction of the necessary probable error of the absolute term, no serious harm will result from the errors of the coefficients. Another device was the construction of tables for finding the coefficients. Such tables relating to Mercury and Venus are found in Vol. II, Part 1, of the Astronomical Papers. These tables are, however, only given for one mean anomaly in each case, and therefore require computations dependent on the value of the other anomaly. They were therefore extended 24, 25] INTRODUCTION OF SECULAR VARIATIONS. 49 to tables of double entry, so that the value of the derivatives of the geocentric Eight Ascension or Declination at any epoch could be taken from the tables at sight. The arguments were the mean anomaly of the planet and the day of the year at which the planet last passed through its perihelion. Introduction of the secular variations. 25. When the equations of condition are formed on the plan just set forth, the unknown quantities will be the corrections to the elements or to the mean longitude at the date of each equation. But every one of the unknown quantities which have been enumerated, the correction of the masses excepted^ is subject to a secular variation. Hence, instead of the unknown quantities heretofore denned, we introduce two others, the one the value of this unknown at some assumed mean epoch, which, for reasons already set forth, must first be determined from the observations; the other the secular variation in a unit of time. The unknown quantities which have been enumerated make twelve for each equation of con- dition. Eleven of these are subject to a secular variation, so that if the secular variations were introduced into the original equations of condition they would each have twenty-three unknown quantities. The following device was employed to reduce to a minimum the work of introducing and determining the secular variations of the various elements : Firstly, the whole time covered by the observations was divided into periods, never exceeding ten years, except when the observations were very few in number, or entitled to but small weight. It was then assumed that no error would arise from supposing the value of the unknown quantity to be the same throughout the period as it was at the mid-epoch of the period. The maximum absolute error thus arising would be the secular variation during half the length of the period, and the mean error the secular variation during one-fourth of the period; but actually the effect of even this error would be almost entirely nullified by the combination of positive and negative coefficients throughout each period. 5690 N ALM 4 50 MERCURY, VENUS, AND MARS. [25 Let us now put x,y, the corrections to the elements at any epoch, T. Let a x+ b y + cz -f . . . ^=n be an equation of condition between these quantities at this epoch. From a system of such equations, extending through a period numbered i, during which #, #, etc., may be considered as constant, we derive normal equations of the form [aa] t x+[db] t y + . . . = [an], which I shall call partial normal equations, and which we might solve so as to obtain the values of x, y, etc. This solu- tion is not, however, necessary. The values of the unknown quantities being really of the general form we may imagine these values substituted in the normal equa- tions (1), the value T, of t for the mean epoch of the period being substituted for t. Let us now suppose that we introduce the quantities # , 2/o, > #', y', . . into the original equations of condition, using for t the value r tj which pertains to the mean epoch of the period. Our equation of condition will thus become ax Q + fy/o 4- + ar t x f + br t y' + . . = n (3) If from a system of conditional equations of this form we form the normal equations for all the unknown quantities, the results will be these : Partial normal equation in x ; [aa] f # -f- [db] t y + . . + rJaaJX + r^ab^y 1 + . . = [aw], (4) 25] INTRODUCTION OF SECULAR VARIATIONS. 51 Partial normal equation in x 1 j T,[aa],a?o + T,[a&],y + . . + T 2 [aa],#' + rf [ab^y 1 + . . = r t [an] t (5) We conclude that the partial normal equations, when the full number of unknown quantities is included, may be derived from those of the form (1) by the following rules. (1) Each partial normal equation in X Q , y , . . . is formed from that in #, y, etc., by adjoining to the first member of the equation the member itself multiplied by r and then changing x, y, . . .to x j XQ'J and, in the products by r, changing x, y, . . . into a?', y', . . . (2) The partial normal equation in a?', y', . . . is formed from the partial equation in x 0j y^ . . . by multiplying all the terms throughout by the factor r. The final or complete normal equations in all the unknown quantities being formed by the addition of the partial normals, the formulae for the coefficients are as follow : For the final equation in X Q [aa] = [aa]! + [aa], + . . . + [aa] n [ab] = [a&]i+ [a&] 2 + . . . + [ab] n [aa]' = n [aa]i + r 2 [aa] 2 + . . . + r n [aa], [an] = [an]^ [an] 2 + . . . + [aw], For the final equation in x' [aa]" = r, a [aa]! + r, 2 [oa] 2 + . . . +T n [aa] n . . . +r n *[ab] n [an]" = n [an]i + T 2 [an] 2 + . . . + r n [an] n The final equations for all the unknown quantities will then be of the form [oa] x + [aft] y + . . + [aa] 1 x' + . . . = [an] ... (8) [aa]'xo+[ab] / y () + . . . +[aa]"0 / + . . . = [an]" 52 MERCURY, VENUS, AND MARS. [25,26 The epoch from which we count the time, r, is arbitrary. An obvious advantage will be gained in counting it from the mid- epoch of all the observations. Then we shall have, by putting w^ w 2 , etc., for the sum of the weights for the different periods : MI r\ + w 2 r 2 + + w n r n = (9) If the observations are then equally distributed around the orbits of the planet and of the Earth it may be expected that the coefficients [.]', [aft]' .... (10) will all nearly or quite vanish. Practically we may expect that as observations are continued through successive revolutions the ratios of these to the other coefficients will approach zero as a limit. We may then divide the normal equations into two sets, one containing the quantities x , y^ etc., and the other #', y', etc. The coefficients (10) being small, the two sets of normals will be nearly independent, and we may omit the terms (10) in the first approximation, and introduce them in one or two successive approximations so far as necessary. The unit of time is also arbitrary. A certain advantage in symmetry will be gained by so choosing it that the mean value of T 3 shall not differ greatly from unity. It was found that twenty-five years was a sufficiently near approximation to be adopted for all three planets. Dates and weights for epochs and periods. 26. As want of space makes impracticable the present publi- cation of the great mass of material worked up, the following particulars have been selected as those most likely to be use- ful in judging and criticising the work. We give three tables, showing the division of the dates of observation into periods, and the weights for each period. The first column of each table contains the number or designation of the period, as found in the manuscript books. The second contains the mean year of the period. The third column shows the time 26] DATES AND WEIGHTS FOR EPOCHS AND PERIODS. 53 of this mean period from the mid-epoch of the observations, which is taken as follows : For Mercury, 1865.0 Venus, The next column contains the sum of the weights of the equations in each period, as used 'in forming the normal equa- tions. These were not, however, the weights actually used in multiplying the coefficients of the equations of condition. Owing to the diversity in the quality of the observations at different times it was not found convenient to reduce the equations at once to a uniform system of weights, and so dif- ferent units of weight were selected for the older observations and for the earlier observations. After the partial normal equations were formed they were multiplied by the factor F, necessary to reduce them to a standard in which the unit of weight should correspond to the mean error The sums of the weights reduced by these factors are shown in the table. In arranging the weights and selecting the factors it should be remarked that a liberal allowance was made at each step for probable constant errors, which results in the given weights being much smaller than they would have been by the theoretical treatment of the original observations. Not- withstanding this allowance the final result seems to show that it was still insufficient, and that the actual weights of the results are less than would follow even from the final ones as given. . The partial normal equations for each period after being' multiplied by the factors F, are added to form the final normal equations as derived from meridian observations. 54 MERCURY, VENUS, AND MARS. [26 WeightSj epochs, and periods of partial normal equations. MERCURY. 1 1 Right Ascension. Declination. Mean year. T Wt (units of 257.) F - Mean year. T (units of 25 y. Wt. F. I 2 3 3i 3-2 4 5 5i 8- 6, 6, 8 9i 9-2 10, I0 2 i n 2 Ii 3 1766.60 1784. 22 1799.81 -3. 9360 -3.2312 2. 6076 26! I * 1 1765-50 1782.99 -3. 9800 3. 2804 O. 2 4.9 1 f 1796.42 1802.37 1809. 1 8 1824. 83 2. 7432 -2.5052 2. 2328 I. 6068 5-0 39-9 52.8 74-1 I I 1809. 53 2. 2188 18.9 i 1818. 79 1825. 80 1835-56 1.8484 1.5680 I. 1776 0.9 34.5 75-o i i i 1833.84 1838. 26 1843. 97 1855-92 1862. 79 1867. 18 1872.64 1877.05 1882. 17 1886. 29 1889. 70 I . 2464 75-3 141.5 281.5 201. 5 189.5 294.5 214.0 204.5 171.5 338. o 176.0 1% 1 1 1 1 1 1 I 0606 1843. 74 1855.90 1863. 10 1867. 12 1872. 62 1877.12 1882. 24 1886.29 1889.82 o. 8504 o. 3640 o. 0760 +o. 0848 4-o. 3048 -j-o. 4848 4-o. 6896 -j-o. 8516 -j-o. 99*28 98.8 83-3 99-8 1 86. o 129.8 129.8 108.2 199.8 109.5 i i i * i | 0.8412 o. 3632 o. 0884 -j-o. 0872 -l-o. 3056 -j-o. 4820 -fo. 6868 4-0.8516 -j-o. 9880 VENUS. I 1755.83 4. 2868 "3 * 1759.69 4. 1324 7.0 i 2 1767.92 3. 8032 19.7 i 1770. 18 -3.7128 IO. O i 3 1781.06 3- 2776 3-7 i I793.25 2. 7900 13.5 i 4 1792.47 2.8212 12.3 i 1806. 73 2. 2508 65.5 * 5 1802. 64 2.4144 23-3 i i8i5-59 1.8964 67.5 i 6 1810. 31 2. 1076 34-0 i 1823.75 1.5700 197.0 i 7 1816. 88 1.8448 42. 7 i 1836. 02 1.0792 762. o 8 1825.55 I.4 9 80 141.0 1844.08 o. 7568 650. o 9 1835.31 I. 1076 339-3 i 1854. 24 o. 354 333-o 10 1843.98 o. 7608 259.3 i 1861.43 o. 0628 749.0 li 1853-51 o. 3796 205.3 i 1868.06 -j-o. 2024 815.0 12 1861. 60 o. 0560 353-7 * 1875-32 -j-o. 4928 692.0 13 1868. 12 +o. 2048 466. o 1 1883. 15 4-o. 8060 819. o i H 1875- 38 +o. 4952 399-5 1 1888.56 4-1.0224 801.0 i 1C 1883. 09 -j-o. 8036 04. c i 16 1888. 67 -j-I. 0268 D T^ J 520. 5 2 | */ *? 26,27] UNKNOWN QUANTITIES OF EQUATIONS. 55 Weights, epochs, and periods of partial normal equations. MARS. Right Ascension. Declination. T3 o 1 Mean year. T (units of 25 y.} Wt. F. Mean year. r (units of 25 y.} Wt. F. , 1757-43 3- 9428 25-3 i 1758.82 -3.8872 8.8 i 2 1770.55 3- 4i8o II. * 11773-79 -3- 2884 8.8 3 1787.82 -2.7272 10. * 1794.48 2. 4608 13.0 i 4 1799.77 2. 2492 20.7 1 1804. 91 -.-2. 0436 47.0 i 5 1811.32 -1.7872 14.7 * ! 1813.00 I. 7200 30.5 1 6 1829. 17 1.0732 60. o i 1828.04 I. Il84 93-o 7 1837- 39 o. 7444 121. O i 1837. 18 o. 7528 371.0 8 1845- 39 o. 4244 76.3 t 1844.95 o. 4420 255-0 9 1853-36 o. 1056 90. o * 1853. 02 o. 1192 245.0 10 1861. 07 -j-o. 2028 114. o i 1860. 94 +0. 1976 306.0 ii 1869. 20 +o. 5280 124. o * 1868. 80 -f o. 5120 197.0 12 1877.71 -j-o. 8684 132. o i 1877. 38 +o. 8552 257.0 J 3 1883. 27 4-i. 0908 91. o i 1883.26 +1.0904 1 60. o 14 1888. 85 + 1.3140 115.5 i 1888. 48 + 1.2992 167.0' Unknown quantities of the equations. 27. For convenience in solving the equations of condition the coefficients of the equations were multiplied by such numerical factors as would reduce their general mean abso- lute value to numbers of approximately the same order of magnitude. Hence, the unknown quantities themselves are not the corrections to the elements, but these corrections divided by the adopted factors. In the case of Mercury the absolute term was also multi- plied by 10, so that effectively the factors in question were reduced to one-tenth part of their value. The unknown quantities of the equations are represented by the symbols of the elements to which they relate inclosed in brackets. For convenience of reference the following table is given, showing the factors used in the case of each planet. In the case of Mercury the column (a) shows the factors by which the differential coefficients were actually multiplied; (b) the factor by which the unknown quantity, as finally found, must be 56 MERCURY, VENUS, AND MARS. [27, 28 multiplied to obtain the correction as expressed in the last column.. In the case of Venus and Mars these factors are the same. Factors by which the unknown quantities are to be multiplied to obtain corrections of the elements. Symbol of unknown. Mercury Factor ior Venus. . Mars. Corr. of element. w (*) [ ] 1 0.1 7 0.3 dm : m Q ( * ] 40 4 5 2 dl 1 JJ 30 3 6 2.5 dJ [Nj 30 3 7 2.5 sin JdR I ] 30 3 3 10^-7 de f * ] 100 10 439 100-r7 d?r 1 * ] 100 2.056 3 1.3323 edn 1 * 1 10 1 4 4 de ["] 6 0.6 2.5 2 de" ["-] 6 0.6 2 2 e"dn" I 1 10 1 1 5 Of [ * ] 10 1 5 5 d [I"] 10 1 4 3 dl" The secular variation of each unknown in 25 years is expressed sometimes by a suffixed 1, sometimes by an accent, thus: [1]' = [l]i = change of [I] in 25 years. 28. It may also be useful to give the values of the principal coefficients in each of the normal equations. They are found in the following table. Were the other coefficients all zero, these numbers would indicate the weights of the different unknown quantities as resulting from the solution. Several of them were greatly diminished by the process of solution. 28, 29] ORDER OF ELIMINATION. 57 Values of the principal diagonal coefficients in the normal equations. I rtercury. Venus. Mars. Symbol c f From oefficien :. From mer. observa- From transits. Sum. From mer. observa- From transits. Sum. mer. observa- tions. tions. tions. mm 5488 o 5488 ^868 2929 8797 17887 11 I0 559 11308 21867 598i 3540 9521 20924 " JJ 15222 1296 16518 13232 7444 20676 28783 : NN ; 14176 2304 16480 I795I 1636 19587 32478 - ee '. 19015 5076 24091 5686 3350 9036 20119 7T 7T 8621 8352 16973 5290 1732 7022 20564 IIOOI 196 11197 11429 3598 15027 31460 '' e" e" '' 9757 508 10265 9586 665 10251 15909 'IT" TT" 9099 261 9360 5836 1895 7731 14911 ' r // r // ^242 o 5242 ' /"/" = Or 13041 542 13583 11031 2349 I338o 15427 aa 13230 13230 335 o 335 25138 r 66 '' 24657 24657 15196 o 15196 53975 11 ' 7014 67155 74169 6005 8983 14988 26689 JJ " 12366 9383 21749 9837 13014 22851 23440 ; NN ; "35 16682 27717 14724 2874 17598 29494 ee ' 15437 29647 45084 5743 8610 '4353 24364 7T7T 6745 493 i 8 56063 4948 4483 943i 27131 ee 8488 1418 9906 8458 6306 14764 25675 : e " e ff " 8409 2937 11346 9805 1682 11487 22947 : TT" TT //= 8439 1513 9952 5242 4805 10047 17356 V'r'/ n 5432 o 5432 '. l " l " ". 11629 3126 I47S5 10677 5667 16344 20655 aa 11400 o 11400 297 o 297 33624 ; 66 \ 18716 o 18716 10772 o 10772 42405 NOTE. The coefficients for Mercury and Venus in this table are given as they were used in the solution, after dropping the units from all the terms of the equations, except those from transits of Mercury. Order of elimination. 29. In dealing with so extensive a system of unknown quantities it is impracticable to investigate the dependence of each upon all the others. It is therefore essential to arrange the unknowns in an order partly that of interdependence and partly that of the liability of each to subsequent change by discussion and adjustment. Hence, the mass of the planet. Mercury or Venus, should be first eliminated, as being that unknown which is least affected by changes in the final values of the other unknowns. The secular variations, as derived 58 MERCURY, VENUS, AND MARS. [29, 30 from meridian observations, are nearly independent of the corrections to the other elements. The solar elements are to be subsequently determined by a combination of the results of the observations of the Sun and of the three planets. Guided by these considerations, the order of elimination was, with some exceptions, as follows : 1. The mass of the disturbing planet. 2. The five elements of the observed planet. 3. The four elements of the Earth's orbit. 4. The corrections to the star-positions for the mid-epoch. 5. The secular variations of the eleven quantities (2), (3), and (4), taken in the same order. Treatment of meridian observations of Mercury. 30. In the case of Mercury the factors of the coefficients of the equations were chosen large enough to admit of the deci- mals being dropped from the products without prejudice to the accuracy of the final result. This was done to facilitate the formation of the normal equations. For the same reason the factors were made so small that the absolute numerical values of the coefficients should generally not exceed 13. As this degree of precision is far short of that usually employed for correcting the elements of a planet, it may be well to set forth the considerations on which it is based. Let any equation of condition as actually used be ax-\- by + cz + . . . =n (a) Let the coefficients a, &, etc., be affected by the mean errors e, ', etc., so that the true equation should be . . . = n This true equation may be written in the form ax + by + . . . = nx e'y . . . (b) We may regard (b) as a rigorous equation, in which the error of the second member is increased by the quantity 30] MERIDIAN OBSERVATIONS OF MERCURY. 5 and the only effect upon tlie precision of the results will be that arising from this increased probable error. Let us esti- mate its magnitude. From an examination of the tables used in finding the coefficients I infer that the probable error of the coefficient of n Avas 1, and that of all the other coefficients 0.6. The mean value of the unknown quantities was gener- ally a small fraction of a second. We conclude, therefore, that the probable or mean value of the error ex fy i . . . would in any case be only a small fraction of a second. More- over, these errors would be purely accidental and not system- atic, since the intervals of time between the equations were generally so long that the coefficients for different equations came from different tables, so that no error from omitted deci- mals in any one equation would enter into the other equations. Now, in view of the necessary systematic errors which affect observations of the planets, there is no hope of approximating to this degree of accuracy in the second members of the equa- tions. Were the observations rigorously correct and the values of the unknown quantities finally determined affected by no error except that arising in this way, they would be many times more accurate than we can hope to make them. The errors might, in fact, be considered unimportant in the present state of astronomy. It has already been remarked that the scale of weights was so taken that the unit of weight should correspond approx- imately to a supposed mean error i 1".0 in the value of each absolute term of an equation of condition, so far as the error could be determined from the discordance of the original observations. The corresponding probable error would be dt 0".65. In the case of Mercury, however, modifications were made which prevents this mean error from corresponding to the unit of weight which would be found from the solutions in the usual way. In the first place, the absolute members were all multiplied by 10; in other words, the decimal point was dropped from tenths of seconds, and no further account taken of it. Secondly, in consequence of the probable error in the coefficients of the normal equations arising from the irnperfec- 60 MERCURY, VENUS, AND MARS. [30 tions of tlie decimals, the final values of these coefficients would be subject to probable errors ranging between 50 and 100 units. In consequence there would be no advantage in retaining the last figure in the normal equations, and it was dropped in all the subsequent solution and discussion of these equations. In dropping the last figure from the absolute term of the normal equations we may consider that we are merely drop- ping the tenths of seconds and that the units are once more expressed in seconds. Thus, considering only the effect of this operation, the unit of weight would correspond to a mean error of 1.0 in units of the absolute term. But in dropping off the last figure from the coefficients we practically reduce the scale of weights, considered as multipliers of the equa- tions, to one-tenth of their former value. On the other hand, in expressing the unknown quantities in terms of the correc- tions to the elements, we divide the multipliers by ten, so that effectively we multiplied the coefficients in the equations of condition, considering the unknown quantities to be defined as on page 56, by 10. Since these coefficients are of the second degree in the normal equations, it follows that the scale of weights has in effect been increased ten fold. Hence the unit of weight for the normal equations between the unknown quantities as finally solved will correspond to the mean error l = 1.0 X VI6 = 3.1 As the mean error is at best a rather indefinite quantity in a case like the present, we may consider its value as 4 units and even then as by no means rigorously determined. Up to the time of writing no attempt has been made to derive rigorously the weights of the unknown quantities from the solution, because in the cases of most of the uukowns such weights would be entirely illusory. The fact is that in solving so immense a mass of equations, we must expect systematic errors to vitiate many of the results. The observations of Mercury, especially of its Eight Ascension, are not made on .a uniform system 5 sometimes the limb is observed, sometimes the apparent center or the center of light. 30, 31] TRANSITS OF MERCURY. 61 An ideally perfect system of reduction would require us to reduce each separate observation with a semidiaineter corre- sponding to the personal equation of the observer. This being entirely impracticable, we must regard the reduction of the observer's semidiameter to that used in the reductions as a probable error. In fact, however, it will be of a systematic character, varying at each point of the relative orbit of Mercury, and going through a cycle of changes impossible to determine in a synodic period of the planet. It is impracti- cable to give even a full discussion of these errors; we shall, however, meet with a proof of their magnitude. Introduction of the equations derived from observed transits of Mercury. 31. The relations between the elements of Mercury and the Earth derived from this source are shown in my Discussion of Transits of Mercury (A. P., Vol. I, Part VI.) On page 447 are found expressions for those linear functions of the corrections to the elements which are determined by the November and May transits, respectively. With a slight change of notation to correspond with that of the present paper, these functions are as follows : V = 1.487 61 - 0.487 dx - 1.137 tie - 1.01 dl" -f 1.19 e"dn" + 1.58 de" W = 0.716 61 + 0.284 drt + 0.896 de - 0.97 61" - 1.11 e"dn" - 1.62 de" The values of V and W being derived from a series of transits extending from 1677 to the present time, enable us to deter- mine both these quantities at some epoch, and their secular variations. The values derived from the transits, together with their mean errors, are found on page 460 of the work in question. Omitting the doubtful factor fc, introduced on account of a possible variability of the Earth's axial rotation, which was not proved by the transits, the values of V and W were found to be as follows : V = 0".90 i 0".31 + ( - 2 // .63 0';.59) (T - 1820) W == + 0".84 0".25 -f (+ 1".84 i 0".60) (T - 1820) 62 MERCURY, VENUS, AND MARS. [31 The mean epoch for the transits is taken as 1820, to which the zero values correspond. The values for 1865.0, the mid epoch for the meridian observations, are, therefore, from the transits alone V = - 2".08 0".41 W = + 1".67 0".37 This, however, is only a first approximation to the quantities which should be introduced. Since the meridian observations help to determine the values of V and W, we should not regard the reductions to 1865.0 as final, but retain the results in the form (a). Another element which is determined from the observed transits of Mercury with greater precision than it can be from meridian observations is the longitude of the node of the orbit relatively to the Sun. In the paper quoted we have put F = (30 -61"} sin i and found from all the transits up to 1881, N = - 0".16 i 0".27 + (0".28 dL 0".62) (T - 1820) (b) The values of Y, W, and N, found from the discussion in question, give rise to six conditional equations, which become completely independent when we take as observed values the secular motions and the absolute values at the mid-epoch of observation. This mid-epoch is not the same for the May and November transits. But I have assumed that no serious error would be introduced by taking 1820.0 as the epoch for all three of the quantities, Y, W, and X. If we substitute for sin i 66 its value in terms of tfJ, etc., namely, Sin idB=- 0.6018 J + 0.796 sin JtfN + 0.721 de (c) and then for tf J, tfN, tff, their values in terms of the unknowns of the equations of condition, we shall have N = - 1.805 [J] + 2.394 [N] + 0.721 [*] - 0.122 [V] (d) 31] TRANSITS OF MERCURY. 63 Similar expressions will be found for the values of Y and W by substituting for the corrections to the elements the unknown quantities of the conditional equations, as already given. Taking 1820.0 as the mid epoch, we may regard the inde- pendent quantities given by the transits of Mercury to be the six following ones : Vo - 1.8 Y x ; W - 1.8 W,; N - V, ; W, 5 N, - Here Y , W , and N indicate values for 1865, the mid-epoch of the meridian observations; and Y 1? W 1? and Nj. the variations in 25 years. The six conditional equations thus found from the transits may be written Yo - 1.8 Yt = - 0".90 0".31 W - 1.8 W t = + 0".84 i 0".25 :N O - 1.8 N! = - 0".16 0".27 Y, = - 0".66 i 0".15 Wi = + 0".46 i 0".15 Ni = + /7 .07 /7 .15 Substituting for Y , YI, etc., their expressions as linear func- tions of the unknowns of the conditional equations, we find the following six equations, which are to be used as conditional equations additional to those given by the meridian observa- tions : 5.95 [1] - 4.87 [TT] - 3.41 [e] - 1.01 [l"\ + 0.71 [n"\ + 0.95 [e"] -1.8)6.95[Z]! - 4.87 [ir]i 3.41 [e],- 1.01 [l"]i+ 0.71 [n"^ + 0.95[e // ] 1 J = -O^O Weight = 250 2.86 [1] + 2.84 [it] + 2.69 [e] - 0.97 [I 11 ] - 0.67 [n"\ - 0.97 [e"} -1.8 {2.86 [I], + 2.84 [w-J! + 2.69 (e}, - 0.97 [l"^ - 0.67 [n"}, -C.97^ 7 '],} = + 7/ .84 Weight = 300 - 1.8 [J] + 2.4 [N] + 0.7 [f] - 0.12 [I"} - 1.8 { - 1.8 [ J] x + 2.4 [NJi + 0.7 [f ]x - 0.12 [/ // ] l } = - 0".16 Weight = 400 64 MERCURY, VENUS, AND MARS. [31 5.95 [1], - 4.87 [TT]! - 3.41 [e], - 1.01 [l"^ + 0.71 [*"]i+ 0.95 = - 0".66 Weight = 700 2.86 [l]t + 2.84 [TT]! + 2.69 [e]i - 0.97 [Z"]i - 0.67 [TT"], - 0.97 [a' 7 ]! = + 0".46 Weight = 700 -1.8 [J], + 2.4 [N] { + 0.7 [e], - 0.12 [I"}, = + 0".07 Weight = 1,600 The weights assigned to these several equations have been determined by the following considerations: We have already found that in the equations of condition from the meridian observations as finally reduced, the scale of weights has so come out as to show a practical mean error for weight unity of about 4". Were this error purely accidental, the weights of the conditional equations derived from the transits would be determined in the same way, from the mean errors assigned to them. But, as a matter of fact, the exist- ence of systematic errors in the meridian observations is shown, as will be subsequently explained, by the large value found for the fictitious quantity 6r 2 . Since observations of transits are made at the point of the relative orbits of Mercury and the Earth, near which meridian observations are rarely available, and are of a higher order of accuracy than meridian observations, it follows from the theory of probabilities that we should assign a larger relative weight to the observations of the transits. How much larger does not admit of being determined with numerical precision. Actually I have taken the weights as if the mean error corresponding to weight unity were between 5 and 6. In the case of the motion of the node a still larger weight has been assigned to the secular variation, from the belief that the accuracy of the determina- tion from transits relative to meridian observations is in this case of a yet higher order of magnitude than in the case of 31, 32] SOLUTION OF EQUATIONS FOR MERCURY. 65 the other elements. Whether this belief is justified or not must be left to the decision of the future astronomer. The first three of the preceding six conditional equations may be treated in a way similar to that adopted for the meridian observations. They express what is supposed to be equivalent to observations of the three quantities V, W, and N in 1820, when r 1.8. Hence, from the partial normals in the six principal unknowns, [e], [>]... [>"], the com- plete normals may be formed by multiplication by r and i* (r = 1.8) in the way set forth in 25. Solutions of the equations for Mercury. 32. In the case of Mercury and Venus, it is desirable to know to what extent the results of the transits diverge from those of the meridian observations. Hence, as already remarked, two solutions of the equations were made, termed A and B. Solution A is that derived from the meridian observations alone. Solution B is that of the normal equations formed from both the meridian observations and the transits. The results of the solutions in the case of Mercury are shown in the following tables. The relation of the unknown quan- tities given in the first columns, A and B, to the corrections of the elements has been shown in a preceding section ( 27). The upper half of the table shows the corrections to the elements; the lower half those of the secular variations. It will be seen that all the values, with a single exception, come out less than a unit. In stating the corrections to the elements, it must be remembered that, owing to the proximity of Mercury to the Sun, the errors of geocentric place are much less than those of the heliocentric elements, so that an error in the latter indicates a proportionally smaller error in the actual observations. For the same reason we must expect a less degree of precision in the elements as finally derived than in the case of the other planets. 5690 N ALM 5 66 MERCURY, VENUS, AND MARS MERCURY. Results of solutions of the normal equations. [32, 33 Unknowns. EM Corrections of elements. Symbol. A. B. Symbol. A. B. ["] o. 1478 o. 1207 O. I 6 m : m o. 0148 O. OI2I ' / o. 1342 0.0752 4- ii o. 537 0.301 : j o. 2436 o. 2299 3- 6] o. 731 o. 690 ;N o. 0227 0. 0201 3- SinJdN -o. 068 o. 061 t -fo. 2074 +o. 2194 i. 6 4-o. 207 4-0.219 e O. I2O2 4-0. 4094 3- 6e o. 361 41.228 ' TT 4-0. 5209 4-0. 2688 10. 6 7T +5- 209 4-2. 689 ' e" -f o. 0669 4-0. 8397 0.6 (5 e f/ -f-o. 040 -f o. 504 V' o. 2248 o. 7027 0.6 e" 6 TT // o. 135 o. 422 >// 4-1. 1240 4-1.0566 2. 6r 4-2. 248 4-2. 113 6 o. 2310 o. 2556 I. 6 0.231 o. 256 /" -0-0354 o. 0897 I. 61" -o. 035 o. 090 a 4-0. 4803 4-0. 4930 I. a -fo. 480 +o- 493 / J o. 2060 o. 0114 o. 1209 -f-o. 0636 1 6. 12. D t # 3. 296 o. 137 i. 935 -f o. 764 N; 4~o. looo 4~o. 0930 12. SInjDt o. 82 ;/?= + 0.0592 Sv - 0.9982 sin * dd I have discussed very fully the observations of the transits of 1761 and 1769 in Astronomical Papers, Vol. n. The final results which I shall use are found on page 404 of that volume. Here I have put. x, correction to A L ; y, correction to /?, the Sun's latitude being supposed to require no correction. The values of x and y for 1769 are distinguished by an accent. I have also represented by z 2 and 2 3 the corrections to the dif- ference of the semidiameters of the Sun and planet, for the respective internal contacts, to which may be added the. un- known but probably nearly constant quantity due to personal error in estimating the time of contact. From their very nature these quantities do not admit of accurate determination, and must therefore be eliminated from the equations. From the observations of internal contact are derived the following four equations : 1761 II; ,87a? + .50 y + z 2 = - 0".07 HI ; + .68 + .73 + 2 3 = - 0".06 1769 II; - .64 a? 7 - .11 y' + z 2 = - 0".27 III; + .84 - .55 + 2 3 = + 0".02 We have here more unknown quantities than equations, so that it is not practicable to determine them all separately. What I have done has been first to assume ^ = 2 3 . This pre- supposes that the distance of centers at the estimated appa- 72 MERCURY, VENUS, AND MARS. [36 rent contact at egress is, in the general mean, the same as at ingress. The result of any error in this hypothesis will be almost completely eliminated from the mean latitude at the two transits, but not from the longitude. Still, the values of x and y can not be separately determined; I have therefore so combined the equations as to obtain mean values of x and y for the two contacts, assuming that this would be the result of supposing these quantities to have the same, values at both epochs. Calling these values x" and y", we have by addition and subtraction, supposing z 2 = 23, - 0.39 a?" + 2.55 y" = 0".12 3.03 #"4- 0.45 y" = 0".30 We thus have* x" = + 0".09 y 1 ' = + 0".06 These corrections are not applicable to the coordinates from LEVERRTER'S . tables as they stand, but to those quantities as corrected by the following amounts : A\ = 4- 0".25 Afi= 4-2".00 * In a second approximation to these quantities, which may be made after the correction to the enteimial motion of the node is determined, we should put, on account of this correction, The solution would then give I have carried through a more careful approximation in a subsequent chapter. 36] EQUATIONS OF CONDITION FROM TRANSITS OF YENUS. 73 We thus find, for the corrections to LEVERRIER'S tables at the epoch 1765.5, d A - L = + 0".09 + 0".25 = + 0".34 3 = _ 0".06 2".00 .= + 1".94 and hence dv = + 0".22 + 0".998 L sin i 6 S = + 1".95 + // .059 6L A still farther modification is required to the tabular longi- tude on account of the correction to the mass of the Earth used by LEVERRIER, and hence to the periodic perturbations in longitude. This correction is + 0".20. We thus have for the correction to the orbit longitude of Venus dv = + 0".02 4- 0".998 6 L For the results of the transits of 1874 and 1882 I have depended entirely on the heliometer measures and photo- graphs made by the German and American expeditions, respectively. The definitive results of the German observa- tions, as worked up by Dr. AUWERS, are found in Vol. V of the German Keports on the Transits.* The American photo- graphic measures of 1874 have not been officially worked up and published, but a preliminary investigation from the data contained in the published measures was made by D. P. TODD, and published in the American Journal of Science, Vol. 21, 1881, page 491. The measures of 1882 have been definitively worked up by HARKNESS, but only the results published. They are found in the report of the Superintendent of the TJ. S. Naval Observatory for the year 1890. The corrections to the geocentric Eight Ascension and Declination of Venus relative to the Sun thus derived are * Die Venus-durchgiinge 1874 und 1882 Bericht uber die Deutsclien Beobachtungen Fuufter Band, Berlin, 1893. 74 MERCURY, VENUS, AND MARS. [36 given in the following table. In taking the mean the weights are not strictly those which would result from the probable errors as assigned, but, in accordance with a general princi- ple, independent results have received a weight more near to equality than would be indicated by the mean errors. 1874: German, 6 E. A. = + 4.77 0.28 American, . . . + 4.14 0.30 Adopted, . . . +4.44 German, 6 Dec. = + 2.28 + 0.10 American, . . . -f 2.50 0.30 Adopted, . . . +2.34 1882: German, 6 E. A. = + 9.03 + 0.12 American, . . . + 9.10 + 0.08 Adopted, . . . +9.07 German, d Dec. = + 2.02 i 0.06 American, . . . +2.02 + 0.08 Adopted, . . . +2.02 We change these results successively to geocentric longi- tude and latitude, heliocentric longitude and latitude, and orbital longitude and latitude. The results of these several changes are as follow: Corr. in geoc. long. Corr. in lat. 1874. + 3''.853 + 2 .724 1882. + 8".077 + 2. 971 Corr. in hel. long. Corr. in hel. lat. -1 .415 + 1 .001 -2 .965 + 1 .091 Corr. in orbital long. Value of sin i 6 -1 .35 -1 .08 -2 .90 -1 .26 37] EQUATIONS FROM TRANSITS OF VENUS. 75 Equations from transits of Venus. 37. The corrections to the heliocentric positions of Yenus and the Earth, as thus found, are now to be expressed in terms of corrections to the elements. The results of this expression are shown in the following equations: Equations given by the corrections to the orbital longitude. I. Epoch, 1765.5; T= 3.90 ; weight = 200 0.992 61 + 1.17 eSn + 1.62 de - 0.976 61" 1.81 e"dn"- 0.85 de" = +0".02 O."15 II. Epoch, 1874.9 ; r = + 0.48; weight = 400 - 0".88/* + 1.009 61 - 1.223 edn - 1.596 6e - 1.030 61" $*" + 0.817 6e" = - 1".35 0".08 III. Epoch, 1882.9; T = + 0.80; weight = 800 0".60,w+ 1.008 61 - 1.146 edn - 1.651 6e - 1.028 ] o. 0708 o. 0834 7- <5 m : m o. 496 0.^584 ' / ' o. 1435 o. 1501 5- 61 -0.718 0-751 " J ' +o. 1156 +o. 1340 6. 6 T +o. 694 +o. 804 N; 4-o. 0164 -j-o. 0106 7- sinJrfN +o. 115 +o. 074 e +o. 0941 -j-o. 1003 3- fit +o. 282 -|-o. 301 7T ] 4-o. 0628 -j-o. 0764 3- e6K 4-o. 1 88 +o. 229 e 4-o. 0246 +0.0271 4- 6e -j-o. 098 4-o. 1 08 ~ e "\ 4-o. 0336 4-o. 0318 2-5 5e" 4-o. 084 +o. 080 ~ir"' o. 0274 0.0212 2. e"tv" o. 055 o. 042 ' a +o. 4742 +o. 4642 I. a +o. 474 +o. 464 i 6 '' o. 0383 -o. 0375 5- 6 o. 192 o. 1 88 */// o. 0768 o. 0743 4- 61" o. 307 -o. 297 ' / ' o. 1846 -o. 1983 20. D t d/ 3. 692 3. 966 :*: +o. 0970 o. 0561 +o. 1088 o. 0594 2 4 . 28. DtJ sinJD t N +2. 328 I-57I +2.611 -1.663 e +o. 1472 +o. 1644 12. Dt e + 1.766 + r -973 TT +0.0555 4-o. 0698 12. ^D t 7r +o. 666 +o. 838 -(-o. 0182 -j-O. O2O2 1 6. Dte +o. 291 +o. 323 "e"\ 4-o. 0283 +0.0317 10. Dt^ x/ +o. 283 7T // +o. 0399 +o. 0506 8. e" Dt TT // +o. 3 J 9 4-o. 405 a o. 0820 -o. 0347 4- D t a o. 328 o. 139 ! <* s 0. 0020 O. OOO2 20. Dt d o. 040 o. 004 L/// - --o. 0562 o. 0662 1 6. EM/" o. 899 i. 059 Mean epoch of correction, 1863.0 Comparison of transits of Venus with meridian observations. 39. To show to what extent the results of the meridian observations differ from those of the observed transits over the Sun, we form the values of the absolute terms of the equations of condition, 37, first by substituting the values A of the corrections, and then the values B. We thus have : 39, 40] EQUATIONS FROX TRANSITS OF VENUS. 77 Residuals in orbital longitude. 1765.5. 1874.9. 1882.0. (a) From meridian obs. alone . - 0".07 - 1".36 - 2".54 ( ft) From combined solution . + 0".04 - 1".43 - 2".78 (y) From transits alone . . . + 0".02 - I". 35 - 2".90 Discordance, (y)-(tf) . . +0".09 + 0".01 - 0".36 Discordance, (y)- (ft) . . - 0".04 + 0.08 -0".12 Residuals in orbital latitude. 1765.5. 1874.9. 1882.9. (a) From meridian obs. alone . + 1".92 - 0".77 - 0".96 (ft) From combined solution . + 2".06 - 0".91 - 1".12 (y) From transits alone . . . + 1".95 - 1".08 - 1".26 Discordance, (y)- () + 0".03 - 0".31 - 0".30 Discordance, (y) (ft) . . - 0".ll - 0".17 - 0".14 It will be seen that the combined solution represents the observations of the transits much better here than in the case of Mercury. Solution of the equations for Mars. 40. As the formation of the normal equations for Mars was approaching its end, a singular discordance among the resid- uals of the partial normal equations for different periods was noticed. On tracing the matter out it appeared that while the correction of the geocentric longitude of LEVERRIER'S tables in 1845 and again in 1892 was quite small, the correction in 1862 was considerable. Now there is an inequality of long period, about forty years, in the mean motion of Mars, depend- ing on the action of the Earth, and having for its argument 15# 7 Sg. This coefficient is of the seventh order in the eccen- tricities, and the terms of the ninth or even of the eleventh order might be sensible in a development in powers of the eccentricities and sines or cosines of multiples of the mean longitudes. The conclusion which I reached was that the the- oretical value of this coefficient was not determined with suffi- cient precision. As the work of solving the equations could not wait for a new determination and a new formation of the absolute terms of the normal equations, it was decided to make an approximate empirical correction to the theory. This was used to correct the absolute terms of the partial normal equa- 78 MERCURY, VENUS, AND MARS. tions for each period, and the solution was then proceeded with. The chances seein to be that by this process the inju- rious effect of the error upon the elements derived from the equations would be inconsiderable; this is, however, a point on which it is impossible to speak with certainty. It is the intention of the writer to recompute the doubtful terms of the perturbations, and, if possible, reconstruct the absolute terms of the normal equations in accordance with the corrected theory. Meanwhile, the present work necessarily rests on the imperfect theory with the approximate empirical corrections, which are as follow : M = 0".30 cos (150' - 8-7 - 223) edn = 0".15 cos (150' 80) As the elements of Mars are derived wholly from meridian observations, only one set of equations of condition was formed. The results of the solution are shown in the following table : MARS. Unknowns. Factors. Corrections of elements. Symbol. Value. Symbol. Value. [ tn/ ~\ . 02278 0-3 6 m : m -o. 007 1 ] . 44854 2. 61 -o. 897 N; + 05479 +.06724 2-5 2-5 SinJJN +o. 137 +o. 1 68 e + .43803 V 6 e +o. 626 TT . 05056 1^-Q 6w o. 722 e +.07474 4- 6s +o. 299 e"\ . 49898 2. be" o. 998 V /x " .42409 2. e' f 6 K" o. 848 a + 18545 5- a +o. 927 ; 6 ' . 04536 5- 6 o. 227 ///' + .05786 3- 61" +o. 174 " / " +. 16605 8. D t d/ +1.328 ; JT + 13408 . 02263 10. 10. D t J SinJDtN +1.341 o. 226 e . 03180 V 1 D t '\ +. 13111 12. D t rf/" +1.573 41] REFERENCE TO THE ECLIPTIC. 79 Reference to the ecliptic. 41. In all the preceding determinations the planes of the orbits are referred to the plane of the Earth's equator, or, to speak more exactly, to a plane through the Sun parallel to the Earth's equator. As in astronomical practice the ecliptic is taken as the fundamental plane, it is necessary to investigate the reduction of the elements from one plane to the other. Let us consider the spherical triangle formed on the celestial sphere by the plane of the orbit, the plane of the ecliptic, and the plane of the Earth's equator. For the sides and opposite angles of this triangle we have Sides: N 6 y Opposite angles : i 180 J e When equatorial coordinates are used, the position of the planet is considered as a function of the three quantities N; J; e (a) When ecliptic coordinates are used, the three corresponding quantities are 0; I-, e (b) Taking the set of quantities (a] as the fundamental parts of the triangle, and expressing the corrections of the other parts as functions of them, we have 6 i = + cos rp6J + sin ip sin JtfN cos Ode sin i d 9 = sin fidJ + cos >/> sin J6N + cos * sin Ode Taking (b) as the fundamental parts, we have for the correc- tions to N and J d J = cos fi6i sin ip sin i$6 + cos sin J#N= sin 6i + cos sin idd cos J sin Ntfs The numerical values assigned to the coefficients in these equations are those corresponding to the mean epoch 1850. The fact that they change somewhat in the course of a hundred years has not been taken account of. The future astronomer will meet with a real difficulty in that the corrections to a 80 MERCURY, VENUS, AND MARS. [41 set of elements at one epoch do not accurately correspond to similar corrections at another epoch. It is impossible to do away rigorously with the difficulty thus arising, except by introducing a more general system of elements than elliptic ones. The error is, happily, not important in the present state of astronomy. The equations in question for the three planets are as follow: Mercury. di = + .799 (5 J + .602 sin J fi X .688 fo sin idO = - .602 6J + .799 sin J d N + .721 fa Venus. di = + .373 6 J + .928 sin J N - .255 fa sin idS .928 d J + .373 sin J d N 4- .967 de Mars. di = .703 6 J + .712 sin J tf N - .664 6s sin id 6 = - .712 3J + .703 sin J 3X + .747 fa For the inverse relations we have Mercury. 3J= .799 (ft, x) as a basis, we may say that the modulus of precision, ft, is nearly always in practice an uncertain quan- tity. Let us then put hi, lit, h 3 , . . . for the possible values of ft, and for the several probabilities that h has these respective values. Then the probability function will become > I I CO t*- TH CO TH O rH IS 00 b- CM rH C5 t CO CO rH C2 - ^- ^ ^ r< * I S S CO ^> I I + 8 g 53 " S I + + I z 9 1O t~ rH I I + ^ CO S ^ 46] NORMAL EQUATIONS FOR SECULAR VARIATIONS. CO O O O CO O b- JO CO O b- H rH rH CM rH rH I + I I + I 1 1 1 + b'38 CO + + cp CQ CO + + i + + + a ^ ^^ co rH O5 tr^" tr** CO t- Tfl b- CO O O C5 SO rH IO O lO rH b- rH C5 b- rH rH + + I I + rH CO CM Ci 8 a ^ 1 o 3 I tJO ti R rH C^ CO 2L S CO CM CO Ci > 6f e "6K" a + discussion, as the well-known determinations of their masses are amply accurate for all our present purposes. Mass of Jupiter. 49. One of the works connected with the present subject has been the determination of the mass of Jupiter from the motions of (33), Polyhymnia. My work on this subject has not yet been printed in full, but I have given in Astronomische Nachrichten y No. 3249 (Bd. 136, S. 130), a brief summary of the results. The mass of Jupiter has been derived not only from the motions of Polyhymnia, but from such other sources as seemed best adapted to give a reliable -result. The following table, tran- scribed from the publication in question, shows the separate results and the conclusions finally reached : Reciprocal of mass of Jupiter from wt All observations of the satellites, 1047.82 1 Action on FATE'S comet (MOL.LER), 1047.79 1 Action on Themis (KRUEGKER), 1047.54 5 Action on Saturn (HiLL), 1047.38 7 Action on Polyhymnia, 1047.34 20 Action on WINNECKE'S comet (v. HAERDTL), 1047.17 10 1047.35 in. e. 0.065 5690 N ALM 7 97 98 MASS OF JUPITER. [49 It will be seen that the result from observations of the- satel- lites has been assigned a very small weight. This course has been indicated by the circumstances. Other conditions being equal, the greater the mass of a planet the less the propor- tionate precision with which that mass can be determined by observations on the satellites. In any case, if the measures of the distances between the satellites and the primary are' in error by a small fraction, or, of their whole amount, then the error of the mass will be in error by the fraction 3 a of its amount. For reasons founded on the construction and use of the heliometer, I doubt whether the absolute measures made with those forms of that instrument which have been used in determining the mass of Jupiter can be relied upon within their three-thousandth part. If so, the determination of the mass of the planet itself would be doubtful by its thousandth part in each separate case. The chance of personal equation between transits of the satellites and the planet vitiates in the same way the results from observed transits of the planet and satellites. Notwithstanding the great refinement of the dis- cussion by KEMPF of observations made at Potsdam, and the care with which he, SCHUR, and others have determined the mass of Jupiter by a discussion of all the^observations of the satellites, I can not conceive that the probable error of any possible result they could derive would be less than 0.3 or 0.4 in the denominator. In this connection the discordances between" the mass of Saturn, found by Prof. HALL and by other observers from observations of the satellites, are worthy of consideration. They lead us to suspect that perhaps it is through good for- tune rather than by virtue of their absolute reliability that determinations of the mass of Jupiter from observations of the satellites have agreed so well. As to the weights assigned to the other results, only the last needs especial mention. The probable error assigned by v. HAERDTL to his result is very much smaller than that which I find for the mean of all the results. But, as remarked in the paper in question, it has received a smaller relative weight than that corresponding to its assigned probable error, because of distrust on my part whether observations on a comet can 49, 50, 51] MASSES OF THE EARTH, MARS, AND JUPITER. 99 be considered as having always been made 011 the center of gravity of a well-defined mass, moving as if that center were a material point subject to the gravitation of the Sun and planets. This distrust seems to be amply justified by our general experience of the failure of comets to move in exact accordance with their ephemerides. I propose to accept the value thus found, Mass of Jupiter = 1 4- 1047.35 as the definitive one to be used in the planetary theories. Mass of Mars. 50. In consequence of the minuteness of the mass of Mars, measures of its satellites, especially the outer one, afford a value of its mass much better than can be derived by its action on the planets. When nearest the earth, the major axis of the orbit of the outer satellite subtends an angle of 70". I can not think that the systematic error to be feared in the best measures, such as those made by Prof. HALL, can be as great as half a second. It therefore Appears to me that the mean error in adopting Prof. HALL'S value of the mass does not exceed its fiftieth part. This is a degree of precision much higher than that of any determination through the action of Mars on another planet. Prof. HALL'S measures of 1892 show a minute increase of the mean distance given by his work of 1877. The result is v>" = + 0.014 These observations, however, were made when the position of the orbit of the satellite was unfavorable to an exact deter- mination of the elements of motion. I have adhered to the original value in the work of the present chapter. Mass of the Earth. 51. I have already pointed out the difficulty in the way of determining the mass of the Earth from its action on the other planets. On tbe other hand, the solar parallax has, in recent years, been determined in various ways with such precision that the mass of the Earth to be used in the plan- 100 MASS OF THE EARTH. [51 etary theories can best be derived from it. The theory of the relation between the mass of the Earth and its distance from the Sun, as given by observations of the seconds pendulum and the length of the sidereal year, is one of the best estab- lished results of celestial mechanics. It is, in effect, the principle on which the lunar theory is constructed. In this theory the disturbing action of the Sun is necessarily a func- tion of the ratio of the mass of the Sun to that of the Earth. But in the accepted theory this ratio is eliminated through the ratio of the lunar month to the sidereal year. From the well-established ratio between the distance of the Moon and the length of the seconds pendulum, the ratio of the masses of the Sun and Earth come out of this theory with great precision. It need not be developed here; it will suffice to give the numerical result, which is that between the ratio M of the mass of the Sun to that of the Earth and the mean equatorial horizontal parallax of the Sun in seconds of arc there exists the relation 7r 3 M = [8.35493] I have derived seven values of the solar parallax by different methods, of which the following are the preliminary results : wt. GILL'S observations of Mars, 1877, 8.780 .020 1 Contact observations, transits of Venus. 8.704 i .018 1 Aberration and velocity of light, 8.798 i .005 16 Parallactic equation of the Moon, 8.799 .007 5 Measures of small planets on GILL'S plan, 8.807 i .007 8 LEVERRIER'S method, 8.818 .030 0.5 Measures of Venus from Sun's center, 8.857 i .022 1 Mean result, n = 8".802 i 0".005 I have provisionally taken this mean as the most probable value of the solar parallax derived from all sources except the mass of the Earth. The above relation then gives M = 332 040 51,52] MASS OF VENUS. 101 Taking for the mass of the Moon 1 4- 81.52, we have for the ratio of the combined masses of the Earth and Moon to the mass of the Sun m" 328 016 a result of which the probable error may be regarded as some- thing more than a thousandth part of its whole amount. Mass of Venus. 52. The mass of Venus adopted in the provisional theory, to which LEVERRIER'S tables were reduced, was .000 002 4885 = 1 +- 401847, which is that of LEVERRIER'S tables of Mer- cury. In the preceding discussions the following three factors of correction to this mass have been found : From observations of the Sun . . .0118 =t .0034 From observations of Mercury . . .0121 =t .0050 From observations of Mars ... .0076 =t . ( ? ) Mean - .0119 .0028 The mean error assigned to the result from observations of the Sun may be regarded as real, because the result is the mean of a great number of completely independent determina- tions, among which no common error is either a priori prob- able or shown by the discordance of the results. In the case of Mercury, however, as already remarked, the effect of systematic errors is such that, although they are almost com- pletely eliminated from the result, the mean error computed in the usual way would be misleading. The weight assigned is therefore largely a matter of judgment. The fact that it was necessary to introduce an empirical correction, with a period of about forty years, into the mean longitude of Mars, vitiates the determination of the mass of Venus from its action on that planet, because one of the prin- cipal terms in the action of Venus on Mars has a period which does not differ from forty years enough to make the determi- nation of the mass independent of this empirical correction. I have therefore assigned no weight to the result. We thus 102 MASS OF MERCURY. [52,53 have for the mass of Venus, as derived from the periodic per- turbations of Mercury and the Earth produced by its action. m' = 1 -^ 406 690 i 1140 Mass of Mercury. 53. The mass of Mercury which I have heretofore adopted, 1 -=- 7 500 000, was rather a result of general estimate than of exact computation. The fact is that the determinations of this mass have been so discordant, and varied so much with the method of discussion adopted, that it is scarcely possible to fix upon any definite number as expressive of the mass. An examination of LEVERRIER'S tables of Venus shows that with the mass of Mercury there adopted (1:3 000 000) Mercury frequently produces a perturbation of more than one second in the heliocentric longitude of Venus. When the latter is near inferior conjunction, the actual perturbation will be more than doubled in the geocentric place, so that the latter might not infrequently be changed by 1", even if the mass of Mer- cury be less than one-half LEVERRIER'S value. It was there- fore to be expected that a fairly reliable value of the mass of Mercury would be obtained from the periodic perturbations of Venus. Eeferriug to 27, it will be seen that the indeterminate mass of Mercury appears in the equations in the form 1+7;, 3000000 From the solution B, 38, the value of /* comes out ^ = _ 0.0834 corresponding to a mass of Mercury of 1 : 7 210 000. But in a subsequent solution of the equations, when the secular vari- ations are determined from theory and substituted in the normal equation for /v, we find ,u = - 0.0889 which gives m = 1 -4- 7 943 000 The work of the present chapter is based on the former value. 53] MASS OF MERCURY. 103 A consideration of the probable error of this result is impor- tant. The fortuitous errors which mostly affect it are of the class which I have termed semi- systematic. Under this term I include that large class of errors which, extending through or injuriously affecting a limited series of observations, cause the probable error of a result to be larger than that given by the solution of the equations, but which, nevertheless, like purely accidental ones, would be eliminated from the mean result of an infinite series of observations. To this class belong the errors arising from personal equation in observing the limb of Venus, or, what is the same thing, a difference between the practical semidiameter corresponding to the observer and that adopted in the reductions. We may suppose that, during a period of several days, when Venus is not far from inferior conjunction, its geocentric position is affected by a perturba- tion produced by Mercury. Through the error alluded to, all the observations made by any one observer, and in fact all that are made anywhere, may be affected by a certain con- stant error in Right Ascension. Near another inferior con- junction the same state of things may be repeated, with the perturbation in the opposite direction. If, now, tne observa- tions were made by the same observer, and under the same circumstances, the personal error would be eliminated from the mean of these two results so far as the mass of Mercury is concerned. But very frequently different observers will have made the observations under the two circumstances, and dif- ferent conditions will have prevailed. Thus, it is only through the general law of averages that we can expect the effect of these fortuitous but systematic errors to be completely elim- inated. That they would be eliminated in the long run is evident from the fact that there can be no permanent rela- tion between the personal equations of the observers and the changes in the action of Mercury upon Venus. Moreover, Venus has been observed with a fair degree of accuracy through more than half a century, and it seems reasonable to suppose that during that time the errors in question would nearly disappear. It is clear from these considerations that the probable error derived from the solution of the equations would be 104 MASS OF MERCURY. [53 entirely misleading. But a probable error which ought to be reliable can be obtained by a process similar to that which I have adopted elsewhere in this paper, namely, dividing up the materials into periods, and determining the probable error from the discordances among the results of the several periods. This probable error will be reliable, because there is no reason why the same error should affect the mass of Mercury through any two periods. I therefore take the partial normal equa- tions in n derived from Eight Ascensions during the several periods, substitute in them the values of the unknown quanti- ties found from solution B, /* excepted, and thus form six- teen partial normal equations in /i. These equations may be changed into the corresponding equations of condition, of weight unity, by dividing each by the square root of the coefficient of the unknown quantity. The residuals then left when the definitive value of the unknown quantity is substi- tuted will be those from whose discordance the probable error may be inferred. The partial normal equations thus found from the Eight Ascensions are as follow: 1750-'62. 44; i= - 38 1830-'40. 5649; i=- 831 1765-'74. 1265 -165 1840-'49. 2913 - 18 1775-'86. 15 - 5 1849->56. 2238 - 49 1787-'96. 209 4- 53 1857->64. 4506 - 129 1796->06. 345 4- 19 1865-'71. 7736 - 265 1806-'14. 439 4-135 1871-79. 7062 761 1814->19. 942 + 2 1879-'86. 4958 - 407 1820-'30. 1786 330 1885-'92. 9561 -1306 Sum: 49668/<= -4095 ^ = _ 0.0824 i .019 The difference between this value of yw, which is obtained only to find the probable error, and that formerly found, arises principally from the omission of the declination equations. The mean error corresponding to weight unity comes out ft = 4".2 53] MASS OF MERCURY. 105 which, as anticipated, is much larger than that which would be given by the discordance of the original observations. This does not mean that the original observations are affected by any such mean error as 4".2, but that the discordances between the 16 values of /* are as great as we should expect them to be if the original observations were absolutely free from systematic error, but affected by purely accidental errors of this mean amount. The results of the solution for the mass of Mercury may be expressed in the form . 10.32 d ~ ' 7 210 000 ' 7 043 000 In all researches which have been made on the motion of ENCKE'S comet by ENCKE, VON ASTEN, and BACKLUND, the determination of this mass has been kept in view. The results are, however, so discordant that, as already remarked, scarcely any definitive result can be derived from them. To this statement there is, however, one apparent exeeption. In an appendix to his very careful and elaborate discussion of WINNECKE'S comet, VON HAERDTL has derived the value of the mass of Mercury from all the return of ENCKE'S comet as worked up by VON ASTEN and BACKLTJND.* The only inter- pretation which I can put upon his result is this : If we regard the acceleration of the comet, which it is supposed results from all the observations made upon it, as non-existent, the following two masses of Mercury are derivable from the obser- vations : 1819-1868, w = 1 4- 5 648 600 i 2000 1871-1885, m = 1 +- 5 669 700 600 000 He also finds, from the motion of WINNECKE'S comet, m = 1 4- 5 012 842 697 863 * Denkschriften der Kaiserlichen Akadeinie der Wissenschaften, Vol. 56, p. 172-175. Vienna, 1889. 106 MASS OF MERCURY. (53,54 and from four equations of LEVERRIER 1 4- 5 514 700 100 000 The consistency of these results seems to me entirely beyond what the observations are capable of giving, and I hesitate to ascribe great weight to them. Moreover, the result implicitly contained in these numbers, that the supposed secular accel- eration of the comet disappears when we attribute the pre- ceding mass to Mercury, merits farther inquiry. The probable density of the planet may form a basis for at least a rude estimate of its probable mass. The fact that the Earth, Yenus, and Mars have densities not very different from each other, while that of the Moon is 0.6 the density of the Earth, leads us to suppose that Mercury, being nearest to the Moon in mass, has probably a slightly greater density. Its diameter at distance unity has been repeatedly measured and found to be 6".6, or, roughly speaking, three-eighths that of the Earth. Were its. density 0.7, its mass would therefore be about 1 : 9,000,000. In view of the fact that the measured diameter is probably somewhat too small, these consider- ations lead us to conclude that the mass is probably between 1:6,000,000 and 1:9,000,000. As the value of the mass to be used in investigating the secular variations, I have adopted v = 0.08 1 08 Mass of Mercury = 7 500 000 Secular variations resulting from theory. 54. In the Astronomical Papers, Vol. V, Part IV, were com- puted the secular variations of the elements of the orbits in question using, as the basis of the work, the values of the THEORETICAL SECULAR VARIATIONS. 107 54] masses whose reciprocals are found in the column A below. In column B are cited the masses which I have decided upon. A B Original Adpofced reciprocal reciprocal of mass. of mass. V Mercury, 7 500 000 6 944 444 + .080 Venus, 410 000 406 750 + .0080 Earth + Moon, 327 000 328 000 -.00305 Mars, 3 093 500 3093500 Jupiter, 1047.88 1047.35 + .00050 Saturn, 3501.6 Uranus, 22756 Neptune, 19 540 In the case of the Earth we have to add the secular varia- tion of the perihelion produced by the non-sphericity of the system Earth + Moon. For the principal term I have found, D t e" d n" = + 0".129 The resulting values of the secular variations, expressed as functions of v, v 1 , v"j v 1 ", are given in the following section: Theoretical secular variations for 1850. Mercury. // // // // // // D t e = + 4.22 +0.00i/+ 2.8 K' + l.lF // -0.1?/ / " = + 4.24 6D t 7Ti =+109.36+0.00 +56.8 +18.8 +0.5 = + 109.76 D t ?: =+ 6.76 -0.04 - 0.6 - 1.4 +0.0 =+ 6.76 sin i Dt ; = 92.12 0.33 -49.3 -12.2 1.2 =92.50 Venus. D t e =- 9.58 1.30^+ 0.0^'- 4.9*/ // 0.2v /// =- 9.67 eDtTTt =+ 0.39-0.81+0.0 -3.9 +0.5 =+ 0.34 D t i =+ 3.43+0.76 + 0.0 + 0.0 -0.3 =+ 3.49 sin iT> t # =-105.92 +0.26 -29.2 -43.2 -1.2 =-106.00 108 THEORETICAL SECULAR VARIATIONS [54 Earth. // // // // // T> t e = - 8.57 -0.12F+ 1.3^' -1.6i'"' = - 8.57 eT> t 7t = + 19.36-0.18 + 5.8 +1.6 =+ 19.39 D t * =- 46.65-0.21 -28.3 -0.7 ==-46.89 Mars. // // // // // // D t e =+ 18.71 +0.03T/4- O.lv'4- 2.1v 7/ =4- 18.71 eDtTf! = + 148.824-0.06 + 4.6 +21.4 = + 148.80 D t i =- 2.34-0.04+12.0 +0.0 +O.OF x// =- 2.25 intD t = -f 0.42 Of the last two equations it may be remarked that the first is that given by Prof. HALL'S original mass of 1877, while the last is derived by Dr. HARSHMAN from HALL'S observations of the outer satellite made during the opposition of 1892. 64] CORRECTION OF MASSES. 123 When we add to the normal equations already formed the products of these last equations by the factors of the unknown quantities, the system of normal equations is as follows : 8266 # - 1563 v 1 - 4991 r" + 140 r"' =+134 -1563 +230831 + 88556 +3455 = +374 -4991 + 88556 +122462 +3750 = -1446 + 140 + 3455 + 3750 +3801 == -26 The solution of these equations gives the following values of the unknown quantities : x = + 0.0071 i .0120 v = + 0.071 i .120 v 1 = + 0.0084 i .0024 V n = _ 0.0177 i .0035 v'"= + 0.0027 i .016 Here again we note that, the Earth aside, the results for the masses are quite satisfactory. The correction to Prof. HALL'S original mass of Mars is so minute and so much less than its probable error that we may consider this value of the mass to be confirmed, and may adopt it as definitive without question. The corrections to the masses of Mercury and Veuus are scarcely changed. The mean residual is reduced to 8 = i 0.91 which is less than the supposed value. We have, therefore, so far as these results go, no reason for distrusting the following value of the solar parallax, which results from that of the mass of the Earth thus derived: 7i = 8".759 ".010 The critical examination and comparison of this and other values of the parallax is the" work of the next two chapters. CHAPTER VII. VALUES OF THE PRINCIPAL CONSTANTS WHICH DEFINE THE MOTIONS OF THE EARTH. The Precessional Constant. 65. The accurate determination of the annual or centennial motion of precession is somewhat difficult, owing to its depend- ence on several distinct elements, and to the probable system- atic errors of the older observations in Right Ascension and Declination. What is wanted is the annual motion of the equinox, arising from the combined motions of the equator and the ecliptic, relative to directions absolutely fixed in space. As observations can not be referred to any line or plane which we know to be absolutely fixed, we are obliged to assume that the general mean direction of the fixed stars remains unchanged, or, in other words, that the stellar system in general has no motion of rotation. This is a safe assumption so far as the great mass of stars of smaller magnitude is concerned. But it is not on such stars that we have the earliest accurate obser- vations. Moreover, observed Right Ascensions of these fainter stars relative to the brighter ones are subject to possi- ble systematic errors, arising from the personal equation being different for brighter and fainter stars. In the case of the stars observed by BRA.DLEY, there is frequently such commu- nity of proper motion among neighboring stars that we can noLbe quite sure that all rotation is eliminated in the general mean. Under these circumstances we have only to make the best use that we can of existing material. We must also remember that observed Right Ascensions are not directly referred to the equinox, but to the Sun, of which the error of absolute mean Right Ascension must be deter- mined. This again can be done only from observed declina- tions, since by definition the equinox is the point at which the Sun crosses the equator. It is also to be noted that the clock stars which are directly compared with the Sun by no 124 65] THE PRECESSIONAL CONSTANT. 125 means include the whole list to be used as absolute points of reference. We therefore have three separate steps in determin- ing completely a correction to the adopted annual precession : (1) The correction to tlie Sun's absolute mean Eight Ascen- sion or longitude. (2) The correction to the general mean Eight Ascension of the clock stars relative to the Sun. (3) The determination of the clock stars relative to the great mass of stars. It goes without saying that the determinations of these three quantities are entirely independent of each other, and that the precision of the result depends on the precision of each sepa- rate determination. The motion of the pole of the equator, on which the luni- solar precession depends, may be determined by observed Declinations quite independently of the Eight Ascensions. A determination of the precession from the latter includes the planetary precession, but as this has to be determined from theory independently of observations, we have, in observed Eight Ascensions and Declinations, two independent methods of determining the motion of the equator. It fortunately happens that the constant of precession is not so closely connected with other constants that a small error in its determination will seriously affect our general con- clusions, or the reduction of places of the fixed stars, because the effect of an error will be nearly eliminated through the proper motions of the fixed stars, or the motions of the planets in longitude. I have therefore satisfied myself with reviewing and combining ,the four best determinations. I pass over in silence the classic determinations of BESSEL and OTTO STRUVE, because the material on which they depend has been incorporated in more recent works. Of these the one which seems entitled to most weight is that of Luowia STRUVE, Bestimmung der Constante der Prcecession, und der eigenen Bewegung des Sonnensy 'stems.* This work was suggested by the completion of AUWERS' re-reduction of BRADLEY'S Obser- vations, and of the Pulkowa standard catalogues for 1845, *Me"moires de PAcademie Impe'riale des Sciences de St. Pe'tersbourg. VII e SSrie. Tome xxxv, No. 3. 126 THE PRECESSIONAL CONSTANT. [65 1855, aiid 1865. It depends entirely on the BRADLEY stars, and the result, when reduced to the most probable equinox, may be regarded as the best now derivable from those stars, or, at least, as not susceptible of any large correction. He, of course, includes in his work the determination of the motion of the solar system relative to the mass of the stars. In addition to this, the possibility of a common rotation of the BRADLEY stars around the axis of the Milky Way is con sidered. This rotation I should be disposed to regard as zero for the present. In place of considering each of the 2,509 stars singly, he divides the celestial sphere into 120 spherical trapezoids, each covering 15 degrees in Declination, and an arc of Right Ascension equal approximately to one hour of a great circle at the equator. The question might be legitimately raised whether a different system of weighting the trapezoids, founded on a consideration and comparison of the proper motions in Eight Ascension and Declination would not have been advis- able. I am, however, fairly confident that no change in this respect would have materially affected the result. With this work of STRUVE I have combined those of BOLTE, DREYER, and NYREN. In the case of the Eight Ascensions it is necessary to reduce all the results to the equinox determined in the last chapter. From this chapter it appears that the standard Eight Ascen- sions with which the reduction of the preceding investigations have been made require a correction to the centennial motion of 4- 0".30. Eeducing each determination to the equinox thus defined, we have the following results for the general preces- sion in Eight Ascension at the epoch 1800 : L. STRUVE, from the comparison of AUWERS-BRADLEY with the modern Pulkowa Eight Ascensions . . . m = 46".050l ; w = 4 DREYER, from the comparison of LALANDE'S Eight Ascensions with those of SCHIELLERUP 46 .0611; w = 2 NYREN, by the comparison of BESSEL'S Eight Ascensions with those of S'CH JELLERUP 46 .0456 ; W = I Mean 46 .0526 65] THE PRECESSIONAL CONSTANT. 127 The weights here assigned are of course a matter of judgment. The general agreement of the results is as good as we could expect. From observed declinations we have L. STRUVE, from the comparison of AUWERS - BRADLEY with modern Pulkowa catalogues w = 20".0495; iv = 2 BOLTE, from the comparison of LA- LANDE'S Declinations with those of SCILJELLERUP 20 .0537 ; w = 1 Mean 20 .0509 We have now to -combine these independent results. I pro- pose to call Precessional Constant that function of the masses of the SUE, Earth, and Moon, and of the elements of the orbits of the Earth and Moon, which, being multiplied by half the sine o twice the obliquity, will give the annual or centennial motion of the pole on a great -circle, and being multiplied by the cosine of tire obliquity will give the lunisolar precession at any time. It is true that this quantity is not absolutely constant, since it will change in the course of time, through the diminution of the Earth's eccentricity. This change is, however, so slight that it can become appreciable only after several centuries. If, then, we put p, the precessional constant, we have, for the annual general precession in Eight Ascension and Declination m = p cos 2 e H sin L cosec s n Y sin cos L being the longitude of the instantaneous axis of rotation of the ecliptic, and H its annual or centennial motion. From the definitive obliquity and masses of the planets adopted hereafter, we find the following values of #, L, and , for 1800 and 1850: lOg H = 1800. 1.67372; 1850. 1.67341 L = 173 2'.31; 173 29'.68 = 23 27.92; 23 27 .53 128 THE PRECESSIONAL CONSTANT. [65 We thus find the following values of p, the unit of time being 100 solar years: From Eight Ascensions, P = 5490.12; w = 2 From Declinations, p = 5489.44; w = 1 Mean, p = 5489 // .89 As the data used in STRUVE'S Investigation may be con- sidered of a more certain kind than those used by the others, we may compare these results with those which follow from STRUVE'S work alone. They are From Eight Ascensions, P = 5489.83 From Declinations, p = 5489.06 Giving double weight to the results from the Eight Ascen sions, the results may be expressed as follows : From STRUVE'S investigation, P = 5489.57 From the other two works, p = 5490.18 Before concluding this investigation, I had adopted as a pre- liminary value P = 5489".78 As this result does not differ from the one I consider most probable, 5489".S9, by more than the probable error of the latter, and diverges from it in the direction of the best deter- mination, I have decided to adhere to it as the definitive value. The centennial value of p is subjected to a secular diminu- tion of 0".00364 per century, owing to the secular diminution of the eccentricity of the Earth's orbit. We therefore adopt p = 5489.78 0.00364 T for a tropical century. p = 5489.90 - 0.00364 T for a Julian century. In the use of p I at first neglected the secular variation, but have- added its effect to the results developed in powers of the time. 66] THE CONSTANT OF NUTATION. 129 Constant of nutation derived from observations. 66. The determination of this constant from observations is extremely satisfactory, owing to the completeness with which systematic errors may be eliminated. If, with a meridian instrument, regular observations are made through a draconitic period, on a uniform plan, upon stars equally distributed through the circle of Eight Ascension, the observations being made daily through more than 12 hours of Eight Ascension, all systematic errors in the determination of the nadir point and all having a diurnal or annual period may be completely eliminated from the constant in question. These conditions are so nearly fulfilled in the observations with the Greenwich: transit circle, and, to a less extent, in those with the Wash- ington transit circle, that the results of the Wdrk with those two instruments alone are entitled to greater weight than has hitherto been supposed. I have, however, discussed quite fully all previous determinations of which it seemed that the probable mean error would be less than 0".10. Eeferring to the volume on the subject to be hereafter pub- lished, the results of the discussion are presented in the fol- lowing table. The weights are assigned on the supposition that weight unity should correspond to a mean error of about 0".07, or to a probable error of /7 .05, this probable value being not entirely a matter of computation from the discord- ance of. the separate results, but, to a certain extent, a matter of judgment. It.must be understood that the results below are not always those given by the authors who are quoted, but that their dis- cussion has,, wherever- possible, been subjected to a revision by the introduction of modern data, or by what seemed to me improved combinations. Thus, NYREN'S equations have been reconstructed on a system slightly different from his, and have been corrected for CHANDLER'S variation of latitude. PETERS'S classical work has also been corrected by the introduction of later data, and by a re-solution of his equations. The Green- wich and Washington results have been derived from the dis- cussion in Astronomical Papers, Vol. II, Part VI. 5690 N ALM 9 130 THE CONSTANT OF NUTATION. [66 Values of the constant of nutation derived from observations. BUSCH, from BRADLEY'S observations with the zenith sector 9.232 1 ROBINSON, from Greenwich mural circles . . 9.22 1 PETERS, from Eight Ascensions of Polaris . 9.214 4 LUND AHL, from Declinations of Polaris . . 9.236 1.5 NYREN, from v Urs. Maj 9.254 3 " " oDraconis 9.242 2.5 " " i Draconis 9.240 4 DE BALL, from WAGNER'S Eight Ascensions of Polaris 9.162 3 DEBALL, from WAGNER'S Declinations of Polaris . 9.213 3 DEBALL, from WAGNER'S Eight Ascensions of51Cephei 9.252 3 DEBALL, from WAGNER'S Declinations of 51 Cephei 9.227 3 DEBALL, from WAGNER'S Eight Ascensions of 6 Urs. Miu 9.208 3 DEBALL, from WAGNER'S Declinations of d Urs. Min 9.263 3 Greenwich North-Polar Distances of South- ern Stars, Series I 9.116 3 Greenwich North-Polar Distances of South- ern Stars, Series II 9.201 3 Greenwich North-Polar Distances of North- ern Stars, Series I 9.204 4 Greenwich North-Polar Distances of North- ern Stars, Series II 9.223 4 Washington Transit Circle, southern stars . 9.217 6 " " u northern stars . 9.177 3 Greenwich, Eight Ascensions of Polaris . . 9.153 2 " Declinations of Polaris . . . . 9.242 2 " Eight Ascensions of 51 Cephei . 9.135 2 " Declinations of 51 Cephei . . . 9.162 2 " Eight Ascensions of 6 Urs. Min. 9.147 2 " Decimations of tfUrs. Min. . . 9.235 2 " Eight Ascensions of A Urs. Min. 9.161 1 " Declinations of A Urs. Min. 9.339 1 Mean . 9.210; wt. = 72 66,67] PRECESSION AND NUTATION. 131 The mean error corresponding to weight unity when derived from the discordance of the results is 0".068, while the estimate was i 0".070. We may therefore put, as the resulj of observation N = 9".210 0".008 Relations betiveen the constants of precession and nutation, and the quantities on which they depend. 67. The formula of precession and nutation have been developed by OPPOLZER with very great rigor and with great numerical completeness as regards the elements of the Moon's orbit, in the first volume of his Bahnbestimmung der Kometen und Planeten, second edition, Leipzig, 1882. What is remarkable about this Avork is that it constantly takes account of the possible difference between the Earth's axis of rotation and its axis of figure, a distinction which has become emphasized by CHANDLER'S discovery since OPPOL- ZER wrote. His theory however fails to take account of the change in the period of the Eulerian nutation produced by the mobility of the ocean and the elasticity of the Earth. But this effect is of no importance in the present discussion. From OPPOLZER'S developments, I have derived the follow- ing expressions, in Avhich the numerical coefficients may be regarded as absolute constants, so accurately determined that no question of their errors need now be considered. These results haA^e been derived quite independently of the similar ones by Mr. HILL in the Astronomical Journal, Yol. XI, which are themselves independent of OPPOLZER'S work. In these formulae we have K, the constant of lunar nutation of the obliquity of the ecliptic, as defined by the equation As = ~N cos &, and expressed in seconds of arc; P, so much of the precession of the equinox on the fixed ecliptic of the date, in seconds of arc and in a Julian year, as is due to the action of the Moon ; P 7 , so much of the same precession as is duejo^j^e^ action of the Sun. ^ * ^l - ^- Of 132 . MASS OF THE MOON. [67, 68 We thus have, luni-solar precession = P + P 7 f, the obliquity of the ecliptic; yu, the ratio of the mass of the Moon to that of the Earth ; A, the mean moment of inertia of the Earth relative to axes passing through its equator; C, the same moment relative to its polar axis. With these definitions we have, General value. Special value for 1850. X = [5.40289J cos s P = [5.975052] cos f P' = [3.72509] c 0-A C = [5.36542! C-A 1 + ~ A = [5.937585] _JL_ C J 1 + /i C ; "^ = [3.68762] C-^. The special values for 1850 are found by putting for the value of the obliquity of the ecliptic for 1850, f = 22 27' 31". 1 The mass of the Moon from the observed constant of nutation. 68. From the two quantities given by observation, N and P 4- P' = po, these equations enable us to determine the two unknown quantities yu and C A As the easiest way of showing the uncertainty of the Moon's mass, arising from uncertainty of the precession and nutation, I give the value of its reciprocal corresponding to different values of these quan- tities in the following table : Reciprocals of the mass of the Moon corresponding to different values of the nutation-constant and luni-solar precession. A i *=" 50.35 50.36 50.37 81.81 81.86 81. 91 81-53 81.58 81.63 81. 25 81.30 81-35 68, 69J THE CONSTANT OF ABERRATION. 133 Taking for the constant of nutation the value just found, N = 9 // .210 ".068 and for the luni-solar precession, lh = 50" .36 ''.006 we have, for the reciprocal of the mass of 'the Moon and its mean error : - = 81.58 0.20 p The Constant of Aberration. 69. In the determination of astronomical constants the inves- tigation of the constant of aberration necessarily takes a very important place, not only on its own account but on account of its intimate connection with the solar parallax. A general determination, founded on all the data available, was therefore commenced by me as far back as 1890, before the fact of the variation of terrestrial latitudes had been well established. The successive discoveries of the law of this variation by CHANDLER required such alterations in the work as it went along that much of it is now of too little value for publication in full. Happily the necessity for a new discussion of the best determinations at Pulkowa has been done away with by the papers of CHANDLER himself in the Astronomical Journal. Quite apart from the disturbing influence of the revolution of the terrestrial pole upon the determination of the constant of aberration, this constant is itself the one of which the deter- mination is most likely to be affected by systematic errors. In this respect it is at the opposite extreme from the constant of nutation. From the very nature of the case it requires a comparison of observations at opposite seasons of the year, when climatic conditions are different. In most cases the determination must even be made at different times of day. The effect of aberration on a star, for example, is generally at one extreme when the star culminates in the morning, and at the other extreme when it culminates in the evening. The culminations at opposite seasons of the year are necessarily 134 THE CONSTANT OF ABERRATION. [69 associated with culminations at opposite times of the day. Moreover, in observations to determine the constant of aber- ration from Declination, the stars which give the largest coeffi- cients are, for the northern hemisphere, tbose near 18 h of Eight Ascension. Any error peculiar to the times or seasons at which these stars are observed will therefore affect the result systematically. Eight Ascensions of close polar stars also lead to a value of this constant. But the same difficulty still exists. In this case the maxima and minima of aberration occur when the star culminates at noon and midnight. Not only is the aspect of the star different at the two culminations, but the effect of any diurnal change in the instrument will be transferred to the final result for the aberration. The prismatic method of LOEWY is free from some of these objections. But its application is extremely laborious, and we have, up to the present time, only two determinations by it, one by LOEWY himself, which is only regarded as preliminary, and one by COMSTOCK, in which a large uncertain correction for personal equation was applied. Under these circumstances the seeking of results derived by methods of the greatest possible diversity is yet more strongly recommended than in the case of the other astronomical con- stants. I have therefore used not only the PULKOWA deter- minations, but all those made elsewhere which it seemed worth while to consider. Notwithstanding the great amount of mate- rial added to NYREN'S paper of 1883, it will be seen that the probable error of the final result at which I have arrived is greater than that which he assigns to his result. This is a natural consequence of combining so many separate determi- nations. The advantage is, however, that the assigned prob- able error is more likely to be the real one. It is not to be supposed that any of the systematic errors already indicated would pertain to all observers and to all instruments. The final outcome should be a result in which the discordances of the separate determinations show the probable values of all the actual errors, both accidental and systematic. Determinations founded on the Eight Ascensions of circum- polar stars are not affected by the motion of the terrestrial 69, 70] THE CONSTANT OF ABERRATION. 135 axis, uor are those founded on declinations of these stars, if only the declinations are observed equally at both culmina- tions. But determinations founded on declinations of stars from upper culmination only are necessarily aifected by this cause. If however the stars on which the determination is based extend through the whole circle of Right Ascension the effect of the cause in question may be wholly eliminated by a suitable treatment of the equations of condition. To practically eliminate the injurious effect it is not even necessary to deter- mine the exact law of variation. In fact, if the stars observed are equally scattered in Eight Ascension, the effect of the varia- tion will be partially eliminated without taking account of it. CHANDLER has shown that there are two periodic terms in the variation of latitude, one having a period of one year, the other of four hundred and twenty-seven days. I may remark that this combination is in accord with my theory developed in the Monthly Notices of the Royal Astronomical Society for March, 1S92. It was there shown that any minute annual change of the position of the principal axis of inertia of the Earth a change which might be produced by the motion of water, ice, and air on its surface would appear as an annual term in the latitude, six times as great as its actual amount, Values of the constant of aberration derived from observations. 70. What I have done since this discovery by CHANDLER has been to reexamine the determinations of the constant of aberration made from time to time, to make such corrections in their bases as seemed necessary, and more especially to determine the correction to be applied to each separate result on account of the periodic term in the latitude. No attempt was made to rework completely the original material, except in the case of the results of the Pulkowa and Washington observations with the prime vertical transit. In the case of the former, however, the preliminary results reached from time to time were so accordant with those of CHANDLER that it is a matter of indifference whether we regard them as belonging to his work or to my own. Owing to the very different estimates placed by the astro- nomical world upon the Pulkowa determinations and 'those 136 THE CONSTANT OF ABERRATION. [70 made elsewhere, I have used the former quite apart from the others. The complete discussion of each separate value is too voluminous for the present publication, and is therefore reserved for a more extended future publication. At pres- ent it appears sufficient to judge the final result by the general discordance of the material on which it rests, rather than by a separate criticism of each particular case. In the exhibit of results which follows it is to be remarked that NYREN'S prime vertical observations do not receive a weight as great, relative to the other Pulkowa determinations, as would be given by their assigned probable errors. The reason of this course is that one can not be entirely confident that the results of any one observer with this instrument are free from constant error arising from differences of personal equation in observing a bright and a faint star. Many of the Pulkowa observations are necessarily made in the morning or evening twilight. In the case of an evening observation the star will therefore be much fainter on account of daylight when it transits over the east vertical than it will when it transits over the west vertical one or two hours later. In the case of morning observations the reverse will be true. It is easy to see that if, in consequence of this difference of aspect, the observer notes the passage of the faint image too late, the effect will be to make the constant of aberration too large. The existence of this form of personal equation, when transits are recorded on the chronograph, is so well known that, had NYREN'S observations been made in this way, I should not have hesitated to ascribe the large values of his aberration constant to this cause. Although it has never been shown that any such personal equation exists when observations are made by eye and ear, as KYREN'S were, yet when we consider that we are dealing with quantities amounting only to one or two huudredths of a second of arc, and that a personal equa- tion of this kind, undiscoverable by ordinary investigation, might affect the result by this minute amount, we can not but have at least a suspicion that his values may be slightly too large from this cause. 70! THE CONSTANT OF ABERRATION. 137 Separate results for the constant of aberration. A. Standard Pulkowa determinations : A b. wt. Observations with Vertical Circle ; Polaris, by n PETERS 20.51 2 Observations with Vertical Circle ; 7 miscellaneous stars, by PETERS 20.47 2 Observations with Vertical Circle 5 1863-1870, Po- laris, by GYLDEN 20.41 2 Observations with Vertical Circle; 1871-1875, Po- laris, by XYREN 20.51 2 Observations with Prime Vertical; 1842-1844, by STRUVE 20.48 4 Observations with Prime Vertical; 1879-1880 by NYREN. . 20.52 6 Observations with Prime Vertical; 1875-1879, by NYREN 20.53 1 Observations with Vertical Circle; 1863-1873, by GYLDEN and NYREN . 20.52 2 WAGNER : Transits of three polar stars .... 20.48 5 From Eight Ascensions of Polaris; 1842-1844, by LINDHAGEN and SCHWEIZER 20.50 2 Mean result: 20 // .493 0".011 This result may be regarded as identical with that found by NYREN in 1882. B. Other determinations: Ab. e wt. AUWERS, from observations with the n zenith sector at Kew 20.53 .12 0.5 AUWERS, from WANSTED observations . 20.46 .12 0.5 PETERS, from BRADLEY'S observations of y Draconis at Greenwich with zenith sector, 1750-1754 20.67 0.5 BESSEL, from Eight Ascensions observed by BRADLEY at Greenwich .... 20.71 i.071 0.5 LINDENAU, from Eight Ascensions of Polaris observed at various observa- tories between 1750 and 1816 . 20.45 .05 3 138 THE CONSTANT OF ABERRATION. | 70 Separate results for the constant of aberration Continued. B. Other determinations Continued. BRINKLEY, from observations of thirteen --/ft. /. stars at Trinity College, Dublin, with 7/ the 8-foot circle 20.46 .10 1 PETERS, from STRUVE'S Dorpat observa- tions of six pairs of circumpolar stars . 20.36 .07 2 RICHARDSON, from observations with the Greenwich mural circles 20.50 .06 3 PETERS, from Right Ascensions of Polaris at Dorpat . 20.41 6 LUNDAHL, from Declinations of Polaris at Dorpat 20.55 5 HENDERSON and MCLEAR, from a 1 and <* 2 Centauri . . . 20.52 .10 1 MAIN, from observations with the Green- wich zenith tube . 20.20 .10 1 DOWNING, from observations of ADra- conis with reflex zenith tube . ... . 20.52 .05 4 XEWCOMB, from observations of <*Lyra3 with the 'Washington prime vertical transit, 1862-1867 20.46 0.4 6 NEWCOMB, from Right Ascensions of Polaris observed with the Washington transit circle, 1866-1867 . 20.55 .05 :J KUSTNER, from observations of pairs of stars by the TALCOTT method . . . 20.46 4 PRESTON, from observations with the- TALCOTT method at Honolulu, 1891- 1892 20.43 .05 4 LOEWY, from his prismatic method . . 20.45 .04 5 COMSTOCK, using LOEWY'S method, slightly modified 20.44 3 KiisTNER, from MARCUSE'S observations, 1889-1890 20.49 .018 4 WANACH, from Pulkowa prime vertical observations . 20.40 .015 4 70, 71] THE LUNAR INEQUALITY. 139 Separate results for the constant of aberration Continued. B. Other determinations Continued. Ab. tvt. From Greenwich Eight Ascensions of polar stars with the transit circle 20.39 3 BECKER, from observations at Strasburg by the TALCOTT method, 1890-1893 20.47 6 DAVIDSON, from similar observations at San Francisco, 1892-1894 20.48 6 Mean result of B : Ab. const. = 20".463 0".013 The two results. A and B, differ by 0".030, a quantity so much greater than their mean errors as to leave room for a suspicion of constant error in one or both means. The Lunar Inequality in the Ear til's motion. 71. The source of this inequality is the revolution of the center of the Earth around the center of mass of the Earth and Moon. The former center describes an orbit which is similar to that of the Moon around the Earth. Since this orbit is not a Keplerian eclipse, but is affected by all the per- turbations of the Moon by the Sun, no such element as a semi- major axis can be assigned to it. Instead of this I take as the principal element of the orbit the coefficient of the sine of the Moon's mean elongation from the sun in the expression for the Sun's true longitude. This element is a function of the solar parallax and of the mass of the Moon, which may be derived from the folio wii>g expression. Let us put yw ; the ratio of the mass of the Moon to that of the Earth ; r,A, /?; the radius vector, true longitude and latitude of the Moon ; r', A',/J'; the same coordinates of the Sun; * ; the linear distance of the Earth's center from the center of mass of the Earth and Moon. 140 THE LUNAR INEQUALITY. [71 We then have, for the perturbations of the Sun's geocentric place due to the cause in question : A log r 1 = - x cos ft cos (A-A 7 ) A\' = - t cos ft sin (A A 7 ) and s I have developed these expressions, putting 7T = 8".848 -.4 and taking for the Moon's coordinates the values found by DELAUNAY. Putting D; the mean value of A A 7 g, g 1 } the mean anomalies of the Moon and Sun, respectively, u'-, the Sun's mean elongation from the Moon's ascending- node; the result for JA 7 is // A\' = 6.533 sin D + 0.013 sin 3 D + 0.179 sin (D + g) 0.429 sin (D g) + 0.174 sin (D g 1 ) - 0.064 sin (D + g') -f 0.039 sin (3 D - g) 0.014 sin (D g g 1 } 0.013 sin 2 u 1 This value of the lunar inequality is substantially identical with that computed from the tables and formula of LEVER- 71] THE LUNAR INEQUALITY. 141 RIER'S solar tables. The development of the numbers there given lead to the value 6".534 of the principal coefficient. We have now to find what value of the coefficient is given by observations. The observations I make use of are (1) all the observations of the Sun's Eight Ascension from early in the century till 1864 ; (2) The heliometer observations of Vic- toria made in 1889 on GILL'S plan and worked up by him. I had intended to use all the observations of the Sun up to the present time. 1 found however that those made after 1864 gave, by comparison with the published ephemerides, inadmissible positive corrections to the coefficient. This cir- cumstance gives rise to a strong suspicion that in the process of interpolating the Right Ascensions of the Sun during at least some years after 1864, the inequality in question was rounded off to the amount of several hundredths of a second. The results were therefore entirely omitted. The results for previous years, when the inequality was computed separately for every day of observation, are: Greenwich, 1820-'64; -.068 3.0 Paris, 1801-'64; -.050 0.8 Konigsburg, 1820-'45; -.054 1.2 Cambridge, 1828->58; -.047 2.0 Dorpat, 1823->38; + .160 0.3 Pulkowa, 1842-'64; -.058 0.5 Washington, 1846-'64: .000 0.2 Mean, JP = 0".048 i 0".018 GILL'S result is given in the Monthly Notices, Royal Astro- nomical Society, for April, 1894 (Vol. LIY, page 350.) It is derived in the following way. In the solar ephemeris which he used for comparison the lunar inequalities were computed rigorously from the coordinates of the Moon, putting n = 8".880 M = 1 -r- 83 To the coefficient P thus arising he found a correction, = + 0".046 142 THE LUNAR INEQUALITY. [71, 72 The above values of n and // give, on the theory just devel- oped, P = 6".400 Thus GILL'S result is, in effect, P-= 6".446 while mine, from observations of the Sun, is 6".533 0".048 = 6".485 I consider that these results are entitled to equal weight, and that we may take, as the result of observation, P = C>".465 i 0".015 Solar parallax from the lunar inequality. 72. With the mass of the Moon already found from the observed constant of nutation, // = 1 : 81.58 (1 i .0025) we may now derive a value of the solar parallax quite inde pendent of all other values. The relation between P, TT, and the mass of the Moon is of the general form // P = A' 7T where fc is a numerical constant, and, for brevity, We have found that the following values correspond to one theory : TT = 8".848; X = 82 5 P = 6". 533 Hence follows log fc = 1.78207 so that we have ^P= [1.78207] n The numerical values P = 6 /7 .465 and // = 82.58 now give 7r = 8 // .818 0".030 73J PARALLAX FROM TRANSITS OF VENUS. 143 Values of the solar parallax derived from measurements of Venus on the face of the Sun during the transits of 1874 and 1882, with the heliometer and photoheliograph. 73. I put these determinations into one class because they rest essentially on the same principle. Both consist, in effect, in measures of the distance between the center of Yenus and the center of the Sun j the latter being denned through the visible limb. , The method is therefore subject to this serjous drawback : that the parallax depends upon the measured differ- ence between arcs which may be from thirty to fifty times as great as the parallax itself, the measures being made in different parts of the earth. The equations of condition given by the American photo graphs of 1874 are found in Part I of Observations of the Transit of Yenus, December 9, 1874 ; Washington, Government Printing Office, 1880. A preliminary solution of these equa- tions, the only one, however, to which they have yet been sub- jected, was published by D. P. TODD, in the American Journal of Science for June, 1881. (Yol. XXT, page 490.) The photographs of 1882 have been completely worked up by Professor HARKNESS, and the results are found in the Eeport of the Superintendent of the Naval Observatory for 1889. The equations derived from the German heliometer measures, with a preliminary discussion of their results, are officially published by Dr. AUWERS, in the Bericht iiber die deutschen Beobachtungen,) Y, p. 710. The separate results for the parallax, with the probable errors assigned by the investigators, are as follows: // // w. W 1874 : Photographic distances, n 8.888 0.040 6 1 Position angles, 8.873 0.060 3 3 Measures with heliometer, 8.876 i 0.042 5 5 1882: Photographic distances, 8.847 0.012 64 6 Position angles, 8.772 i 0.050 4 4 Measures with heliometer, 8.879 0.025 1C 10 Under w is given a system of weights proportionally deter- mined from the probable errors as assigned. Using this sys- tem, the mean result is n =8".854 i ".016 144 PARALLAX FROM TRANSITS OF VENUS. [73 I conceive, however, that these relative weights do not cor- respond to the actual precision of the measures. The very small probable error assigned by Prof. HARKNESS to the result of the photographic distances of 1882 does not include the probable error of the angular value of the unit of distance on the plate, which may arise from a number of sources, includ- ing the possible deviation of the mirror of the instrument from a perfect plane. From this error the position angles are entirely free. I have, therefore, assigned another set of weights, w', which seem to me to correspond more nearly to the facts. The result of this system is 7t = 8".857 i ".016 This mean error is derived from the individual discordances, and not from comparisons with the values of the parallax otherwise determined. As there may be a fortuitous agree- ment among the separate values, another estimate may be made on the basis of the total mean error derived by AUWERS, which includes all known sources of error. He finds 6 = ".032 for the combined heliometer results, to which I have assigned weight 15. Hence, for the total weight 29, we have = - ".025 in the final equations of the original paper. After each result is given the mean .error with which it is affected, as deter- mined by the investigation in question. When thus treated, the equations which I have given on pages 391-398 of the paper referred to give the following normal equations for tfc, the indeterminates & 2 and & 3 being retained as such in order to show their final effect on the result. // // 1761. II 5 8.5 do = + 0.76 - 18.5 fc 2 0.78 III; 41.7 dc = - 2.81 - 19.2 k, 1.30 1769. II; 44.8 dc = - 8.00 - 104.1 fc 2 i 1.95 III; 12.1 do = + 0.31 - 16.0 fc 3 0.70 In order to vary the proceeding as much as possible from that of the former investigation, I now express dc in terms of dX and 6fi, which, for the time being, I take as the corrections to the heliocentric longitude and latitude of Venus referred to the Earth, and these again in terms of dv and sin 166, which latter, for brevity, I call u. The first transformation is made with the coefficients of p. 71, where we have put x and y for 6\ and d/3, and the last by the equations // 6X = 6v + 0.06 u 8p = u 0.06 v Putting Ui for the value of u in 1765, we have, in consequence of the known change in the motion of the node, // In 1761; u = Ui + 0.11 In 1769; u = ^ 0.11 80] MOTION OF THE NODE OF VENUS. 161 We thus have the four equations which follow for determining 6v and HI, the former being supposed the same at the times of the two transits. - .84 6v - .55 M! + * a = + 0.15 - 2.2 & 2 i 0.09 + .73 - .69 + z 3 = + 0.01 - 0.5 fc 3 0.03 - .69 + .73 + 2 2 = - 0.10 - 2.3 A" 2 0.04 + .81 + .60 + s 3 = + 0.10 - 1.3 & 3 0.06 Eliminating z. 2 and 3 by subtracting the first equation from the third, and the second from the fourth, we have .15 6v + 1.28 % = - o'.25 -~ o'.l A; 2 i 0.10 .08 <* + 1.29 ui = + 0.09 - 0.8 fc 3 0.07 We thus have for Ui the value m = - 7/ .04 - 0.08 6v - 0.03 A: 2 - 0.36 fc 3 i O x/ .05 dv can not be determined independently of z 2 and 3 . Assum- ing these quantities to be equal, we have already found it to be only /7 .02, and may therefore, to determine its probable effect upon the result by assigning to it the value In the former paper I have found for k 2 and & 3 the values fc 2 = + 0.040 i 0.040 Jc 3 = - 0.034 0.040 A preliminary correction of + 2 /7 .02 having been applied to the tabular orbital latitude, we have, for the epoch 1765.5, sin id 6 = + 1".99 i 0".06 Combining this result with that of the transits of 1874 and 1882, we have the following results, which are compared with those of meridian observations : // Transits of Yenus alone ...... sin i D t $6 = 2.82 Meridian observations alone .... " 2.45 Combined solution ........ _ 2.71 Adjusted with other* results (46) . . . 2.73 Adopted ........... 2.77 5690 N ALM - 11 162 MOTION OF THE NODE OF VENUS. [80 The adopted result is the one which seems the most probable. For the final probable error we are to include that of the pre- cession and of the Sun's longitudes at the two epochs. We may estimate the combined value of these at i 1", correspond- ing to an error of 0".06 in sin i D t 66. Thus we have sin i D t 66 = 2". 77 i // .084 I conceive this mean error to be as real as any that can be determined in astronomy. This conviction rests upon the fact (1) that the systematic errors affecting the four contacts are shown to be small by the general minuteness of the four values of dc; (2) that whatever systematic errors may affect the formation or disappearance of the thread of light are almost completely eliminated from the mean of the transits of 1761 and 1769 by the method in which the observations have been combined. The accordance of the observations of external contact made at the same transits strengthens this view. The equation thus derived takes the place of the sixth equation of 63 and should have twice the weight there assigned. As the mass of the Earth determined by the secu- lar variations rests mainly on this equation, I shall first con- sider it alone. Expressing the theoretical secular variation of sin i66 in terms of the above observed value, we find that the observed motion of the node of Yenus gives the equation 0".26 v 29". 2 v 1 43".2 v" = + 0".48 i // .084 (a) which gives for v" the value v" = 0.0 ill 4- 0.006 v - 0.676 v 1 i .0019 The value of the solar parallax for v" is 8" .811. Hence, for the value expressed in terms of the corrections to the assumed masses of Yeuus and Mercury, this equation gives n = 8".778 + 0".020 r 1".98V We have found from the periodic perturbations // // v - _ 0.055 i .25 v 1 = + 0.0080 i .0025 80] SOLAR PARALLAX. 163 Whence, // // Y" = - 0.0168 i .0029 n = 8.762 i .0086 This result of observation, errors and unknown actions aside, Fcan not suppose to be affected by any other mean error than that here assigned. We have now to consider how far this result may be recon- ciled with the others by changes in the masses of Mercury and Venus. No admissible change in the former could greatly affect the result. The question then arises whether the dis- crepancy may not be due to an error in the concluded mass of Venus. In making so large a change in this element, we meet with insuperable difficulties. The observed motion of the ecliptic, which is a fairly well-determined quantity, indi- cates a still further increase of this mass. We may put this difficulty in another form. The observed motion of the node of Venus is a relative one, consisting in the combined effect of the motion of the ecliptic around an axis at right angles to the node of Venus, and an absolute motion of the orbit of Venus around nearly the same axis. This motion of the ecliptic depends mainly on the mass of Venus ; the absolute motion Of the orbit of Venus mainly on that of the Earth. If, now, we determine the motion of the ecliptic from observation, we shall find that the relative motion of the orbit of Venus still unac- counted for is yet greater than we have supposed it to be, and should therefore find a yet smaller mass of the Earth than that heretofore concluded. The determination of the mass of Venus already made from observations of the Sun and Mercury seems to admit of no doubt. We can not conceive that the mean of fifteen deter- minations, made during one hundred and thirty years, at dif- ferent observatories, which determinations are so separated as to be entirely independent of each other, can be affected by any considerable common error. The entire accordance of the result thus reached from the periodic perturbations produced by Venus with that from a combination of all the secular variations, as shown in Chapter VI, strengthens the result yet further. Unknown actions and possible defects of theory 164 SYSTEMATIC ERRORS OF PARALLAX. [tO, M aside, it seems to me that the value of the solar parallax derived from this discussion is less open to doubt from any known cause than any determination that can be made. Possible systematic errors in determinations of the parallax. 81. We have now to return to the other values, in order to see to what extent they may be affected by systematic error. I have already excused myself from discussing the validity of the assumed relation between the constant of aberration and the velocity of light, because there is nothing valuable to be said on the subject, and have alluded to the possible sources of systematic error in the Pulkowa determinations of aberra- tion. It is worthy of attention here that the very best of these determinations, that of NYR^N with the prime vertical transit, in resp,ect to the care with which it was made, and the general accordance of the entire work throughout, gives a result most accordant with that under consideration. In fact, to the value 8". 77 of the solar parallax corresponds the value 20 // .55 of the constant of aberration, which is larger by only // .02 than the result of NYREN'S best determinations. A.S for miscellaneous determinations of the constant, it is to be remembered that the corrections applied to a part of the separate values on account of the Chandlerian inequality of latitude are somewhat doubtful, and the general mean mav have been affected by a few huudredths of a second in conse- quence. It is not, however, possible to determine the amount of the correction, except by an exhaustive rediscussion of the whole of the original observations, and even then the result would still be doubtful. Next in the order of weight we have the results of measures on the minor planets with the heliometer, on GILL'S plan. I have already remarked upon the possible error in such obser- vations arising from the probable difference of color between the planet and the star. A hypothetical estimate of the amount of this error is worth attempting. Let us assume that in the case of a minor planet the mean of the visible spec- trum corresponds to the line D, and that in the case of a star the same mean is halfway between the lines D and E. 81] SYSTEMATIC ERRORS OF PARALLAX. 165 The index of refraction of air has been determined inde- pendently by KETTLER and LORENTZ for the different rays. The mean of their results for the rays D and E is For D, n = 1.000 2920 ForE, n = 1.000 2940 These results are accordant in giving a dispersion between these two lines equal to about .0037 of the total refraction. We have hypothetically taken the extreme possible difference . between planet and star to be one-half of this. At an altitude of 45, where the refraction is about 60", the error would be 0".ll. At an altitude of 30 the error would be 0".20. We are thus led to the noteworthy conclusion : If the difference between the spectra of a minor planet and a comparison star is such that the means of their respective visible spectra, or the apparent amounts of their respective refractions, differ by one- tenth of the space between D and E, an error of 0" .02 or 0" .03 may be produced in the apparent parallax of the planet. The question thus arising maybe readily settled by measures with the heliometer. The distances of pairs of stars differing as widely as possible in color should be measured at different altitudes, when one is nearly above or below the other, in order to see what difference of refraction depending on the color is indicated. A colored double star, such as ft Oygni, might also be used for the same purpose. The minor planets are of different colors. I am not aware of any evidence that Victoria or Sappho differ in color from the average of the stars, but 1 believe that Iris is somewhat yellow, or reddish. Kow, in this connection, it is a significant fact that the parallax found from observations of Iris, 8".82o, is the largest by GILL'S method. I have already remarked that the value of the solar parallax derived from the parallactic equation of the Moon is one of which the probable mean error is subject to uncertainty. While it is true that the value may be smaller than that we have assigned, we must also admit that it may be much larger. The probable error of the determination by the lunar equa- tion of the Earth is larger than that of any other method. At 166 RESULTS FOR THE SOLAR PARALLAX. [82 the same time I do not think that it is liable to systematic error, and we must therefore regard the mean error assigned as real. Results for the solar parallax after making allowance for prob- able systematic errors. 82. Let us now see whether we can reach a satisfactory result by making a liberal allowance for the more or less probable sources of systematic error just pointed out. The modifications we make in the weights formerly assigned are these: We reduce the weight of GILL'S Ascension result to one-half, owing to the uncertainty arising from the color of the planet Mars. We retain the Pulkowa determinations of the constant of aberration with their full weight, but reduce the weight of the miscellaneous determinations. In the case of the parallactic inequality, we reduce the weight for the reasons already given. We omit Iris from the determination from the minor planets. We also reduce to one- half its former value the relative weight assigned to measures of Venus on the Sun, on the theory that the actual mean error must be larger than that given by the discordance of results. Our combination will then be as follows : wt. From the motion of the node of Venus .... n = 8.708 10 From GILL'S Ascension observations .... 8.780 1 From the Pulkowa constant of aberration . . . 8.793 40 From contacts of Venus with the Sun's limb . . 8.794 3 From heliometer observations on Victoria and Sappho 8.799 5 From the parallactic inequality of the Moon . . 8.794 10 From miscellaneous determinations of the con- stant of aberration 8.806 10 From the lunar inequality in the motion of the Earth 8.818 1 From measures on Venus in transit 8.857 1 Mean result, ignoring the first ; 8".7965i .0045 This mean result still differs from that given by the motion of the node of Venus by more than five times the probable error of the latter, and is yet farther from the combined result 82] RESULTS FOR THE SOLAR PARALLAX. 167 of all the secular variations, so that no reconciliation is brought about. The embarrassing question which now meets us is whether we have here some unknown cause of difference, or whether the discrepancy arises from an accidental accumulation of fortuitous errors in the separate determinations. We have already discussed the former hypothesis, and have been unable to find any reasonably probable cause of abnormal action. The motion of the planes of the orbits is that which is least likely to deviate from theory, because it is independent of all forms of action depending upon distance from the Sun, or upon the velocity of the planet. An examination and comparison of all the results shows one curious feature: the unanimity with which the secular varia- tions speak against the large value of the solar parallax, or of the mass of the Earth, as the one quantity at fault. The adopted motion of the node of Venus is sustained not only by the meridian observations, but by the external contacts at the transits of 1761 and 1769, and, weakly, by a comparison of the transits of 1874 and 1882. If we determine the correction of the mass of the Earth from other secular variations than that of the node of Venus, by the equations of 63, we have, after eliminating the masses of Mercury and Venus, v" = -0.029; p. e. .018 If, instead of eliminating these values, we put v = + .08; v 1 = + .0080; we have v" = -0.026; p. e. i .014 In each case the value of the parallax is yet smaller than that found from the motion of the node of Venus. I have already remarked that the observed motion of the ecliptic indicates an increase of the mass of Venus. The question thus takes the form, whether it is possible that the mean of the eeven determinations of the solar parallax TT = 8".797 i ".0035 168 DEFINITIVE ADJUSTMENT. [82, 83 can with reasonable possibility be in error by aii amount the correction of which would bring it within the range of adjust- ment of the other quantities. From what has already been said of the systematic errors to which every one of the determinations may be liable, it is evident that we should have no difficulty in accepting the necessary reduction of each of the separate values. The improbability which meets us is not so much the amount of the individual errors of the determinations as the fact that seven of the eight independent determinations should all be largely in error in the same direction.* Still, under the cir- cumstances, we must admit this possibility, and make what seems to be the best adjustment of all the results. Definitive adjustment. 83. In making the definitive adjustment I shall proceed on the supposition that no correction is necessary to the adopted mass of Mars. I also go on the principle that no result is to be rejected on account of doubt or discordance, except when it is affected with a well-established cause of systematic error, and shows a large deviation in the direction in which this cause would act. At the same time it will be admissible to diminish the weights in special cases, on account of causes of systematic error which we know to exist, although we can not determine the directions in which they would act ; and also on account of deviations so wide as to show that the probable error of the result must have been greatly underestimated. Proceeding on this plan, we might reweight the last eight results for the solar parallax, so as to get a result slightly different from 8". 797. But 1 doubt whether such a reweight- ing would not involve an objectionable bias. We might diminish the weight of the result given by the Pulkowa constant of aberration on the ground that no one method should have so preponderating a weight as this has. If we did so the result might be increased to 8".800. We * For a very searching criticism of the systematic errors with which the determinations of the solar parallax may be affected, reference may be made to the first two articles by Dr. DAVID GILL, in Vol. I of The Observa- tory. 83] DEFINITIVE ADJUSTMENT. 169 might very largely increase the relative weight assigned to the heliometer observations on Victoria and Sappho, but no admissible increase would appreciably change the result. We might also diminish the relative weight of the largely dis- cordant result derived from measures of Venus during transit. But as, by throwing out this result altogether, we should only diminish the mean by ".001, it is scarcely worth while to do so. Altogether no rediscussion of the relative weights seems necessary. On the other hand, the weight which we assign to the mean result will enter as a very important factor into the final adjustment. This is a point on which it is impossible to reach a positive numerical conclusion by any mathematical process. If, as one extreme case, we consider that the mean error of each separate result corresponds to i0 7/ .03 for weight unity, we shall have a mean error of rt".0035 for the value 8". 797. The result will not be very different if we determine the mean error from the discordance of the eight separate results. On the other hand, if we include the deviation of the result given by the motion of the node of Venus, the mean error for weight unity will be increased to i 0".0045. The latter is undoubt- edly the most logical course, so long as we proceed on the hypothesis that the deviations of the final adjustment can all be explained as due to fortuitous errors. If we include a com- parison with the results of all the secular variations we shall have a yet larger mean error. To show the result of assigning one weight or the other I shall make two solutions, A and B, in one of which a less and in the other a greater weight will be assigned. To the value 8".797 i .005 or .007 of the solar parallax corresponds r" = - 0.049 i .0016 or .0025 According as we assign one weight or the other to this result, we may take as the corresponding equation of condition of weight unity (A): 400^' =-2.0 r (BH 600," = -2.9 W 170 DEFINITIVE ADJUSTMENT. [83 The masses of Venus and Mercury, determined by methods independently of the secular variations, also enter as conditions into the adjustment. I have, however, made a revision of the preliminary adjustment given in 64, the latter being based on the results of 32-38; whereas it is better to use the defini- tive results of the combination used in 46. For the mass of Mercury the result found in 53 by the last combination is The values of the denominator corresponding to the mean limits here assigned are 5 890 000 and 12 210 000 These limits are so wide as to include all admissible results for the mass of Mercury. Moreover, we can not definitely say that the value (6) of this mass is markedly greater or less than that given by the weighted mean of all other results, since we might so weight the latter as to give a result greater or less without transcending the bounds of judicious judgment. I conceive, therefore, that we are justified in reducing the mean error to i 0.26, which will give as the equation of condition r= - 0.055 i 0.25 and hence 40 x = - 0.22 i 1 (c) When, in the normal equation for the mass of Venus, given by the observations on Mercury, we substitute the values of the secular variations found from the general combination of 46, the result is v 1 = 0.0114 Combining this with the result from the Sun, we have v 1 = - 0.0117 In view of the fact that the mass derived from observations of Mercury may be affected by systematic errors of the kind 83] DEFINITIVE ADJUSTMENT. 171 shown and discussed in 53, the mean error formerly assigned to this result should be somewhat diminished. The result is 406 600 From this we have v' = + 0.0084 .0030 For the equation of condition of weight unity I take 330 v' = + 2.8 (d) With these equations of condition we have to combine the eleven equations of 63, which we use unchanged, except that we double the weight assigned, to the sixth equation, that derived from the motion of the node of Venus, on account of the smaller probable error of the result of our preceding redis- cussion, and use the value of the absolute term found in 80. If we accept the view that all the perihelia move according to the same law of gravitation toward the Sun, namely, that expressed by HALL'S hypothesis, then the value of the quan- tity 6 in the formula expressing the law of gravitation is so well determined by the motions of Mercury that it becomes legitimate to use the observed motions of the perihelia of the other three planets as equations of condition. But since it is not impossible that the minor planets between Mars and Jupiter may have an appreciable influence on the motion of the perihelion of Mars, it is a question whether we should not exclude that motion from the equations. The conditional equations given by the motions of the three perihelia in question are found by comparing the results of 46, 54, and 61. They are 40 x + v 1 + 20 v" = + 1.0 -14+46 +0 = - 0.3 (e) 2 - 13 +61 = + 0.7 The conditional equations to be combined are the eleven equations of 63, the sixth of which is to have double weight^ and the six equations (a), (c), (d), and (e). 172 DEFINITIVE ADJUSTMENT. [83 The normal equations to which we are thus led are the following, which show the results of the four combinations we may make according as we use (A) or (B) for the equation given by the mass of the Earth, and omit or include the third equation (d), which is given by the motion of the perihelion of Mars. (a.) Including the motion of the perihelion of Mars. 9 607,r 7 147 7' ; 11 335j/" = + 220 = -587 = - 3388 (A) = - 4328 (B) 7 147 + 267 174 + 168 727 11 335 + 168 727 + 406 300 1 1 335 , + 168 727 + 606 300 (/?.) Omitting the motion of the perihelion of Mars. 9 603# - 7 12lv' - 11 457 v" = + 219 - 7 121 + 267 003 + 169 520 = - 578 - 11 457 + 169 520 + 402 578 = - 3431 (A) - 11 457 + 169 520 + 602 578 = - 4371 (B) The results of the solutions in the four cases are: Aa A// B B/? X -f 0.0147 + 0.0142 + 0.0161 + 0.0158 V -j- 0.147 + 0.142 -f 0.161 + 0.158 V 1 + 0.004 34 + 0.004 60 + 0.003 10 + 0.003 i>.") v" 0.009 73 0.01005 - 0.007 70 0.007 87 1 -V- m 6 539 000 6 567 000 6 460 000 6 477 000 1 + m' 408 230 408 120 408 730 408 670 , 7T 8".783 8 // .782 8".789 8".788 I conceive that if the secular variations, especially the motion of the node of Venus, are not affected by any unknown cause, some mean between these should be regarded as the most probable solution. The result does not, however, bring about a satisfactory reconciliation. We still find ourselves confronted by this embarrassing dilemma: Either there is something abnormal in connection with the node of Venus, due to an unknown cause acting on the planet, to some extraordinary errors in the observations or their reduction, or to some error in the theory on which the discussion is based, or the deter- 83, 84, 85] ADOPTED PARALLAX AND MASSES. 173 ruinations of the solar parallax are nearly all in error in one direction by amounts which are, in more than one case, quite surprising. Possible causes of the observed discordances. 84. Two possible causes of discordance may be suggested, one of which has not been touched upon at all in the preceding chapters, and one perhaps inadequately. As to the hypothesis of non-sphericity of the Sun, considered in 56, it may be remarked that Dr. HARTZER shows that an ellipticity of the Sun sufficient to produce the observed motion of the perihelion of Mercury would cause a direct motion of 5".l in the motion of the node of Venus. This would correspond to a change of 0".30 in the value siniD t # and would therefore go far toward reconciling the discrepancy. But it is easy to see that this cause would produce a secular motion of 2".6 in the inclina- tion of Mercury. We have seen that the observed motion of the inclination already exceeds the theoretical motion by 0".38; so that introducing the hypothesis of ellipticity of the Sun we should have a discrepancy of about S^.O between theory and observation. This conclusion alone seems fatal to the theory, which otherwise has been shown to be scarcely tenable. The other possible cause is an inequality of long period ; especially one depending on the argument \3l" 81' which has a period of about two hundred and forty three years. A very simple computation shows that the coefficient of this term is only of the order of magnitude (V'.Ol. It is a curious coincidence that if we had neglected to add the mass of the Moon to that of the Earth, in computing the secular variations, the discrepancy would not have existed. Adopted values of the doubtful quantities. 85. The practical question which has been before the writer in working out the preceding results is : What values of the constants should be used in the tables of the celestial motions of which the results of this discussion are to form the basis ? Should we aim simply at getting the best agreement with obser- vations by corrections more or less empirical to the theory ? It seems to me very clear that this question should be answered in the negative. No conclusions could be drawn from future 174 ADOPTED PARALLAX AND MASSES. [85 comparisons of such tables with observations, except after reducing the tabular results to some consistent theory. The imposition of such a labor upon the future investigator is not to be thought of. Moreover, there is no certainty that the tables which would best represent past observations would also best represent future ones. Our tables must be founded on some perfectly consistent theory, as simple as possible, the elements of which shall be so chosen as best to represent the observations. In choosing the theory and its constants we have again a certain range. If we accept the necessity of assuming the secular variations of the orbits of Mercury and Venus to be affected by the action of unknown masses of matter, then the simplest course to adopt is to construct our theory on the sup- position of a planet or group of planets between Mercury and Venus. It seems to me that the introduction of the action of such a group into astronomical tables would not be justifiable. The more I have reflected upon the subject the more strongly seems to me the evidence that no such group can exist, and, indeed, that whatever anomalies exist can not be due to the action of unknown masses of matter. Besides, the six elements of such a group would constitute a complication in the tabular theory. On the other hand, it did not seem to me best that we should wholly reject the possibility of some abnormal action or some defect between the assumed relations of the various quanti- ties. What I finally decided on doing was to increase the theo- retical motion of each perihelion by the same fraction of the mean motion, a course which will represent the observations without committing- us to any hypothesis as to the cause of the excess of motion, though it accords with the result of HALL'S hypothesis of the law of gravitation ; to reject entirely the hypothesis of the action of unknown masses, and to adopt for the elements what we might call compromise values between those reached by the preceding adjustment and those which would exist if there is abnormal action. The exigency of hav- ing to prepare the tables required me to reach a conclusion on this subject before the final revision of the preceding discus- 85,86] FUTURE DETERMINATIONS. 175 sion, so that the numbers used are not wholly based upon it. The conclusions I have reached are these: Since, if there is nothing abnormal in the theory, the solar parallax is probably not much larger than 8".780, and if there is anything abnormal it is probably as large as 8".795 or even 8' '.800, we may adopt the value 8". 790 as one which is almost certainly too large on the one hypothesis and too small on the other, and which is therefore best adapted to afford a decision of the question. For the mass of Venus I took, as an intermediate value, m ' =1-1-408000 For the mass of Mercury I took 1 4- 6,000,000 Actually it seems that this mass is larger than the most prob- able one on either hypothesis, though not without the range of easy possibility. With these values the outstanding difference between theory and observation in the centennial motion of the node of Venus is A sin i D t = 0".25 If this difference arises wholly from the error of the theory, then between the transits of 1874 and 2004 the accumulated error would amount to 0".32 in the heliocentric latitude, and about 0".8 in the geocentric latitude. Unless an improvement is made in the method of determining the position of Venus by observation, the twentieth century must approach its end before this difference can be detected. Bearing of future determinations on the question. 86. The following shows the influence which subsequent determinations of the principal elements will have upon our judgment as to the solution of the dilemma. The changes in the second column will, by emphasizing the discordance between the results, tend to confirm the hypothesis of an abnormal defect in the theory, while the opposite ones, in the last column, will tend to reconcile theory and observation : USUVBRSITf 176 FUTURE DETERMINATIONS. [86 Element or quantity. Change tending to confirm the dis- cordance between theory and observation. Change tending to reconcile exist- ing theory with observation. The solar parallax. Increase. Diminution. Longitude of the node of Mercury. Increase. Diminution. Longitude of the node of Venus. Increase. Diminution. Constant of aberration. Diminution. Increase. Mass of Venus. Increase. Diminution. Mass of Mercury. Diminution. Increase. Secular diminution of the obliquity. Diminution. Increase. Among these constants are some the values of which can scarcely be decisively obtained except by observations con- tinued through half a century, or even through the whole twentieth century, unless improvements are made in our pres- ent methods of observing. The improvement of others, however, is quite within the reach of the astronomy of the present time. Among these the constant of aberration and the solar parallax have the first place. The more accurate determination of these quanti ties thus assumes an importance which may justify some sug- gestions on the subject. The observations made on the European continent for the detection and study of the variations of latitude have been executed with such precision that we might look to them for a marked improvement in the determination of the constant of aberration, were it not for a single circumstance. In the gen- eral average few are made after midnight, while the maxima and minima of aberration occur in the morning and evening. The extension of the system into the early morning therefore seems desirable. Although these observations may scarcely equal in accuracy those made by NYREN, with the prime 86] FUTURE DETERMINATIONS. 177 vertical transit, they have the advantage of not requiring so long a period for a complete observation. The great disad- vantage of the prime vertical instrument is that unless a star culminates within a few minutes of the zenith, an hour, or even several hours, will be required for the completion of a determination, which may thus be made impossible by the 'advent of daylight. It may be remarked in this connection that the northern latitudes of the European observatories are favorable to the determination of the aberration-constant. LOEWY'S method has over all others the great advantage of being independent of the direction of the vertical. But its application, and the reduction of the observations made with, it, are laborious in a high degree. So far as practical astronomy has yet developed, the best method of directly measuring planetary parallax, and there- fore the only one to be considered, is that of GILL. It there- fore seems desirable that measures by this method should be continued. At the same time it is very necessary that the spectra of the small planets to be used should be carefully studied photometrically, and that the probable influence of coloration upon the measures should be investigated. The necessity of completing the present work, and of pro- ceeding immediately to the construction of tables founded upon the adopted elements, prevent the author's awaiting the mature judgment of astronomers upon the embarrassing ques- tions thus raised. The regret with which he accepts this necessity is weakened by the consideration that even if the solar parallax which he has adopted requires the largest cor- rection to which it can reasonably be supposed subject, namely, one of 0".015, reducing the value of this constant to 8". 775, the effect of the error will not be prejudicial to the astronomy of the'immediate future. More important will be the error /x .035 in the constant of aberration. Yet a long-continued series of observations will be necessary to establish even the existence of such an error, and should it prove detrimental in any astronomical work the evil will be easily remedied by a slight correction. 5690 N ALM 12 CHAPTER IX. DERIVATION OF RESULTS. Ulterior corrections to the motions of the perihelion and mean . longitude of Mercury. 87. In 32 and 46 we have reached three values of the correction to the tabular motion of the perihelion of Mercury. Of these the first rests on meridian observations alone, the second on the combination of meridian observations with trans- its, and the third is derived by substituting in the eliminating equations the corrections to the solar elements and their secular variations which result from observations. The three values thus reached are 9".54, 1".01, and + 6".34. The pro- gressive divergence of these values, taken in connection with the discrepancy pointed out in 33, leads us to distrust the influence of the meridian observations upon the motion of the perihelion. Under these circumstances I deem it advisable to make such final corrections to the motions in mean longitude and mean anomaly as will best satisfy all the observed transits over the disk of the Sun. In doing this I am enabled to intro- duce the results of a preliminary discussion of the transits of 1891 and 1894. By combining the observations of these two transits with those of the older ones I derive the following values of the functions Y and W defined in 31 : // // Y = -1.93- 3.03 T W= + 1.50 + 2.04 T The preliminary theory, so far as yet investigated, gives for the values of this quantity, // // Y = - 2.44 3.40 T W = + 1.38 + 1.3GT 178 87, 88] PERIHELION OF MERCURY. 179 Equating these values to the corresponding linear functions of the corrections to /, TT, and their secular motions, we have the equations, // // 0.72 SI + 0.28 67f = + 0.12 + 0.68 T + 1 .49 - 0.49 = + 0.51 + 0.37 T We find, from these equations, // // 61 = +0.26 + 0.56 T Sn = - 0.24 + 0.97 T The preliminary values to which these corrections are appli- cable are // // 61 = +0.04- 1.33 T 6jr= + 5.83 + 6.34 T The definitive values thus become 61 = + 0.30- 0.77 T tf TT = + 5.59+ 7.31 T Definitive elements of the f out inner planets for the epoch 1850, as inferred from all the data of observation. 88. We have made a fourth solution of the normal equations which give the corrections to the elements of each planet by substituting in those equations the definitive values of all the other quantities, including the values of the secular variations derived from theory. In making this substitution for Mercury, however, the ulterior corrections just found were not applied. The values of the unknowns resulting from this solution are shown in the first column of the next table. From these numbers are derived the definitive elements for 1850, 'by the following processes: (a.) By multiplying the unknowns by the appropriate factor given in 27, we have the corrections of the tabular elements at the mid- epoch of observations for each planet. These cor- rections are found in the second column. (/?.) The preceding corrections are to be reduced from the respective mid-epochs to 1850. This reduction is found by 180 DEFINITIVE QUANTITIES. [88 multiplying the definitive correction to the tabular secular variation by the elapsed interval, and is shown in the third column. (y) We next have the value of the tabular elements for the fundamental epoch 1850, January 0, Greenwich mean noon. These numbers are those of LEVERRIER'S tables, with the following modifications: (d) The reduction from 1850, January 1, Paris noon, to January 0, Greenwich noon (f) The corrections to LEVERRIER'S values of the eccen- tricity and perihelion which are necessary to represent those terms in the perturbations of the mean longitude which depend only upon the sine and cosine of the mean anomaly. The theory is more symmetrical in form when all such terms are included with those of the elliptic motion. In LEVERRIER'S tables they have the following values: Mercury 5 6v = 4- 0.030 sin I - 0.111 cos I Venus; +0.010 +0.037 Earth; 0.067 -0.098 Mars; +1.061 +0.718 These terms of the longitude may be represented by the follow- ing corrections to the elements: Mercury; de = + 0.058 dn r= 0.0 Venus; -0.012 +2.3 Earth; +0.054 +1.4 Mars; +0.613 -1.0 Applying these corrections d and e to LEVERRIER'S tabular quantities, we have the values of the tabular elements as given in the fourth column. Then applying the preceding correc- tions we have the definitive values given in the last column. In some cases this derivation is modified. Instead of using the correction to the perihelion, mean longitude and mean motion of Mercury given by the unknown quantities of the 88] ELEMENTS FOR 1850. 181 equations, we have used the values for 1850 derived from the discussion of the preceding section. The quantities which give the position of the node and inclination have been treated in the same way as their secular variations. The symbols J and N indicate values of the unknown quantities related to the corrections of the elements J and N. These unknowns are then changed to corrections of the elements by the factors of 27, and these again to correc- tion of the inclination and node by the equations of 41. In the case of the node of Venus two values are given. The value (a) is that which follows immediately from the normal equations. If we carry forward the position of the node just derived to the mean epoch of the last two transits of Venus, we find a discrepancy amounting to 2".04 in the longitude, corresponding to a difference of 0".121 in the heliocentric lati- tude. This is considerably larger than the probable error of the results of the observations of the transits. It may, there- fore, be questioned whether the latter are not entitled to a greater relative weight than that assigned, owing to the prob- able systematic errors of the meridian observations. A second value (b) has therefore been derived from the observations of the transits alone. In subsequent investigations we may choose between these two values. Formation of definitive elements of the four inner planets, for tlit\ epoch 1850 7 January 0, Oreemvich mean noon. Mercury. Unknown of Corr. of Red. to Tabular Concluded equations. element. 1850. element. element. // // // // n -.0940 - 0.77 0.0 538106654.49 538106653.72 e - .0741 - 0.222 - 0.005 42 409.088 42 408.861 n + .6763 -1- 5.59 75 7 13.78 75 7 19'!s7 t .0402 + 0.30 323 11 23.53 323 11 23.83 i -.2762J-- 0.64 - 0.07 7 7.71 7 7.00 d -.0001N+ 3.88 - 0.27 46 33 8.63 46 33 12.24 182 DEFINITIVE QUANTITIES. [88,89 Formation of definitive elements, etc. Continued, Venus. Unknown of Corr. of Ked. to Tabular Concluded equations, element. 1850. element. element. n - .1783 - 3.57 210669165.04 210669161.47 e + .1463 0.439 - 0.165 1 411.522 1 411.796 129 7t + .0835 + 36.6 16.4 z _ .1330 - 0.67 + 0.46. 243 i + .0968 J + 0.31 + 0.12 3 0(a)+ .0126 N- 9.39 + 6.63 75 0(b) -20.36 +15.56 Earth. 27 57 23 19 14.3 129 44.34 243 34.83 3 52.21 75 75 27 57 23 19 19 34.5 44.13 35.26 49.45 47.41 1.10 0.12 2.4 0.15 0.02 129 602 767.84 129 602 766.74 3 459.334 100 21 43.4 23 27 31.83 99 48 18.72 Mars. - .1094 - 0.88 68 910 105.38 - .1088 - 0.155 + 0.058 19 237.101 + .1663 + 2.38 + 0.02 333 3 459.454 100 21 41.0 23 27 31.68 99 48 18.74 68 910 104.50 19 237.004 - .4029 - 0.81 + 0.05 83 - .0507 J + 0.18 - 0.01 1 + . 1135 N+ 6.56 +1.34 48 17 9 51 23 52.47 16.92 2.28 53.02 333 83 1 48 17 9 51 24 54.87 16.16 2.45 0.92 Definitive values of the secular variations. 89. The definitive values of the secular variations, as inferred from the adopted theories and the concluded values of the masses, are shown in the following table, which gives in detail the parts of which each quantity is made up. The first four lines of the table give the values of the secular variations as they result from the investigations found in Vol. V, Part IV, of the Astronomical Papers, after correcting the mass of each planet by its appropriate factor. The motion of the perihelion first given, denoted by D t n\, is measured along the plane of the orbit itself. The numbers 89 1 SECULAR VARIATIONS. 183 given being divided by the corresponding values of the eccen- tricity we have the motion of the perihelion itself along the plane. The symbols i and # represent the inclinations and longitudes of the nodes referred at each epoch to the ecliptic and equinox of 1850, regarded as fixed. The motions of these elements are next to be referred to the fixed ecliptic of the date. So referred, they are designated as D? i and D? 6. The transformations to the latter quantities are made by comput- ing an approximate value of the motion of the node due to the motion of the ecliptic alone along the plane of the orbit regarded as fixed. If we put ,, the inclination of the fixed orbit of the planet at any epoch TO to the moving ecliptic at any time; 61, the longitude of the corresponding node, h; F, the distance from the node Q t to the instantaneous rota- tion axis of the orbit at the epoch T ; we shall have D t v = H" cosec i\ sin (L" #1) (a] If we compute v and H from the equations H sin VQ = sin i D? H cos r = D? i and then find Av by integrating the value (a) of D t r from 1850 to the date we shall have sin i D? # = H sin ( V Q + A v] D i = H cos (v Q -f Av) The change of D t ^ between 1850 and the extreme epochs has been found so nearly uniform that it was sufficient to multiply its value at the mid-epoch (1675 or 1975) by 2.5 to obtain Av. Next, we have the changes in i and due to the motion of the ecliptic, represented by T>]i and Df0, and computed by the formula D l t i= H f/ GOS(r /f -0) sin i D[ 6 = H" cos i sin (v" 0) 184 DEFINITIVE QUANTITIES. [89 The planetary precession due to the motion of the ecliptic is here omitted, to be afterwards included in the general preces- sion. The sum of the two motions gives the actual variation at each epoch, referred to a fixed equinox. The motion of 6 itself thus found is increased by the general precession, which gives the motion of 6 at each epoch. The motion of the perihelion to be actually used in the tables is equal to the motion of the node from the mean equinox, plus the increase of the arc of the orbit between the node and perihelion. The adopted value of this quantity is found by increasing the motion of n\ by the following quantities: 1. The change due to the motion of the plane of the orbit. 2. The change due to the motion of the ecliptic. The formulae for these two quantities are (1) ; d] D t n = tan J i$m i D? d (2) 5 ^ + 0.08 6621.51 + 0.07 6623.96 + 0.06 6626.25 D t 2776.39 2776.87 2776.63 Secular acceleration of the mean motions. 90. The mean motions of the planets, like that of the Moon, are subject to a secular acceleration arising from the secular vari- ations of the elements of the orbits. The following formulae for this acceleration are formed by differentiating the known 90] SECULAR ACCELERATIONS. 187 expressions for the variation of the longitude of the epoch in the theory of the variation of elements. The notation is that of Astronomical Papers, Vol. V, Part IV. We compute for the action of an outer on an inner planet: A = D <*\ } B = - (D - D 2 2 D 3 ) c ( > 8 W- - (2 - 9 D + 3 D 2 + 4 D 3 ) c^ 8 Then D; = w' a n D t { A and A. The spherical triangle P E E gives the follow- ing equations: sin A, sin 77! sin 7I sin k " sin e i sin e During a period of several centuries the quantities k and A are so small that no distinction is necessary between them and their sines. We may therefore put A = k sin 77 t cosec 1 = k sin 77 cosec e (b) We also have, from the law of motion of the pole of the equator, D t t = n sin A D t fy = n cos A cosec e\ 102] MOTION OF THE EQUATOR. 199 As the value of 81 does not change by 0".6 from one epoch to another, we may, without appreciable error, use f for ^ in the formulae (b) and (c). To use these equations, we first obtain k and 77! from the secular motion of the ecliptic, while n is com- puted for any epoch from the formula (a). We then easily develop the values of s-i and ip in powers of the time by the equations (c). The values of n have no reference to any special coordinates. From the table ot 100 it will be seen that we may put n = 2004".79 - 2".13 r' r' being counted from 1850. To find the value of III in each case, we remark that the instantaneous values of L given in 100 show that the instan- taneous node, or intersections of two consecutive ecliptics, moves with so near an approach to uniformity that we may take for the actual node between the ecliptics of any two epochs TI and r 2 the mean of the instantaneous nodes for those two epochs. For example, let it be required to find the value of 77j for the node of the ecliptic of 2100 on that of 1850. We have For 2100 For 1850, referred to eq. of 2100 Concluded value of 77i ... L = 175 46.63 L = 176 59.13 77! = 176 22.9 As the basis of our work we have computed the required quantities for the zero ecliptics of 1600, 1850, and 2100, respectively. The values of k and 77i for the ecliptics of two hundred and fifty years before and after these epochs are as follows : Zero epoch. 1600 1850 2IOO 250 Y + 2 5 OY k n, k n, ff -118.48 118.07 -117.64 / 1 68 20.0 170 36.7 172 53-4 // + 118.07 + 117.64 + 117-23 / 174 5-9 176 22. 9 178 39.9 200 DEFINITIVE QUANTITIES. [102 Changing the. unit of time to two hundred and fifty years, the equations (a) (b) and (c) give the following values of the derivatives of fi and : Zero-epoch. 250 Y +250Y 250 Y +250Y 1600 _ 1.4636 + 0.7400 12600.33 12573.65 1850 -1.1768 +0.4527 12603.44 12576.65 2100 -0.8898 +0.1665 12606.57 12579.71 At the respective epochs T> T \ vanishes, and Dr# has the values of the luuisolar precession in longitude ( 100). Developing in powers of r we have- the following results: Zero-epoch. o / // // // 1600; e l = 23 29 28.69-+ 0.5509 r 2 - 0.1206 r 3 1850; ei = 23 27 31.68 + 0.4074 - 0.1207 2100; t = 23 25 34.56 + 0.2641 - 0.1206 1600; # = 12587.00 T - 6.67 r 2 1850; # = 12590.05 -6.70 2100; # = 12593.14 -6.72 // // 1600; A = 45.28 r - 14.83 T* 1850; A = 33.52 - 14.86 2100; A = 21.75 -14.88 These values of Si and # completely fix the position of the equator at the time T relative to the zero ecliptic and equinox. For the reduction of coordinates from one epoch to another we must express the position of the equator at the time r. We consider the triangle PE P , of which the sides and opposite angles are designated Sides, fi Opposite angles, 90 - C 90 , # If, in the Gaussian relations between the parts of this triangle, we put sin A (f, Q } = A (fi e) = A z/ 102] MOTION OF THE EQUATOR. 201 and regard the cosine of this angle as unity, we have tan J (C + Ci) = cos J (e t + e ) tan ip If we develop the differences between the tangent and the arc we find from these equations r + c, = ^ cos A (fi + f ) (1 - where we put z for the approximate value of C Ci For the inclination 6 of the mean equator of the epoch r to the zero equator, we have the equation sin 9 = cos and then, by developing in powers of 6 and ^, we find = ip sin f (1 + 4 C 2 ) (1 i ^ 2 cos 2 f ) We thus find Zero-epoch. // // // 1600; C + Ci = 11543.79 T - 6.12 r 2 + 0.57 r 3 1850; 11549.44 - 6.L4 +0.57 2100; 11555.12 -6.16 +0.58 1600; C - Ci = 45.29 r - 9.92 r 2 1850; 33.53 -9.93 2100; 21.76 - 9.94 1600; e = 5017.30 T - 2.66 r 2 - 0.64 r 3 1850; 5011.97 - 2.67 - 0.64 21CO; 5006.64 -2.67 -0.65 To show the significance of the preceding quantities, con sider once more the spherical quadrangle P E EP. Let these 202 DEFINITIVE QUANTITIES. [102 letters represent the positions of the poles on the celestial sphere at any two epochs. In this quadrangle we shall have Angle E P E = 90 - Ci Angle E P P = 90 - C + A SideP P = # Let S be the position of a star on the celestial sphere. Its polar distances at the two epochs will be P S and P S and its Eight Ascensions will be determined by the angles P and P of the triangle S P P. Thus, if the Eight Ascension and Declination of S are given for one epoch, we can find it for the other epoch by the solu- tion of the triangle S P P when we have given the values of the quantities 0, Ci, and -f A. To find the values of these quantities from the preceding formula, let T be the zero-epoch, expressed in calendar years, and let -c be the interval between the two epochs, taken posi- tively when the zero-epoch is the earlier one, and negatively when it is the later one. We interpolate the coefficients of r and its powers from the preceding formula to the epoch T. Then by substituting the value of r in the formula we shall have the values of the required quantities, and hence the data for reducing the position of S from one epoch to the other. UNIVERSITY OF CALIFORNIA LIBRARY, BERKELEY THIS BOOK IS DUE ON THE LAST DATE STAMPED BELOW Books not returned on time are subject to a fine of 50c per volume after the third day overdue, increasing to $1.00 per volume after the sixth day. Books not in demand may be renewed if application is made before expiration of loan period. MAT REC'D LD DEC 19 1957 otf JUN 2 8EC.CB.ltOf 5T8 1948 75m-7,'30 U.C.BERKELEY LIBRARIES COL13S63L7 UNIVERSITY OF CALIFORNIA LIBRARY