THEORETICAL ASTRONOMY RELATING TO THE MOTIONS OF THE HEAVENLY BODIES REVOLVING AROUND THE SUN IN ACCORDANCE WITH THE LAW OF UNIVERSAL GRAVITATION EMBRACING A SYSTEMATIC DERIVATION OF THE FORMULAE FOR THE CALCULATION OF TIlE GEOCENTRIC AND HELIOCENTRIC PLACES, FOR THE DETERIINATION OF THE ORBITS OF PLANETS AND COMETS, FOR THE CORRECTION OF APPROXIMATE ELEMENTS, AND FOR THE COM'PUTATION OF SPECIAL PERTURBATIONS; TOGETHER WITH THE THEORY OF THE COMBINATION OF OBSERVATIONS AND THE METHOD OF LEAST SQUARES. ZltZ1 lumtmervral'taampl)es anlt 3uxmliary FTal el BY JAMES C. WATSON DIRECTOR OF THE. OBSERVATORY AT ANN ARBOR, AND PROFESSOR OF ASTRONOMY IN THE UNIVERSITY OF MICHIGAN - PHILADELPHIA J. B. LIPPINCOTT & CO. LONDON: TRUBNER & CO. 1868 Entered, according to Act of Congress, in the year 1868, by J. B. LIPPINCOTT & CO., in the Clerk's Office of the District Court of the United States for the Eastern District of Pennsylvania. PREFACE. THE discovery of the great law of nature, the law of gravitation, by NEWTN, prepared the way for the brilliant achievements which have distinguished the history of astronomical science. A first essential, however, to the solution of those recondite problems which were to exhibit the effect of the mutual attraction of the bodies of our system, was the development of the infinitesimal calculus; and the labors of those who devoted themselves to pure analysis have contributed a most important part in the attainment of the high degree of perfection which characterizes the results of astronomical investigations. Of the earlier efforts to develop the great results following from the law of gravitation, those of EULER stand pre-eminent, and the memoirs which he published have, in reality, furnished the germ of all subsequent investigations in celestial mechanics. In this connection also the names of BERNOUILLI, CLAIRAUT, and D'ALEMBERT deserve the most honorable mention as having contributed also, in a high degree, to give direction to the investigations which were to unfold so many mysteries of nature. By means of the researches thus inaugurated, the great problems of mechanics were successfully solved, many beautiful theorems relating to the planetary motions demonstrated, and many useful formulae developed. It is true, however, that in the early stage of the science methods were developed which have since been found to be impracticable, even if not erroneous; still, enough was effected to direct attention in the proper channel, and to prepare the way for the more complete labors of LAGRANGE and LAPLACE. The genius and the analytical skill of these extraordinary men gave to the progress of Theoretical Astronomy the most rapid strides; and the intricate investigations which they successfully performed, served constantly to educe new discoveries, so that of all the problems relating to the mutual attraction of the several planets 3 4 PREFACE. but little more remained to be accomplished by their successors than to develop and simplify the methods which they made known, and to introduce such modifications as should be indicated by experience or rendered possible by the latest discoveries in the domain of pure analysis. The problem of determining the elements of the orbit of a comet moving in a parabola, by means of observed places, which had been considered by NEWTON, EULER, BOSCOVICH, LAMBERT, and others, received from LAGRANGE and LAPLACE the most careful consideration in the light of all that had been previously done. The solution given by the former is analytically complete, but far from being practically complete; that given by the latter is especially simple and practical so far as regards the labor of computation; but the results obtained by it are so affected by the unavoidable errors of observation as to be often little more than rude approximations. The method which was found to answer best in actual practice, was that proposed by OLBERS in his work entitled Leichteste und bequemste Methode die Bahn eines Cometen zu berechnen, in which, by making use of a beautiful theorem of parabolic motion demonstrated by EULER and also by LAMBERT, and by adopting a method of trial and error in the numerical solution of certain equations, he was enabled to effect a solution which could be performed with remarkable ease. The accuracy of the results obtained by OLBERS'S method, and the facility of its application, directed the attention of LEGENDRE, IVORY, GAUSS, and ENCKE to this subject, and by them the method was extended and generalized, and rendered applicable in the exceptional cases in which the other methods failed. It should be observed, however, that the knowledge of one element, the eccentricity, greatly facilitated the solution; and, although elliptic elements had been computed for some of the comets, the first hypothesis was that of parabolic motion, so that the subsequent process required simply the determination of the corrections to be applied to these elements in order to satisfy the observations. The more difficult problem of determining all the elements of planetary motion directly from three observed places, remained unsolved until the discovery of Ceres by PIAZZI in 1801, by which the attention of GAUSS was directed to this subject, the result of which was the subsequent publication of his Theoria Motus Corporum Coelestium,-_a most able work, in which he gave to the world, in a finished form, the results of many years of attention PREFACE. 5 to the subject of which it treats. His method for determining all the elements directly from given observed places, as given in the Theoric Motus, and as subsequently given in a revised form by ENCKE, leaves scarcely any thing to be desired on this topic. In the same work he gave the first explanation of the method of least squares, a method which has been of inestimable service in investigations depending on observed data. The discovery of the minor planets directed attention also to the methods of determining their perturbations, since those applied in the case of the major planets were found to be inapplicable. For a long time astronomers were content simply to compute the special perturbations of these bodies from epoch to epoch, and it was not until the commencement of the brilliant researches by HANSEN that serious hopes were entertained of being able to compute successfully the general perturbations of these bodies. By devising an entirely new mode of considering the perturbations, namely, by determining what may be called the perturbations of the time, and thus passing from the undisturbed place to the disturbed place, and by other ingenious analytical and mechanical devices, he succeeded in effecting a solution of this most difficult problem, and his latest works contain all the formulae which are required for the cases actually occurring. The refined and difficult analysis and the laborious calculations involved were such that, even after HANSEN'S methods were made known, astronomers still adhered to the method of special perturbations by the variation of constants as developed by LAGRANGE. The discovery of Astrcea by HENCKE was speedily followed by the discovery of other planets, and fortunately indeed it so happened that the subject of special perturbations was to receive a new improvement. The discovery by BOND and ENCKE of a method by which we determine at once the variations of the rectangular co-ordinates of the disturbed body by integrating the fundamental equations of motion by means of mechanical quadrature, directed the attention of HANSEN to this phase of the problem, and soon after he gave formule for the determination of the perturbations of the latitude, the mean anomaly, and the logarithm of the radius-vector, which are exceedingly convenient in the process of integration, and which have been found to give the most satisfactory results. The formulae for the perturbations of the latitude, 6 PREFACE. true longitude, and radius-vector, to be integrated in the same manner, were afterwards given by BRjNNOW. Having thus stated briefly a few historical facts relating to the problems of theoretical astronomy, I proceed to a statement of the object of this work. The discovery of so many planets and comets has furnished a wide field for exercise in the calculations relating to their motions, and it has occurred to me that a work which should contain a development of all the formule required in determining the orbits of the heavenly bodies directly from given observed places, and in correcting these orbits by means of more extended discussions of series of observations, including also the determination of the perturbations, together with a complete collection of auxiliary tables, and also such practical directions as might guide the inexperienced computer, might add very materially to the progress of the science by attracting the attention of a greater number of competent computers. Having carefully read the works of the great masters, my plan was to prepare a complete work on this subject, commencing with the fundamental principles of dynamics, and systematically treating, from one point of view, all the problems presented. The scope and the arrangement of the work will be best understood after al examination of its contents; and let it suffice to add that I have endeavored to keep constantly in view the wants of the computer, providing for the exceptional cases as they occur, and giving all the formule which appeared to me to be best adapted to the problems under consideration. I have not thought it worth while to trace out the geometrical signification of many of the auxiliary quantities introduced. Those who are curious in such matters may readily derive many beautiful theorems from a consideration of the relations of some of these auxiliaries. For convenience, the formula are numbered consecutively through each chapter, and the references to those of a preceding chapter are defined by adding a subscript figure denoting the number of the chapter. Besides having read the works of those who have given special attention to these problems, I have consulted the Astronomische Nachrichtein, the Astronomical Journal, and other astronomical periodicals, in which is to be found much valuable information resulting from the experience of those who have been or are now actively engaged in astronomical pursuits. I must also express my obligations to the publishers, PREFACE. 7 Messrs. J. B. LIPPINCOTT & CO., for the generous interest which they have manifested in the publication of the work, and also to Dr. B. A. GOULD, of Cambridge, Mass., and to Dr. OPPOLZER, ofVienna, for valuable suggestions. For the determination of the time from the perihelion and of the true anomaly in very eccentric orbits I have given the method proposed by BESSEL in the Monatliche Correspondenz, vol. xii.,-the tables for which were subsequently given by BRtNNOW in his Astronomical Notices,-and also the method proposed by GAUSS, but in a more convenient form. For obvious reasons, I have given the solution for the special case of parabolic motion before completing the solution of the general problem of finding all of the elements of the orbit by means of three observed places. The differential formula and the other formule for correcting approximate elements are given in a form convenient for application, and the formule for finding the chord or the time of describing the subtended arc of the orbit, in the case of very eccentric orbits, will be found very convenient in practice. I have given a pretty full development of the application of the theory of probabilities to the combination of observations, endeavoring to direct the attention of the reader, as far as possible, to the sources of error to be apprehended and to the most advantageous method of treating the problem so as to eliminate the effects of these errors. For the rejection of doubtful observations, according to theoretical considerations, I have given the simple formula, suggested by CHAUVENET, which follows directly from the fundamental equations for the probability of errors, and which will answer for the purposes here required as well as the more complete criterion proposed by PEIRCE. In the chapter devoted to the theory of special perturbations I have taken particular pains to develop the whole subject in a complete and practical form, keeping constantly in view the requirements for accurate and convenient numerical application. The time is adopted as the independent variable in the determination of the perturbations of the elements directly, since experience has established the convenience of this form; and should it be desired to change the independent variable and to use the differential coefficients with respect to the eccentric anomaly, the equations between this function and the mean motion will enable us to effect readily the required transformation. 8 PREFACE. The numerical examples involve data derived from actual observations, and care has been taken to make them complete in every respect, so as to serve as a guide to the efforts of those not familiar with these calculations; and when different fundamental planes are spoken of, it is presumed that the reader is familiar with the elements of spherical astronomy, so that it is unnecessary to state, in all cases, whether the centre of the sphere is taken at the centre of the earth, or at any other point in space. The preparation of the Tables has cost me a great amount of labor, logarithms of ten decimals being employed in order to be sure of the last decimal given. Several of those in previous use have been recomputed and extended, and others here given for the first time have been prepared with special care. The adopted value of the constant of the solar attraction is that given by GAuss, which, as will appear, is not accurately in accordance with the adoption of the mean distance of the earth from the sun as the unit of space; but until the absolute value of the earth's mean motion is known, it is best, for the sake of uniformity and accuracy, to retain GAUss's constant. The preparation of this work has been effected amid many interruptions, and with other labors constantly pressing me, by which the progress of its publication has been somewhat delayed, even since the stereotyping was commenced, so that in some cases I have been anticipated in the publication of formulae which would have here appeared for the first time. I have, however, endeavored to perform conscientiously the self-imposed task, seeking always to secure a logical sequence in the development of the formulTe, to preserve uniformity and elegance in the notation, and to elucidate the successive steps in the analysis, so that the work may be read by those who, possessing a respectable mathematical education, desire to be informed of the means by which astronomers are enabled to arrive at so many grand results connected with the motions of the heavenly bodies, and by which the grandeur and sublimity of creation are unveiled. The labor of the preparation of the work will have been fully repaid if it shall be the means of directing a more general attention to the study of the wonderful mechanism of the heavens, the contemplation of which must ever serve to impress upon the mind the reality of the perfection of the OMNIPOTENT, the LIVING GOD! OBSERVATORY, ANN ARBOR, June, 1867. CONTENTS. THEORETICAL ASTRONOMY. CHAPTER I. INVESTIGATION OF THE FUNDAMENTAL EQUATIONS OF MOTION, AND OF THE FORMULAE FOR DETERMINING, FROM KNOWN ELEMENTS, THE HELIOCENTRIC AND GEOCENTRIC PLACES OF A HEAVENLY BODY, ADAPTED TO NUMERICAL COMPUTATION FOR CASES OF ANY ECCENTRICITY WHATEVER. PAGE Fundamental Principles......................................................................... 15 Attraction of Spheres....................................................................... 19 Motions of a System of Bodies................................................................. 23 Invariable Plane of the System................................................................ 29 Motion of a Solid Body........................................................................... 31 The Units of Space, Time, and Mass........................................................ 36 Motion of a Body relative to the Sun............................................... 38 Equations for Undisturbed Motion............................................................ 42 Determination of the Attractive Force of the Sun....................................... 49 Determination of the Place in an Elliptic Orbit.......................................... 53 Determination of the Place in a Parabolic Orbit........................................ 59 Determination of the Place in a Hyperbolic Orbit....................................... 65 Methods for finding the True Anomaly and the Time from the Perihelion in the case of Orbits of Great Eccentricity.................................................... 70 Determination of the Position in Space...................................................... 81 Heliocentric Longitude and Latitude......................................................... 83 Reduction to the Ecliptic........................................................................ 85 Geocentric Longitude and Latitude..................................................... 86 Transformation of Spherical Co-ordinates.................................................. 87 Direct Determination of the Geocentric Right Ascension and Declination........ 90 Reduction of the Elements from one Epoch to another................................ 99 Numerical Examples............................................................................ 103 Interpolation........................................................................................ 112 Time of Opposition............................................................................... 114 9 10 CONTENTS. CHAPTER II. INVESTIGATION OF THE DIFFERENTIAL FORMULAE WHICH EXPRESS THE RELATION BETWEEN THE GEOCENTRIC OR HELIOCENTRIC PLACES OF A HEAVENLY BODY AND THE VARIATIONS OF THE ELEMENTS OF ITS ORBIT. PAGE Variation of the Right Ascension and Declination....................................... 118 Case of Parabolic Motion....................................................................... 125 Case of Hyperbolic Motion.................................................................... 128 Case of Orbits differing but little from the Parabola...................................... 130 Numerical Examples.............................................................................. 135 Variation of the Longitude and Latitude.................................................... 143 The Elements referred to the same Fundamental Plane as the Geocentric Places 149 Numerical Example...................................................................... 150 Plane of the Orbit taken as the Fundamental Plane to which the Geocentric Places are referred................................................................ 153 Numerical Example................................................................................ 159 Variation of the Auxiliaries for the Equator............................................... 163 CHAPTER III. INVESTIGATION OF FORMULAE FOR COMPUTING THE ORBIT OF A COMET MOVING IN A PARABOLA, AND FOR CORRECTING APPROXIMATE ELEMENTS BY THE VARIATION OF THE GEOCENTRIC DISTANCE. Correction of the Observations for Parallax................................................ 167 Fundamental Equations................................................................. 169 Particular Cases........................................................................... 172 Ratio of Two Curtate Distances............................................................. 178 Determination of the Curtate Distances...................................................... 181 Relation between Two Radii-Vectores, the Chord joining their Extremities, and the Time of describing the Parabolic Arc..............1...8....................... 184 Determination of the Node and Inclination................................................ 192 Perihelion Distance and Longitude of the Perihelion................................... 194 Time of Perihelion Passage..................................................................... 195 Numerical Example......................................................................... 199 Correction of Approximate Elements by varying the Geocentric Distance........ 208 Numerical Example............................................................................ 213 CHAPTER IV. DETERMINATION, FROM THREE COMPLETE OBSERVATIONS, OF THE ELEMENTS OF THE ORBIT OF A HEAVENLY BODY, INCLUDING THE ECCENTRICITY OR FORM OF THE CONIC SECTION. Reduction of the Data............................................................................. 220 Corrections for Parallax................................................................. 223 CONTENTS. 11 PAGE Fundamental Equations............................................................................ 225 Formule for the Curtate Distances............................................................ 228 Modification of the Formulae in Particular Cases.......................................... 231 Determination of the Curtate Distance for the Middle Observation.................. 236 Case of a Double Solution........................................................................ 239 Position indicated by the Curvature of the Observed Path of the Body............ 242 Formule for a Second Approximation.................................................. 243 Formule for finding the Ratio of the Sector to the Triangle........................... 247 Final Correction for Aberration......................................................... 257 Determination of the Elements of the Orbit................................................ 259 Numerical Example................................................................................ 264 Correction of the First Hypothesis........................................................... 278 Approximate Method of finding the Ratio of the Sector to the Triangle........... 279 CHAPTER V. DETERMINATION OF THE ORBIT OF A HEAVENLY BODY FROM FOUR OBSERVATIONS, OF WHICH THE SECOND AND THIRD MUST BE COMIPLETE. Fundamental Equations.2........................................................................ 282 Determination of the Curtate Distances...................................................... 289 Successive Approxim ations...................................................................... 293 Determination of the Elements of the Orbit............................................... 294 Numerical Example................................................................................ 294 Method for the Final Approximation...................................................... 307 CHAPTER VI. INVESTIGATION OF VARIOUS FORMULA FOR THE CORRECTION OF THE APPROXIMATE ELEMENTS OF THE ORBIT OF A HEAVENLY BODY. Determination of the Elements of a Circular Orbit...................................... 311 Variation of Two Geocentric Distances...................................................... 313 Differential Formul.............................................................................. 318 Plane of the Orbit taken as the Fundamental Plane...................................... 320 Variation of the Node and Inclination...................................................... 324 Variation of One Geocentric Distance...................................................... 328 Determination of the Elements of the Orbit by means of the Co-ordinates and V elocities....................................................................................... 332 Correction of the Ephemeris.................................................................... 335 Final Correction of the Elements........................................................ 338 Relation between Two Places in the Orbit................................................. 339 Modification when the Semi-Transverse Axis is very large............................ 341 Modification for Hyperbolic Motion..................................................... 346 Variation of the Semi-Transverse Axis and Ratio of Two Curtate Distances..... 349 12 CONTENTS. PAGE Variation of the Geocentric Distance and of the Reciprocal of the Semi-Transverse Axis....................................................................................... 352 Equations of Condition............................................................ 353 Orbit of a Comet.................................................................................... 355 Variation of Two Radii-Vectores............................................ 357 CHAPTER VII. METHOD OF LEAST SQUARES, THEORY OF THE COMBINATION OF OBSERVATIONS, AND DETERMINATION OF THE MOST PROBABLE SYSTEM OF ELEMENTS FROM A SERIES OF OBSERVATIONS. Statement of the Problem....................................................................... 360 Fundamental Equations for the Probability of Errors................................... 362 Determination of, the Form of the Function which expresses the Probability... 363 The Measure of Precision, and the Probable Error...................................... 366 Distribution of the Errors........................................................................ 367 The Mean Error, and the Mean of the Errors.............................................. 368 The Probable Error of the Arithmetical Mean............................................ 370 Determination of the Mean and Probable Errors of Observations................... 371 Weights of Observed Values........................................................................ 372 Equations of Condition............................................................................ 376 Normal Equations.................................................................................. 378 Method of Elimination.......................................................................... 380 Determination of the Weights of the Resulting Values of the Unknown Quantities................................................................................................ 386 Separate Determination of the Unknown Quantities and of their Weights........ 392 Relation between the Weights and the Determinants..................................... 396 Case in which the Problem is nearly Indeterminate...................................... 398 Mean and Probable Errors of the Results................................................... 399 Combination of Observations................................................................... 401 Errors peculiar to certain Observations...................................................... 408 Rejection of Doubtful Observations............................................................ 410 Correction of the Elements....................................................................... 412 Arrangement of the Numerical Operations.................................................. 415 Numerical Example.....................................4.............................. 418 Case of very Eccentric Orbits................................................................... 423 CHAPTER VIII. INVESTIGATION OF VARIOUS FORMULAE FOR THE DETERMINATION OF THE SPECIAL PERTURBATIONS OF A HEAVENLY BODY. Fundamental Equations......................................................................... 426 Statement of the Problem......................................................................... 428 Variation of Co-ordinates.............................................................. 429 CONTENTS. 13 PAGE Mechanical Quadrature...........................4................................ 433 The Interval for Quadrature.................................................................. 443 Mode of effecting the Integration.............................................................. 445 Perturbations depending on the Squares and Higher Powers of the Masses...... 446 Numerical Example................................................................................ 448 Change of the Equinox and Ecliptic......................................................... 455 Determination of New Osculating Elements................................................ 459 Variation of Polar Co-ordinates........................................................ 462 Determination of the Components of the Disturbing Force........................... 467 Determination of the Heliocentric or Geocentric Place............................... 471 Numerical Example.............................................................................. 474 Change of the Osculating Elements............................................................ 477 Variation of the Mean Anomaly, the Radius-Vector, and the Co-ordinate...... 480 Fundamental Equations.................................................................... 483 Determination of the Components of the Disturbing Force............................ 489 Case of very Eccentric Orbits........................................................ 493 Determination of the Place of the Disturbed Body.................................. 495 Variation of the Node and Inclination..................................................... 502 Numerical Example............................................................................... 505 Change of the Osculating Elements........................................................... 510 Variation of Constants............................................................................ 516 Case of very Eccentric Orbits.................................................................. 523 Variation of the Periodic Time................................................................ 526 Numerical Example............................................................................. 529 Formulne to be used when the Eccentricity or the Inclination is small.............. 533 Correction of the Assumed Value of the Disturbing Mass.............................. 535 Perturbations of Comets................................. 536 Motion about the Common Centre of Gravity of the Sun and Planet............... 537 Reduction of the Elements to the Common Centre of Gravity of the Sun and Planet............................................................................................ 538 Reduction by means of Differential Formulae............................................. 540 Near Approach of a Comet to a Planet...................................................... 546 The Sun may be regarded as the Disturbing Body........................................ 548 Determination of the Elements of the Orbit about the Planet........................ 550 Subsequent Motion of the Comet............................................... 551 Effect of a Resisting Medium in Space...................................................... 552 Variation of the Elements on account of the Resisting Medium..................... 554 Method to be applied when no Assumption is made in regard to the Density of e the Ether......................................... 556 f14 CONTENTS. TABLES. PAGE I. Angle of the Vertical and Logarithm of the Earth's Radius.............. 561 II. For converting Intervals of Mean Solar Time into Equivalent Intervals of Sidereal Time.................................................................. 563 III. For converting Intervals of Sidereal Time into Equivalent Intervals of Mean Solar Time............................................................. 564 IV. For converting Hours, Minutes, and Seconds into Decimals of a Day... 565 V. For finding the Number of Days from the Beginning of the Year...... 565 VI. For finding the True Anomaly or the Time from the Perihelion in a Parabolic Orbit............................................................. 566 VII. For finding the True Anomaly in a Parabolic Orbit when v is nearly 180~ 611 VIII. For finding the Time from the Perihelion in a Parabolic Orbit........... 612 TX. For finding the True Anomaly or the Time from the Perihelion in Orbits of Great Eccentricity............................................................ 614 X. For finding the True Anomaly or the Time from the Perihelion in Elliptic and Hyperbolic Orbits.................................................. 618 XI. For the Motion in a Parabolic Orbit................................. 619 XII. For the Limits of the Roots of the Equation sin (z'- ) mvo sin4 a'.. 622 XIII. For finding the Ratio of the Sector to the Triangle........................... 624 XIV. For finding the Ratio of the Sector to the Triangle........................... 629 XV. For Elliptic Orbits of Great Eccentricity............................. 632 XTVI. ForHyperbolic Orbits................................................................ 632 XVI. For Hyperbolic Orbits. 632 XVII. For Special Perturbations........................................................... 633 XVIII. Elements of the Orbits of the Comets which have been observed......... 638 XIX. Elements of the Orbits of the Minor Planets................................... 646 XX. Elements of the Orbits of the Major Planets................................... 648 XXI. Constants, &c............................................................................. 649 EXPLANATION OF THE TABLES............................................................... 651 APPENDIX. —Precession............................................................................ 657 Nutation........................................................................... 658 A berration.......................................................................... 659 Intensity of Light............................................................... 660 Numerical Calculations......................................................... 662 THEORETICAL ASTRONOMY. CHAPTER I. INVESTIGATION OF THE FUNDAMENTAL EQUATIONS OF MOTION, AND OF THE FORMULE FOR DETERMINING, FROM KNOWN ELEMENTS, THE HELIOCENTRIC AND GEOCENTRIC PLACES OF A HEAVENLY BODY, ADAPTED TO NUMERICAL COMIPUTATION FOR CASES OF ANY ECCENTRICITY WHATEVER. 1. THE study of the motions of the heavenly bodies does not require that we should know the ultimate limit of divisibility of the matter of which they are composed,-whether it may be subdivided indefinitely, or whether the limit is an indivisible, impenetrable atom. Nor are we concerned with the relations which exist between the separate atoms or molecules, except so far as they form, in the aggregate, a definite body whose relation to other bodies of the system it is required to investigate. On the contrary, in considering the operation of the laws in obedience to which matter is aggregated into single bodies and systems of bodies, it is sufficient to conceive simply of its divisibility to a limit which may be regarded as infinitesimal compared with the finite volume of the body, and to regard the magnitude of the element of matter thus arrived at as a mathematical point. An element of matter, or a material body, cannot give itself motion; neither can it alter, in any manner whatever, any motion which may have been communicated to it. This tendency of matter to resist all changes of its existing state of rest or motion is known as inertia, and is the fundamental law of the motion of bodies. Experience invariably confirms it as a law of nature; the continuance of motion as resistances are removed, as well as the sensibly unchanged motion of the heavenly bodies during many centuries, affording the 15 16 THEORETICAL ASTRONOMY. most convincing proof of its universality. Whenever, therefore, a material point experiences any change of its state as respects rest or motion, the cause must be attributed to the operation of something external to the element itself, and which we designate by the word force. The nature of forces is generally unknown, and we estimate them by the effects which they produce. They are thus rendered comparable with some unit, and may be expressed by abstract numbers. 2. If a material point, free to move, receives an impulse by virtue of the action of any force, or if, at any instant, the force by which motion is communicated shall cease to act, the subsequent motion of the point, according to the law of inertia, must be rectilinear and uniform', equal spaces being described in equal times. Thus, if s, v, and t represent, respectively, the space, the velocity, and the tisme, the measure of v being the space described in a unit of time, we shall have, in this case, s vt. It is evident, however, that the space described in a unit of time will vary with the intensity of the force to which the motion is due, and, the nature of the force being unknown, we must necessarily compare the velocities communicated to the point by different forces, in order to arrive at the relation of their effects. We are thus led to regard the force as proportional to the velocity; and this also has received the most indubitable proof as being a law of nature. Hence, the principles of the composition and resolution of forces may be applied also to the composition and resolution of velocities. If the force acts incessantly, the velocity will be accelerated, and the force which produces this motion is called an accelerating force. In regard to the mode of operation of the force, however, we may consider it as acting absolutely without cessation, or we may regard it as acting instantaneously at successive infinitesimal intervals represented by dt, and hence the motion as uniform during each of these intervals. The latter supposition is that which is best adapted to the requirements of the infinitesimal calculus; and, according to the fundamental principles of this calculus, the finite result will be the same as in the case of a force whose action is absolutely incessant. Therefore, if we represent the element of space by ds, and the element of time by dt, the instantaneous velocity will be ds V dt' which will vary from one instant to another. FUNDAMENTAL PRINCIPLES. 17 3. Since the force is proportional to the velocity, its measure at any instant will be determined by the corresponding velocity. If the accelerating force is constant, the motion will be uniformly accelerated; and if we designate the acceleration due to the force byf, the unit off being the velocity generated in a unit of time, we shall have v =ft. If, however, the force be variable, we shall have, at any instant, the relation =dv fdt' the force being regarded as constant in its action during the element of time dt. The instantaneous value of v gives, by differentiation, dv d2s dt dt2' and hence we derive d2s =dt; (1) - d/;2 so that, in varied motion, the acceleration due to the force is measured by the second differential of the space divided by the square of the element of time. 4. By the mass of the body we mean its absolute quantity of matter. The density is the mass of a unit of volume, and hence the entire mass is equal to the volume multiplied by the density. If it is required to compare the forces which act upon different bodies, it is evident that the masses must be considered. If equal masses receive impulses by the actioii of instantaneous forces, the forces acting on each will be to each other as the velocities imparted; and if we consider as the unit of force that which gives to a unit of mass the unit of velocity, we have for the measure of a force F, denoting the mass by M, F =-Mv. This is called the quantity of motion of the body, and expresses its capacity to overcome inertia. By virtue of the inert state of matter, there can be no action of a force without an equal and contrary reaction; for, if the body to which the force is applied is fixed, the equilibrium between the resistance and the force necessarily implies the development of an equal and contrary force; and, if the body be free to move, in the change of state, its inertia will oppose equal and 2 1 8'THEORETICAL ASTRONOMY. contrary resistance. Hence, as a necessary consequence of inertia, it follows that action and reaction are simultaneous, equal, and contrary. If the body is acted upon by a force such that the motion is varied, the accelerating force upon each element of its mass is represented by dt, and the entire motive force F is expressed by dv F-MA dt' 211 being the sum of all the elements, or the mass of the body. Since ds — dt' this gives d2'S which is the expression for the intensity of the motive force, or of the force of inertia developed. For the unit of mass, the measure of the force is d2s dt2' and this, therefore, expresses that part of the intensity of the motive force which is impressed upon the unit of mass, and is what is usually called the accelerating force. 5. The force in obedience to which the heavenly bodies perform their journey through space, is known as the attraction of gravitation; and the law of the operation of this force, in itself simple and unique, has been confirmed and generalized by the accumulated researches of modern science. Not only do we find that it controls the motions of the bodies of our own solar system, but that the revolutions of binary systems of stars in the remotest regions of space proclaim the universality of its operation. It unfailingly explains all the phenomena observed, and, outstripping observation, it has furnished the means of predicting many phenomena subsequently observed. The law of this force is that every particle of matter is attracted by every other particle by a force which varies directly as the mass and inversely as the square of the distance of the attracting particle. This reciprocal action is instantaneous, and is not modified, in any degree, by the interposition of other particles or bodies of matter. It is also absolutely independent of the nature of the molecules themselves, and of their aggregation. ATTRACTION OF SPHERES. 19 If we consider two bodies the masses of which are n and m', and whose magnitudes are so small, relatively to their mutual distance p, that we may regard them as material points, according to the law of gravitation, the action of m on each molecule or unit of mI will be - and the total force on n'l will be P2P mt~ The action of tm' on each molecule of m will be expressed by 2-, and its total action by m' -7The absolute or moving force with which the masses m and m' tend toward each other is, therefore, the same on each body, which result is a necessary consequence of the equality of action and reaction. The velocities, however, with which these bodies would approach each other must be different, the velocity of the smaller mass exceeding that of the greater, and in the ratio of the masses moved. The expression for the velocity of n', which would be generated in a unit of time if the force remained constant, is obtained by dividing the absolute force exerted by m by the mass moved, which gives -7 and this is, therefore, the measure of the acceleration due to the action of in at the distance o. For the acceleration due to the action of w' we derive, in a similar manner, p2 6. Observation shows that the heavenly bodies are nearly spherical in form, and we shall therefore, preparatory to finding the equations which express the relative motions of the bodies of the system, determine the attraction of a spherical mass of uniform density, or varying from the centre to the surface according to any law, for a point exterior to it. If we suppose a straight line to be drawn through the centre of the sphere and the point attracted, the total action of the sphere on the point will be a force acting along this line, since the mass of the sphere is symmetrical with respect to it. Let din denote an element 20 THEORETICAL ASTRONOMY. of the mass of the sphere, and p its distance from the point attracted; then will dm express the action of this element on the point attracted. If we suppose the density of the sphere to be constant, and equal to unity, the element dm becomes an element of volume, and will be expressed by dm = dx dy dz; x, y, and z being the co-ordinates of the element referred to a system of rectangular co-ordinates. If we take the origin of co-ordinates at the centre of the sphere, and introduce polar co-ordinates, so that x r cos sP cos 0, y = r cos p sin 0, z - r sin.p, the expression for dmn becomes dm r- 2 cos p dr dy dO; and its action on the point attracted is r2 cos p dr drp dO df= If we suppose the axis of z to be directed to the point attracted, the co-ordinates of this point will be' = O, y' = O, z' - a, a being the distance of the point from the centre of the sphere, and, since p (x -x')2 + (y _ y') + (Z )2, we shall have p2 = a2 2ar sin p +- r. The component of the force df in the direction of the line a, joining the point attracted and the centre of the sphere, is df cos y, where r is the angle at the point attracted between the element dmn and the centre of the sphere. It is evident that the sum of all the components which act in the direction of the line a will express the total action of the sphere, since the sum of those which act perpen ATTRACTION OF SPHERES. 21 dicular to this line, taken so as to include the entire mass of the sphere, is zero. But we have a -- z + p cos r, and hence a-rSilm o a - r sin (p cos = P The differentiation of the expression for p2, with respect to a, gives dp a — r sin p - cos r. da p - Therefore, if we denote the attraction of the sphere by A, we shall have, by means of the values of df and cos r, r cos ( dr d d -O dp dA - 2' a P da or d dA = -r2 cos p dr d dO d -P. da The polar co-ordinates r, (o, and 0 are independent of a, and hence d 2 cos 0 dr do dO dA — P da Let us now put dV=- cos c dr d dO (2) p and we shall have dV A = da' Consequently, to find the total action of the sphere on the given point, we have only to find V by means of equation (2), the limits of the integration being taken so as to include the entire mass of the sphere, and then find its differential coefficient with respect to a. If we integrate equation (2) first with'reference to 0, for which p is constant, between the limits 0 = 0 and 0 = 2r, we get V= 22ffr~ cos dr This must be integrated between the limits 5= + - 7r and o = - ~ir; 22 THEORETICAL ASTRONOMY. but since p is a function of (p, if we differentiate the expression for p2 with respect to op, we have r cos dp =- - P dp, a and hence V-~ffr dr dp. Corresponding to the limits of jp we have p a- r, and p a + r; and taking the integral with respect to p between these limits, we obtain -g__ 4~ f:r2 dr. Integrating, finally, between the limits r = 0 and r r,, we get irr V- 4 I';, being the radius of the sphere, and, if we denote its entire mass by m, this becomes a, Therefore, dV m da aG from which it appears that the action of a homogeneous spherical mass on a point exterior to it, is the same as if the entire mass were concentrated at its centre. If, in the integration with respect to r, we take the limits r' and r", we obtain A 4 = (v"3 - r'3) 3 2a and, denoting by m0 the mass of a spherical shell whose radii are r" and r', this becomes A 2a"' Consequently, the attraction of a homogeneous spherical shell on a point exterior to it, is the same as if the entire mass were concentrated at its centre. The supposition that the point attracted is situated within a spherical shell of uniform density, does not change the form of the FUNDAMENTAL PRINCIPLES. 23 general equation; but, in the integration with reference to p, the limits will be p =r + a, and,o r - a, which give V= - 44xr dr; and this being independent of a, we have dV A - -0. da Whence it follows that a point placed in the interior of a spherical shell is equally attracted in all directions, and that, if not subject to the action of any extraneous force, it will be in equilibrium in every position. 7. Whatever may be the law of the change of the density of the heavenly bodies from the surface to the centre, we may regard them as composed of homogeneous, concentric layers, the density varying only from one layer to another, and the number of the layers may be indefinite. The action of each of these will be the same as if its mass were united at the centre of the shell; and hence the total action of the body will be the same as if the entire mass were concentrated at its centre of gravity. The planets are indeed not exactly spheres, but oblate spheroids differing but little from spheres; and the error of the assumption of an exact spherical form, so far as relates to their action upon each other, is extremely small, and is in fact compensated by the magnitude of their distances from each other; for, whatever may be the form of the body, if its dimensions are small in comparison with its distance from the body which it attracts, it is evident that its action will be sensibly the same as if its entire mass were concentrated at its centre of gravity. If we suppose a system of bodies to be composed of spherical masses, each unattended with any satellite, and if we suppose that the dimensions of the bodies are small in comparison with their mutual distances, the formation of the equations for the motion of the bodies of the system will be reduced to the consideration of the motions of simple points endowed with forces of attraction corresponding to the respective masses. Our solar system is, in reality, a compound system, the several systems df primary and satellites corresponding nearly to the case supposed; and, before proceeding with the formation of the equations which are applicable to the general case, we will consider, at first, those for a simple system of bodies, considered as points and subject to their mutual actions and the action of the forces which correspond to the 24 THEORETICAL ASTRONOMY. actual velocities of the different parts of the system for any instant. It is evident that we cannot consider the motion of any single body as free, and subject only to the action of the primitive impulsion which it has received and the accelerating forces which act upon it; but, on the contrary, the motion of each body will depend on the force which acts upon it directly, and also on the reaction due to the other bodies of the system. The consideration, however, of the variations of the motion of the several bodies of the system is reduced to the simple case of equilibrium by means of the general principle that, if we assign to the different bodies of the system motions which are modified by their mutual action, we may regard these motions as composed of those which the bodies actually have and of other motions which are destroyed, and which must therefore necessarily be such that, if they alone existed, the system would be in equilibrium.?We are thus enabled to form at once the equations for the motion of a system of bodies. Let in, in', m"( &c. be the masses of the several bodies of the system, and x, y, z, x', y', z', &c. their coordinates referred to any system of rectangular axes. Further, let the components of the total force acting upon a unit of the mass of n, or of the accelerating force, resolved in directions parallel to the co-ordinate axes, be denoted by X, Y, and Z, respectively, then will'nX, mi Y, mZ, be the forces which act upon the body in the same directions. The velocities of the body mi at any instant, in directions parallel to the co-ordinate axes, will be dx dy dz dt' dt' dt' and the corresponding forces are dx dy dz cdt adt dt By virtue of the action of the accelerating force, these forces for the next instant become dx dy dz nm -- + mXdt, m +- m Ydt, m dt + nmZdt, which may be written respectively: MOTION OF A SYSTEM OF BODIES. 25 d dx md + rXdt, dy dy dy c_ 4 mdd - __d 7 dt dt dt m dt +n md- dt —md — dt- + mZdt. The actual velocities for this instant are dx dx dy dy dz dz E + drd d dt, d t + d dt, dt dt dt' dt..dt' and the corresponding forces are dx dx dy dy dz dz m -- m+ d - + nd md d-' dt dt dt dt dt? dt Comparing these with the preceding expressions for the forces, it appears that the forces which are destroyed, in directions parallel to the co-ordinate axes, are dx -md d- +- mvXdt, - mda d- m Ydt, (3) dt - md-~ -- mZdt. dt In the same manner we find for the forces which will be destroyed in the case of the body m': dx' - mid d -+ i'X'dt, -- m'd - + n' Y'dt, dzt m'd d- + m'Z'dt; and similarly for the other bodies of the system. According to the general principle above enunciated, the system under the action of these forces alone, will be in equilibrium. The conditions of equilibrium for a system of points of invariable but arbitrary form, and subject to the action of forces directed in any manner whatever, are 2X, = O Y, O O, 2Z, - O, xz( Y - Xy) --, z (Xt - Z,) = O, (ZY - Y,-=l 0; in which X, Y,, Z,, denote the components, resolved parallel to the 26 THEORETICAL ASTRONOMY. co-ordinate axes, of the forces acting on any point, and x, y, z, the co-ordinates of the point. These equations are equally applicable to the case of the equilibrium at any instant of a system of variable form; and substituting in them the expressions (3) for the forces destroyed in the case of a system of bodies, we shall have d2x zm~ - taX =- 0, dt' dt'y d2z dt2 Zndt -- L -Z_0, (4) (dt" Y dt- ) n (Y- Xy).0, t( d Y d2y __ m( Y dt2 dzt)- m (Xz - Yz) (y- -- (z) y ~ 0; which are the general equations for the motions of a system of bodies. 8. Let x,, y,, z,, be the co-ordinates of the centre of gravity of the system, and, by differentiation of the equations for the co-ordinates of the centre of gravity, which are i Zmx _my imz 2S~ m' y~,- 2m' we get'E d2x vm d2y vm d2z dx rndt-2 d2y, dt'z, __d__ cdt m i dt2 2mr dt2 2m Introducing these values into the first three of equations (4), they become d2x, 2mX d2y, 2m Y dz, ZnmZ dt2 min dt2 2mn dt2 n' ( from which it appears that the centre of gravity of the system moves in space as if the masses of the different bodies of which it is composed, were united in that point, and the forces directly applied to it. If we suppose that the only accelerating forces which act on the bodies of the system, are those which result from their mutual action, we have the obvious relation: mX - rm'X', m Y- - mn' Y', nZ - m'Z', MOTION OF A SYSTEM OF BODIES. 27 and similarly for any two bodies; and, consequently, 2mX = O, m Y- O, 2mZ = 0; so that equations (5) become 0, d dt,' t o d 0 t2 -0. dt2 dt~ dd2 Integrating these once, and denoting the constants of integration by c, c', c", we find, by combining the results, dx+ dy,2 dz + _ 2 2 + - c2 +.c2 dt2 and hence the absolute motion of the centre of gravity of the system, when subject only to the mutual action of the bodies which compose it, must be uniform and rectilinear. Whatever, therefore, may be the relative motions of the different bodies of the system, the motion of its centre of gravity is not thereby affected. 9. Let us now consider the last three of equations (4), and suppose the system to be submitted only to the mutual action of the bodies which compose it, and to a force directed toward the origin of coordinates. The action of m' on m, according to the law of gravitation, is expressed by ~2, in which p denotes the distance of m from mn'. To resolve this force in directions parallel to the three rectangular axes, we must multiply it by the cosine of the angle which the line joining the two bodies makes with the co-ordinate axes respectively, which gives XmW' (~ x) n__' (y' — y) mn' (' - z) f t P Po Further, for the components of the accelerating force of m on m/, we have X' - m (x - x') y - m (y - Y) Z - m (z - z') XI~?'?~' ~ ~' m ( Y - Xy) + ml' ( 7VX - Xy ) = 0, and generally m (Y. - Xy) = 0. (6) 28 THEORETICAL ASTRONOMY. In a similar manner, we find Zm (Xz - Zx) 0, (7) Zmn (Zy - Yz) = 0. These relations l11 not be altered if, in addition to their reciprocal action, the bodies of the system are acted upon by forces directed to the origin of co-ordinates. Thus, in the case of a force acting upon m, and directed to the origin of co-ordinates, we have, for its action alone, Yx =Xy, Xz= - Zx, Z = YZ, and similarly for the other bodies. Hence these forces disappear from the equations, and, therefore, when the several bodies of the system are subject only to their reciprocal action and to forces directed to the origin of co-ordinates, the last three of equations (4) become m ( d Y dt2 ), Z dt2.dt" ) I dz d2y\ C, the integration of which gives 2m (xdy - ydx) = cdt, mn (zdx - xdz) - ddt, (8) mn (ydz - zdy) - c"dt, c, c', and c" being the constants of integration. Now, xdy -ydx is double the area described about the origin of co-ordinates by the projection of the radius-vector, or line joining n with the origin of co-ordinates, on the plane of xy during the element of time dt; and, further, zdx - xdz and ydz - zdy are respectively double the areas described, during the same time, by the projection of the radius-vector on the planes of xz and yz. The constant c, therefore, expresses the sum of the products formed by multiplying the areal velocity of each body, in the direction of the co-ordinate plane xy, by its mass; and c', c", express the same sum with reference to the co-ordinate planes xz and yz respectively. Hence the sum of the areal velocities of the several bodies of the system about the origin of co-ordinates, each multiplied by the corresponding mass, is constant; and the sum of the areas traced, each multiplied by the corresponding mass, is proportional to the time. If the only forces which operate, are those INVARIABLE PLANE. 29 resulting from the mutual action of the bodies which compose the system, this result is correct whatever may be the point in space taken as the origin of co-ordinates. The areas described by the projection of the radius-vector of each body on the co-ordinate planes, are the projections, on these planes, of the areas actually described in space. We may, therefore,conceive of a resultant, or principal plane of projection, such that'he,sum of the areas traced by the projection of each radius-vector.on this plane, when projected on the three co-ordinate planes, each being multiplied by the corresponding mass, will be respectively equal to the first members of the equations (8). Let a, 9, and 7 be the angles which this principal plane makes with tye co-ordinate planes xy, xz, and yz, respectively; and let S denote the, sum of the areas traced on this plane, in a unit of time, by the projection of the radius-vector of each of the bodies of the system, each area being multiplied by the corresponding mass. The sum S will be found to be a maximum, and its projections on the co-ordinate planes, corresponding to the element of time dt, are S cosa dt, S cos dt, S cos dt. Therefore, by means of equations (8), we have c -= cos a, C, - cos, " cos r, and, since cos2a + cos2 + cos2r = 1, S2_ C2 + c'2 + Ct"2 Hence we derive c c cos a os0" cosr = — C2 + c02 + cr/2 These angles, being therefore constant and independent of the time, show that this principal plane of projection remains constantly parallel to itself during the motion of the system in space, whatever may be the relative positions of the several bodies; and for this reason it is called the invariable plane of the system. Its position with reference to any known plane is easily determined when the velocities, in directions parallel to the co-ordinate axes, and the masses and co-ordinates of the several bodies of the system, are known. The values of c, c', c" are given by equations (8), and 30 THEORETICAL ASTRONOMY. hence the values of a,?, and r, which determine the position of the invariable plane. Since the positions of the co-ordinate planes are arbitrary, we may suppose that of xy to coincide with the invariable plane, which gives cos 8 = 0 and cos r 0, and, therefore, c' = 0 and c" = 0. Further, since the positions of the axes of x and y in this plane are arbitrary, it follows that for every plane perpendicular to the invariable plane, the sum of the areas traced by the projections of the radii-vectores of the several bodies of the system, each multiplied by the corresponding mass, is zero. It may also be observed that the value of S is constant whatever may be the position of the co-ordinate planes, and that its value is necessarily greater than that of either of the quantities in the second member of the equatity. 2 - C + e'2 + c"]2, except when two of them are each equal to zero. It is, therefore, a maximum, and the invariable plane is also the plane of maximum areas. 10. If we suppose the origin of co-ordinates itself to move with uniform and rectilinear motion in space, the relations expressed by equations (8) will remain unchanged. Thus, let x,, y,, z, be the coordinates of the movable origin of co-ordinates, referred to a fixed point in space taken as the origin; and let xw, y0, -, x0', y,', z'o, &c. be the co-ordinates of the several bodies referred to the movable origin. Then, since the co-ordinate planes in one system remain always parallel to those of the other system of co-ordinates, we shall have X= xi + -, Y - Y + Yo, Z = Z,.+ and similarly for the other bodies of the system. Introducing these values of x, y, and z into the first three of equations (4), they become ( d2x, d2x) ) ( d, d2y, d)d The condition of uniform rectilinear motion of the movable origin gives - __ - __ ___ 0 d\_ __ ~ dtY —, d~= O, dt- dt. dt2 MOTION OF A SOLID BODY. 31 and the preceding equations become d2x 2 0-~27nX==0, -2 - 2xY=- O, d~ — mY-O, (9) 8d2z Im dt- mZ - 0. Substituting the same values in the last three of equations (4), observing that the co-ordinates x,, y,, z, are the same for all the bodies of the system, and reducing the resulting equations by means of equations (9), we get n O ( Yo ~ dt ) ~ ( Y. - Xy) = O0, ( ) d2X0 d22zo fn Yo d - - zo ~dt2 - zm (Zy0 - Yzo) = O. Hence it appears that the form of the equations for the motion of the system of bodies, remains unchanged when we suppose the origin of co-ordinates to move in space with a uniform and rectilinear motion. 11. The equations already derived for the motions of a system of bodies, considered as reduced to material points, enable us to form at once those for the motion of a solid body. The mutual distances of the parts of the system are, in this case, invariable, and the masses of the several bodies become the elements of the mass of the solid body. If we denote an element of the mass by dA, the equations (5) for the motion of the centre of gravity of the body become dt2 dty dtz d dm% =f= J'Xdmy, d=' fYdmn, mnd —2 rZdm, (11) the summation, or integration with reference to dm, being taken so as to include the entire mass of the body, from which it appears that the centre of gravity of the body moves in space as if the entire mass were concentrated in that point, and the forces applied to it directly. If we take the origin of co-ordinates at the centre of gravity of the body, and suppose it to have a rectilinear, uniform motion in space, and denote the co-ordinates of the element dm, in reference to this origin, by x0, yo, z, we have, by means of the equations (10), 32 THEORETICAL ASTRONOMY. d( Y — -- d dim - (Yxo -Xyo) dm - 0, fz0 dtr — Yo dt2 / f( d t2 o d2" ) dm -X - Zx0) dm = 0 (12) the integrao dtion wih r(Zye o YZo) t i d t the integration with respect to dcm being taken so as to include the entire mass of the body. These equations, therefore, determine the motion of rotation of the body around its centre of gravity regarded as fixed, or as having a uniform rectilinear motion in space. Equations (11) determine the position of the centre of gravity for any instant, and hence for the successive instants at intervals equal to dt; and we may consider the motion of the body during the element of time dt as rectilinear and uniform, whatever may be the form of its trajectory. Hence, equations (11) and (12) completely determine the position of the body in space,-the former relating to the motion of translation of the centre of gravity, and the latter to the motion of rotation about this point. It follows, therefore, that for any forces which act upon a body we can always decompose the actual motion into those of the translation of the centre of gravity in space, and of the motion of rotation around this point; and these two motions may be considered independently of each other, the motion of the centre of gravity being independent of the form and position of the body about this point. If the only forces which act upon the body are the reciprocal action of the elements of its mass and forces directed to the origin of coordinates, the second terms of equations (12) become each equal to zero, and the results indicated by equations (8) apply in this case also. The parts of the system being invariably connected, the plane of maximum areas, or invariable plane, is evidently that which is perpendicular to the axis of rotation passing through the centre of gravity, and therefore, in the motion of translation of the centre of gravity in space, the axis of rotation remains constantly parallel to itself. Any extraneous force which tends to disturb this relation will necessarily develop a contrary reaction, and hence a rotating body resists any change of its plane of rotation not parallel to itself. We may observe, also, that on account of the invariability of the mutual distances of the elements of the mass, according to equations (8), the motion of rotation must be uniform. 12. We shall now consider the action of a system of bodies on a MOTION OF A SOLID BODY. 33 distant mass, which we will denote by M. Let x,, y, z0, x o,', z0', &c. be the co-ordinates of the several bodies of the system referred to its centre of gravity as the origin of co-ordinates; x,, y,, and z, the co-ordinates of the centre of gravity of the system referred to the centre of gravity of the body M. The co-ordinates of the body m, of the system, referred to this origin, will therefore be X = x, -- X0, y= y, +-yo, -, +- Z, and similarly for the other bodies of the system. If we denote by r the distance of the centre of gravity of m from that of M1, the accelerating force of the former on an element of mass at the centre of gravity of the latter, resolved parallel to the axis of x, will be mx r3 and, therefore, that of the entire system on the element of M, resolved in the same direction, will be mx r We have also'' (X- + xo) 2 + (y, + yo) 2 + (, +- Z) 2, and, if we denote by r, the distance of the centre of gravity of the system from M, r,2 x,2 + y,2 + Z-2. Therefore _3 r= (X + ox) (r,2 + 2 (, Xo + y, yo + z,o) + r2) We shall now suppose the mutual distances of the bodies of the system to be so small in comparison with the distance r, of its centre of gravity from that of AI, that terms of the order r02 may be neglected; a condition which is actually satisfied in the case of the secondary systems belonging to the solar system. Hence, developing the second factor of the second member of the last equation, and neglecting terms of the order r2, we shall have x +x x 3x, (x,0 + y, Yo + Z, o) r3 r3 3 t5 and mxIV Tv?,m 2mxo 3x, }=, + -m - 3~- (x,2mxo + y,2myO + z,omzO). r,3 r 3 r,3, 3 34 THEORETICAL ASTRONOMY. But, since x, Yo, 0, are the co-ordinates in reference to the centre of gravity of the system as origin, we have ZmXo -= 0, -Zmy O, mzo = 0, and the preceding equation reduces to IrnX ICm 2 -mx r3 r 3 In a similar manner, we find ny TZin rmZ TZin r3 r,3t r3 r,3 The second members of these equations are the expressions for the total accelerating force due to the action of the bodies of the system on M1 resolved parallel to the co-ordinate axes respectively, when we consider the several masses to be collected at the centre of gravity of the system. Hence we conclude that when an element of mass is attracted by a system of bodies so remote from it that terms of the order of the squares of the co-ordinates of the several bodies, referred to the centre of gravity of the system as the origin of co-ordinates, may be neglected in comparison with the distance of the system from the point attracted, the action of the system will be the same as if the masses were all united at its centre of gravity. If we suppose the masses n, n', m", &c. to be the elements of the mass of a single body, the form of the equations remains unchanged; and hence it follows that the mass 1M is acted upon by another mass, or by a system of bodies, as if the entire mass of the body, or of the system, were collected at its centre of gravity. It is evident, also, that reciprocally in the case of two systems of bodies, in which the mutual distances of the bodies are small in comparison with the distance between the centres of gravity of the two systems, their mutual action is the same as if all the several masses in each system were collected at the common centre of gravity of that system; and the two centres of gravity will move as if the masses were thus united. 13. The results already obtained are sufficient to enable us to form the equations for the motions of the several bodies which compose the solar system. If these bodies were exact spheres, which could be considered as composed of homogeneous concentric spherical shells, the density varying only from one layer to another, the action of MOTION OF A SYSTEM OF BODIES. 35 each on an element of the mass of another would be the same as if the entire mass of the attracting body were concentrated at its centre of gravity. The slight deviation from this law, arising from the ellipsoidal form of the heavenly bodies, is compensated by the magnitude of their mutual distances; and, besides, these mutual distances are so great that the action of the attracting body on the entire mass of the body attracted, is the same as if the latter were concentrated at its centre of gravity. Hence the consideration of the reciprocal action of the single bodies of the system, is reduced to that of material points corresponding to their respective centres of gravity, the masses of which, however, are equivalent to those of the corresponding bodies. The mutual distances of the bodies composing the secondary systems of planets attended with satellites are so small, in comparison with the distances of the different systems from each other and from the other planets, that they act upon these, and are reciprocally acted upon, in nearly the same manner as if the masses of the secondary systems were united at their common centres of gravity, respectively. The motion of the centre of gravity of a system consisting of a planet and its satellites is not affected by the reciprocal action of the bodies of that system, and hence it may be considered independently of this action. The difference of the action of the other planets on a planet and its satellites will simply produce inequalities in the relative motions of the latter bodies as determined by their mutual action alone, a.nd will not affect the motion of their common centre of gravity. Hence, in the formation of the equations for the motion of translation of the centres of gravity of the several planets or secondary systems which compose the solar system, we have simply to consider them as points endowed with attractive forces corresponding to the several single or aggregated masses. The investigation of the motion of the satellites of each of the planets thus attended, forms a problem entirely distinct from that of the motion of the common centre of gravity of such a system. The consideration of the motion of rotation of the several bodies of the solar system about their respective centres of gravity, is also independent of the motion of translation. If the resultant of all the forces which act upon a planet passed through the centre of gravity, the motion of rotation would be undisturbed; and, since this resultant in all cases very nearly satisfies this condition, the disturbance of the motion of rotation is very slight. The inequalities thus produced in the motion of rotation are, in fact, sensible, and capable of being indicated by observation, only in the case of the earth and moon. It has, indeed, 36 THEORETICAL ASTRONOMY. been rigidly demonstrated that the axis of rotation of the earth relative to the body itself is fixed, so that the poles of rotation and the terrestrial equator preserve constantly the same position in reference to the surface; and that also the velocity of rotation is constant. This assures us of the permanency of geographical positions, and, in connection with the fact that the change of the length of the mean solar day arising from the variation of the obliquity of the ecliptic and in the length of the tropical year, due to the action of the sun, moon, and planets upon the earth, is absolutely insensible, -amounting to only a small fraction of a second in a million of years,-assures us also of the permanence of the interval which we adopt as the unit of time in astronomical investigations. 14. Placed, as we are, on one of the bodies of the system, it is only possible to deduce from observation the relative motions of the different heavenly bodies. These relative motions in the case of the comets and primary planets are referred to the centre of the sun, since the centre of gravity of this body is near the centre of gravity of the system, and its preponderant mass facilitates the integration of the equations thus obtained. In the case, however, of the secondary systems, the motions of the satellites are considered in reference to the centre of gravity of their primaries. We shall, therefore, form the equations for the motion of the planets relative to the centre of gravity of the sun; for which it becomes necessary to consider more particularly the relation between the heterogeneous quantities, space, time, and mass, which are involved in them. Each denomination, being divided by the unit of its kind, is expressed by an abstract number; and hence it offers no difficulty by its presence in an equation. For the unit of space we may arbitrarily take the mean distance of the earth from the sun, and the mean solar day may be taken as the unit of time. But, in order that when the space is expressed by 1, and the time by 1, the force or velocity may also be expressed by 1, if the unit of space is first adopted, the relation of the time and the mass-which determines the measure of the forcewill be such that the units of both cannot be arbitrarily chosen. Thus, if we denote by f the acceleration cue to the action of the mass m on a material point at the distance a, and by f the acceleration corresponding to another mass am' acting at the same distance, we have the relation 91~ f _ nm f',-'X MOTION RELATIVE TO THE SUN. 37 and hence, since the acceleration is proportional to the mass, it may be taken as the measure of the latter. But we have, for the measure off, d2s f~ dt2 Integrating this, regarding f as constant, and the point to move from a state of rest, we get - s _.ft2. (13) The acceleration in the case of a variable force is, at any instant, measured by the velocity which the force acting at that instant would generate, if supposed to remain constant in its action, during a unit of time. The last equation gives, when t 1, f= 2s; and hence the acceleration is also measured by double the space which would be described by a material point, from a state of rest, during a unit of time, the force being supposed constant in its action during this time. In each case the duration of the unit of time is involved in the measure of the acceleration, and hence in that of the mass on which the acceleration depends; and the unit of mass, or of the force, will depend on the duration which is chosen for the unit of time. In general, therefore, we regard as the unit of mass that which, acting constantly at a distance equal to unity on a material point free to move, will give to this point, in a unit of time, a velocity which, if the force ceased to act, would cause it to describe the unit of distance in the unit of time. Let the unit of time be a mean solar day; k2 the acceleration due to the force exerted by the mass of the sun at the unit of distance; andf the acceleration corresponding to the distance r; then will k2 r2 and k2 becomes the measure of the mass of the sun. The unit of mass is, therefore, equal to the mass of the sun taken as many times as k2 is contained in unity. Hence, when we take the' mean solar day as the unit of time, the mass of the sun is measured by k2; by which we are to understand that if the sun acted during a mean solar day, on a material point free to move, at a distance constantly equal to the mean distance of the earth from the sun, it would, at the end of that time, have communicated to the point a velocity which, if 38 THEORETICAL ASTRONOMY. the force did not thereafter act, would cause it to describe, in a unit of time, the space expressed by k2. The acceleration due to the action of the sun at the unit of distance is designated by k2, since the square root of this quantity appears frequently in the fornulhe which will be derived. If we take arbitrarily the mass of the sun as the unit of mass, the unit of time must be determined. Let t denote the number of mean solar days which must be taken for the unit of time when the unit of mass is the mass of the sun. The space which the force due to this mass, acting constantly on a material point at a distance equal to the mean distance of the earth from the sun, would cause the point to describe in the time t, is, according to equation (13), s i- kt. But, since t expresses the number of mean solar days in the unit of time, the measure of the acceleration corresponding to this unit is 2s, and this being the unit of force, we have k2t2 1; and hence k Therefore, if the mass of the sun is regarded as the unit of mass, the number of mean solar days in the unit of time will be equal to unity divided by the square root of the acceleration due to the force exerted by this mass at the unit of distance. The numerical value of k will be subsequently found to be 0.0172021, which gives 58.13244 mean solar days for the unit of time, when the mass of the sun is taken as the unit of mass. 15. Let x, y, z be the co-ordinates of a heavenly body referred to the centre of gravity of the sun as the origin of co-ordinates; r its radius-vector, or distance from this origin; and let mn denote the quotient obtained by dividing its mass by that of the sun; then, taking the mean solar day as the unit of time, the mass of the sun is expressed by k, and that of the planet or comet by mk2. For a second body let the co-ordinates be x', yi, z'; the distance from the sun, r'; and the mass, m'k2; and similarly for the other bodies of the system. Let the co-ordinates of the centre of gravity of the sun referred to any fixed point in space be C, V, t, the co-ordinate planes being parallel to those of x, y, and z, respectiVely; then will the MOTION RELATIVE TO THE SUN. 39 acceleration due to the action of n on the sun be expressed by ~2r and the three components of this force in directions parallel to the co-ordinate axes, respectively, will be 2X 2 Y 2 mk ~ mk~ Y mkw_ r3 r3 7 r3 The action of m' on the sun will be expressed by m'lc2 m'2 k2 m', r3 r" and hence the acceleration due to the combined and simultaneous action of the several bodies of the system on the sun, resolved parallel to the co-ordinate axes, will be,2,v, _ 2 Illy _2_mz r3 r7 r3' The motion of the centre of gravity of the sun, relative to the fixed origin, will, therefore, be determined by the equations d$ MX d 2, Zmy rn p Z n dt2 ~- — 3-, dt2.d2 3 (14 Let p denote the distance of ra from m'; p' its distance from mi", adding an accent for each successive body considered; then will the action of the bodies mn', n", &c. on m be k2.~ - p2 of which the three components parallel to the co-ordinate axes, respectively,' are k2m2xt kyt zt k -- 2 p^__3,p k^Wn -X'p3 3 The action of the sun on m, resolved in the same manner, is expressed by Px k2y k2z r 3' r3. r3' which are negative, since the force tends to diminish the co-ordinates x, y, and z. The three components of the total action of the other bodies of the system on m are, therefore, 40 THEORETICAL ASTRONOMY. kx - -A~ + k2 lnd (zt X) ky m'(y-y) r-T- + ~.zY' — y)P3 kz + k2fn/ (z' z) r3 P0 and, since the co-ordinates of m referred to the fixed origin are + x, +4 Y + Z, the equations which determine the absolute motion are d2 d4v k2x 2m' (X' - x) dlrl d J ~ _ ~y),^ dt + d2 +,-r 3k' k ) +t + + k2yy- (15) d2C d2Z k~z m' (Z' -- z) dt2 + + --- z ( a the symbol of summation in the second members relating simply to the masses and co-ordinates of the several bodies which act on m, d2~ d2'7 d2. exclusive of the sun. Substituting for dt2 d-' and their values (it dt2 d(t given by equations (14), we get d2I X 1 X X X d + k2 ( m) = - ~ \) = l2y + k2(11 + W -y~ k'2f'( y — ~ \(16) dt2 + -(l + -) r3 3 r3 dz+ k2 (1 + _n) 3 k2m ( 2 z -z ) dt2 r \p3r I Since x, y, z are the co-ordinates of m relative to the centre of gravity of the sun, these equations determine the motion of m relative to that point. The second members may be put in another form, which greatly facilitates the solution of some of the problems relating to the motion of rn. Thus, let us put i-' 1 x'+yy. +- zz l -' X-"+ +yy"+ \ (&, l + in \ p r13 /+ r3 + 1+ --- ~ +T pi- r —+' (17) and we shall have for the partial differential coefficient of this with respect to x, d\, _ m' 1 dp x'\ " / I dp' x" \ \ dx- l- - n\ p2 d r'3 + - + — bl pt2 d- r.3. + MOTION RELATIVE TO THE SUN. 41 But, since 2 - (x - X)2 + (y - y)2 + (Z - )2, p12 (XI' - )2 + (y y)2 + (z - (Z)2, we have dp x' x- dp' x" x dx p' dx'' and hence we derive dQg' I xx x' \ " { x" x x"\ \ dx) 1 +m\ rI 3 1 +m)m\ rp )+...&c. or ( _7 \ I( _ _ - _ (1+ ) d fo p 3- +, dx! p) We find, also, in the same manner, for the partial differential coefficients with respect to y and z, (+n)( dQ ) ( Y, — Y ) -- Z,' (1 +'i) ( d{ - ) 2m' Z — z- ~ 3 The equations (16), therefore, become d2X 22 \ dV+~k(l +k m)y h k'(l +m ddy dt 2 ) ( 1 )(r3 =k(1 +)(dS P), d-r 2(1 +)4 (1+)( (18) dt2 + k2 + ) 3 k(1 + )( d ) It will be observed that the second members of equationu (16) express the difference between the action of the bodies mn', mn", &c. on m and on the sun, resolved parallel to the co-ordinate axes respectively. The mutual distances of:the planets are such that these quantities are generally very small, and we may, therefore, in a first approximation to the motion of m relative to the sun, neglect the second members of these equations; and the integrals which may then be derived, express what is called the undisturbed motion of m. By means of the results thus obtained for the several bodies successively, the approximate values of the second members of equations (16) may be found, and hence a still closer approximation to the actual motion of m. The force whose components are expressed by the second members of these equations is called the disturbing force; 42 THEORETICAL ASTRONOMY. and, using the second form of the equations, the function l2, which determines these components, is called the perturbing function. The complete solution of the problem is facilitated by an artifice of the infinitesimal calculus, known as the variation of parameters, or of constants, according to which the complete integrals of equations (16) are of the same form as those obtained by putting the second members equal to zero, the arbitrary constants, however, of the latter integration being regarded as variables. These constants of integration are the elements which determine the motion of m relative to the sun, and when the disturbing force is neglected the elements are pure constants. The variations of these, or of the co-ordinates, arising from the action of the disturbing force are, in almost all cases, very small, and are called the perturbations. The problem which first presents itself is, therefore, the determination of all the circumstances of the undisturbed motion of the heavenly bodies, after which the action of the disturbing forces may be considered. It may be further remarked that, in the formation of the preceding equations, we have supposed the different bodies to be free to move, and, therefore, subject only to their mutual action. There are, indeed, facts derived from the study of the motion of the comets which seem to indicate that there exists in space a resisting medium which opposes the free motion of all the bodies of the system. If such a medium actually exists, its effect is very small, so that it can be sensible only in the case of rare and attenuated bodies like the comets, since the accumulated observations of the different planets do not exhibit any effect of such resistance. But, if we assume its existence, it is evidently necessary only to add to the second members of equations (16) a force which shall represent the effect of this resistance,which, therefore, becomes a part of the disturbing force,-and the motion of m will be completely determined. i16. When we consider the undisturbed motion of a planet or comet relative to the sun, or simply the motion of the body relative to the sun as subject only to the reciprocal action of the two bodies, the equations (16) become d2 k2(l+ -0, dt2 r3 dt + k2(1: + ) 0, (o, d+ 2- i1 + a) = 0. dt2z +2( +m) - 3O. MOTION RELATIVE TO THE SUN. 43 The equations for the undisturbed motion of a satellite relative to its primary are of the same form, the value of k2, however, being in this case the acceleration due to the force exerted by the mass of the primary at the unit of distance, and m the ratio of the mass of the satellite to that of the primary. The integrals of these equations introduce six arbitrary constants of integration, which, when known, will completely determine the undisturbed motion of m relative to the sun. If we multiply the first of these equations by y, and the second by x, and subtract the last product from the first, we shall find, by integrating the result, xdy - ydx dt c being an arbitrary constant. In a similar manner, we obtain xdz - zdx ydz - zdy dt dt ~ If we multiply these three equations respectively by z, y, and x, and add the products, we obtain cz -c y + C"x - 0. This, being the equation of a plane passing through the origin of co-ordinates, shows that the path of the body relative to the sun is a plane curve, and that the plane of the orbit passes through the centre of the sun. Again, if we multiply the first of equations (19) by 2dx, the second by 2dy, and the third by 2dz, take the sum and integrate, we shall find +dx'+d +d'+2k(+ ( + J xd + ydy + zdz 0. dt2 d But, since r2 = x - y2 +'2, we shall have, by differentiation, rdr xdx + ydy - zdz. Therefore, introducing this value into the preceding equation, we obtain dX2 + dy2 + d 2(1 + m) + h (20) A be-i +a r = o, (to dtb r h being an arbitrary constant. 44 THEORETICAL ASTRONOMY. If we add together the squares of the expressions for c, c', and cl", and put c2 + e2 + Gc2 = 4f2, we shall have (x' + y2 + Z ) (dx2 + dy2 + -dz') (xdx + ydy + zdz)2 42 dt' dt or dt dt~ - 4f' 2 dx2- + dy + dz2 _r2dr2 4f 21) If we represent by dv the infinitely small angle contained between two consecutive radii-vectores r and r + dr, since dx2 + cdy + dz2 is the square of the element of path described by the body, we shall have dx2 + dy + dz2 - dr2 + r'2dV2 Substituting this value in the preceding equation, it becomes r2dv 2fdt. (22) The quantity r2dv is double the area included by the element of path described in the element of time dt, and by the radii-vectores r and r + dr; and f, therefore, represents the areal velocity, which, being a constant, shows that the radius-vector of a planet or comet describes equal areas in equal intervals of time. From the equations (20) and (21) we find, by elimination, dt -rdr (23) l/2rk2 (1 +- m) -- hr2 — 4f( Substituting this value of dt in equation (22), we get dv- - 2fdr (24) rl/2rk2 (1 +- ) - hr2 — 4f2 which gives, in order to find the maximum and minimum values of r, dr rV/2rk(1 + m) -hr-24f2 dv- 2f or 2rk2(1 + n) - hr2- 4f2 = 0. Therefore k2(1 +M) 4f2,4(1 + m)2 h + h h2 and 1 k2l(1+m)) 4f2 k+(1 + m) h h are, respectively, the maximum and minimum values of r. The MOTION RELATIVE TO THE SUN. 45 points of the orbit, or trajectory of the body relative to the sun, corresponding to these values of r, are called the apsides; the former, the aphelion, and the latter, the perihelion. If we represent these values, respectively, by a (I + e) and a ( - e), we shall have h = ( a; 4f- ak2 (1 + m) (1 - e2) =k2p (1 + n), in which p a (1-e2). Introducing these values into the equation (24), it becomes l/p dr e dr dv - ri2r -2-p 1_ 1 1 ) the integral of which gives -1 1(p v==w-+cos - 1 e r co being an arbitrary constant. Therefore we shall have r - 1 cos (v - ) from which we derive 1 + e cos (v - )' which is the polar equation of a conic section, the pole being at the focus, p being the semi-parameter, e the eccentricity, and v - ( the angle at the focus between the radius-vector and a fixed line, in the plane of the orbit, making the angle wo with the semi-transverse axis a. If the angle v - o.is counted from the perihelion, we have ) = 0, and,= P (25) 1 + e cos v The angle v is called the true anonmaly. Hence we conclude that the orbit of a heavenly body revolving around the sun is a conic section wqith the sun in one of the foci. Observation shows that the planets revolve around the sun in ellipses, usually of small eccentricity, while the comets revolve either in ellipses of great eccentricity, in parabolas, or in hyperbolas, a circumstance which, as we shall have occasion to notice hereafter, greatly 46 THEORETICAL ASTRONOMY. lessens the amount of labor in many computations respecting their motion. Introducing into equation (23) the values of h and 4f2 already found, we obtain dt i/a rdr k 1/1 + i n I/a e2- (a — r) which may be written 3 a-\ - r~ __ -~ dt -- - a ae k ]/1 + l-t /] a r )' ae or dt a / e-4 ) de -)\ kV 1 I \ 1l (a - )2 - ( a-r 2 the integration of which gives t b 1/1' t 1 +4/ cos — Ce) ae- ) + C. (26) In the perihelion, r a (1 - e), and the integral reduces to t' - C; therefore, if we denote the time from the perihelion by t0, we shall have t-~1 -(-a cos — ( e- 1 - ) (27) ^ kVl-m\ \ inae -' ae' In the aphelion, r = a (1 + e); and therefore we shall have, for the time in which the body passes from the perihelion to the aphelion, to =- I, or 2cr - 2 a k / 1 -'mn T being the periodic time, or time of one revolution of the planet around the sun, a the semi-transverse axis of the orbit, or mean distance from the sun, and wT the semi-circumference of a circle whose radius is unity. Therefore we shall have = k 47 ( m (28) k2 (I + M)'> MOTION RELATIVE TO THE SUN. 47 For a second planet, we shall have a"3 4r72 _ 472 k2 (1 + A'); and, consequently, between the mean distances and periodic times of any two planets, we have the relation (1- -)- rn> ~r2 3 (1 + m') r"'3 (2 If the masses of the two planets m and m' are very nearly the same, we may take 1 + m -1 + m'; and hence, in this case, it follows that the squares of the periodic times are to each other as the cubes of the mean distances from the sun. The same result may be stated in another form, which is sometimes more convenient. Thus, since tab is the area of the ellipse, a and b representing the semai-axes, we shall have - =f areal velocity; and, since b2 = a e (I - e2), we have 3 1 1 3 Cra2a2 (I - e2) r //p f a - = - which becomes, by substituting the value of r already found, f= k /p (1 + m). (30) In like manner, for a second planet, we have I,- Tl, /p, (i + n); and, if the masses are such that we may take 1 + mn sensibly equal to 1 + in', it follows that, in this case, the areas described in equal times, in different orbits, are proportional to the square roots of their parameters. 17. We shall now consider the signification of some of the constants of integration already introduced. Let i denote the inclination of the orbit of m to the plane of xy, which is thus taken as the plane of reference, and let a be the angle formed by the axis of x and the line of intersection of the plane of the orbit with the plane of xy; then will the angles i and a determine the position of the plane of 48 THEORETICAL ASTRONOMY. the orbit in space. The constants c, c', and c", involved in the equation cz - c'y + Cx O, are, respectively, double the projections, on the co-ordinate planes, xy, xz, and yz, of the areal velocity f; and hence we shall have c = 2f cos i. The projection of 2f on a plane passing through the intersection of the plane of the orbit with the plane of xy, and perpendicular to the latter, is 2f sin i; and the projection of this on the plane of xz, to which it is inclined at an angle equal to 2, gives c' = 2f sin i cos g. Its projection on the plane of yz gives " == 2fsin i sin Q. Hence we derive z cosi - y sin i cos Q +- x sin i sin g = 0, (31) which is the equation of the plane of the orbit; and,.by means of the value of f in terms of p, and the values of c, c', c", we derive, also, dY dx _ — yd- kV/p (1 +qm) cos/, dz dx x-t - z- = k/p (1 + ) cos sini, (32) dz dy y-t Z -- k I/p (1 +- n) sin g2 sin i. These equations will enable us to determine Q, i, and p, when, for any instant, the mass and co-ordinates of m, and the components of its velocity, in directions parallel to the co-ordinate axes, are known. The constants a and e are involved in the value of p, and hence four constants, or elements, are introduced into these equations, two of which, a and e, relate to the form of the orbit, and two, g -and i, to the position of its plane in space. If we measure the angle v- to from the point in which the orbit intersects the plane of xy, the constant o will determine the position of the orbit in its own plane. Finally, the constant of integration C) in equation (26), is the time MOTION RELATIVE TO THE SUN. 49 of passage through the perihelion; and this determines the position of the body in its orbit. When these six constants are known, the undisturbed orbit of the body is completely determined. Let V denote the velocity of the body in its orbit; then will equation (20) become V2- k2 (1 + m) ( ) At the perihelion' r is a minimum, and hence, according to this equation, the corresponding value of V is a maximum. At the aphelion, V is a minimum. In the parabola, a = o, and hence V-= k i/l +m J2 \ r which will determine the velocity at any instant, when r is known. It will be observed that the velocity, corresponding to the same value of r, in an elliptic orbit is less than in a parabolic orbit, and that, since a is negative in the hyperbola, the velocity in a hyperbolic orbit is still greater than in the case of the parabola. Further, since the velocity is thus found to be independent of the eccentricity, the direction of the motion has no influence on the species of conic section described. If the position of a heavenly body at any instant, and the direction and magnitude of its velocity, are given, the relations already derived will enable us to determine the six constant elements of its orbit. But since we cannot know in advance the magnitude and direction of the primitive impulse communicated to the body, it is only by the aid of observation that these elements can be derived; and therefore, before considering the formulse necessary to determine unknown elements by means of observed positions, we will investigate those which are necessary for the determination of the heliocentric and geocentric places of the body, assuming the elements to be known. The results thus obtained will facilitate the solution of the problem of finding the unknown elements from the data furnished by observation. 18. To determine the value of k, which is a constant for the solar system, we have, from equation (28), 3 27r a2 k - 1/V +m 4 50 THEORETICAL ASTRONOMY. In the case of the earth, a — 1, and therefore 27 k - ___ In reducing this formula to numbers we should properly use, for r, the absolute length of the sidereal year, which is invariable. The effect of the action of the other bodies of the system on the earth is to produce a very small secular change in its mean longitude corresponding to any fixed date taken as the epoch of the elements; and a correction corresponding to this secular variation should be applied to the value of r derived from observation. The effect of this correction is to slightly increase the observed value of r; but to determine it with precision requires an exact knowledge of the masses of all the bodies of the system, and a complete theory of their relative motions,-a problem which is yet incompletely solved. Astronomical usage has, therefore, sanctioned the employment of the value of k found by means of the length of the sidereal year derived directly from observation. This is virtually adopting as the unit of space a distance which is very little less than the absolute, invariable mean distance of the earth from the sun; but, since this unit may be arbitrarily chosen, the accuracy of the results is not thereby affected. The value of r from which the adopted value of k has been computed, is 365.2563835 mean solar days; and the value of the combined mass of the earth and moon is 1 -354710' Hence we have log r 2.5625978148; log vl + m= 0.0000006122; log 27r =0.7981798684; and, consequently, log k 8.2355814414. If we multiply this value of k by 206264.81, the number of seconds of arc corresponding to the radius of a circle, we shall obtain its value expressed in seconds of arc in a circle whose radius is unity, or on the orbit of the earth supposed to be circular. The value of k in seconds is, therefore, log k = 3.5500065746. 2,r The quantity expresses the mean angular motion of a planet in a mean solar day, and is usually designated by,p. We shall, therefore, have MOTION RELATIVE TO THE SUN. 51 k~ 1 ~ m -- 1 +, (33) a2 for the expression for the mean daily motion of a planet. Since, in the case of the earth, 1/1 + m differs very little from 1, it will be observed that k very nearly expresses the mean angular motion of the earth in a mean solar day. In the case of a small planet or of a comet, the mass m is so small that it may, without sensible error, be neglected; and then we shall have -- -. (34) a" For the old planets whose masses are considerable, the rigorous expression (33) must be used. 19. Let us now resume the polar equation of the ellipse, the pole being at the focus, which is a(1 - e) r - + e cosv If we represent by ( the angle included between the conjugate axis and a line drawn from the extremity of this axis to the focus, we shall have sin v -= e; and, since a (1 - e2) is half the parameter of the transverse axis, which we have designated by p, we have r P 1 -- sin pO cos v The angle 5p is called the angle of eccentricity. Again, since p a (1 - e2) -= a cos2 (, we have a cos2 ( r= (35) 1 + sin CP cos v It is evident, from this equation, that the maximum value of r in an elliptic orbit corresponds to v = 1800, and that the minimum value of r corresponds to v = 0. It therefore increases from the perihelion to the aphelion, and then decreases as the planet approaches the perihelion. 52 THEORETICAL ASTRONOMY. In the case of the parabola, o - 90~, and sinp = e = 1; consequently, 1 -+ cos v But, since 1 + cos v = 2 cos2,v, if we put q -- p, we shall have r =cos, (36) in which q is the perihelion distance. In this case, therefore, when v ~ 1800, r will be infinite, and the comet will never return, but course its way to other systems. The angle (p cannot be applied to the case of the hyperbola, since in a hyperbolic orbit e is greater than 1; and, therefore, the eccentricity cannot be expressed by the sine of an arc. If, however, we designate by 4 the angle which the asymptote to the hyperbola makes with the transverse axis, we shall have e cos 4 - 1. Introducing this value of e into the polar equation of the hyperbola, it becomes p COS s p cos (7) 2 cos. (v +- ) cos (v-)' ( It appears from this formula that r increases with v, and becomes infinite when I + e cos v - 0, or cos v =- cos, in which case v - 180~ -- -: consequently, the maximum positive value of v is represented by 180~ -4, and the maximum negative value by - (180~ ~ ). Further, it is evident that the orbit will be that branch of the hyperbola which corresponds to the focus in which the sun is placed, since, under the operation of an attractive force, the path of the body must be concave toward the centre of attraction. A body subject to a. force of repulsion of the same intensity, and varying according to the same law, would describe the other branch of the curve. The problem of finding the position of a heavenly body as seen from any point of reference, consists of two parts: first, the determination of the place of the body in its orbit; and then, by means of this and of the elements which fix the position of the plane of the PLACE IN THE ORBIT. 53 orbit, and that of the orbit in its own plane, the determination of the position in space. In deriving the formulae for finding the place of the body in its orbit, we will consider each species of conic section separately, commencing with the ellipse. 20. Since the value of a - r can never exceed the limits - ae and + ae, we may introduce an auxiliary angle such that we shall have a - r - cos E. ae This auxiliary angle E is called the eccentric anomaly; and its geometrical signification may be easily known from its relation to the true anomaly. Introducing this value of into the equation ae (27) and writing t - T in place of to, T being the time of perihelion passage, and t the time for which the place of the planet in its orbit is to be computed, we obtain kill ~ m 3I + (t - T) -E - e sin E. (38) a But k1 I+ m But ~ — = mean daily motion of the planet ~- p; therefore a 4 (t - T) - E -e sinE. The quantity p (t- T) represents what would be the angular distance from the perihelion if the planet had moved uniformly in a circular orbit whose radius is a, its mean distance from the sun. It is called the mean anomaly, and is usually designated by M. We shall, therefore, have M= I (t — T), M= EE- e sin E. (39) When the planet or comet is in its perihelion, the true anomaly, mean anomaly, and eccentric anomaly are each equal to zero. All three of these increase from the perihelion to the aphelion, where they are each equal to 180~, and decrease from the aphelion to the perihelion, provided that they are considered negative. From the perihelion to the aphelion v is greater than E, and E is greater than lM. The same relation holds true from the aphelion to the perihelion, if we regard, in this case, the values of v, E, and M3 as negative. As soon as the auxiliary angle E is obtained by means of the mean motion and eccentricity, the values of r and v may be derived. For 54 THEORETICAL ASTRONOMY. this purpose there are various formulae which may be applied in practice, and which we will now develop. The equation a -r --- os CE, ae gives r a (1 - e cos E). (40) This also gives -- ae a cos E - ae, or - a cos E - ae, which, by means of equation (25), reduces to r cos v a cos E- ae. (41) If we square both members of equations (40) and (41), and subtract the latter result from the former, we get r2 sin2 v - a (1 - e2) sin2 E, or r sin v = a/1 e2 sin E =- b sin E. (42) By means of the equations (41) and (42) it may be easily shown that the auxiliary angle E, or eccentric anomaly, is the angle at the centre of the ellipse between the semi-transverse axis, and a line drawn from the centre to the point where the prolongation of the ordinate perpendicular to this axis, and drawn through the place of the body, meets the circumference of the circumscribed circle. Equations (40) and (41) give r(1 F cos v) = a(1 - e) (1 - cos E). By using first the upper sign, and then the lower sign, we obtain, by reduction, //r sin v - 1/a(1 - e) sin'E, 1/r7 cos v - /a(l -e) cos'E, (43) which are convenient for the calculation of r and v, and especially so when several places are required. By division, these equations give tan /1 - e tan V - tan E. (44) PLACE IN THE ORBIT. 55 Since e = sin q', we have 1 — e 1 - sin A(4 5o ~ tan'.(45~ - ^). I + e I + sin n Consequently, tan!E -- tan (45~ -. ) tan v.i (45) Again, 1/ -- e =-/1 -+ sin ( 1/1 - 2 sin cos (, which may be written V/1 + e = 1/sin2 - + cos2 q- + 2 sin 9 cos, or 1/1 e = sin s + cos 4. In a similar manner we find 1/1 - e - sin 1 D + cos. From these two equations we obtain 1/1 +- e + 1/1 -e=-2 cos v, 1/1 + e —1/1 - e 2 sin l, (46) which are convenient in many transformations of equations involving e or (p. Equation (42) gives r sin v p sin v sin E - - - -; b b (1 + e cos v) but p -a cos2 a, and b a cos D, hence sin v cos o sin v sin E=- (47) acos - 1 - e cosv Equation (41) gives r cosv - ae p cos v cos E ~ + e, oa a (- + e cos v) or -P cos v + ae + ae2 cos v cos --; a (1 + e cos v) and, putting a cos2 o instead of p, and sin (p for e, we get cos v +c cos E 1 cos (48) f we m y e cos v If we lmultiply the first of equations (43) by coslE, and the 56 THEORETICAL ASTRONOMY. second by sin E, successively add and subtract the products, and reduce by means of the preceding equations, we obtain sin 1 (v - E) =- - cos,Ip sin E, sin ~ (v -E)= - sin Lo sin E. (49) The perihelion distance, in an elliptic orbit, is given by the equation q = a ( - e). 21. The difference between the true and the mean anomaly, or v- M, is called the equaction of the centre, and is positive from the perihelion to the aphelion, and negative from the aphelion to the perihelion. When the body is in either apsis, the equation of the centre will be equal to zero. We have, from equation (39),;, E 1 M-+ e sinE. Expanding this by Lagrange's theorem, we get dF(M) ed 2 dF (M) e' F(E) - F(M) +- sin MdF (M - d( sin dF1M) ) e ^ 7 id, iMF(M) e +l d+ ( s dF( i +)3 +Mn M _ (50) +dM 2 diil 1d3d Let us now take, equation (40), a2 d F(E)- (1 — e cosE)2- r 2 and, consequently, F(M) = (1 - e cos M) Therefore we shall have -2 3 - (1 - e cos M) — 2e2 sin2 M(1 - e cos M) -e d (sin31(1- e os M) -.... Expanding these terms, and performing the operations indicated, we get ~2 2 = _ 1 + 2e cos M+ - (6 cos2 M- 4 sin2 M) (6 cos 36 sin2 M cos + (16 cos33 M 36 sin M cos M) +..., PLACE IN THE ORBIT. 57 which reduces to a2 e2 2-1+2e cosM+-+ (1+5 cos2M)+ (13 cos3M+3 cosM)+.... (51) r c 2 4 Equation (22) gives 2fdt r2 and, since f= kVp (1 + m), we have dv k/p(+ dt (52) ~~~~~or~~r or.2 dv- =/1 /- e. adt =/ e2 d. a2 But hc, and therefore a 2 d dv i (1 e2 dt /.) dM. 2 r By expanding the factor 1/l -- e2, we obtain /i-e 1 2_- 1e2 4i and hence dv (1. ~...) dM. Substituting for a its value from equation (51), and integrating, we get, since v = 0 when M=- 0, v- M=2e sin M-+ e sin 2M1+ 2 (13 sin 3M-3 sin M) +... (53) which is the expression for the equation of the centre to terms involving e3. In the same manner, this series may be extended to higher powers of e. When the eccentricity is very small, this series converges very rapidly; and the value of v- M for any planet may be arranged in a table with the argument M. For the purpose, however, of computing the places of a heavenly body from the elements of its orbit, it is preferable to solve the equations which give v and E directly; and when the eccentricity is 58 THEORETICAL ASTRONOMY. very great, this mode is indispensable, since the series will not in that case be sufficiently convergent. It will be observed that the formula which must be used in obtaining the eccentric anomaly from the mean anomaly is transcendental, and hence it can only be solved either by series or by trial. But fortunately, indeed, it so happens that the circumstances of the celestial motions render these approximations very rapid, the orbits being usually either nearly circular, or else very eccentric. If, in equation (50), we put F(E) -E, and consequently F(M) -, we shall have, performing the operations indicated and reducing, E = M + e sin M+ -e2 sin 211 + &c. (54) Let us now'denote the approximate value of E computed from this equation by E0, then will Eo + Eo - E, in which AEo is the correction to be applied to the assumed value of E. Substituting this in equation (39), we get M-= E + - aE, - e sinEo - e cos EoaEo; and, denoting by Mo the value of M corresponding to E0, we shall also have MO =Eo- e sin E. Subtracting this equation from the preceding one, we obtain 1 -e cos E It remains, therefore, only to add the value of aEo found from this formula to the first assumed value of E, or to B0, and then, using this for a new value of E,, to proceed in precisely the same manner for a second approximation, and so on, until the correct value of E is obtained. When the values of E for a succession of dates, at equal intervals, are to be computed, the assumed values of E0 may be obtained so closely by interpolation that the first approximation, in the manner just explained, will give the correct value; and in nearly every case two or three approximations in this manner will suffice. Having thus obtained the value of E corresponding to 31 for any instant of time, we may readily deduce from it, by the formulae already investigated, the corresponding values of r and v. In the case of an ellipse of very great eccentricity, corresponding to the orbits of many of the comets, the most convenient method of PLACE IN THE ORBIT. 59 computing r and v, for any instant, is somewhat different. The manner of proceeding in the computation in such cases we shall consider hereafter; and we will now proceed to investigate the formulae for determining r and v, when the orbit is a parabola, the formulse for elliptic motion not being applicable, since, in the parabola, a = o, and e 1. 22. Observation shows that the masses of the comets are insensible in comparison with that of the sun; and, consequently, in this case, nz-0 and equation (52), putting for p its value 2q, becomes k/ 2q dt - r2dv, or kl/2 dt= 22 dv which may be written k_ = t (I + tan2 v) sec2 ~vdv- (1 + tan2 -v) d tan Bv. 1/'2 qu Integrating this expression between the limits T and t, we obtain k(t-T) tanv + - tan3, v (55) i/2q2 which is the expression for the relation between the true anomaly and the time from the perihelion, in a parabolic orbit. Let us now represent by Tr the time of describing the arc of a parabola corresponding to v 90~; then we shall have ko 4 3 - 3' ~~~~3or 4 _3k 4 qf 1/2 To 3k Now, - is constant, and its logarithm is 8.5621876983; and if we take q -1, which is equivalent to supposing the comet to move in a parabola whose perihelion distance is equal to the semi-transverse axis of the earth's orbit, we find log r, 2.03987229, or r= 109.61558 days; that is, a comet moving in a parabola whose perihelion distance 60 THEORETICAL ASTRONOMY. is equal to the mean distance of the earth froin the sun, requires 109.61558 days to describe an arc corresponding to v- 90~. Equation (55) contains only such quantities as are comparable with each other, and by i t - T, the time from the perihelion, may be readily found when the remaining terms are known; but, in order to find v from this formula, it will be necessary to solve the equation of the third degree, tan 1v being the unknown quantity. If we put x =tan 1v, this equation becomes X-+ 3x - a -0, in which a is the known quantity, and is negative before, and positive after, the perihelion passage. According to the general principle in the theory of equations that in every equation, whether complete or incomplete, the number of positive roots cannot exceed the number of variations of sign, and that the number of negative roots cannot exceed the number of variations of sign, when the signs of the terms containing the odd powers of the unknown quantity are changed, it follows that when a is positive, there is one positive root and no negative root. When a is negative, there is one negative root and no positive root; and hence we conclude that equation (55) can have but one real root. We may dispense with the direct solution of this equation by forming a table of the values of v corresponding to those of t- T in a parabola whose perihelion distance is equal to the mean distance of the earth from the sun. This table will give the time corresponding to the anomaly v in any parabola, whose perihelion distance is q, by multiplying by q2, the time which corresponds to the same anomaly in the table. We shall have the anomaly v corresponding to the time t- T by dividing t - T by q', and seeking in the table the anomaly corresponding to the time resulting from this division. A more convenient method, however, of finding the true anomaly from the time, and the reverse, is to use a table of the form generally known as Barker's Table. The following will explain its construction: Multiplying equation (55) by 75, we obtain 75k 3 (t- T) - 75 tan -v + 25 tans Av. 1/ q2 Let us now put M= 75 tan ~v + 25 tan3 V, PLACE IN THE ORBIT. 61 75k and Co -= -, which is a constant quantity; then will v2 C (t- T) M. q The value of Co is log C - 9.9601277069. Again, let us take qC - which is called the mean daily motion in the parabola; then will M- m (t - T) = 75 tan ~v + 25 tan3 Mv. If we now compute the values of HM corresponding to successive values of v from v- 0~ to v 180~, and arrange'them in a table with the argument v, we may derive at once, from this table, for the time (t - T) either M when v is known, or v when M — m (t- T) is known. It may also be observed that when t- T is negative, the value of v is considered as being negative, and hence it is not necessary to pay any further attention to the algebraic sign of t - T than to give the same sign to the value of v obtained from the table. Table VI. gives the values of M for values of v from 0~ to 1800~, with differences for interpolation, the application of which will be easily understood. 23. When v approaches near to 180~, this table will be extremely inconvenient, since, in this case, the differences, between the values of MI for a difference of one minute in the value of v increase very rapidly; and it will be very troublesome to obtain the value of v from the table with the requisite degree of accuracy. To obviate the necessity of extending this table, we proceed in the following manler:Equation (55) may be written k (t- T) _= I tan8'v (1 + 3 cot2 Iv); 1/2 T and, multiplying and dividing the second member by (1 + cot2 ~v)% we shall have k (t- T ) 1 -+ 3 cot2 Iv V(tq T)__ I tan8 Iv (1 + cot2 -v)3 1 7- ctv)3 3 37^. 3 f' L cot, V)/ e~~~tt ~Z~y * 62 THEORETICAL ASTRONOMY. 2 But 1 + co sin v tan and consequently 2 sin v tan IV k (t- T) 8 1 + 3 cot2 2v 1/2 q 3 sin v ( l+cot2 v)' Now, when v approaches near to 180~, cot Iv will be very small, and the second factor of the second member of this equation will nearly 1. Let us therefore denote by w the value of v on the supposition that this factor is equal to unity, which will be strictly true when v 180~, and we shall have, for the correct value of v, the following equation: V W + A0, A being a very small quantity. We shall therefore have 3 tan I (w + A)+ tan' (w+ A), and, putting tan 1w -, and tan h = x, we get, from this equation, (1 + 0)33 0 + x (0 + x)3 03 1 - x (1 -- x)' Multiplying this through by 03 (1 - Ox), expanding and reducing, there results the following equation: 1 + 302= 30 (1 + 402 + 204 + 06) X - 302 (1 + 402 + 240 + 6) X2 + 03 (2 + 602 + 304 + 06) X3. Dividing through by the coefficient of x, we obtain 1 + 302 2 + 02 (2 + 682 + 30 + 06) x3 3~ (1 + 402 + 204 + 06) 3 (1 + 402 + 204 + 6) Let us now put 1 + 302 30 (1 + 402 + 204 + 0) y; then, substituting this inthe preceding equation, inverting the series and reducing, we obtain finally + (4 + 18 + 904 + 56) + &C - y + y +3 (1 + 402+ 2 + + o6) But tan A0 = x, therefore A, = 2x - x3 +..... PLACE IN THE ORBIT. 63 Substituting in this the value of x above found, and reducing, we obtain -- 2 - 32* -+- 166 t+ 100s AO (2y 2y2+ 32(1+402+ 160+ 0 y3 + &c. For all the cases in which this equation is to be applied, the third term of the second member will be insensible, and we shall have, to a sufficient degree of approximation, Ao = 2y + 20y2. Table VII. gives the values of a0, expressed in seconds of arc, corresponding to consecutive values of w from w = 155~ to w 180~. In the application of this table, we have only to compute the value of M precisely as for the case in which Table VI. is to be used, namely, M= m (t —T); then will w be given by the formula ^200 sin w -- M20' since we have already found k (t- T) 8 1/ 2 q 3sinw'.or i 8q V/2 200 sin w- 3 (t -T)k M' Having computed the value of w from this equation, Table VII. will furnish the corresponding value of a0; and then we shall have, for the correct value of the true anomaly, V -w +- A0, which will be precisely the same as that obtained directly from Table VI., when the second and higher orders of differences are taken into account. If v is given and the time t- T is required, the table will give, by inspection, an approximate value of A~ using v as argument, and then w is given by w =- V - Ao. 64 THEORETICAL ASTRONOMY. The exact value of A is then found from the table, and hence we derive that of w; and finally t- T from 200 q t - T CO sin3w 24. The problem of finding the time t - T when the true anomaly is given, may also be solved conveniently, and especially so when v is small, by the following process:Equation (55) is easily transformed into 3k (t- T) _ sin v (3 2 sin, -q= c-"OS'v ^ (3 - 2sin ^v), 2 qV2 Cos from which we obtain, since q =r cos2'v, 3^ (t~-T) _3( sin 2v ) 4 ( sin g 2 r0 V V~ Let us now put sin v sin x - ~-, and we have 3k (t T) _ - 3 sin x -4 sin3 x - sin 3x. 2 rConsequently, 2 3 3 t-T —k r sin 3x, which admits of an accurate and convenient numerical solution. To facilitate the calculation we put sin 3x N - sin v' the values of which may be tabulated with the argument v. When v = 0, we shall have N= -V' 2, and when v = 90, we have N- 1; from which it appears that the value of N changes slowly for values of v from 0~ to 90~. But when-v= 180, we shall have N= c, and hence, when v exceeds 90~, it becomes necessary to introduce an auxiliary different from N. We shall, therefore, put in this case, N' = N sin v = sin 3x; PLACE IN THE ORBIT. 65 from which it appears that N'- 1 when v =90~, and that N'= i/2 when v = 180~. Therefore we have, finally, when v is less than 90~, 2 3 t - T Nr2 sin v, 3k and, when v is greater than 90~, 2 3 t- T-= N'r2, 3k 2 in which log 3= 1.5883272995, from which t - T is easily derived t3iC when v is known. Table VIII. gives the values of N, with differences for interpolation, for values of v from v = 0~ to v 90~, and the values of N' for those of v from v- 90~ to v- 180~. 25. We shall now consider the case of the hyperbola, which differs from the ellipse only that e is greater than 1; and, consequently, the formulae for elliptic and hyperbolic motion will differ from each other only that certain quantities which are positive in the ellipse are negative or imaginary in the hyperbola. We may, however, introduce auxiliary quantities which will serve to preserve the analogy between the two, and yet to mark the necessary distinctions. For this purpose, let us resume the equation p cos % 2 cos - (v + A) cos 2 ( - )' When v 0, the factors cosI(v + A) and cos(v - ) in the denominator will be equal; and since the limits of the values of v are 180~-~ and -(180~ - ), it follows that the first factor will vanish for the maximum positive value of v, and that the second factor will vanish for the maximum negative value of v, and, therefore, that, in either case, r oo. In the hyperbola, the semi-transverse axis is negative, and, consequently, we have, in this case, p - a (e 1), or a p cot2. We have, also, for the perihelion distance, q a(e -1). Let us now put tanF= tan- + -,' (56) 5 66 THEORETICAL ASTRONOMY. which is analogous to the formula for the eccentric anomaly E in an ellipse; and, since e -, we shall have cos e -1 1 -cos 2 e +- 1 + cosa 2n and, consequently, tan -F = tan Iv tan 57. (57) We shall now introduce an auxiliary quantity a, such that 1 + tan;'F a tan (450~ +,F) - 1 +- tan. F' 1 - tan IF' whence we derive tan F- 1 (58) and also cos I (v - =OS. ~V w/ (59) cos ~ (v +,) This last equation shows that a- 1 when the comet is in its perihelion; a =oo when v =180~ —; and =Owhenv -(180~ - ). 2 tan'IF Since tan F -- t2 F' we shall have 1 - tan2 4F tan F= ( - (60),+1 Squaring this equation, adding 1 to both members, and reducing we obtain cos F= (61) Replacing a in this equation by its value from equation (59), we get 1 cos2 ~ (v + ) +os2 ( - ) cos F 2 cos (v + ) cos, (v-+)' or 1 1 + cos v cos _ (e + cos v) cos S cos F 2 cos (v + ) cos (v - ) 2 cos I (v + +) cos ~(v - which reduces to 1 __r(e+cos v) cosF-~ ~ (p2) PLACE -IN THE ORBIT. 67 If we add +T 1 to both members of this equation, we shall have 1 -t- cos F r(e - 1) (1 -+ cos v) cos F p Taking first the upper sign, and then the lower sign, and reducing, we get l/rsin v (e ) sin F, l/cos F Vr cos = f oV ) cos a 1. (63) These equations for finding r and v, it will be observed, are analogous to those previously investigated for an elliptic orbit. These equations give, by division, tan -! v -- tan -F, tang Je - 1 tan ~F, which is identical with the equation (56), and may be employed to verify the computation of r and v. j Multiplying the last of equations (63) by the first, putting for e2- 1 its value tan2 4, and reducing, we obtain r sin v atan, tan F==a tan —. (64) Further, we have p cos v ar (e + cos v) r cos v - = ae -- I + e cos v' p which, combined with equation (62), gives rcosv_ —a ) e -a( 2e- -- ~ (65) If we square these values of r sin v and r cos v, add the results together, reduce, and extract the square root, we find r —a -1 =Ja 2e - + — ~ (66) We might also introduce the auxiliary quantity a into the equations (63); but such a transformation is hardly necessary, and, if at all desirable, it can be easily effected by means of the formule which we have already derived. 68 THEORETICAL ASTRONOMY. 26. Let us now resume the equation cos (v - 4) cos (v + 4)' Differentiating this, regarding ~ as constant, we have sin dv -2 cos ( + ), and, dividing this equation by the preceding one, we get ds _ sin d_ a 2 cos - (v + ) cos (v-) But p cos 2 cos (v + 4) cos (v -)' consequently, da r tan 4 dv, which gives r2dv - pr da. a tan + Substituting this value of r2dv in equation (22), and putting instead of 2f its value kv/p, from equation (30), the mass being considered as insensible in comparison with that of the sun, we get klp dt pr d. atan Then, substituting for r its value from equation (66), and for p its value a tan2 ~, we have k/p dt == a2tan4 ( e (I + - da. Integrating this between the limits T and t, we obtain kl/ (t- T) atan ( -e ( —— I loge a), (67) in which loge a is the Naperian or hyperbolic logarithm of a. Since l/p -l/~a tan', if we put k = a PLACE IN THE ORBIT. 69 in which v is the mean daily motion; and if we also put - (t-T) - N0, in which No corresponds to the mean anomaly M in an ellipse, we shall have, from equation (67), N - (1 ( -!) loge1 a. (68) If we multiply both members of this equation by =0.434294482, the modulus of the common system of logarithms, and put y- -NA (t- T), a" we shall have x —;e -A A - log,; wherein log A = 9.6377843113, and log Ak - 7.8733657527. Let us now introduce F into this formula; and for this purpose we have tan Fi ^ ), and also log =- log tan (45~ + -F). Therefore we obtain N= eA tan F- log tan (450 +.F). (69) This equation will give, directly, the time t- T from the perihelion, when a, e, and Fare known; but, since it is transcendental, in the solution of the inverse problem, that of finding the true anomaly and radius-vector from the time, the value of F can only be found by successive approximations. If we differentiate the last equation, regarding N and F as variable we get dN= (e - cos F) dF. cos2F Hence, if we denote an approximate value of F by F,, and the corresponding value of N by N, the correction aF, to the assumed value of F may be computed by the formula F (N - N,) cos2F, (e - cosF,) 70 THEORETICAL ASTRONOMY. This correction being applied to F,, a nearer approximation to the true value of F will be obtained; and by repeating the operation there results a still closer approximation. This process may be continued until the exact value of F is found, and, when several successive places are required, the first assumed value may be estimated, in advance, so closely that a very few trials will suffice. In practice, however, cases will rarely occur in which this formula will be applied, since the probability of hyperbolic motion is small, and, whenever any positive indication of an eccentricity greater than I has been found to exist, it has only been after a very accurate series of observations has been introduced as the basis of the calculation. For a majority of the cases which do really occur, the most accurate and convenient method of finding r and v will be explained hereafter. 27. If we consider the equation M= E e sin E, we shall see that, when logarithms of six or seven decimals are used, the error which may exist in the determination of E when lMk and e are given, will increase as e increases, but in a much greater ratio; and, when the eccentricity becomes nearly equal to that of the parabola, the error may be very great. In the case of hyperbolic motion, also, the numerical solution of equation (69), when e — 1 is very small, and with the ordinary logarithmic tables, becomes very uncertain. This can only be remedied, when equations (39) and (69) are employed, by using more extended logarithmic tables; and when the orbit differs only in an extremely slight degree from a parabola, even with the most extended logarithmic tables which have been constructed, the error may be very large. For this reason we have recourse to other methods, which will give the required accuracy without introducing inconveniences which are proportionally great. We shall, therefore, now proceed to develop the formulae for finding the true anomaly in ellipses and hyperbolas which differ but little from the parabola, such that they will furnish the required accuracy, when the exact solution of equations (39) or (69) with the logarithmic tables in common use is impossible. For this purpose, let us resume equation (22), which, by substituting for 2f its value klVp, the mass of the comet being neglected in comparison with that of the sun, becomes k I/p dt == r2dv, PLACE IN THE ORBIT. 71 or k-V/P dt~( p~ ~dv k p dt - (1 + e cos )2' Let us now put i = tan v, and we shall have 1 -u2 d 2du cos -- u; dv 1 - Substituting these values in the preceding equation, and putting 1-e 1 q- e i, we get d- 2p2 (l + 2)du k pdt (1-+ e)2 (1 + i,2)2 or, since p=q (1 e), kl / + e dt (1 + u) du 2 - (1 + i2)' Let us now develop the second member into a series. This may be written thus: dut (1 + iu2) ( + j2)- 2; and developing the last factor into a series, we obtain (1 + iu2)- = 1- 2iu2 + 3i2u4 4-i36 + &c. Consequently, (1 -+ 2) (1 + iu2)-2 1 + u2 - 2i (+a2 + u') + 3i2 (u +- 6) - 4i (u6 t ) - +....; Multiplying this equation through by du, and integrating between the limits T and t, the result is k(t- T)l/l u +,U - 2i (1, + O5) + 3i2 (5 + i7) 2qg - 4i3 (u7 + I99) + &c. L. (70) In the case of the parabola, e = 1 and i =0, and this equation becomes identical with (55). Let us now put k (t -T) 1/ 1 + e U+ 3 (71) 2q - 3u (71) 2q 72 THEORETICAL ASTRONOMY. and also U= tan; then the angle Vwill not be the true anomaly in the parabola, but an angle derived from the solution of a cubic equation of the same form as that for finding the parabolic anomaly; and its value may be found by means of Table VI., if we use for M the value computed from 75k (t - T) <1 + e 2 2 Let U be expanded into a series of the form U = -+ i +- i2 + yr3 +.... which is evidently admissible, a,, r,.... being functions of u and independent of i. It remains now to determine the values of the coefficients A, r,, &c., and, in doing so, it will only be necessary to consider terms of the third order, or those involving i3, since, for nearly all of those cases in which the eccentricity is such that terms of the order i4 will sensibly affect the result, the general formula already derived, with the ordinary means of solution, will give the required accuracy. We shall, therefore, have U + U3 u, + ai + Xi2 + + ( + + + + + Y, ) or, again neglecting terms of the order 4, u+ 1 U3 == U + - u + i ( + u) a + 2 (uf2 + (1+ uf2) ) + i3 (i3 + 2uoaj + (1 + t2) r). But we have already found, (70), k(t -T)1/Te __ U u+ I U3 = u + l u -3 2i (Q + at&) 2qY + 3i2 (I u5 + Ju7) 4i3 (u 1 + _ 9). Since the first members of these equations are identical, it follows, by the principle of indeterminate coefficients, that the coefficients of the like powers of i are equal, and we shall, therefore, have (a + Mb2) ffi = 1a -d 1Q5) t2 + (1 + in+ ) 2P + 3 (5U5 + ue), 3' + 2uNfa + (1 + i)2). 4 (7 + 1 9). From the first of these equations we find PLACE IN THE ORBIT. 73 3 ~1 5 2 (u + U ) The second equation gives 3(L5 t+ 1,7) _ uQ2 1 +u2 or, substituting for a its value just found, and reducing, 3 (I5U + 3 —7 U- + 3 L9 + 547 11) (i +(1+u) We have also - 4 (-uI7 + -'1) - a ~3- 20,ue 1 + to and hence, substituting the values of a and 9 already found, and reducing, we obtain finally 4 (1u7 + 1 2 9 8I + 1 017 4 1 16t13 + T223 t15+ 2 7) r — (+1 -(+ t2)' Again, we have tan U- tan (I +- i -+ i2 + ri3). Developing this, and neglecting terms of the order i, we get tan' U= tan'u + (1+ ( +a + 2 (1 + -2)2 ('2i2 + 2afii3) 2 1 + (1 3 3i3' Now, since = tan v and U= tan g V, we shall have V — + 1+ (ai + (ii + ri3) ( + 2 i2 + 2cafis) + u- ) 3i3 or 2. 2/5 2 2U 7 -V + 1 + U2' + L2 (1+ ) ) (1 + U2 ( + i2)2 ( + 2))3 4 2r 4cL u 2('~1) Substituting in this equation the values of a, P, and r already found, and reducing, we obtain finally 4'U3 4t + + 4. 6 + TO ~ T 38 +1 I 3 V v -+ i +.3 1 0 V v — (1 +i) (1 + U) 8 7 + 52 8 1 2638411 4 13 5128 15+ 904 T 83 T ( 1 14 +I -- I 3 1 8 _ o 4787 5 3. (73) (] U26 74 THEORETICAL ASTRONOMY. This equation can be used whenever the true anomaly in the ellipse or hyperbola is given, and the time from the perihelion is to be determined. Having found the value of V, we enter Table VI. with the argument V and take out the corresponding value of 3I; and then we derive t - T from t _T__M 32 -- co \+ e in which log C0 - 9.96012771. For the converse of this, in which the time from the perihelion is given and the true anomaly is required, it is necessary to express the difference v - V in a series of ascending powers of i, in which the coefficients are functions of U. Let us, therefore, put u= U + ai + l'i2 + r'i3 + &c. Substituting this value of u in equation (70), and neglecting terms multiplied by i4 and higher powers of i, we get k(t- T)V1 +- e U+ U3+_ ('(1 + U2) _U3 2_ _U5)i 3 2q2 + (I'(1 + U2) + U' 2- 2 U2U ( + U2) + 3 U5 + 3 Ui) 2 + (r' (1 + U2) +' a3 + 2 Ua''+- 3U4a' (I + U2)- 2' U2 (1 + U2) -4 U3t2 - 2 Ud2- - 4 U - U9) i3. But, since the first member of this equation is equal to U I- U3, we shall have, by the principle of indeterminate coefficients,' (I + U) 2- U3 - U -O, (1 + U2)+ ud2-2 (+ U + U5 + 3 U' 0, 71 (I + U2) + I U~3 U2 (I + U2) r'(l + U2) + "'+ 2Ua1'' + 3U4'd(1 + U2) - 2'U2(1 + U2) -4 U3a' 2 Ua 2- 4 U7 4 U9 0. From these equations, we find 2 U3 +'U5 " + U' ( 1+ U2)3 29U + 1 7928 US 10328 U+" + U + U135 66 + U 17 _ 3 1 283 3 1 28 3a 15 JI f 48 Ua 1 57 r~ (1+ U2)5 If we interchange v and V in equation (72), it becomes, writing at, 3', r' for a, p, r, PLACE IN THE ORBIT. 75 ^,V 2ar2 ( 2/ 2d"U., V -+ U 1 U'I (1 + U2)2' 2'_ 4 aWi4/U +2 (U" — ) d~3) i.' I+ U ( (1 + U2)2 (1 / U2)' Substituting in this equation the above values of a', I', and y', and reducing, we obtain, finally, 4T3_+ 4 U5 22U 5+J 598U L 86 o U + 17 U 2 v 3 T-t- 3 ^ -a T5 ~.5rT T05, 2-175 -2' (1 + U')2' (1 + U2)4 f84 U7 9752 U9+ 37328 Ul+ 1 f48rU13+ 17 6 78 U15 184 U17 + 3 1a T 835 t 141T75 1 5756 I785 I 7875 3 1 (74) (1 +- U2)6 by means of which v may be determined, the angle Vbeing taken from Table VI., so as to correspond with the value of M derived from M=(t-T)q3 \ Equations (73) and (74) are applicable, without any modification, to the case of a hyperbolic orbit which differs but little from the parabola. In this case, however, e is greater than unity, and, consequently, i is negative. 28. In order to render these formulae convenient in practice, tables may be constructed in the following manner:Letx - v or V, and tan -x = 0, and let us put A 03+ 405 100(1 + 02)2 B 2 05+ 5 807 + %09 + _ T9 8 10000 (1 + 02)4 62 W5 + 984 6 6507 + "1109 + l+75 +'B5 5 3.1 85 0 11 1oooo (1- o+ 2)4 1000000 (I + 02)6 80' + 58 + 2638401 + 464 13+ 28015 + 9048 wherein s expresses the number of seconds corresponding to the length of arc equal to the radius of a circle, or logs 5.31442513. We shall, therefore, have: When x = V, v = + A (100i) + B (100i)2 + C(00i)3; 76 THEORETICAL ASTRONOMY. and, when x = v, v= v - A (100i) + B' (100) - C' (100i). Table IX. gives the values of A, B, B', C, and C' for consecutive values of x from x = 00 to x = 1490, with differences for interpolation. When the value of v has been found, that of r may be derived from the formula q(1 + e) 1 - e cos v Similar expressions arranged in reference to the ascending powers / /2 \2 of (1- e) or of (( ) 1 I ) may be derived, but they do not con2 ) verge with sufficient rapidity; for, although ( ( I -1 ) is less than i,yet the coefficients are, in each case, so much greater thah those of the corresponding powers of i, that three terms will not afford the same degree of accuracy as the same number of terms in the expressions involving i. 29. Equations (73) and (74) will serve to determine v or t - T in nearly all cases in which, with the ordinary logarithmic tables, the general methods fail. However, when the orbit differs considerably from a parabola, and when v is of considerable magnitude, the results obtained by means of these equations will not be sufficiently exact, and we must employ other methods of approximation in the case that the accurate numerical solution of the general formulae is still impossible. It may be observed that when E or F exceeds 50~ or 60~, the equations (39) and (69) will furnish accurate results, even when e differs but little from unity. Still, a case may occur in which the perihelion distance is very small and in which v may be very great before the disappearance of the comet, such that neither the general method, nor the special method already given, will enable us to determine v or t- T with accuracy; and we shall, therefore, investigate another method, which will, in all cases, be sufficiently exact when the general formulae are inapplicable directly. For this purpose, let us resume the equation k(t- T) - esinE, 8 PLACE IN THE ORBIT. 77 which, since q = a (- e), may be written k (t T) V/1- e 1 1 1 9e. 3 -- / _ (9E - sin E) + - (E - sinE). 2 10 1 If we put E-sin E ^9E - sinE' we shall have k (t -T)/1 — e 20i/A _ 1 1 +9e 2q2 * 9E- + sin E- 3'5 (1 e) Let us now put B 9E + sin E 201/A and tan'2w 1 + 9e_ A 5 (1- e) then we have k(t-T) 1/ (1 + 96) -~ ~T) - V T 1 9e) tan w + 1 tan' w. (75) 3 B T/2 When B is known, the value of w may, according to this equation, be derived directly from Table VI. with the argument M 75k(t T) /v(1 +9e) 3B 1/2 q and then from w we may find the value of A. It remains, therefore, to find the value of B; and then that of v from the resulting value of A. Now, we have 2 -tan 1E sin E -- 1 + tan"2E' and if we put tan2 E - T, we get sin E = 2 =_ 2r:2 ( T- + 2 — ~ + &c.). sin E=1 - We have, also, E =2 tan' =2 2 (1 - + 12 - T3 + &c.). 3 7 5? t ~j 78 THEORETICAL ASTRONOMY. Therefore, 15 (E - in E) 2+1 (10 - 6+0o2 + 9_ t _ + &c.), and 9E + sin E-= 2r~ (10 - 12r + 144T2 16 -+ 84-_ &c.) Hence, by division, 15 E- sinE A=r - 42 +43 _ 192r4 2 + 1744375 19Eq sinE --? 3 -c 1 0 8 99 8 85T6 + &C.; and, inverting this series, we get A 1 -_ ~ _ A + TA' +.A1 + _ 89fA4+ _ ~84 -&. &c. which converges rapidly, and from which the value of ~ may be found. Let us now put A I 2~ C2 then the values of C may be tabulated with the argument A; and, besides, it is evident that as long as A is small C2 will not differ much from 1 + 4A. Next, to find B, we have -A ^ (1 210 46-2 I041 - + 161 4 - &c.), and hence i(9E- +sinE) B 6 753r2 + 336874- &c.; -^-= 1 + T -,i +,i3 - &c.; V77 625 9007 1/A from which we easily find B = 1 + 3A+ a, + 3 337 A4 + &c. If we compare equations (44) and (56), we get tan -E=/= V 1 tan F. Hence, in the case of a hyperbolic orbit, if we put tan2 FF-=', we must write -' in place of r in the formulhe already derived; and, from the series which gives A in terms of r, it appears that A is in this case negative. Therefore, if we distinguish the equations for PLACE IN THE ORBIT. 79 hyperbolic motion from those for elliptic motion by writing A', B', and C' in place of A, B, and C, respectively, we shall have 1 A' t1 +4A'+ At 8 2.. 3+ 2 87 441 At+ &c., C~2 ~I =,at- ga'+,?gA~2 — L5.1,3 7't6 A~a 13 4 4 &C B' 1 + -f3 A2 -_ -2 HA3 +- 3-34 75 —14 - &c. Table X. contains the values of log B and log C for the ellipse and the hyperbola, with the argument A, from A= 0 to A- 0.3. For every case in which A exceeds 0.3, the general formulae (39) and (69) may be conveniently applied, as already stated. The equation tan v = 1-e+ e tan JE gives tan2 == _ e A C2 or, substituting the value of A in terms of w, tan v - Ctan C t1a 9e) (76) The last of equations (43) gives r cos v - q cos E +t +I -tan2 -E Hence we derive r - 77) (1 + AC2)cos o (77) The equation for v in a hyperbolic orbit is of precisely the same form as (76), the accents being omitted, and the value of A being computed from 5 (e —1) A- = 1 tan9 w. (78) For the radius-vector in a hyperbolic orbit, we find, by means of the last of equations (63), (1A AC2) cos2 v (79) When t T is given and r and v are required, we first assume B = I, and enter Table VI. with the argument M (t - T). (1 +9e) B' 80 THEORETICAL ASTRONOMY. in which log C0 = 9.96012771, and take out the corresponding value of w. Then we derive A from the equation A 5(1-e) tan 1 -+9e tanw in the case of the ellipse, and from (78) in the case of a hyperbolic orbit. With the resulting value of A, we find from Table X. the corresponding value of log B, and then, using this in the expression for M, we repeat the operation. The second result for A will not require any further correction, since the error of the first assumption of B 1 is very small; and, with this as argument, we derive the value of log C from the table, and then v and r by means of the equations (76) and (77) or (79). When the true anomaly is given, and the time t - T is required, we first compute r from 1 —e =. + e tan2 v, in the case of the ellipse, or from e-1 T =e l tan' v, e + 21 in the case of the hyperbola. Then, with the value of T as argument, we enter the second part of Table X. and take out an approximate value of A, and, with this as argument, we find logB and log C. The equation C2 will show whether the approximate value of A used in finding log C is sufficiently exact, and, hence, whether the latter requires any correction. Next, to find w, we have tan 1 1+9e tanlwz —.w t2 - C (1 be)' and, with w as argument, we derive M from Table VI. Finally, we have t-T= MBq (80) Co / -j (1 + 9e) by means of which the time from the perihelion may be accurately determined. POSITION IN SPACE. 81 30. We have thus far treated of the motion of the heavenly bodies, relative to the sun, without considering the positions of their orbits in space; and the elements which we have employed are the eccentricity and semi-transverse axis of the orbit, and the mean anomaly at a given epoch, or, what is equivalent, the time of passing the perihelion. These are the elements which determine the position of the body in its orbit at any given time. It remains now to fix its position in space in reference to some other point in space from which we conceive it to be seen. To accomplish this, the position of its orbit in reference to a known plane must be given; and the elements which determine this position are the longitude of the perihelion, the longitude of the ascending node, and the inclination of the plane of the orbit to the known plane, for which the plane of the ecliptic is usually taken. These three elements will enable us to determine the co-ordinates of the body in space, when its position in its orbit has been found by means of the formulae already investigated. The longitude of the ascending node, or longitude of the point through which the body passes from the south to the north side of the ecliptic, which we will denote by, is the angular distance of this point from the vernal equinox. The line of intersection of the plane of the orbit with the fundamental plane is called the line of nodes. The angle which the plane of the orbit makes with the plane of the ecliptic, which we will denote by i, is called the inclination of the orbit. It will readily be seen that, if we suppose the plane of the orbit to revolve about the line of nodes, when the angle i exceeds 180~, g will no longer be the longitude of the ascending node, but will become the longitude of the descending node, or of the point through which the planet passes from the north to the south side of the ecliptic, which is denoted by 3, and which is measured, as in the case of g, from the vernal equinox. It will easily be understood that, when seen from the sun, so long as the inclination of the orbit is less than 90~, the motion of the body will be in the same direction as that of the earth, and it is then said to be direct. When the inclination is 90~, the motion will be at right angles to that of the earth; and when i exceeds 90~, the motion in longitude will be in a direction opposite to that of the earth, and it is then called retrograde. It is customary, therefore, to extend the inclination of the orbit only to 90~, and if this angle exceeds a right angle, to regard its supplement as the inclination of the orbit, noting simply the distinction that the motion is retrograde. 6 82 THEORETICAL ASTRONOMY. The longitude of the perihelion, which is denoted by wr, fixes the position of the orbit in its own plane, and is, in the case of direct motion, the sum of the longitude of the ascending node and the angular distance, measured in the direction of the motion, of the perihelion from this node. It is, therefore, the angular distance of the perihelion from a point in the orbit whose angular distance back from the ascending node is equal to the longitude of this node; or it may be measured on the ecliptic from the vernal equinox to the ascending node, then on the plane of the orbit from the node to the place of the perihelion. In the case of retrograde motion, the longitudes of the successive points in the orbit, in the direction of the motion, decrease, and the point in the orbit from which these longitudes in the orbit are measured is taken at an angular distance from the ascending node equal to the longitude of that node, but taken, from the node, in the same direction as the motion. Hence, in this case, the longitude of the perihelion is equal to the longitude of the ascending node diminished by the angular distance of the perihelion from this node. It may, perhaps, seem desirable that the distinctions, direct and retrograde motion, should be abandoned, and that the inclination of the orbit should be measured from 0~ to 180~, since in this case one set of formulse would be sufficient, while in the common form two sets are in part required. However, the custom of astronomers seems to have sanctioned these distinctions, and they may be perpetuated or not, as may seem advantageous. Further, we may remark that in the case of direct motion the sum of the true anomaly and longitude of the perihelion is called the true longitude in the orbit; and that the sum of the mean anomaly and longitude of the perihelion is called the mean longitude, an expression which can occur only in the case of elliptic orbits. In the case of retrograde motion the longitude in the orbit is equal to the longitude of the perihelion minus the true anomaly. 31. We will now proceed to derive the formule for determining the co-ordinates of a heavenly body in space, when its position in its orbit is known. For the co-ordinates of the position of the body at the time t. we have x — r cos v, y - r sin v, POSITION IN SPACE. 83 the line of apsides being taken as the axis of x, and the origin being taken at the centre of the.sun. If we take the line of nodes as the axis of x, we shall have x r cos (v +- W), y = r sin (v+ W), w being the arc of the orbit intercepted between the place of the perihelion and df the node, or the angular distance of the perihelion from the node. No'w, we have o) — Tr- in the case of direct motion, and w= -ir in the case of retrograde motion; and hence the last equations become x = r cos (v z 7i =F 2) y = r sin (v ~ z =T ) the upper and lower signs being taken, respectively, according as the motion is direct or retrograde. The arc v +~ T - 7 - u is called the argunment of the latitude. Let us now refer the position of the body to three co-ordinate planes, the origin being at the centre of the sun, the ecliptic being taken as the plane of xy, and the axis of x, in the line of nodes. Then we shall have Xr = COS U, y' = = r sin u cos i, z' - r sin i sin i. If we denote the heliocentric latitude and longitude of the body, at the time t, by b and I, respectively, we shall have x' - r co os b cos (I ), y' = r cos b sin (I- g ), z' r sin b, and, consequently, cos u = cos b cos (I1- ), ~ sin u cos i. cos b sin (I -), (81) sin u sin i = sin b. From these we derive tan (I - ) -- = tan u cos i, tan b = ~ tan i sin ( — S ), (82) which serve to determine I and b, when 2, i, and i are given. Since 84 THEORETICAL ASTRONOMY. cos b is always positive, it follows that I - 2 and u must lie in the same quadrant when i is less than 90~; but if i is greater than 90~, or the motion is retrograde, 1 - Q and 360 - u will belong to the same quadrant. Hence the ambiguity which the determination of I~- g by means of its tangent involves, is wholly avoided. If we use the distinction of retrograde motion, and consider i always less than 90~, 1 - 2 and — u will lie in the same quadrant. 32. By multiplying the first of the equations (81) by sinu, and the second by cos u, and combining the results, considering only the upper sign, we derive cos b sin ( - 1+ - ) 2 sin u cos ut sin2 /i, or cos b sin (u -1 + 2 ) sin 2u sin2 /i. In a similar manner, we find cos b cos (u - 1-t + ) = cos2 + sin2u cos i, which may be written cos b cos (u - + g) -=(1 u)+1 (1 - cos 2u) cos i, or cos b cos (u - I+ ) = (1 + cos ) + (1 cos ) cos 2u; and hence cos b cos (u - 1 - + ) = cos2 i + sina i cos 2u. If we divide this equation by the value of cos b sin (u - 1 -+ g) already found, we shall have tan' Ai sin 2it tan (u -- + ) + t2 is 2. (83) +. I-ttan2 i cos2u( The angle u -1 + 2 is called the reduction to the ecliptic; and the expression for it may be arranged in a series which converges rapidly when i is small, as in the case of the planets. In order to effect this development, let us first take the equation n sin x tan y _ e t 1- n+ cos x Differentiating this, regarding y and n as variables, and reducing, we find dy sin x d — 1 + 2n1 cos x - n2' POSITION IN SPACE. 85 which gives, by division, or by the method of indeterminate coefficients, =d_ sin x - n sin 2x + n2 sin 3 - n3 sin 4x + &c. dn Integrating this expression, we get, since y = 0 when x = 0, y nsin X ~- n2 sin 2x -+ 13 nsin 3x - In4 sin 4x +...., (84) which is the general form of the development of the above expression for tan y. The assumed expression for tan y corresponds exactly with the formula for the reduction to the ecliptic by making n - tan2 i and x = 2X; and hence we obtain u - -]- = tan2 i sin 2u - tan4 i sin 4u + 1 tan6 1i sin 61 -4 tan8 -i sin 8u + 1 tan1~ Ii sin 10i - &c. (85) When the value of i does not exceed 100 or 12~, the first two terms of this development will be sufficient. To express u — 1 + in seconds of arc, the value derived from the second member of this equation must be multiplied by 206264.81, the number of seconds corresponding to the radius of a circle. If we denote by Re the reduction to the ecliptic, we shall have I - tu + ~ -R, v- +~ —Re. But we have v -= M the equation of the centre; hence I - M + - + equation of the centre - reduction to the ecliptic, and, putting L -1 + 7z = mean longitude, we get I L + equation of centre - reduction to ecliptic. (86) In the tables of the motion of the planets, the equation of the centre (53) is given in a table with M as the argument; and the reduction to the ecliptic is given in a table in which i and u are the arguments. 33. In determining the place of a heavenly body directly from the elements of its orbit, there will be no necessity for computing the reduction to the ecliptic, since the heliocentric longitude and latitude may be readily found by the formulae (82). When the heliocentric place has been found, we can easily deduce the corresponding geocentric place. Let x, y, z be the rectangular co-ordinates of the planet or comet referred to the centre of the sun, the plane of xy being in the ecliptic, 86 THEORETICAL ASTRONOMY. the positive axis of x being directed to the vernal equinox, and the positive axis of z to the north pole of the ecliptic. Then we shall have x-r cos b cos I, y r cos b sin 1, z - r sin b. Again, let X, Y, Zbe the co-ordinates of the centre of the sun referred to the centre of the earth, the plane of XY being in the ecliptic, and the axis of X being directed to the vernal equinox; and let 0 denote the geocentric longitude of the sun, R its distance from the earth, and Z its latitude. Then we shall have X-= R cos 2 cos 0, Y- R cos sin O, Z - R sin Z. Let x', y', zt be the co-ordinates of the body referred to the centre of the earth; and let a and 9 denote, respectively, the geocentric longitude and latitude, and J, the distance of the planet or comet from the earth. Then we obtain x'- cos ft cos X, y'- = cos f sin, (87) Z _ J sin f3. But, evidently, we also have x'=-x +X, y'=y + Y, z'=z + Z, and, consequently, A cos f1 cos r co cos b cos I -+c R cos cos 0, A cos 3 sin = r cos b sin I + R cos Z sin 0, (88) A sin f - r sin b + R sin 2. If we multiply the first of these equations by cos 0, and the second by sin 0, and add the products; then multiply the first by sin 0, and the second by cos 0, and subtract the first product from the second, we get cos t cos (A- 0) - r cos b cos ( - 0) + R cos 2, A cos ft sin ( - ) - r cos b sin (I - ), (89) A sin r sin b + R sin. It will be observed that this transformation is equivalent to the supposition that the axis of x, in each of the co-ordinate systems, is POSITION IN SPACE. 87 directed to a point whose longitude is 0, or that the system has been revolved about the axis of z to a new position for which the axis of abscissas makes the angle'( with that- of the primitive system. We may, therefore, in general, in order to effect such a transformation in systems of equations thus derived, simply diminish the longitudes by the given angle. The equations (89) will determine 2, /, and A when i, b, and I have been derived from the elements of the orbit, the quantities R, 0, and 2 being furnished by the solar tables; or, when A, /, and 2 are given, these equations determine 1, b, and r. The latitude 2 of the sun niever~ exceeds - 0'.9, and, therefore, it may in most cases be neglected, so that cos 2 1 and sin 2'-0, and the last equations become os cos s ( - 30) - r cos b cos ( - ) + R, cosS sin (A - )= r cos b sin (- 0), (90) A sill r sin b. If we suppose the axis of x to be directed to a point whose longitude is 2, or to the ascending node of the planet or comet, the equations (88) become J co Cs (- ) cosu + c eos ( ) c os + - 2 ), A cos i sill (A — ) =- r sin u cos i + R cos I sin (( - S ), (91) J sin /1 - r sin u sini + R sin 2, by means of which, and 2 may be found directly from 2, i, r, and u. If it be required to determine the geocentric right ascension and declination, denoted respectively by a and 8, we may convert the values of / and 2 into those of o( and 8. To effect this transformation, denoting by e the obliquity of the ecliptic, we have cos 8 cos a cos f cos A, cos S sin a -"y/cos f sin Ajcos — Sin sin e, sin c S cos n sin ) sin sin cos e. Let us now take r —-- n sin N - sin f, n cos N=- cos f sin, and we shall have cos 8 cos a COS / cos 2, cos 6 sin a = n cos (-N+ s), sin = - sin (N +- ). 88 THEORETICAL ASTRONOMY. Therefore, we obtain tan5 tac -cos(N — + t) Y t 1 an tan a s(N tan (92) sin A cos N tan s tan (N +- ) sin a. We also have cos (N + E) cos S sin a cos N cos i sin A' which will serve to check the calculation of a and 8. Since cos 8 and cos are always positive, cos a and cos must have the same sign, and thus the quadrant in which a is to be taken, is determined. For the solution of the inverse problem, in which o and 8 are given and the values of A and f are required, it is only necessary to interchange, in these equations, a and A, 8 and /, and to write - e in place of e. 34. Instead of pursuing the tedious process, when several places are required, of computing first the heliocentric place, then the geocentric place referred to the ecliptic, and, finally, the geocentric right ascension and declination, we may derive formula which, when certain constant auxiliaries have once been computed, enable us to derive the geocentric place directly, referred either to the ecliptic or to the equator. We will first consider the case in which the ecliptic is taken as the fundamental plane. Let us, therefore, resume the equations X - r cos u, y' -+ r sin u cos i, z' - r sin sin i, in which the axis of x is supposed to be directed to the ascending node of the orbit of the body. If we now pass to a new system x, y,, — the origin and the axis of z remaining the same,-in which the axis of x is directed to the vernal equinox, we shall move it back, in a negative direction, equal to the angle 2, and, consequently, = x' cos g - y' sin Q, y x' sin + -y' cos, Therefore, we obtain x r (cos u cos G + sin u cos i sin g ), y - r (+- sin u cos i cos + cos u sin ), (93) z -- r sin u sin i, POSITION IN SPACE. 89 which are the expressions for the heliocentric co-ordinates of a planet or comet referred to the ecliptic, the positive axis of x being directed to the vernal equinox. The upper sign is to be used when the motion is direct, and the lower sign when it is retrograde. Let us now put cos gQ -sin a sin A, -t cos i sin 2 - sin a cos A, sin a - sin b sin B, ~ cos i cos a - sin b cos B, in which sin a and sin b are positive, and the expressions for the coordinates become x r sin a sin (A + u), y r sin b sin (B + u), (95) z - r sin i sin u. The auxiliary quantities a, b, A, and B, it will be observed, are functions of 2 and i, and, in computing an e5la, i are constant so long as these elements are regarded as constant. They are called the constcats for the ecliptic. To determine them, we have, from equations (94), cot A- +- tan g, cos i, cot B =-~ cot g cos i, cos a sin s sin a in A sin b - sinB sinA smnB the upper sign being used when the motion is direct, and the lower sign when it is retrograde. The auxiliaries sin a and sin b are always positive, and, therefore, sin A and cos 2, sin B and sin 2, respectively, must have the same signs. The quadrants in which A and B are situated, are thus determined. From the equations (94) we easily find cos a - sini sin,, / cosb -- sin i cos g. (96) If we add to the heliocentric co-ordinates of the body the co-ordinates of the sun referred to the earth, for which the equations have already been given, we shall have x + X_= zd cos cos, y -- Y -- A cosj sin 2, (97) z + Z = J sin i, 90 THEORETICAL ASTRONOMY. which suffice to determine 2, &, and A. The values of a and 8 may be derived from these by means of the equations (92). 35. We shall now derive the formulae for determining a and 8 directly. For this purpose, let x, y, z be the heliocentric co-ordinates of the body referred to the equator, the positive axis of x being directed to the vernal equinox. To pass from the system of coordinates referred to the ecliptic to those referred to the equator as the fundamental plane, we must revolve the system negatively around the axis of x, so that the axes of z and y in the new system make the angle e with those of the primitive system, e being the obliquity of the ecliptic. In this case, we have x" x y" = y cos e - z sin e, " = y sin +- z cos e. Substituting for x, y, and z their values from equations (93), and omitting the accents, we get x - r cosu cos =F r sin u cos i sin., y r cossin cos r sin u (= cos i cos c cos+s - sin i sin e), (98) z =r cosu sin Q sin s+ r sinu (=t co i cos 2 sin -- sin i cos e). These are the expressions for the heliocentric co-ordinates of the planet or comet referred to the equator. To reduce them to a convenient form for numerical calculation, let us put cos a - sin a sin A, == cos i sin a = sin a cos A, sin a cos = sin b sin B,. (99) ~ cos i cos C cos e - sin i sin e - sin b cos B, sin a sin -- sin c sin C, - cos i cos a sin -- sin i cos e _ sin c cos C; and the expressions for the co-ordinates reduce to x - r sin a sin (A + u), y = r sin b sin (B + u), (100) z = r sin c sin (C + u). The auxiliary quantities, a, b, c, A, B, and C, are constant so long as a2 and i remain unchanged, and are called constants for the equator. It will be observed that the equations involving a and A, regarding the motion as direct, correspond to the relations between the parts of a quadrantal triangle of which the sides are i and a, the POSITION IN SPACE. 91 angle included between these sides being that which we designate by A, and the angle opposite the side a being 90~ - 2. In the case of b and B, the relations are those of the parts of a spherical triangle of which the sides are b, i, and 90~ -+ E B being the angle included by i and b, and 180~ - 2 the angle opposite the side b. Further, in the case of c and C, the relations are those of the parts of a spherical triangle of which the sides are c, i, and e, the angle C being that included by the sides i and c, and 180~ - 2 that included by the sides i and s. We have, therefore, the following additional equations: cos a - sin i sin R, cos b -- cos g sin i cos e - cos i sin e, (101) cos c - - cos a sin i sin +- cos i cos e. In the case of retrograde motion, we must substitute in these 180~ - i in place of i. The geometrical signification of the auxiliary constants for the equator is thus made apparent. The angles a, b, and c are those which a line drawn from the origin of co-ordinates perpendicular to the plane of the orbit on the north side, makes with the positive coordinate axes, respectively; and A, B, and C are the angles which the three planes, passing through this line and the co-ordinate axes, make with a plane passing through this line and perpendicular to the line of nodes. In order to facilitate the computation of the constants for the equator, let us introduce another auxiliary quantity E0, such that sin i - eo sin E,, ~cos i cos eo cos Eo, eo being always positive. We shall, therefore, have tan i tanE0 tan i cos Since both eo and sini are positive, the angle Eo cannot exceed 180~; and the algebraic sign of tan E0 will show whether this angle is to be taken in the first or second quadrant. The first two of equations (99) give cot A = - tan Q cos i; and the first gives cos g sin a -- -. sin A 92 THEORETICAL ASTRONOMY. From the fourth of equations (99), introducing e0 and E0, we get sin b cos B = eO cos Eo cos e - e sin Eo sine - e0 cos (Eo +- e). But sin b sin B = sin Q cos e; therefore cots eo cos (E,0+e) ^ 4 cos i cos (Eo + ) sin g cos e tan a cos Eo cose We have, also, sin QS cose sin b sinB In a similar manner, we find cot C= cos i sin (Eo + e) tan a cos Eo sin e and sin a sin e sin c -- sin C The auxiliaries sin a, sin b, and sin c are always positive, and, therefore, sill A and cos 2, sin B and sin g, and also sin C and'sin g., must have the same signs, which will determine the quadrant in which each of the angles A, B, and C is situated. If we multiply the last of equations (99) by the third, and the fifth of these equations by the fourth, and subtract the first product from the last, we get, by reduction, sin b sin c sin (C- B) =- sin i sin g~. But sin a cos A - + cos i sin g; and hence we derive sin b sine sin (C- B) --- +- tan i, sin a cos A which serves to check the accuracy of the numerical computation of the constants, since the value of tani obtained from this formula must agree exactly with that used in the calculation of the values of these constants. If we put A' = A - 7r, B' = B, and' = C_- f g2, the upper or lower sign being used according as the motion is direct or retrograde, we shall have POSITION IN SPACE. 93 - r sil a sin (A' + v), y - sin b sin (B' + v), (102) z r sin sin ( C' + v), a transformation which is perhaps unnecessary, but which is convenient when a series of places is to be computed. It will be observed that the formule for computing the constants a, b, c, A, B, and C, in the case of direct motion, are converted into those for the case in which the distinction of retrograde motion is adopted, by simply using 180~ - i instead of i. 36. When the heliocentric co-ordinates of the body have been found, referred to the equator as the fundamental plane, if we add to these the geocentric co-ordinates of the sun referred to the same fundamental plane, the sum will be the geocentric co-ordinates of the body referred also to the equator. For the co-ordinates of the sun referred to the centre of the earth, we have, neglecting the latitude of the sun, X R cos Q, Y — R sin ( cos e, Z, — R sin O sin z -- Y tan e, in which R represents the radius-vector of the earth, 0 the sun's longitude, and s the obliquity of the ecliptic. We shall, therefore, have x + X — J cos cos a, y+ y J cos sin, (103) z + Z = z sin S, which suffice to determine a, A, and J. If we have regard to the latitude of the sun in computing its geocentric co-ordinates, the formulhe will evidently become X- R cos O cos C, Y=- R sin 0 cos Z cos s -R sin 2 sin, (104) Z- R sin 0 cos 2 sin s -- R sin 2 cos e, in which, since Z can'niever exceeds - 0,.9, cos 2 is very nearly equal to 1, and sin 2 = 2. The longitudes and latitudes of the sun may be derived from a solar ephemeris, or from the solar tables. The principal astronomical ephemerides, such as the Berliner Astronomisches Jahrbuch, the Nautical Almanac, and the American Ephemeris and Nautical Al 94 THEORETICAL ASTRONOMY. manac, contain, for each year for which they are published, the equatorial co-ordinates of the sun, referred both to the mean equinox and equator of the beginning of the year, and to the apparent equinox of the date, taking into account the latitude of the sun. 37. In the case of an elliptic orbit, we may determine the coordinates directly from the eccentric anomaly in the following manner:The equations (102) give, accenting the letters a, b, and c, x - r cos v sin a' sin A' + r sin v sin a' cos A', y = cos v sin b' sin B' - r sin v sih b' cos B', z - r cos v sin c' sin C' + r sin v sin c' cos C'. Now, since r cos v=- a cos E- ae, and r sin v = a cos To sinE, we shall have x = a sin at sin A' cos E - ae sin a' sin A' + a cos p sin a' cos A' sin E, y a= c sin b' sin B' cos E -.ae sin b' sin B' + a cos sin b cos B' sin E, z = a sin C' sin C' cos E- ae sin c' sin C' + a cos ( sin c' cos C' sin E. Let us now put a cos c sin a' cos A' — A cos Lx, a sin a' sin sn L,, - ae sin a' sin A' - e2R sin Lx - v.; a cos no sin b' cos B' = cos Ly, a sin b sin B' - A sin Ly, -- ae sin b' sin B' - e- e sin Ly = vy; a cos ~ sin c' cos C' = A, cos Lz, a sin c' sin C' 2= sin L,, - ae sin c' sin C' - - e2z sin L. -- vY; in which sin a', sin b', and sin c' have the same values as in equations (102), the accents being added simply to mark the necessary distinction in the notation employed in these formulae. We shall, therefore, have x - Rx sin (Lx + E) + v, y = A, sin (+y +- E) + vy, (105) z - A sin (Lz + E) + vz. By means of these formulae, the co-ordinates are found directly from the eccentric anomaly, when the constants 2,, 2y, Z, Lx, Ly, L, vx, iy, and Pz, have been computed from those already found, or from a, 6, c, A, B, and C. This method is very convenient when a great POSITION IN SPACE. 95 number of geocentric places are to be computed; but, when only a few places are required, the additional labor of computing so many auxiliary quantities will not be compensated by the facility afforded in the numerical calculation, when these constants have been determined. Further, when the ephemeris is intended for the comparison of a series of observations in order to determine the corrections to be applied to the elements by means of the differential formulae which we shall investigate in the following chapter, it will always be advisable to compute the co-ordinates by means of the radius-vector and true anomaly, since both of these quantities will be required in finding the differential coefficients. 38. In the case of a hyperbolic orbit, the co-ordinates may be computed directly from F, since we have r cos v - a (e - sec F), r sin v _ a tan 4 tan F; and, consequently, x = ae sin a' sin A' - a sec F sin a' sin A' +- a tan 4 tan F sin a' cos A', y = ae sin b' sin B' - a sec F Y s b' sin B' + a tan 4 tan F sin b' cos B', z = ae sin c' sin C' - a sec F sin c' sin C' + a tan ~ tan F sin c' cos C'. Let us now put ae sin a' sin A' -A, - a sin a' sin A' p-, a tan 9 sin c' cos A' - vx; ae sin b' sin B' = y, - a sil b' sin B' = y, a tan 4 sin b' cos B' =v y; ae sin g sin C' - A, - a sin c' sin C' = -Z, a tan 4 sin c' cos C' --. Then we shall have x -- -+ l/, sec F + x tan F, y = Ay + -y sec F + v, tan F, (106) z -= A + z, sec F + Yz tan F. In a similar manner we may derive expressions for the co-ordinates, in the case of a hyperbolic orbit, when the auxiliary quantity a is used instead of F. 39. If we denote by,', 2', and i' the elements which determine the position of the orbit in space when referred to the equator as the 96 THEORETICAL ASTRONOMY. fundamental plane, and by to0 the angular distance between the ascending node of the orbit on the ecliptic and its ascending node on the equator, being measured positively from the equator in the direction of the motion, we shall have' U= — +-' -+ -Oo To find at and il, we have, from the spherical triangle formed by the intersection of the planes of the orbit, ecliptic, and equator with the celestial vault, cos i' cos i cos E - sin i sin e cos S, sin i' sin Q' = sin i sin g, a.-& c.9.. sin i' cos a' o s s i sin +- sin i cos e cos.'i Let us now put n sin N- cos i; n cos N — sin i cos 2,and these equations reduce to cos i' -- n sin (N - e),. sin i sin' =sin sin Q,'sin i' cos' = n cos (N- ); - from which we find tan N- cot i, cos N tan N co -^, tan os (V- tan, cos Q, cos (N - e) (107) coti' tan (N- e) cos g2'. (107) Since sin i is always positive, cos N and cos a must have the same signs. To prove the numerical calculation, we have sin i cos g cos N sin i' cos 2' cos (N )' the value of the second member of which must agree with that used in computing g'. In order to find wo, we have, from the same triangle, sin w0 sin i' = sin S sin e, Ho cos w sin i' = cos s sin i + sin e cos i cos g. Let us now take - m sin M= cos e, m cos M=- sin e cos S;; and we obtain POSITION IN SPACE. 97 cot M - tan s cos R, cos IM tan % (3G M tan, (108) 0 Cos (M i) and, also, to check the calculation, sin e cos cos M sin i' cosO — o cos (M- i) If we apply Gauss's analogies to the same spherical triangle, we get cos-.i' sin (g' + %W) sin cos. (i -), cos o) cos cos (' + ) _ cos cos + (109) i' sni' sin s (n' ~ ) - sin.} sin (i - ), sin ci' cos (~ Loo) = cos sin ( + ). The quadrant in which 2 ( + wo) or ( - o0) is situated, must be so taken that sin 2i' and cos -i' shall be positive; and the agreement of the values of the latter two quantities, computed by means of the value of li' derived from tan s', will serve to check the accuracy of the numerical calculation. For the case in which the motion is regarded as retrograde, we must use 180 — i instead of i in these equations, and we have, also,,o - =T -- + - to We may thus find the elements n', a i, and i', in reference to the equator, from the elements referred to the ecliptic; and using the elements so found instead of r, 2, and i, and using also the places of the sun referred to the equator, we may derive the heliocentric and geocentric places with respect to the equator by means of the formula already given for the ecliptic as the fundamental plane. If the position of the orbit with respect to the equator is given, and its position in reference to the ecliptic is required, it is only necessary to interchange Q and S', as well as i and 180~- i', e remaining unchanged, in these equations. These formulae may also be used to determine the position of the orbit in reference to any plane in space; but the longitude Q must then' be measured from the place of the descending node of this plane on the' ecliptic. The value of g,, therefore, which must be used in.the solution of the equations is, in this case, equal to the longitude of the ascending node of the orbit on the ecliptic diminished by the longitude of the descending node of the new plane of reference on the ecliptic. The quantities P', i' and wo will have the same signification in reference 7 98 THEORETICAL ASTRONOMY. to this plane that they have in reference to the equator, with this distinction, however, that a' is measured from the descending node of this new plane of reference on the ecliptic; and e will in this case denote the inclination of the ecliptic to this plane. 40. We have now derived all the formulae which can be required in the case of undisturbed motion, for the computation of the heliocentric or geocentric place of a heavenly body, referred either to the ecliptic or equator, or to any other known plane, when the elements of its orbit are known; and the formulae which have been derived are applicable to every variety of conic section, thus including all possible forms of undisturbed orbits consistent Avith the law of universal gravitation. The circle is an ellipse of which the eccentricity is zero, and, consequently, M —v - u, and - - a, for every point of the orbit. There is no instance of a circular orbit yet known; but in the case of the discovery of the asteroid planets between Mars and Jupiter it is sometimes thought advisable, in order to facilitate the identification of comparison stars for a few days succeeding the discovery, to compute circular elements, and from these an ephemeris. The elements which determine the form of the orbit remain constant so long as the system of elements is regarded as unchanged; but those which determine the position of the orbit in space, T, g, and i, vary from one epoch to another on account of the change of the relative position of the planes to which they are referred. Thus the inclination of the orbit will vary slowly, on account of the change of the position of the ecliptic in space, arising from the perturbations of the earth by the other planets; while the longitude of the perihelion and the longitude of the ascending node will vary, both on account of this change of the position of the plane of the ecliptic, and also on account of precession and nutation. If Tc, g, and i are referred to the true equinox and ecliptic of any date, the resulting heliocentric places will be referred to the same equinox and ecliptic; and, further, in the computation of the geocentric places, the longitudes of the sun must be referred to the same equinox, so that the resulting geocentric longitudes or right ascensions will also be referred to that equinox. It will appear, therefore, that, on account of these changes in the values of wT, 2, and i, the auxiliaries sina, sinb, sin c, A, B, and C, introduced into the formula for the coordinates, will. not be constants in the computation of the places for a series of dates, unless the elements are referred constantly, in the calculation, to a fixed equinox and ecliptic. It is customary, there POSITION IN SPACE. 99 fore, to reduce the elements to the ecliptic and mean equinox of the beginning of the year for which the ephemeris is required, and then to compute the places of the planet or comet referred to this equinox, using, in the case of the right ascension and declination, the mean obliquity of the ecliptic for the date of the fixed equinox adopted, in the computation of the auxiliary constants and of the co-ordinates of the sun. The places thus found may be reduced to the true equinox of the date by the well-known formule for precession and nutation. Thus, for the reduction of the right ascension and declination from the mean equinox and equator of the beginning of the year to the apparent or-true equinox and equator of any date, usually the date to which the co-ordinates of the body belong, we have a- =f+ g sin(G+ a)tan (, A= g cos (G + a), for which the quantitiesf, g, and G are derived from the data given either in the solar and lunar tables, or in astronomical ephemerides, such as have already been mentioned. The problem of reducing the elements from the ecliptic of one date t to that of another date t' may be solved by means of equations (109), making, however, the necessary distinction in regard to the point from which 2 and I2' are measured. Let 0 denote the longitude of the descending node of the ecliptic oft' on that of t, and let a denote the angle which the planes of the two ecliptics make with each other, then, in the equations (109), instead of ag we must write a - 0, and, in order that R' shall be measured from the vernal equinox, we must also write Q' - in place of 2'. Finally, we must write q instead of s, and ace for ow0 which is the variation in the value of co in the interval t' - t on account of the change of the position of the ecliptic; then the equations become cos i' sin (g' + A- ) - sin 1 (g - ) cos (i - ), Cos i' cos ~ (a' - Q + A@) = cos ( S - ) cos ~ ( +, sin -}i' sin 1(. Bo) A sin -1 ( 0) sin ] (i ), (111) sin,i' cos (a'-O- A c) = cos (2 -- O) sin. (i + ). These equations enable us to determine accurately the values of g', i', and Aco, which give the position of the orbit in reference to the ecliptic corresponding to the time t', when 0 and' are known. The longitudes, however, will still be referred to the same mean equinox as before, which we suppose to be that of t; and, in order to refer 100 THEORETICAL ASTRONOMY. them to the mean equinox of the epoch t', the amount of the precession in longitude during the interval' - t must also be applied. If the changes in the values of the elements are not of considerable magnitude, it will be unnecessary to apply these rigorous formulae, and we may derive others sufficiently exact, and much more convenient in application. Thus, from the spherical triangle formed by the intersection of the plane of the orbit and of the planes of the two ecliptics with the celestial vault, we get sin'C cos ( - 0) - cos i' sin i + sin i' cos i cos aw, from which we easily derive sin (' -i) = sin - cos ( 6 0) -+ 2 sin i cos i sin2 GAi. (112) We have, further, sin ao sin i' = sin 7 sin ( - ), or sin (g ) -10) sin Aw =sill sin i' (113) We have, also, from the same triangle, sin a^ cos i' -cos ( - ) sin (' -0) + sin (2 - ) cos (g' -a ) cos 7, which gives sin (2' - 2) - sin Aow cosi' - 2 sin (g2 - ) cos (' - a) sin2., or sin (~'- )=i n n sin (- ) cot i' - 2 sin ( -) cos (i'- ) sin2 -7. (114) Finally, we have - C - 2'- gr - + Ac. Since C is very small, these equations give, if we apply also the precession in longitude so as to reduce the longitudes to the mean equinox of the date', sin (2 -0) sin i i' =- i,- n+ cos ( a -- 0) +- 4 sin 2 i,'- +( —) 1 sin( — )coti' —sin2(g-), (115) - r + sin(-tan - s2( -; dl 2 = V_+0(t'-t) +7) sin(P2 - 6) tani i —O sin2(gb -6); dt 8 POSITION IN SPACE. 101 in which d is the annual precession in longitude, and in which s 206264".8. In most cases, the last terms of the expressions for i', 2', and w', being of the second order, may be neglected. For the case in which the motion is regarded as retrograde, we must put 1800 - i and 180 - i', instead of i and i', respectively, in the equations for acw i', and Q'; and for 7t', in this case, we have i' A ~ ^ = f' a ^ - ac, which gives a: -;. dl V" =r -+- (t' -t) d sin (g -0) tan i' —- sin 2(g -). dt' If we adopt Bessel's determination of the luni-solar precession and of the variation of the mean obliquity of the ecliptic, we have, at the time 1750 + v, - 50".21129 +- 0."0002442966r, dt - 0".48892- 0.000006143T, dt and, consequently, - = (0."48892 - 0."000006143r) ( - t); and in the computation of the values of these quantities we must put = (t + t) - 1750, t and t' being expressed in years., The longitude of the descending node of the ecliptic of the time t on the ecliptic of 1750.0 is also found to be 351~ 36' 10" - 5".21 (t - 1750), which is measured from the mean equinox of the beginning of the year 1750. The longitude of the descending node of the ecliptic of t' on that of t, measured from the same mean equinox, is equal to this value diminished by the angular distance between the descending node of the ecliptic of t on that of 1750 and the descending node of the ecliptic of t' on that of t, which distance is, neglecting terms of the second order, 5".21 (' - 1750); and the result is 351~ 36' 10" - 5".21 (t- 1750) - 5".21 (t' - 1750), or 351~ 36' 10" 10".42 (t - 1750) - 5".21 ('- t). 102. THEORETICAL ASTRONOMY. To reduce this longitude to the mean equinox at the time t, we must add the general precession during the interval t -1750, or 50".21 (t - 1750), so that we have, finally, 0 - 351~ 36' 10" + 39".79 (t - 1750) -- 5".21 (t' - t). When the elements w, Q, and i have been thus reduced from the ecliptic and mean equinox to which they are referred, to those of the date for which the heliocentric or geocentric place is required, they may be referred to the apparent equinox of the date by applying the..,nutation in longitude. Then, in the case of the determination of the right ascension and declination, using the apparent obliquity of the ecliptic in the computation of the co-ordinates, we directly obtain the place of the body referred to the apparent equinox. But, in computing a series of places, the changes which thus take place in the elements themselves from date to date'induce corresponding changes in the auxiliary quantities a, b, c, A, B, and C, so that these are no longer to be considered as constants, but as continually changing their values by small differences. The differential formule for the computation of these changes, which are easily derived from the equations (99), will be given in the next chapter; but they are perhaps unnecessary, since it is generally most convenient, in the cases which occur, to compute the auxiliaries for the extreme dates for which the ephemeris is required, and to interpolate their values for intermediate dates. It is advisable, however, to reduce the elements to the ecliptic and mean equinox of the beginning of the year for which the ephemeris is required, and using the mean obliquity of the ecliptic for that epoch, in the computation of the auxiliary constants for the equator, the resulting geocentric right ascensions and declinations will be referred to the same equinox, and they may then be reduced to the apparent equinox of the date by applying the corrections for precession and nutation. The places which thus result are free from parallax and aberration. In comparing observations with an ephemeris, the correction for parallax is applied directly to the observed apparent places, since this correction varies for different places on the earth's surface. The correction for aberration may be applied in two different modes. We may subtract from the time of observation the time in which the light from the planet or comet reaches the earth, and the true place for this reduced time is identical with the apparent place for the time NUMERICAL EXAMPLES. 103 of observation; or, in case we know the daily or hourly motion of the body in right ascension and declination, we may compute the motion during the interval which is required for the light to pass from the body to the earth, which, being applied to the observed place, gives the true place for the time of observation. We may also incl1de -th, aberration directly in the ephemeris by using the time t — ~497"-4: in computing the geocentric places for the time t, or by subtracting from the place free from aberration, computed for the time t, the motion in c and 8 during the interval 497.784, in which expression d is the distance of the body from the earth, and 497.78 the number of seconds in which light traverses the mean distance of the earth from the sun. It is customary, however, to compute the ephemeris free from aberration and to subtract the time of aberration, 497".78J, from the time of observation when comparing observations with an ephemeris, according to the first method above mentioned. The places of the sun used in computing its co-ordinates must also be free from aberration; and if the longitudes derived from the solar tables include aberration, the proper correction must be applied, in order to obtain the true longitude required. 41. EXAMPLES.-We will now collect together, in the proper order for numerical calculation, some of the principal formulae which have been derived, and illustrate them by numerical examples, commencing with the case of an elliptic orbit. Let it be required to find the geocentric right ascension and declination of the planet Eurynome @, for mean midnight at Washington, for the date 1865 February 24, the elements of the orbit being as follows:Epoch - 1864 Jan. 1.0 Greenwich mean time. M= 1~ 29' 40".21 - 44 20 33.09 Ecliptic and Mean _- - 206 42 40.13 2-206 42 40.13 Equinox, 1864.0. i- 4 36 50.51) -=- 11 15 51.02 log a 0.3881319 log — = 2.9678088 /i. 928".55745 When a series of places is to be computed, the first thing to be done is to compute the auxiliary constants used in the expressions for the co-ordinates, and although but a single pl:ace is required in the problem proposed, yet we will proceed in this manner, in order to 104 THEORETICAL ASTRONOMY. exhibit the application of the formulse. Since the elements a, Q, and i are referred to the ecliptic and mean equinox of 1864.0, we will first reduce them to the ecliptic and mean equinox of 1865.0. For this reduction we have t 1864.0, and t'= 1865.0, which give l= 50".239, 0- 352~ 51' 41", r 0".4882. dt - Substituting these values in the equations (115), we obtain i'- i- Ai 0".40, A -- + 53".61, As +- 50".23; and hence the elements which determine the position of the orbit in reference to the ecliptic of 1865.0 are - 44~ 21' 23".32, g -206~ 43' 33".74, i= 4~ 36' 50".11. For the same instant we derive, from the American Ephemeris and Nanltical Almanac, the value of the mean obliquity of the ecliptic, which is e 23~ 27' 24".03. The auxiliary constants for the equator are then found by means of the formulae tan i cot A -tang cos i, tan Eo - cos i o','!' cos i cos (o,+ E) cot B - tan a cos Eo cos cos i sin (E + e) cot C * tan Q cos E0 sine cos a sin C cos. sin a sine sin a sin b,i sin b - s in - sin A',inB' sin C The angle Eo is always less than 180~, and the quadrant in which it is to be taken, is indicated directly by the algebraic sign of tan E0. The values of sin a, sin b, and sin c are always positive, and, therefore, the angles A, B, and C must be so taken, with respect to the quadrant in which each is situated, that sin A and cos 2, sin B and sin 2, and also sin C and sin 2, shall have the same signs. From these we derive A - 296~ 39' 5".07, log sin a = 9.9997156, B =205 55 27.14, log sin b = 9.9748254, C= 212 32 17.74, log sin c 9.5222192. Finally, the calculation of these constants is proved by means of the formula NUMERICAL EXAMPLES. 105 sin b sin c sin ( C- B) tan i sin a cos A which gives log tan i 8.9068875, agreeing with the value 8.9068876 derived directly from i. Next, to find r and m.' The date 1865 February 24.5 mean time at Washington reduced to the meridian of Greenwich by applying the difference of longitude, 5h 8"" 118.2, becomes 1865 February 24.714018 mean time at Greenwich. The interval, therefore, from the epoch for which the mean anomaly is given and the date for which the geocentric place is required, is 420.714018 days; and multiplying the mean daily motion, 928".55745, by this number, and adding the result to the given value of VM, we get the mean anomaly for the required place, or M/=- 1~ 29' 40".21 + 108~ 30' 57".14 1100 0' 37".35. The eccentric anomaly E is then computed by means of the equation M E- e sin E, the value of e being expressed in seconds of arc. For Eurynome we have log sin p=- log e= 9.2907754, and hence the value of e expressed in seconds is log e = 4.6052005. By means of the equation (54) we derive an approximate value of E, namely, E, =119~ 49' 24", the value of e expressed in seconds being loge2=- 3.895976; and with this we get C.? C......i..- j:.: i M, — E e sin E, 110~ 6' 50". Then we have AE M- I- M 372".7 aE - cosE - -— 7- =l - 339~ 7, 1 -e osE0) 1.097 which gives, for a second approximation to the value of E, O 119~ 43' 44".3. This gives M3 = 1100 0' 36".98, and hence 0".37 AE- 1 - 197 +0".34. 1.097 106 THEORETICA.I ASTRONOMY. Therefore, we have, for a third approximation to the value of E, E- 119~ 43' 44".64, which requires no further correction, since it satisfies the equation between 3 and E. I - To find r and v, we have V r e)sin e) s, Vr cos, ~v 1/a(1 - e) cos ME. The values of the first factors in the second members of these equations are: log Va(l- + e) 0.2328104, and log /a(1 -e) 0.1468741; and we obtain v 129~ 3' 50".52, logr = 0.4282854. Since z - 197~ 37' 49/.58, we have u = v + - X a 326~ 41' 40".10. The heliocentric co-ordinates in reference to the equator as the fundamental plane are then derived from the equations x r sin a sin (A -+ t), y r sin b sin (B + u), z =r sin c sin(C + u), which give, for Eurynome, x — 2.6611270, y - +0.3250277, z = 0.0119486. The Americacn Nautical Almanac gives, for the equatorial co-ordinates of the sun for 1865 February 24.5 mean time at Washington, referred to the mean equinox and equator of the beginning of the year, X- + 0.9094557, Y=- 0.3599298, Z - 0.1561751. Finally, the geocentric right ascension, declination, and distance are given by the equations Y Y z Z+ z z -Z Z z- Z tan a Y ~ —, tan s -= sin a = - cos a, s = a x- + X' y+Y X sin 6' the first form of the equation for tan 3 being used when sin a is greater than cos a. The value of J must always be positive; and 8 cannot exceed -- 90~, the minus sign indicating south declination. Thus; we obtain NUMERICAL EXAMPLES. 107 -- 181~ 8' 29".29, a - 4~ 42' 21".56, log A = 0.2450054. To reduce a and 8 to the true equinox and equator of February 24.5, we have, from the Nautical Alcmnac, f= + 16".80, log g =1.0168, G =45~ 16'; and, substituting these values in equations (110), the result is Aa — + 17".42, -- 7".17. Hence the geocentric place, referred to the true equinox and equator of the date, is a 181~ 8' 46".71, 4~ 42' 28".73, log = 0.2450054. WVhen only a single place is required, it is a little more expeditious to compute r from r - a(1 - e cosE), and then v - E from sin - (v - E) - si sin si E. Thus, in the case of the required place of Eurynome, we get log r = 0.4282852, v - E — 9 20' 5".92, v - 129~ 3' 50".56, agreeing with the values previously determined. The calculation may be proved by means of the formula sin - (v + E) - \- cos (P sin E. In the case of the values just found, we have 1(v + E) - 124~ 23' 47".60, log sin (v + E)- 9.9165316, while the second member of this equation gives log sin - (v + E) -- 9.9165316. In the calculation of a single place, it is also very little shorter to compute first the heliocentric longitude and latitude by means of the equations (82), then the geocentric latitude and longitude by means of (89) or (90), and finally convert these into right ascension and declination by means of (92). When a large number of places are to be computed, it is often advantageous to compute the heliocentric 108 THEORETICAL ASTRONOMY. co-ordinates directly from the eccentric anomaly by means of the equations (105). The calculation of the geocentric place in reference to the ecliptic is, in all respects, similar to that in which the equator is taken as the fundamental plane, and does not require any further illustration. The determination of the geocentric or heliocentric place in the cases of parabolic and hyperbolic motion differs from the process indicated in the preceding example only in the calculation ofr and v. To illustrate the case of parabolic motion, let t - T 75.364 days; log q - 9.9650486; and let it be required to find r and v. First, we compute m from m -- in which log Co- 9.9601277, and the result is log m = 0.0125548. Then we find M from _-/ m (t T), which gives log M- 1.8897187. From this value of log M1I we derive, by means of Table VI., v - 79~ 55' 57".26. Finally, r is found from r — V Cos2 -eV which gives logr = 0.1961120. For the case of hyperbolic motion, let there be given t- T== 65.41236 days; = 370 35' 0".0, or log e 0.1010188; and log o - 0.6020600, to find r and v. First, we compute N from N — (- T), a2 in which log - 9.6377843, and we obtain log - 8.7859356; N= 0.06108514. The value of F must now be found from the equation N eA tan 4F- log tan (45~ + - F). NUMERICAL EXAMPLES. 109 If we assume F== 30~, a more approximate value may be derived from N+- log tan 60~ tanF,which gives F,= 280 40' 23", and hence N, 0.072678. Then we compute the correction to be applied to this value of F, by means of the equation AF -- (N — N,) cos2, ((e - cosF) 8, wherein s - 206264".8; and the result is AF,- 4.6097 (N- N,)s - 3 3' 43".0. Hence, for a second approximation to the value of F, we have F, = 25~ 36' 40".0. The corresponding value of Nis N,= 0.0617653, and hence AF, - 5.199 (N-,) s - 12' 9".4. The third approximation, therefore, gives F, -25~ 24' 30".6, and, repeating the operation, we get F — 25 24' 27".74. which requires no further correction. To find r, we have cos F ) which gives log r 0.2008544. Then, v is derived from tan vv - cot 1 tan F, and we find v 67~ 3' 0".0. When several places are req.uired, it is convenient to compute v and r by means of the equations 1/r sin sin- siF, V/cos F cr cos -IV ~ Cos F F. /cos F 110 THEORETICAL ASTRONOMY. For the given values of a and e we have log V/a(e + 1) = 0.4782649, log (/a(e - 1)- 0.0100829, and hence we derive v = 67~ 2' 59".92, log r = 0.2008545. It remains yet to illustrate the calculation of v and r for elliptic and hyperbolic orbits in which the eccentricity differs but little from unity. First, in the case of elliptic motion, let t- T= 68.25 days; e = 0.9675212; and log q- 9.7668134. We compute I from 2^ 2 M-= (t-T) O wherein log Co= 9.9601277, which gives log M 2.1404550. With this as argument we get, from Table VI., V= 101~ 38' 3".74, and then with this value of V as argument we find, from Table IX., A - 1540".08, B 9".506, C- 0".062. Then we have log i log 1 - 8.217680, and from the equation v - V-+ A (100i) + B (100i)2 + C(100i)3, we get v- V+ 42' 22".28 + 25".90 + 0".28 = 102~ 20' 52".20. The value of r is then found from q( +e) 1 + e cos v' namely, log r 0.1614051. We may also determine r and v by means of Table X. Thus, we first compute M from C (t - T). (1 + 9e) a'2 B Assuming B = 1, we get log M- 2.13757, and, entering Table VI. with this as argument, we find w -101~ 25'. Then we compute A from A 5(1 - te)an 1 + 9e NUMERICAL EXAMPLES. 111 which gives A 0.024985. With this value of A as argument, we find, from Table X., log B - 0.0000047. The exact value of M is then found to be log M= 2.1375635, which, by means of Table VI., gives w -101~ 24' 36".26. By means of this we derive A - 0.02497944, and hence, from Table X., log C= 0.0043771. Then we have tan 1v — C tan -w ( which gives v - 102~ 20' 52".20, agreeing exactly with the value already found. Finally, r is given by r C2= OIq (1- A G) cos2v' from which we get log r -0.1614052. Before the time of perihelion passage, t- T is negative; but the value of v is computed as if this were positive, and is then considered as negative. In the case of hyperbolic motion, i is negative, and, with this distinction, the process when Table IX. is used is precisely the same as for elliptic motion; but when table X. is used, the value of A must be found from 5 (e - 1) A = tau'1w (1 + 9e) and that of r from rr~q (1-A C2) cos2 v' the values of log B and log C being taken from the columns of the table which belong to hyperbolic motion. In the calculation of the position of a comet in space, if the motion 112 THEORETICAL ASTRONOMY. is retrograde and the inclination is regarded as less than 90~, the distinctions indicated in the formulae must be carefully noted. 42. When we have thus computed the places of a planet or comet for a series of dates equidistant, we may readily interpolate the places for intermediate dates by the usual formulae for interpolation. The interval between the dates for which the direct computation is made should also be small enough to permit us to neglect the effect of the fourth differences in the process of interpolation. This, however, is not absolutely necessary, provided that a very extended series of places is to be computed, so that the higher orders of differences may be taken into account. To find a convenient formula for this interpolation, let us denote any date, or argument of the function, by a + nqco and the corresponding value of the co-ordinate, or of the function, for which the interpolation is to be made, by f (a + no)). If we have computed the values of the function for the dates, or arguments, a -, a, a + w, a + 2-, &c., we may assume that an expression for the function which exactly satisfies these values will also give the exact values corresponding to any intermediate value of the argument. If we regard n as variable, we may expand the function into the series f(a + nw) =f(a) + An + Bn + C3 + &c. (116) and if we regard the fourth differences as vanishing, it is only necessary to consider terms involving n3 in the determination of the unknown coefficients A, B, and C. If we put n successively equal to -, 0, 1, and 2, and then take the successive differences of these values, we get I. Diff. II. Diff. III. Diff. f (a-) =f(a)-A -+B CA-B+ f(ao) =f(a) A -B+ 2B f(a+w) f(a)+A +B +C A+B+C 2B+6 f(a - 2) ==f (a) 2A + 4B + A8 C + + If we symbolize, generally, the difference f (a + no) (-f + (n- 1) w) by f' (a +- (n — ) w), the difference f (a + (n + ) w) -f' (a + (n — 2) w) by f" (a +- no), and similarly for the successive orders of differences, these may be arranged as follows: Argument. Function. I. Diff. II. Diff. III. Diff. a - f (a - o) a f(a)' (a -- 1) "f (a) a+u f+u) f ('+(a + ) f( ) f"'(a+iW) a +2 w f(a + 2)' (( + o) ) a + 2,, f/(a + 2o)' (f + i') INTERPOLATION. 113 Comparing these expressions for the differences with the above, we get C-f"' ( + 1), B - "(a), A.=f (a + -) f (a) - (a) - (a + s), which, from the manner in which the differences are formed, give C- (a" (a + ) -f" (a)), B = f" (a), A =f(a + w) -f (a) - f" (a) - (f" ( + w) -f (a)). To find the value of the function corresponding to the argument a -- -W, we have n =, and, from (116), f(a + J) =/(a) + -A + -B + I C. Substituting in this the values of A, B, and C, last found, and reducing, we get f( + - o) - (fa + a) +f (a)) - ( (f" (a + w) +- (a))), in which only fourth differences are neglected, and, since the place of the argument for n 0 is arbitrary, we have, therefore, generally, f( + (n + 1) ) -= (f(a + (n + 1) ) +f(a + nw)) ( (f" (a + (n + 1) w) -f" (a + nw))). (117) Hence, to interpolate the value of the function corresponding to a date midway between two dates, or values of the argument, for which the values are known, we take the arithmetical mean of these two known values, and from this we subtract one-eighth of the arithmetical mean of the second differences which are found on the same horizontal line as the two given values of the function. By extending the analytical process here indicated so as to include the fourth and fifth differences, the additional term to be added to equation (117) is found to be + 3 (f iv (a + (A + 1) a+) +f'i (a + nw))), and the correction corresponding to this being applied, only sixth differences will be neglected. It is customary in the case of the comets which do not move too rapidly, to adopt an interval of four days, and in the case of the asteroid planets, either four or eight days, between the dates for which the direct calculation is made. Then, by interpolating, in the case of an interval w, equal to four days, for the intermediate dates, we obtain a series of places at intervals of two days; and, finally, inter8 114 THEORETICAL ASTRONOMY. polating for the dates intermediate to these, we derive the places at intervals of one day. When a series of places has been computed, the use of differences will serve as a check upon the accuracy of the calculation, and will serve to detect at once the place which is not correct, when any discrepancy is apparent. The greatest discordance will be shown in the differences on the same horizontal line as the erroneous value of the function; and the discordance will be greater and greater as we proceed successively to take higher orders of differences. In order to provide against the contingency of systematic error, duplicate calculation should be made of those quantities in which such an error is likely to occur. The ephemerides of the planets, to be used for the comparison of observations, are usually computed for a period of a few weeks before and after the time of opposition to the sun; and the time of the opposition may be found in advance of the calculation of the entire ephemeris. Thus, we find first the date for which the mean longitude of the planet is equal to the longitude of the sun increased by 180~; then we compute the equation of the centre at this time by means of the equation (53), using, in most cases, only the first term of the development, or v - M — 2e sin M, e being expressed in seconds. Next, regarding this value as constant, we find the date for which L + equation of the centre is equal to the longitude of the sun increased by 180~; and for this date, and also for another at an interval of a few days, we compute u, and hence the heliocentric longitudes by means of the equation tan (I- ~ ) - tan u cos i. Let these longitudes be denoted by I and 1', the times to which they correspond by t and t', and the longitudes of the sun for the same times by ( and 0'; then for the time to, for which the heliocentric longitudes of the planet and the earth are the same, we have t —t + t(o- ) ( (e- t), or (118) i' - 180~ - 0' to - e + ( - - )- ) (t' — t), the first of these equations being used when 1 180~ - 0 is less TIME OF OPPOSITION. 115 than 1' - 180~ - ). If the time to differs considerably from t or t', it may be necessary, in order to obtain an accurate result, to repeat the latter part of the calculation, using to for t, and taking t' at a small interval from this, and so that the true time of opposition shall fall between t and t'. The longitudes of the planet and of the sun must be measured from the same equinox. When the eccentricity is considerable, it will facilitate the calculation to use two terms of equation (53) in finding the equation of the centre, and, if e is expressed in seconds, this gives 5 e' v -M -2e sinM+ - - sin 2M, 4 s s being the number of seconds corresponding to a length of arc equal to the radius, or 206264".8; and the value of v- M will then be expressed in seconds of arc. In all cases in which circular arcs are involved in an equation, great care must be taken, in the numerical application, in reference to the homogeneity of the different terms. If the arcs are expressed by an abstract number, or by the length of arc expressed in parts of the radius taken as the unit, to express them in seconds we must multiply by the number 206264.8; but if the arcs are expressed in seconds, each term of the equation must contain only one concrete factor, the other concrete factors, if there be any, being reduced to abstract numbers by dividing each by s the number of seconds in an arc equal to the radius. 43. It is unnecessary to illustrate further the numerical application of the various formulae which have been derived, since by reference to the formulae themselves the course of procedure is obvious. It may be remarked, however, that in many cases in which auxiliary angles have been introduced so as to render the equations convenient for logarithmic calculation, by the use of tables which determine the logarithms of the sum or difference of two numbers when the logarithms of these numbers are given, the calculation is abbreviated, and is often even more accurately performed than by the aid of the auxiliary angles. The logarithm of the sum of two numbers may be found by means of the tables of common logarithms. Thus, we have log (a + b) -log (+ a log b (+ If we put log tan x =- (log b - log a), 116 THEORETICAL ASTRONOMY. we shall have log (a + b) =log a - 2 log cos x, or log (a + b) log b - 2 log sin x. The first form is used when cos x is greater than sin x, and the second form when cos x is less than sin x. It should also be observed that in the solution of equations of the form of (89), after tan (2- )) —using the notation of this particular case-has been found by dividing the second equation by the first, the second members of these equations being divided by cos (2 - 0) and sin (2R- 0), respectively, give two values of J cos 3, which should agree within the limits of the unavoidable errors of the logarithmic tables; but, in order that the errors of these tables shall have the least influence, the value derived from the first equation is to be preferred when cos ( - o) is greater than sin (R-(0), and that derived from the second equation when cos (2i- ) is less than sin (2 - 0). The value of D, if the greatest accuracy possible is required, should be derived from dcos / when / is less than 450, and from A sin when /i is greater than 45~. In the application of numbers to equations (109), when the values of the second members have been computed, we first, by division, find tan ( q+o- W) and tan ( - w0); then, if sin ( + wo0) is greater than cos 1 (Q' + wo), we find cos ~i' from the first equation; but if sin (2/ + -oo) is less than cos (g + (w0), we find cos ~i' from the second equation. The same principle is applied in finding sin -i' by means of the third and fourth equations. Finally, from sin -i' and cos 1i we get tan ii', and hence i'. The check obtained by the agreement of the values of sin -i' and cos i', with those computed from the value of i' derived from tan i', does not absolutely prove the calculation. This proof, however, may be obtained by means of the equation sin i' sin Q' -sin i sin Q, or by sin i' sin o -= sin e sin a. In all cases, care should be taken in determining the quadrant in which the angles sought are situated, the criteria for which are fixed either by the nature of the problem directly, or by the relation of the algebraic signs of the trigonometrical functions involved. DIFFERENTIAL FORMULBE. 117 CHAPTER II. INVESTIGATION OF THE DIFFERENTIAL FORMULA WHICH EXPRESS THE RELATION BETWEEN THE GEOCENTRIC OR HELIOCENTRIC PLACES OF A HEAVENLY BODY AND THE VARIATION OF THE ELEMENTS OF ITS ORBIT. 44. IN many calculations relating to the motion of a heavenly body, it becomes necessary to determine the variations which small increments applied to the values of the elements of its orbit will produce in its geocentric or heliocentric place. The form, however, in which the problem most frequently presents itself is that in which approximate elements are to be corrected by means of the differences between the places derived from computation and those derived from observation. In this case it is required to find the variations of the elements such that they will cause the differences between calculation and observation to vanish; and, since there are six elements, it follows that six separate equations, involving the variations of the elements as the unknown quantities, must be formed. Each longitude or right ascension, and each latitude or declination, derived from observation, will furnish one equation; and hence at least three complete observations will be required for the solution of the problem. When more than three observations are employed, and the number of equations exceeds the number of unknown quantities, the equations of condition which are obtained must be reduced to six final equations, from which, by elimination, the corrections to be applied to the elements may be determined. If we suppose the corrections which must be applied to the elements, in order to satisfy the data furnished by observation, to be so small that their squares and higher powers may be neglected, the variations of those elements which involve angular measure being expressed in parts of the radius as unity, the relations sought may be determined by differentiating the various formulae which determine the position of the body. Thus, if we represent by 0 any co-ordinate of the place of the body computed from the assumed elements of the orbit, we shall have, in the case of an elliptic orbit, 0 f (r, i, Mo,,,, ), 118 THEORETICAL ASTRONOMY. Mo being the mean anomaly at the epoch T. Let 0' denote the value of this co-ordinate as derived directly or indirectly from observation; then, if we represent the variations of the elements by AT, a^, Ai, &c., and if we suppose these variations to be so small that their squares and higher powers may be neglected, we shall have do do do do do do o A + d —a (1) ~ d1I dQ 0 do do The differential coefficients d,, &c. ust now be derived from d d must now b derived rom the equations which determine the place of the body when the elements are known. We shall first take the equator as the plane to which the positions of the body are referred, and find the differential coefficients of the geocentric right ascension and declination with respect to the elements of the orbit, these elements being referred to the ecliptic as the fundamental plane. Let x, y, z be the heliocentric co-ordinates of the body in reference to the equator, and we have 0 =f(x, y, z), or do do do do-d d + - dy + dz. dx dy dz Hence we obtain do do dx do dy do dz -d7r d. d7r.dy d7r dz'; and similarly for the differential coefficients of 0 with respect to the other elements. We must, therefore, find the partial differential coefficients of 8 with respect to x, y, and z, and then the partial differential coefficients of these co-ordinates with respect to the elements. In the case of the right ascension we put 0- a, and in the case of the declination we put 0 = 8. 45. If we differentiate the equations x + X — a cos 8 cos a, y -+ Y-= cos 8 sin a, z + Z = A sin 8, regarding X, Y, and Z as constant, we find DIFFERENTIAL FORMULLE. 119 dx - cos a cos 8 dJ -- J sin a cos S dCa - cos a sinl d, dy sin c cos 8 dA + A cos a cos 8 du, - A sin a sin 3 d3, dz - sin a dJ + J cos 3 dS. From these equations, by elimination, we obtain sin ac COS a cosa S d x+ - ady, (3) cos sin a sin cos 8 d8 - - dx dy + - dz. Therefore, the partial differential coefficients of a and ~ with respect to the heliocentric co-ordinates are du sin a da cos o sin a cos 6 ~ - -' dx a' dx A da cos a dd sin a sin ( cos ~ (4) dy A' dy () de dS cos cos 6 -w- -- 0,~ dz dz A Next, to find the partial differential coefficients of the co-ordinates x, y, z, with respect to the elements, if we differentiate the equations (100)1, observing that sin a, sin b, "sin c, A, B, C are functions of g and i, we get dx. dr + x cot(A + u) du + d + dx di, r dy di dy Y dr + y cot (B ud+ ) dd + d ddi, rd r -I- z cot (C+ u) du {+ I d - dIi. d Cdv dC dx d To find the expressions for -, d-I &c., we have the equations d2' di x - r cos u cos - r sin u sin Q cos i, y = r cos u sin Q cos e + r sin u cos Q cosi cos - r sin u sin i sin e, z = r cos u sin 2 sine - r sin u cos gQ cos i sine -- r sinu sin i cos, which give, by differentiation, dx - - =- r cos u sin 9 - r sin u cos cos i, d2 -- == r cos u cos O cos e - r sin u sin Q cos i cos, dg 120 THEORETICAL ASTRONOMY. dz = r cos u cos g sin - r sin u sin g2 cos i sin e, d2 dx -- r sin u sin sin i, dy i- r sin u cos g sin i cos e -r sin u cos i sin, dz. -- r sin u cos Rg sin i sin e -+- r sin u cos i cos e. di The first three of these equations immediately reduce to dx ddy dz d - YS x cos - z sine; (5) d2'' d2 ^ dQ ^ and since cos a - sin S sin i, cos b = - cos Q sin i cos e- cos i sin e, cos = - cos Q sin i sin e -- cos i cos, we have, also, dx dy dz — = r sin u cos a, - r sin u cos b, d r sin u cos c. di di di Further, we have du v - dv - ddQ, and hence, finally, dx =- dr + x cot(A + u) dv - x cot (A + u) dr + (- x cot(A + u) - y cos e - z sin e) d q + r sin u cos a di, dy - Ydr + y cot(B + u) dv + y cot(B- + u) dr ( r (6) + (- y cot(B + u) +x cos e) d2 -+r sinu cos b di, dz =- Zdr cot ( dv cot(C ) z ct(C + u) d, + (- z cot (C - +) - + x sin e) d - + r sin u cos di. These equations give, for the partial differential coefficients of the heliocentric co-ordinates with respect to the elements, dx dx dydy d= - = x cot (A + u), d -d =y cot(B + u), d7r dv cot (r dz dz cot(c+ U); dr - - DIFFERENTIAL FORMULE. 121 d- - cot (A+Zu)-y cos — zsine, d~ — y cot (B+-u)-x cos e, dz d = z cot(C + u) + x sin e; dg dx. dy i dz d — r -sin u cos a, - r sin u cos b, d - r sin u cos c; (7) di n di di dx x dy y dz z dr r dr r' dr r When the direct inclination is greater than 90~, if we introduce the distinction of retrograde motion, we have du = dv - dr+ d-2, and hence dx - dx ct, dy dy __ d — ~ -- - x cot (A 0+ u), dr- dv —y cot (B + u), da - dv da dv dz dz dz - -z cot(C+ t); (8) j= -- dv dx d dy dy dy dz d d2 ~ dv y cos - z sin d, + xco d dv+ sin d - - d - o, d -- dv - i dx dy dz The expressions dr' dr dr T he expressions for -y- d- and ~ remain unchanged; and we have, also, dx dy. dz -di -r rsincos a - rslntcosb, z- r slnu cosc. (9) ad^ snos di di It is advisable, in order to avoid the use of two sets of formule, in part, to regard the motion as direct and the inclination as susceptible of any value from 0~ to 1800. If the elements which are given are for retrograde motion, we take the supplement of i instead of i; and if we designate the longitude of the perihelion, when the motion is considered as being retrograde, by (wv), we shall have 7- 2 - (7r). If we introduce, as one of the elements of the orbit, the distance of the perihelion from the ascending node, we have du = dv + do, and, hence, dx dx dy dy dx dx _ cot(A + ) dy), — y cot (B 4- u), d d d co t ( u)dv do( dv dz dz do - d — zcot( C - ). (10) dw dv. 122 THEORETICAL ASTRONOMY. dx dy dz The values of -, -, and - must, in this case, be found by means dQ' dQ' d2 of the equations (5). By means of these expressions for the differential coefficients of the co-ordinates x,y, z, with respect to the various elements, and those given by (4), we may derive the differential coefficients of the geocentric right ascension and declination with respect to the elements 62, i, and rr or (o, and also with respect to r and v, by writing successively a and 8 in place of 0, and 2, i, &c., in place of z in the equation (2). The quantities r and v, however, are functions of the remaining elements p, M0, and p; and we have. d dr +dr dM + dr dv dv v I- dM, dA= d9p + dM dTo + d,~ Da. Therefore, the partial differential coefficients of x, with respect to the elements p:, Mo, and I, are dx. d dx dr d dv d'p dr d dv d' dx dx dr dx dv dMo ~dr dM, + dv'dM' (1) dx dx dr dx dv d- dr d+ v dv l The expressions for the partial differential coefficients in the case of the co-ordinates y and z are of precisely the same form, and are obtained by writing, successively, y and z in place of x. The values of dx dx dy dy dz dz d, -d' d -,d and - are given by the equations (7), and dr dv dr dv dr dv dr dv dr d dvdr dv when the expressions for d) d,and -d ave beend d d-' d-p dMo' dM-' dn dp found, the partial differential coefficients of the heliocentric co-ordinates with respect to the elements o(, HM, and / will be completely determined, and hence, by means of (2), making the necessary changes, the differential coefficients of a and a with respect to these elements. 46. If we differentiate the equation M E- e sin E, DIFFERENTIAL FORMULi.M 123 we shall have dM- dE(1 - e cos E) -cos op sin E dy. r r But, since 1 e cos E -, and cos ( sin E= - sin v, this reduces to'a'a r r dM- - dE - -sin v dr, a a or dE - dM - sin v dp. If we take the logarithms of both members of the equation tan Iv - tan.E tan (450 + ~ ), and differentiate, we find dv dE d__ 2 sin v cos v 2 sin Ecosos E 2 sin (45~ + - ) cos (450 + ) which reduces to sin v sin v dv - - dE+ dy'. sin E cos 5y Introducing into this equation the value of dE, already found, and. r sin vu.^ replacing sin E by r sn we get i a cos?a2= COS d sin v acos, ) dv F=~ -dM-{ ~+ 1 d.' r2 cos 5p r But since a cos2 == p, and - 1 + sin qp cos v, this becomes A2 COS r d cos + 2 + tan p cos v sin v dp. (12) If we differentiate the equation r a (1 - e cos E), we shall have - dr - da +- ae sin E dE - a V c cos E d; a and substituting for dE its value in terms of dMi and dp, the result is dr = - -da + a tan sinvdM+ (ae sinEsinv-acosy cosE)d(p. (13) a 124 THEORETICAL ASTRONOMY. sin v cos cos v + e lNow, since sin E; o, and cosE E= ~, we shall have 1 + e cos v, I - e cos v ae cos p5 sin2 v a cos p (cos v + e) ae sin E sin v a cos E cos E- c1 -+ e ecosv 1 cos which reduces to ae sin E sin v - a cos p cos E - a cos 9 cos v Hence, the expression for dr becomes dr =- da -+ a tan e sin v dM — a cos v cos v dp. (14) Further, we have H= =Mo +- (t- T), T being the epoch for which the mean anomaly is M, and kVll + m a2 Differentiating these expressions, we get dM=-dM, + (t - T) di, da 2 d/j a F and substituting these values in the expressions for dr and dv, we have, finally, dr = a tan S sin v dMo + (a tan so sin v (t T) ~- ) d/ -a cos cos v de, (15) d C acos a Ca cos_^,, 2 i tan< cosvsinvd. r a2 cos o From these equations for dr and dv we obtain the following values of the partial differential coefficients:dr dv 2 -- a cos cos v,sinv, C~~dsoT9^~~ d9'- cos p dr dv a2 cos st d-a tan sin v, (16) dM o dM0 2 dr 2r dv aa cos T d/ = a tan lo sin v (t- T) 3- 206264.8, d — r (t - T). dfi 3fiL d/~ r2 DIFFERENTIAL FORMULA.M 125 dr It will be observed that in the last term of the expression for we have supposed / to be expressed in seconds of arc, and hence the factor 206264.8 is introduced in order to render the equation homogeneous. 47. The formulae already derived are sufficient to find the variations of the right ascension and declination corresponding to the variations of the elements in the case of the elliptic orbit of a planet; but in the case of ellipses of great eccentricity, and also in the cases of parabolic and hyperbolic motion, these formulae for the differential coefficients require some modification, which we now proceed to develop. First, then, in the case of parabolic motion, sin ip - 1, and instead of MO and, we shall introduce the elements T and q, the differential coefficients relating to 7r, 2, and i remaining unchanged from their form as already derived. If we differentiate the equation k(t — T).I..t =/ ) - (tan v + i tan Iv), regarding T, q, and v as variable, we shall have d,T/ - / ~t (t — T) T)17) rV21/ 2q Instead of q, we may use logq, and the equation will, therefore, become k kT2q 3k T)(t - T)/2q d —-l/2qT (t- )/ d log q, (18) in which is the modulus of the system of logarithms. in which 20 is the modulus of th e system of logarithms. 126 TTHEORETICAL ASTRONOMY. If we take the logarithms of both members of the equation r q cos2 1vV and differentiate, we find r dr -= dq + r tan -v dv. q Introducing into this equation the value of dv from (17), we get 1 t t \ 1/2qtan1,^ dr= r- 3k-(t~ — T) tan' \dq tan dT. (19) q r2 2IIq r k_ (t - T) Now, since k(t ) q (tan 2v -+ tan3 ~v), and q = r cos2 v, we have v2q 1 3k (t - T) tan. 1 * 1- 3kt- T/anv2 _(1 + tan2 Iv -3 sin2 v - sin2 2v tan2 Iv) q r21/ 2q r cos V,. r We also have kV1/ 2q t 2q cos2 1v tan Iv k sin v tan - ~~r 4 1/~~~~ 2q Therefore, equation (19) reduces to k sin v dr - cos v dq dT. (20) 1/2q If we introduce d log q instead of dq, this equation becomes _cosv k sin v dr- os d log q -2_ dT. (21) From the equations (17), (18), (20), and (21), we derive dr ksin v dv kl/2q dT 1/ 2q dT r2 dr dv 3k (t - T) Cosv,~ (22) dq dq - 1/ 2 dr __ qcosv dv 3k (t -T) 1/ 2q dlog q o dlogq 2A r and then we have, for the differential coefficients of x with respect to T and q or log q, DIFFERENTIAL FORMULE. 127 dx dx dr dx d dx d dr dx dv dT dr dT dv dT' dq dr' dq + dv dq dx x dr dx dv dlogq- dr' dlogq v' dlogq' and similarly for the differential coefficients of y and z with respect to these elements. The expressions for the partial differential coefficients of x, y, and z, respectively, with respect to r and v are the same as already found in the case of elliptic motion. We shall thus obtain the equations which express the relation between the variations of the geocentric places of a comet and the variation of the parabolic elements of its orbit, and which may be employed either to correct the approximate elements by means of equations of condition furnished by comparison of the computed place with the observed place, or to determine the change in the geocentric right ascension and declination corresponding to given increments assigned to the elements. 48. We may also, in the case of an elliptic orbit, introduce T, q, and e instead of the elements po, Mo, and /u. If we differentiate the expression q= a( -e), we shall have da = - dq +- de. q q We have, also, M= l/1 + m a- (t - T), in which T is the time of perihelion passage, and dMi= -k / + a- dT — 1/ + m a- (t T) da. Hence we derive dMl - - kV/l - I m a d T- kl1 + m (t - T) dq Of -kq___ q k/l+ma(t T)de. q Substituting this value of dM in equation (12), replacing sin o( by e, and reducing, we get d kp (1 ) dTT-a kp (L -T) dq r2 o.2 -(~ k~7~ —~~r(t-T j _ i~ 1 ld (23) ( 2p (+ - - 1 sin v I ede. (23) qr 7) 128 THEORETICAL ASTRONOMY. In a similar manner, by substituting the values of da and dM in equation (14), and reducing, we find drz- kVl/ +m Cdr- e sin v dT Vp r kl/l +m(t-T) A 2.e sin)dq +-~1r ~',- \~'- \^~-e sin v dq + Pq — cos v -- k(1 (t - T) de. (24) + P q q e i —~,de' (24) These equations, (23) and (24), will furnish the expressions for the dv dv dv dr dr I dr partial differential coefficients dT d- d- - d-, and -de which are dT dq de' d T dq de required in finding the differential coefficients of the heliocentric coordinates with respect to the elements T, q, and e, these quantities being substituted for MI, A#, and (n, respectively, in the equations (11). 49. When the orbit is a hyperbola, we introduce, in place of /I0, 6u, and qp, the elements T, q, and ~. If we differentiate the equation N,- e tan F loge tan (450 +.F), we shall have dN, c o -F - I + tan F de, ~ ( cos F which is easily transformed into r dF, __tan _ dA==-' cos. + tan F t d, ~ cos F cos or dF a atan sinF r tan FdN r cos Let us now take the logarithms of both members of the equation tan F == tan Xv tan, and differentiate, and we shall have dF sin v dv - sin v s. d+. sin F sin dF Introducing into L;s equation the value of sin already found, we get =a sinv asinv tan { sin v dv a - ~ ~. +. d.i r tan ~ r cos t Sin DIFFERENTIAL FORMULAE. 129 But, since r sin v = a tan i tan F, and p =- a tan2 4, this reduces to Adv - a dNV - + 1-. (25) r2 TO I - r sinv r If we differentiate the equation r=a(~ ~ e 1 ) \cos F f we get r tan2 dF a tan, dr da + ae tan2F + d. a sin +- cos F cos dF7 Substituting in this equation the value of sin we obtain d r + a2e tan F dN a2etan2 F_ a\tan, a r r cosFF cos4d which is easily reduced to - sinuv r ae dr=r daa d +C c da ae).* a sin c ro c os F C F sin But, since r ae a cos F cos2 F cos this reduces to r, a sinv pa i \ d4 dr=- da d~N, + ~pa e a sin dN r cos Fjsin' or r sin v Cos V dr= -da + a -. dNo + p d%. (26) a sinm sin{ Now, since q = (e- 1), we have d q d - a tan d, dq _ - da d, a cos{ or a a da - dq - p' d{. q q cos We have, also, N k- (t - T), and hence dN - ka- 3 dT- aka — (t- T) da. By substituting the value of da, this becomes do kA T qm - kaq T _ _ka-(t- T ) T) dNO =.~ka~ ~2dT~ ~^~-dq -0~-~ ~ d9. q aq cos % 9 130 THEORETICAL ASTRONOMY. Substituting this value of dN0o in equation (25), and reducing, we obtain dv -- d T k p T) dq r2 qr ( qr2 (\r l /sins (27) In a similar manner, substituting in equation (26) the values of da and dNo, and reducing, we get dr l sinv 3 dT~(r 3k(t -T) sinv )d:- ~ —7' cos )dq Vp cos 4 2 V2q c 1 os4 /cos4/., p(t — T). sin v r )p) (28) C ( cos -v p\-~ (28) t~\ dl/c T iq cos q sin s'( The equations (27) and (28) will furnish the expressions for the partial differential coefficients of r and v with respect to the elements T, q, and 4, required in forming the equations for cos 8 doc and dS. It will be observed that these equations are analogous to the equations (23) and (24), and that by introducing the relation between e and 4, and neglecting the mass, they become identical with them. We might, indeed, have derived the equations (27) and (28) directly from (23) and (24) by substituting for e its value in terms of 4'; but the differential formulae which have resulted in deriving them directly from the equations for hyperbolic motion, will not be superfluous. 50. It is evident, from an inspection of the terms of equations (23), (24), (27), and (28) which contain de and de, that when the value of e is very nearly equal to unity, the coefficients for these differentials become indeterminate. It becomes necessary, therefore, to develop the corresponding expressions for the case in which these equations are insufficient. For this purpose, let us resume the equation k (t T) (I + - u 4-+ ~u - 2i(Iu' + q 5-u) + 3i2 (u5 ( + u) - &c, 2q p 1-e in which u =tan Av, and i = Then, since - i <(1 e) - (1- e)2 + — e &., e 1le) e)2 &ha we shall have DIFFERENTIAL FORMULaE. 131 k(t )- + T)(~ - t -t5) ( - e) 1/2 q+ (32u - o3' + 38u7) ( - e)2 + &c. (29) dv If it is required to find the expression for d in the case of the variation of the elements of parabolic motion, or when 1 - e is very small, we may regard the coefficient of 1 - e as constant, and neglect terms multiplied by the square and higher powers of 1 - e. By differentiating the equation (29) according to these conditions, and regarding u and e as variable, we get 0 = (1 +- u2) du- ( u- - Iu5) de; and, since du- (1 + qu) dv, this gives dv lu - 13 ~- ut5 de (1 + u2)2 The values of the second member, corresponding to different values of v, may be tabulated with the argument v; but a table of this kind dv is by no means indispensable, since the expression for de may be changed to another form which furnishes a direct solution with the same facility. Thus, by division, we have d - 2 - 9 +9 3 de — — ~ + -J - T (1 q- u2)2' and since, in the case of parabolic motion, k (t - T) I r2 = 2 + 2)2 3 +r2q2(1 +- u)2, this becomes dv 9 k(t-T) /- 2 tan v. (31) de o r2 If we differentiate the equation q (1 -e)3 1 + e cos v' regarding r, v, and e as variables, we shall have dr __ 2r in2 v re sin v dv _~ _ (i-i.- e21 + - ^ (i ^ g) ~ ^(32) de - q ( 1 q- e)2 - q(e)'de —' 132 THEORETICAL ASTRONOMY. In the case of parabolic motion, e- I, and this equation is easily transformed into dr- r tan v (tan + 2 (33) de de (33) Substituting for d its value from (31), and reducing, we get de dr o (t - T) sin - Ltan2Iv. (34) de^- =. sin v + -r tan (4 dv dr The equations (31) and (34) furnish the values of - and de to be used in forming the expressions for the variation of the place of the body when the parabolic eccentricity is changed to the value 1 + de. When the eccentricity to which the increment is assigned differs but dv little from unity, we may compute the value of - directly from d e equation (30). A still closer approximation would be obtained by dv using an additional term of (29) in finding the expression for -; but a more convenient formula may be derived, of which the numerical application is facilitated by the use of Table IX. Thus, if we differentiate the equation v V+ A (100i) + B (100o)2 + ( (00i)3, regarding the coefficients A, B, and C as constant, and introducing the value of i in terms of e, we have dv dV 200A 400B 0 600 de de s(1 + e)' s(1 - e)' s8(1 q- e) in which s- 206264.8, the values of A, B, and C, as derived from dV the table, being expressed in seconds. To find V, we have k (t- T)l/ 1 + etn V tan3 - - tan.~ +i tan 7 2q~ which gives, by differentiation, k (t-T) de dV 2q' c/ 1 + e cos4 V; and if we introduce the expression for the value of M used as the argument in finding V by means of Table VI., the result is DIFFERENTIAL FORMULIE. 133 dV Mcos4TV de 75(1 e)' Hence we have dv Mcos4V7 200A 400B (100i) 6000 ( de - 75 (1 + e) ~(l + e)2 s (l + e)2 s(1 + e) dv by means of which the value of de is readily found. When the eccentricity differs so much from that of the parabola that the terms of the last equation are not sufficiently convergent, dv the expression for, which will furnish the required accuracy, may dle be derived from the equations (75), and (76),. If we differentiate the first of these equations with respect to e, since B may evidently be regarded as constant, we get dw _ k (t -T) cos4 4w q' B+ (36) de-1" V/2qt *B1/i,(1 -9e) (6 If we take the logarithms of both members of equation (76),, and differentiate, we get dv dC dw 4de T+ - - __(37)' sinv C sin w (1 + e) (1 + 9 e) To find the differential coefficient of C with respect to e, it will be sufficient to take - 1- 4A C0 which gives dC - 2 C2 dA. The equation 5 (1 -- e) (1 - 9e) t gives 50 A_9_dw_ dA - - tan2 — w de + t and w; (1 + 9e) " tan wcos2 ^w and hence we obtain / dC 20 A.C dC — (12 9e) tan w de + -4 - dw. Substituting this value in equation (37), we get dv 20 C. C sin v dw 4 sin v d-e ~ (1 +9 )2 t + sinw * de ( +e)(l1 + 9e) 134 THEORETICAL ASTRONOMY. dw and substituting, finally, the value of dew- we obtain dv k (t — T) C2 sin v cos' uw 20 C. de 1/2 i (+1 - 9e) zw (1- 9e)2 4 sin v (1 + e) (1 + 9e)' which, by means of (76)1, reduces to dv 9 k(t- T) 2 sin v cos2 w 8 tan - (v de ~ 1/2 B/(1+9e) tan w (1 e) (1 9e) If we introduce the quantity M which is used as the argument in finding w by means of Table VI., this equation becomes dv 9 MGi cos2 sin 8 tan Iv Mco C2SV __ — _ (39) de 2 (I + 9e) 75 tan w (1e) 9 e)' This equation remains unchanged in the case of hyperbolic motion, the value of C being taken from the column of the table which cordv responds to this case.; and it will furnish the correct value of d in de all cases in which the last term of equation (23) is not conveniently dr applicable. The value of d is then given by the equation (32). When the eccentricity differs very little from unity, we may put B= 1, and tan Iw = tan 1 v (l + 9e), COos2 w =- C2 cos2 v. Then we shall have M cos2 w. 2k (t -- T) 4 ~^2 Csmv= ~, = cos4 1w. 75 tan 1 1/2 q3 The equation = (1 + A C2) coS2 V= (1 + A) cos2 I, gives r -- (1 + QA) cos4 w C cos4 w. r2 Hence we derive Mcos2w2 sn (t- T) p p C 2 e 75 tan -w r~ C (1 q ( e)' NUMERICAL EXAMPLES. 135 If we substitute this value in equation (39), and put C2 (1 + e)= 2, we get A __ 9 k ]c 8 tan __ ~dv 9 kl/2P (t) 8tanTv - (40) de 2(1+9e) r (1 + e) (1 + 9e)' and when e 1, this becomes identical with equation (31). 51. EXAMPLES.-We will now illustrate, by numerical examples, the formulae for the calculation of the variations of the geocentric right ascension and declination arising from small increments assigned to the elements. Let it be required to find for the date 1865 February 24.5 mean time at Washington, the differential coefficients of the right ascension and declination of the planet Eurynome ~ with respect to the elements of its orbit, using the data and results given in Art. 41. Thus we have =-181~ 8' 29".29, — 40 42' 21".56, log J= 0.2450054, log r 0.428285, v 129~ 3' 50".5, u = 326~ 41' 40".1, A - 2960 39' 5".0, B - 205~ 55' 27".1, C- 212~ 32' 17".7, log sin a - 9.999716, log sin b 9.974825, log sin c 9.522219, log x 0.425066,, log y - 9.511920, log z 8.077315, =23~ 27' 24".0, t - T 420.714018. First, by means of the equations (4), we compute the following values:log cos 8 d =- 8.054308, log d - 8.66899, dx dx l dct / log cos -- 9.75499, log -= 6.968348, dy n) dyI log - 9.753529. dz Then we find the differential coefficients of the heliocentric co-ordinates, with respect to iT, i, i, v, and r, from the formule (7), which give dx d 49199/, dy dy log d =7.876A3 log -8.830941, log 9.2229, dx dy_ dz log -di 8.726364, -lo - 9.687577, log -= 0.142443,, log dx d9.996780 log dz og dr 9.996780,, log 9.083635, log — 7.649030. 136 THEORETICAL ASTRONOMY. dx dy dz In computing the values of di -, and d-, those of cosa, cos b nld cos may generally be obtained with sufficient accuracy from sina, sin b, and sine. Their algebraic signs, however, must be strictly attended to. The quantities sin a, sin b, and sin c are always positive; and the algebraic signs of cos a, cos b, and cos c are indicated at once by the equations (101)1, from which, also, their numerical values may be derived. In the case of the example proposed, it will be observed that cos a and cos b are negative, and that cos c is positive. dla da To find the values of cos a d and d-, we have, according to equation (2), da dao dx da d__ dr dx d dd dy dr da da dx d8 dy d dz: d-' dx dy di d dr' which give du d: da d 8 cos -- cos -- - 1.42345, - d - 0.48900. dn dv dO dv In the case of, i, and r, we write these quantities successively in place of r in the equations (41), and hence we derive dcu da cos 8 d - = 0.03845, - - 0.09533, dQ d Z cos d da cos = 0di.2 -0 - 0.78993, du, da cos d- - 0.08020, d -+ 0.04873. dr dr Next, from (16), we compute the following values:dr dr dr log -= 0.179155, log dr = 9.577453, log d = 2.376581~, dv dv dv log = 0.171999, logd- 9.911247, log = 2.535234. dx dx We may now find We may now find d-, dloN &c. by means of the equations (11), du da and thence the values of cos a da d &c.; but it is most convenient d-p dy' do du da da to derive these values directly from cos 8 d, cos a-, -, and -, dr dv dr dv in connection with the numerical values last found, according to the NUMERICAL EXAMPLES. 137 equations which result from the analytical substitution of the expressions fordx dy dz. sios d - &c., in equation (2), writing successively p, Mo0, and i in place of 7r. Thus, we have dc da dr dc dv c os cos -- + cos a' ds dr dy dv do' da da dr +d dv (42) d* dr dr dv dv and similarly for Ml4 and,, which give cos - = + 1.99400, - 0.65307, dyo d cos -a = + 1.13004, d - 0.38023, dum, dM, cos 8 d - +507.264, d - 179.315. Therefore, according to (1), we shall have cos A - +1.42345Azr- 0.03845A ~ - 0.27641Ai +1.99400As + 1.13004lMO + 507.264a,u, a^ - 0.48900A7 - 0.09533^ ~ - 0.78993i - 0.65307Ao -0.38023AMo~- 179.315Aa. To prove the calculation of the coefficients in these equations, we assign to the elements the increments AMo + 10", AT-r -20", Aa -10" Ai-+ 10", a=, _ + 10", A L + -0".01, so that they become Epoch =1864 Jan. 1.0 Greenwich mean time. Mo- 1" 29' 50".21 r — 44 20 13.09) =-206 42 30.13 Mean Equinox 1864.0 i — 4 37 0.51 J -= 11 16 1.02 log a = 0.3881288 - 928.56745 With these elements we compute the geocentric place for 1865 February 24.5 mean time at Washington; and the result is -= 181~ 8' 34".81, = - 4~ 42' 30".58, log = 0.2450284, 138 THEORETICAL ASTRONOMY. which are referred to the mean equinox and equator of 1865.0. The difference between these values of a and ~ and those already given, as derived from the unchanged elements, gives a= =- + 5".52, cos 8 AM = + 5".50, - = 9".02, and the direct substitution of the assumed values of aWr, ad, Ai, &c. in the equations for cos 8 hao and AS, gives cos 8 Am - + 5".46, 9".29. The agreement of these results is sufficiently close to show that the computation of the differential coefficients has been correctly performed, the difference being due chiefly to terms of the second order. When the differential coefficients are required for several dates, if we compute their values for successive dates at equal intervals, the use of differences will serve to check the accuracy of the calculation; but, to provide against the possibility of a systematic error, it may be advisable to calculate at least one place directly from the changed elements. Throughout the calculation of the various differential coefficients, great care must be taken in regard to the algebraic signs involved in the successive numerical substitutions. In the example given, we have employed logarithms of six decimal places; but it would have been sufficient if logarithms of five decimals had been used; and such is generally the case. It will be observed that the calculation of the coefficients of AT, A 2, and Ai is independent of the form of the orbit, depending simply on the position of the plane of the orbit and on the position of the orbit in this plane. Hence, in the case of parabolic and hyperbolic orbits, the only deviation from the process already illustrated is in the computation of the coefficients of the variations of the elements which determine the magnitude and form of the orbit and the position of the body in its orbit at a given epoch. In all du, dC, d8 d_ cases, the values of cos,' cos,- and are determined as dv dr dv dr already exemplified. If we introduce the elements T, q, and e) we shall have d dda, dr dc dv cos 8 d-T -+ cos a.. d-os dT ldrd d+T dv drI d8 dJ dr dd dv, adT deren + dv eff' r and similarly for the differential coefficients with respect to q and e. NUMERICAL EXAMPLES. 139 dr cdv dr dv dr dv The mode of calculating the values of d-r, dv dr' d' and A depends on the nature of the orbit. In the case of passing from one system of parabolic elements to another system of parabolic elements, the coefficients of he vanish. dr dv To illustrate the calculation of - dT- &c. in the case of parabolic motion, let us resume the values t-T= 75.364 days, and logq = 9.9650486, from which we have found log r - 0.1961120, v = 790 55' 57".26. Then, by means of the equations (22), we find dr d., log _= 8.095802, log = 9.242547, dv dv log dT 7.976397, log -0.064602,. If, instead of dq, we introduce d log q, we shall have dr dv logd lg = 9.569812, log 0.391867 dlogq ~d log q From these, by means of (43), we obtain the differential coefficients of Oc and ~ with respect to T and q or log q. The same values are also used when the variation of the parabolic eccentricity is taken dv into account. But in this case we compute also - from equation dedr (31) and r from (33) or (34), which give, for v = 790 55' 571.3, de e log -8.147367., log -9.726869. dr dv In the case of very eccentric orbits, the values of d, T, &c. are found from dv k p dr k d-T 2~- 7~d T - _ ~ e sin v, (44) dT r2 dT - - dAv k/Pt dr.r k (t- T) ~dv=- ~P (t - T), _ 3 e sin v dq B qr2 dq q e sin dr r r2 e sinv dv dq + p'dq the mass being neglected. 140 THEORETICAL ASTRONOMY. To illustrate the application of these formula, let us resume the values, t- T= 68.25 days, e 0.9675212, and log q 9.7668134, from which we have found (Art. 41) v = 102~ 20' 52".20, log r 0.1614052. Hence we derive logp - 0.0607328, and dv dr log T 7.943137n, log =8.180711,n dv dr log - = 0.186517~, log d — 0186517. dq dq If we wish to obtain the differential coefficients of v and r with respect to log q instead of q, we have dv q d dr q dr dlogq O dq dlogq Aodq in which 20 is the modulus of the system of logarithms. dv Then we compute the value of de by means of the equation (30). (35), (39), or (40). The correct value as derived from (39) is dv -e- 0.24289. de The values derived from (35), omitting the last term, from (40) and from (30), are, respectively, - 0.24440, - 0.24291, and - 0.23531. The close agreement of the value derived from (40) with the correct value is accidental, and arises from the particular value of v, which is here such as to make the assumptions, according to which equation (40) is derived from (39), almost exact. dr Finally, the value of de may be found by means of (32), which gives dr =+ 0.70855. de, When, in addition to the differential coefficients which depend on the elements T, q, and e, those which depend on the position of the orbit in space have been found, the expressions for the variation of the geocentric right ascension and declination become NUMERICAL EXAMPLES. 141 dalda da dal cos a =os cS A,r- cos d A ^ +cos d, Ai + cos d- A T dc-r dg di dT dc du + cos o- q a+ cos - d e, A= A7T +- AR +d-g Ai+'^AT+q-Ag+-A e. d8r d2 dT dq de If we introduce log q instead of q, the terms containing q become du ~ d8 respectively cos d A log q and d A log q. It should be d log q d log q observed that if ATr, Ag, and ai are expressed in seconds, in order that these equations may be homogeneous, the terms containing AT, aq, and Ae must be multiplied by 206264.8; but if arr, ^2, and Ai are expressed in parts of the radius as unity, the resulting values of cos 8 AO and AJ must be multiplied by 206264.8 in order to express them in seconds of arc. The most general application of the equations for cos a ac and A8 in terms of the variations of the elements is for the cases in which the values of cos a Ao and of a8 are already known by comparison of the computed place of the body with the observed place, and in which it is required to find the values of ar,, A i, &c., which, being applied to the elements, will make the computed and the observed places agree. When the variations of all the elements of the orbit are taken into account, at least six equations thus derived are necessary, and, if more than six equations are employed, they must first be reduced to six final equations, from which, by elimination, the values of the unknown quantities aTr, A, t&c. may be found. In all such cases, the values of Ao and Ad, as derived from the comparison of the computed with the observed place, are expressed in seconds of arc; and if the elements involved are expressed in seconds of arc, the coefficients of the several terms of the equations must be abstract numbers. But if some of the elements are not expressed in seconds, as in the case of T, q, and e, the equations formed must be rendered homogeneous. For this purpose we multiply the coefficients of the variations of those elements which are not expressed in seconds of arc by 206264.8. Further, it is generally inconvenient to express the variations A T,, and ae in parts of the units of T, q, and e, respectively; and, to avoid this inconvenience, we may express these variations in terms of certain parts of the actual units. Thus, in the case of T, we may adopt as the unit of AT the nth part of a mean solar day, and the coefficients of the terms of the equations for cos 8 aa and Ad which involve aT 142 THEORETICAL ASTRONOMY. must evidently be divided by n. In the same manner, it appears that if we adopt as the unit of Aq the unit of the mth decimal place of its value expressed in parts of the unit of q, we must divide its coefficient by 10%, and similarly in the case of Ae, so that the equations become da da dac s dac cos 8 a -= Cos d Ar -+- cosa A -a + cosa d- i -A- - cos 8a A T dT dQ, di nit dT S d, s d + - cos ~ q+ 1-bcos - he, (45) dd ds da s ds da dA + d dT A dq s ds -- ~e, n which s 206264.8. When log q is introduced in place of q, the coefficients of A log q are multiplied by the same factor as in the case of aq, the unit of A log q being the unit of the mth decimal place of the logarithms. The equations are thus rendered homogeneous, and also convenient for the numerical solution in finding the values of the unknown quantities An, A, Ai, AT, &c. When AT, Aq, and corrections to be applied to the corresponding elements arein, Os, and 1e0u' In the same manner, we may adopt as the unknown quantity, instead of the actual variation of any one of the elements of the orbit, n times that variation, in which case its coefficient in the equations must be divided by n. The vale of ab, derived by taking the difference between the computed and the observed place, is affected by the uncertainty necessarily incident to the determination of a by observation. The unavoidable error of observation being supposed the same in the case of a as in e wen epreed n the case of when expressed in parts of the same unit, it is evident that an error of a given magnitude will produce a greater apparent error in a than in a, since in the case of a it is measured on a small circle, of which the radius is cos 8; and hence, in order that the difference between computation and observation in a and 8 may have the same influence in the determination of the corrections to be applied to the elements, we introduce cos 8t a instead of a s. The same principle is applied in the case of the longitude and of all corresponding spherical co-ordinates. DIFFERENTIAL FORMUL;E. 143 52. The formule already given will determine also the variations of the geocentric longitude and latitude corresponding to small increments assigned to the elements of the orbit of a heavenly body. In this case we put s 0, and compute the values of A, B, sin a, and sinb by means of the equations (94)1. We have also C= 0, sin c - sin i, and, in place of a and 8, respectively, we write 2 and /. But when the elements are referred to the same fundamental plane as the geocentric places of the body, the formulae which depend on the position of the plane of the orbit may be put in a form which is more convenient for numerical application. If we differentiate the equations x'r cosu cos r sinu sin m cosi, y' = r cos u sin Q + r sin u cos cos i, z _ r sin u sin i, we obtain x' dx' -=- dr - r (sin u cos 2 +- cos u sin Q cos i) du r (cosu sin g + sin u cos os s i) d q + r sin u sin 2 sin i di, dy' = cdr - r (sin u sin Q - cos u cos cos i) du + r (cos u cos - sin u sin 2 cos i) d2 - r sin u cos g sin i di, (46) dz' =- dr + r cos u sin i du + r sin u cos i di, r in which x', y', z' are the heliocentric co-ordinates of the body in reference to the ecliptic, the positive axis of x being directed to the vernal equinox. Let us now suppose the place of the body to be referred to a system of co-ordinates in which the ecliptic remains as the plane of xy, but in which the positive axis of x is directed to the point whose longitude is 2; then we shall have dx.= dx' cos Q + dy' sin Q, dy - dx' sin S, + dy' cos g^, dz - dz', and the preceding equations give dx - dr- r sin u du -r sin u cos i da, dy ==dr +rr cos u cos i du-+r cos u d - r sinu sini di, (47) dz -- dr + r cos u sin i du + r sin u cos i di. r 144 THEORETICAL ASTRONOMY. This transformation, it will be observed, is equivalent to diminishing the longitudes in the equations (46) by the angle 8g through which the axis of x has been moved. Let X,, Y,, Z, denote the heliocentric co-ordinates of the earth referred to the same system of co-ordinates, and we have x x,= J cos cos (A - 2), y Y, - cos sin ( - ), z + Z, = D sin fi, in which 2 is the geocentric longitude and P the geocentric latitude. In differentiating these equations so as to find the relation between the variations of the heliocentric co-ordinates and the geocentric longitude and latitude, we must regard PQ as constant, since it indicates here the position of the axis of x in reference to the vernal equinox, and this position is supposed to be fixed. Therefore, we shall have dx =cos p cos (A- ) da- sin f cos (A - ) df- J cos sin (A- ) dA, dy= cosfi sin ( -2) d- A sinf sin (A -g) d3 +- cosf cos(A —g) dA, dz -sin d dJ +- cosI dP, from which, by elimination, we find cos dA - sin (A- ) d cos ( -)dy, d A d sin cos ( -2 ) sin sin (A - ) cos ~~ d~x~ - dy- -dz. These equations give cd sin (A - 2) df sin f cos (A - ) cos = ~ cod'A dx - C d cos (Ay - ) df sin sin (-2) (48) dy d A dA dfp cos f cos f - 0, - cosdz~ dz D If we introduce the distance co between the ascending node and the place of the perihelion as one of the elements of the orbit, we have du - dv + dw, and the equations (47) give dx x dy y dz z d —-cosu, -— d sinu cosi, d=-= sinu sini; dr r dr r dr r dx dx d dy dy. d d - -= — -- r sin u, d - - rcos ucos, d - =- rcosu sin i dv dw dv dw dv do DIFFERENTIAL FORMULAE. 145 dx - dy dz d — =-r sin it cos i, d —r cos u,0 (49) d2 ~ ddb ~ d-0 dx dy dz — 0, d -r sin u sin i, - r sin u cos i. di di di If we introduce w, the longitude of the perihelion, we have du dv + di - d d, and hence the expressions for the partial differential coefficients of the heliocentric co-ordinates with respect to w and a become dx. dy dz d sn d- r os u cos, d r cos sin i; dir dr dr dx dy dz0) -g si = 2' sin si, d- 2r cos Si2, - = r cosu sin i When the direct inclination exceeds 90~ and the motion is regarded as being retrograde, we find, by making the necessary distinctions in regard to the algebraic signs in the general equations, dx. dy. dz di, - d r sin u sin i, d - r sin u cos; (51) dx dx dx dy and the expressions for d-' dv d', &c. are derived directly from (49) by writing 180~ - i in place of i. If we introduce the longitude of the perihelion, we have, in this case, du = -dv — d + d, and hence dx. dy dz d- r sin u, — =-r OS u Co, - - r cos d sin i; dv dy Ao (52) dx. 2.. dy n2 i dz dgQ d, 2r scosu sinin, d -r cosu sin i. But, to prevent confusion and the necessity of using so many formulae, it is best to regard i as admitting any value from 0~ to 180~, and to transform the elements which are given with the distinction of retrograde motion into those of the general case by taking 180~ - i instead of i, and 2g2 - n instead of T, the other elements remaining the same in both cases. 53. The equations already derived enable us to form those for the differential coefficients of A and: with respect to r, v, 2, i, and w or 7r, by writing successively i and f in place of 0, and 2, i, &c. in 10 146 THEORETICAL ASTRONOMY. place of 7r in equation (2). The expressions for the differential coefficients of r and v, with respect to the elements which determine the form of the orbit and the position of the body in its orbit, being independent of the position of the plane of the orbit, are the same as those already given; and hence, according to (42) and (43), we may derive the values of the partial differential coefficients of i and /9 with respect to these elements. The numerical application, however, is facilitated by the introduction of certain auxiliary quantities. Thus, if we substitute the values given by (48) and (49) in the equations d dA dx d2 dy cos f - d = — c os d d - +cosf d- d dv dx aa dv df_ df dx d3 dy d1 dz dv dx d dy dv dz dv' and put cos i cos (A- g) = A0 sin A, sin (A - A) = Ao cos A, sin i n sin (n5, -sin (A - 2) cos i -Z n cos N, in which AO and n are always positive, they become dA da r c os cos d = ~_ Ao sin (A -+ u), d- d3f r (sin f cos (A - ) sin u + n cos u sin (N + -) ). av - w J Let us also put n sin (N+ p) B, sin B, 54 sin A cos (2 - ~) = Bo cos B, and we have d2 d2 r cos -- = cos -- Ao sin (A + — ), dv dw ( dv d'- r Bo sin (B +- u). d dd 3 The expressions for cos d- and -d give, by means of the same auxiliary quantities, d- Ao cosd cos (A + u), ^d __ B (56) d — B cos (B + u). dr In the same manner, if we put DIFFERENTIAL FORMULE. 147 cos ( - g~)= C0 sin C, cos i sin (R- P2) =C Cos C;(57) cos i - Do sin D, sin ( - g) sin i = D cos D; we obtain d2 r cos id g- A C sin ( C + u), db _ Ao singi cos (A +- ); (58) dI r. cos sin i== — sin si cos (A- ), di d O d-l 4 ss sin (.D+p). If we substitute the expressions (55) and (56) in the equations dA dA dr dA dv cos /- Co = cos - * + cos P -, df dr dr d dv d dp dr d,p dv do dr do dv dso' and put dr - d _==f sin F = a cos (o cos v, (59) r - j cosF (- - + tan cos v r sin v, d14p " cos / we get cos/5 ( + cos d = f A sin (A+ F +u), t (60) dfBo d f, Bo sin (B - F+ q). In a similar manner, if we put dr d, = g sin G = - a tan s sin v, dv a2 cos P r-o =g cos G= Td0M r (61) - d - h sin H -( a tan sinv (t-T)- - 206264.8), r da =h cos r (t- T), ~dA^~ r 148 THEORETICAL ASTRONOMY. we obtain cos, dE g Ao sin (A + G + u- ), cos # - d- g- Bo sin (B + G + u); dH0 dMAz1 (62) cos d -- -- Ao sin (A + H+ u), df h Bo sin (B +- 1 + +). The quadrants in which the auxiliary angles must be taken are determined by the condition that A,, Bo Co f, g, and h are always positive. 54. If the elements T, q, and e are introduced in place of Mo, p, and j, we must put dr dv f sin F - d' fcos F- r -, dr dv gsin G= — d g cos G=r dT' (63) dr Av h sin _H -d hcos -=r -d dq' dq' and the equations become cos l d Ao sin (A+ F-+ -u), de -- -- B sin (B + F+ u); deA cos i.-g Ao sin (A + G + u), cost - d — s dq A 0 nA)(64) -p cs g- BA sin (A + + u6); dA h cod Bo sin (B + H-ff+- ) dq A In the numerical application of these formule, the values of the second members of the equations (63) are found as already exemplified for the cases of parabolic orbits and of elliptic and hyperbolic orbits in which the eccentricity differs but little from unity. In the same manner, the differential coefficients of i and f with respect to any other elements which determine the form of the orbit may be computed. NUMERICAL EXAMPLES. 149 In the case of a parabolic orbit, if the parabolic eccentricity is supposed to be invariable, the terms involving e vanish. Further, in the case of parabolic elements, we have dr k sin v dv gsin G —dT- 1/2q - - r tan.),vdT' dv gcos G =rdT dT which give tan G- -- tan'v. Hence there results G-180~ —'v, and g —=k., which is the expression for the linear velocity of a comet moving in a parabola. Therefore, cos fl d A -/2 A~ sin (A + ut - u v), _d _ k V2 (65) dT ~ Bo sin (B +u - ~v). For the case in which the motion is considered as being retrograde, 180~ - i must be used instead of i in computing the values of A0, A, n, N, CQ, and C, and the equations (55), (56), and the first two of (58), remain unchanged. But, for the differential coefficients with respect to i, the values of Do and D must be found from the last two of equations (57), using the given value of i directly; and then we shall have d_ r cos = - sin i sin u cos (A - ), (66) d - Do sin sin (D ). di A 55. EXAMPLES.-The equations thus derived for the differential coefficients of A and 9 with respect to the elements of the orbit, referred to the ecliptic as the fundamental plane, are applicable when any other plane is taken as the fundamental plane, if we consider A and f as having the same signification in reference to the new plane that they have in reference to the ecliptic, the longitudes, however, being measured from the place of the descending node of this plane on the ecliptic. To illustrate their numerical application, let it be required to find the differential coefficients of the geocentric right ascension and declination of Eurynome 0 with respect to the elements of its orbit referred to the equator, for the date 1865 February 24.5 mean time at Washington, using the data given in Art. 41. 150 THEORETICAL ASTRONOMI. In the first place, the elements which are referred to the ecliptic must be referred to the equator as the fundamental plane; and, by means of the equations (109),, we obtain Q'- 3530 45' 35".87, i'= 19~ 26' 25".76, wo 212~ 32' 17".71, and' -- w+ 0- = 50~ 10' 7".29, which are the elements which determine the position of the orbit in space when the equator is taken as the fundamental plane. These elements are referred to the mean equinox and equator of 1865.0. Writing a and a in place of 2 and A, and 2', i', wi in place of S, i, and w, respectively, we have Ao sinA cos (a,-') cos i', Ao cos A = sin (a -'); n sin N sini', n cosN- -cos i' sin (a-'); Bo sin B n sin (N + 3), B cos B z sin 8 cos (a~ -'); C sin C — cos ( - g'), Co cos C-sin (a -') cos'; Do sin D = cos i', Do cos D sin i' sin (a -'); f sin F = a cos' cos v, f cos F- + tan cosv) r sin v; g sin G= - a tan p sin v, a cos Gp - h sin H= ( a tan p sinv (t - T) - - 206264.8), a2 cos ( h cos H -- o (t - T). r The values of AO, n, B, (C,, Do, f, g, and h must always be positive, thus determining the quadrants in which the angles A, B, &c. must be taken; and these equations give log A,= 9.97497, A 262~ 10' 40", log BO 9.52100, B= 75 48 35, log C= 9.99961, C- 263 2 6, logDo= 9.97497, D= 92 35 47, logf =0.62946, F=339 14 0, log g =0.34593, G =350 11 16, log h =2.97759, H=- 14 30 48, u' v + -w'- 179~ 13' 58". NUMERICAL EXAMPLES. 151 Substituting these values in the equations (55), (58), (60), and (62), and writing a and 8 instead of 2 and j, and u' in place of u, we find cos a - - + 1.4235, d - 0.4890, dw' dw cos a d +- - 1.5098, d -+- 0.0176, da dd cos 8 I- + 0.0067,, + 0.0193, di' 0.0193, du d8 cos d -- + 1.9940, - 0.6530, d^ d d cos d -- +1.1300, d- - 0.3802, dlu4 dM8 cos d - + 507.25, - - 179.34; and hence cos 3 Aa = + 1.4235 Ao' + 1.5098 A'g' + 0.0067 Ai' + 1.9940 ao + 1.1300 AM +- 507.25 Aa/, a - - 0.4890 Aw' + 0.0176 A 2' + 0.0193 i' - 0.6530 A^ - 0.3802 AMo - 179.34 a/. If we put av'= - 6".64, A 2' - 14".12, i' - 8".86, -+ 10", M 10", +Am + 1- 0".01, we get cos 8 A +- 5".47, A - 9".29; and the values calculated directly from the elements corresponding to the increments thus assigned, are cos -- + 5".50, ^ - 9".02. The agreement of these results is sufficiently close to prove the calculation of the coefficients in the equations for cos 8 Aa and A^. When the values of Aoc, A a, and Ai' are small, the corresponding values of A^o, A, and Ai may be determined by means of differential formulae. From the spherical triangle formed by the intersection of the planes of the orbit, ecliptic, and equator with the celestial vault, we have cos s i cos' cos + sin i' sin e cos g', sin i cos g = - cos i' sin e + sin i' cos E cos', i sin i sin 2 == sin i' sin g', (67) sin i sin w =- sin' sin e, sin i cos wo =co s e sin i' - sin cos i' cos g, 152 THEORETICAL ASTRONOMY. from which the values of 2, i, and (o may be found from those of g' and i. If we differentiate the first of these equations, regarding e as constant, and reduce by means of the other given relations, we get di = cos wv di' - sin w sin i'd Q'. (68) Interchanging i and 180~ - i', and also a and t', we obtain di' = cos Wo di - sin w sin i d 2. Eliminating di from these equations, and introducing the value sin i' sin 2 sin i sin " the result is sin c, s di'. d - C.sin'os tDo d ~ ~ ~s (69) sin PI sinl If we differentiate the expression for cos w0 derived from the same spherical triangle, and reduce, we find dwo =- cos i dp - cos i' d'. Substituting for dQ its value given by the preceding equation, and reducing by means of sin 2' cos i' =sin a cos w0 cos i - cos g sin wo, we get sin w in %w dwto Sill cos 8g dP' -. cos i di'. (70) sinl sin i The equations (68), (69), and (70) give the partial differential coefficients of 2, i, and oo with respect to 2' and i', and if we suppose the variations of the elements, expressed in parts of the radius as unity, to be so small that their squares may be neglected, we shall have sin (% sin w ao - cos 2A'-. cosi Ai', sin' sni sin, sin., A~= sinb g S cost agg'....~Ai', (71) si 2 I "` Q' sin i Ai sin w0 sin' A Q' + cos w Ai', AZ t At- Atoo. If we apply these formulae to the case of Eurynome, the result is Aw o- 4.420A P' + 6.665Ai', A = — 3.488A 2' + 6.686Ai', Ai = - 0.179g' - 0.843Ai'; DIFFERENTIAL FORMULE. 153 and if we assign the values A _- 14".12, ai - 8".86, Aw' - 6".64, we get ao0 =- + 3".36, A - 10".0 i 0.0, i -]- 10.0, A 10.0, and, hence, the elements which determine the position of the orbit in reference to the ecliptic. The elements ow',', and i' may also be changed into those for which the ecliptic is the fundamental plane, by means of equations which may be derived from (109)t by interchanging 2 and 2' and 1800 -' and i. 56. If we refer the geocentric places of the body to a plane whose inclination to the plane of the ecliptic is i, and the longitude of whose ascending node on the ecliptic is Q, —which is equivalent to taking the plane of the orbit corresponding to the unchanged elements as the fundamental plane,-the equations are still further simplified. Let x', y', z' be the heliocentric co-ordinates of the body referred to a system of co-ordinates for which the plane of the unchanged orbit is the plane of xy, the positive axis of x being directed to the ascending node of this plane on the ecliptic; and let x, y, z be the heliocentric co-ordinates referred to a system in which the plane of xy is the plane of the ecliptic, the positive axis of x being directed to the point whose longitude is 2. Then we shall have dx' dx, dy' — dy cos i + dz sin i, dz' - dy sini -+ dz cos i. Substituting for dx, dy, and dz their values given by the equations (47), we get. X' dx' - dr- r sin u du- r sin u cos i d, r dy' - dr + r cos u du + r cos u cos i d2, dz' =- dr -r cos u sin i d2 +- r sin u di. It will be observed that we have, so long as the elements remain unchanged, X' = r cos u, y' =r sin u, 0, 154 THEORETICA.L ASTRONOMY. and hence, omitting the accents, so that x, y, z will refer to the plane of the unchanged orbit as the plane of xy, the preceding equations give dx - cos u dr- r sin u du- r sin u cos i dQ, dy = sin u dr + r cos u du -- r cos u cosi d 2, dz - r cosu sini dQ - r sin di. The value of w is subject to two distinct changes, the one arising from the variation of the position of the orbit in its own plane, and the other, from the variation of the position of the plane of the orbit. Let us take a fixed line in the plane of the orbit and directed from the centre of the sun to a point the angular distance of which, back from the place of the ascending node on the ecliptic, we shall designate by a; and let the angle between this fixed line and the semitransverse axis be designated by X. Then we have X/ - + 0. The fixed line thus taken is supposed to be so situated that, so long as the position of the plane of the orbit remains unchanged, we have a=- 2, — = 7 But if the elements which fix the position of the plane of the orbit are supposed to vary, we have the relations d6a cosid a, dw = dX - cos i dQ, (72) dn -= d + (1 - cos i) d =- dX + 2 sin's i d a. Now, since u v + w, we have u= - + -, and du - d + dv - d -_ dv + d% - cos i d. Substituting this value of du in the equations for dx, dy, dz, they reduce to dx cos u dr - r sin u dv- r sin u dz, dy - sin u dr + r cos u dv + r cos u dZ, (73) dz -- r cos u sin i d2 + r sin u di. The inclination is here supposed to be susceptible of any value from 0~ to 180~, and if the elements are given with the distinction of retrograde motion we must use 180~ - i instead of i. Let us now denote by 0 the geocentric longitude of the body measured in the plane of the unchanged orbit (which is here taken as the DIFFERENTIAL FORMULAE. 155 fundamental plane) from the ascending node of this plane on the ecliptic, and let the geocentric latitude in reference to the same plane be denoted by ^. Then we shall have x +- X- A cos C cos 0, y -- Y= J cos 7 sin 0, z +- Z = a sin 77, in which X, Y Z are the geocentric co-ordinates of the sun referred to the same system of co-ordinates as x, y, and z. These equations give, by differentiation, dx cos r cos 0 dDA A sin,^cos 0 dr - A cos -q sin 0 dO, dy= cos sin 0 dJ - A sin - sin 0 dv +- A cos - cos 0 dO, dz - sin f dJA - A cos V dr; and hence we obtain sin O cos 0 cos j dO - ~j dx.+ a - dy, sin cos 0 sin V sin 0 cos lr — dx- dy+ a dz. These give do sin o dO cos o do cos - - cos, cos - 0; dx J'dy A dz (74) dr, sin - cos dr sin sinO dv cos 7. dx u' d' d and from (73) we get dx dy s dz -C=GOSU d, - sin u,. = O; dr -cosu, dr dr dx dx dy dy dz dz...-.-... r sin, i -n- r cos u, =- 0; dv - d dv d dv d 7 (75) dx dy dz (75 -d d= 0, -- -r cos u sin i; ddb d2 dQ dx dy dz r sin. di di di r s Substituting the values thus found, in the equations do do dx dO dy cos = cos dx dv + cos f -d' dv dx dv dy dv drv d-r dx + dr dy dr d dv dx dv dy dv dz dv' 156 THEORETICAL ASTRONOMY. we get do do r cos d -- os ) - cos (O - t), d7 d7) r.(76) d-q d-q r. dvr - d ~ sin 7 sin (O -u). In a similar manner, we derive cos 7 sin ( ), d- _ 1 sin Cos (0 -~ z), dQ d^ ^, dr. dO d r cos = 0, d cos x sin cosu, (77) do dv r cos0 di -o d7 c+os s sin u. di' If we introduce the elements p, MO, and /, which determine r and v, we have, from dO do dr dO dv cos v -- cos d' cos - dSt dr dyp dv d. dr- drl dr d) dv d=o dr dp dv dA' if we introduce also the auxiliary quantitiesf and F, as determined by means of the equations (59), os f d cos (- -F), sin sin (0 -u —F). (78) ls cos do -s Finally, using the auxiliaries g, h, G, and H, according to the equations (61), we get Co. ( do C g sin V sin (0 - u - cos cos (G), dMo sin in ( uG), dMoo - - dM o - ~~dM~li0 zi~~ dlH0~~ ~(79) dO h dr) h.s( cos 0 d. cos (O - - H), d- =- sinm sin (O - u H). If we express r and v in terms of the elements T, q, and e, the values of the auxiliaries f, g, h, F, &c. must be found by means of,(64); and, in the same manner, any other elements which determine the form of the orbit and the position of the body in its orbit, may be introduced. The partial differential coefficients with respect to the elements having been found, we have do do d o do cos ] aO -= cos ) -- A2' q- cos 7 -~ zAx q- cos V 7j AM0 q- cos ) d7] 1, df) df) d7) d7) d7 i. d7A dQb di c-d di dM dAL DIFFERENTIAL FORMULaE. 157 from which it appears that, by the introduction of X as one of the elements of the orbit, when the geocentric places are referred directly to the plane of the unchanged orbit as the fundamental plane, the variation of the geocentric longitude in reference to this plane depends on only four elements. 57. It remains now to derive the formulae for finding the values of 7 and 0 from those of 2 and f. Let x, yo, z0 be the geocentric coordinates of the body referred to a system in which the ecliptic is the plane of xy, the positive axis of x being directed to the point whose longitude is Q; and let xO', yO, z0' be the geocentric co-ordinates of the body referred to a system in which the axis of x remains the same, but in which the plane of the unchanged orbit is the plane of xy; then we shall have Xw o J= = cos f cos ), cos cos 0, O = A cos f3 sin (A- ^), yo' = J cos 7 sin 0, z = A sin f,, z' = A sin V, and also Yo = yo cos i + zo sin i, o - yo sin i -+ z cos i. Hence we obtain cos 7 cos 0 = cos f cos (A - g ), cos v sin 0 cos f sin (A- ) cos i - sin f sin i, (80) sin ] - - cos sin ( - ) sin i + sin /i cos i. These equations correspond to the relations between the parts of a spherical triangle of which the sides are i, 900 -', and 90~ - f, the angles opposite to 90~ — ^ and 90~ - f being respectively 90~ + (2 - 2) and 90~ - 0. Let the other angle of the triangle be denoted by r, and we have cos sin -sin i cos (. - ), cos v cos - sin i sin (A - ) sin f + cos i cos. 8 The equations thus obtained enable us to determine r, 0, and r from 2 and f9. Their numerical application is facilitated by the introduction of auxiliary angles. Thus, if we put n sin N- sin fi, n cos N= cos f sin (A - ), 158 THEORETICAL ASTRONOMY. in which n is always positive, we get cos - cos 0 = cos cos (2 - ), cos ~ sin o n cos (N - i), (83) sin f = n sin (N - i), from which ^ and 0 may be readily found. If we also put n' sin N' - cos i, n' cos N' si i sin (A - ), () we shall have cot N' =tan i sin (A- g), -cos N' tan r (N' + i) cot (A- ). (85) If r is small, it may be found from the equation sini cos (t 2 -b) sin n cos - (86) cos -/ The quadrants in which the angles sought must be taken, are easily determined by the relations of the quantities involved; and the accuracy of the numerical calculation may be checked as already illustrated for similar cases. If we apply Gauss's analogies to the same spherical triangle, we get sin (450 - ~) sin (45~ - (0 + r)) cos (45~ + ( ))sin (45~ -- ( + i)), sin (450 - ) cos (45~ - (d + r)) sin (45~ + ~ (A - 2)) sin (45~ - - ( - i)), cos (45~ - ) sin (45~ - ( r)) (87) cos(45~ + - (A - )) cos (45~ - (- + i)), cos (45~ - V) cos (45~ - (o r)) sin (45~ + ( ( - 2)) cos (45~ - ( -)) from which we may derive I, 0, and r. When the problem is to determine the corrections to be applied to the elements of the orbit of a heavenly body, in order to satisfy given observed places, it is necessary to find the expressions for cos a AO and Ar in terms of cos / a/2 and A/. If we differentiate the first and second of equations (80), regarding 2 and i (which here determine the position of the fundamental plane adopted) as constant, eliminate the terms containing dq from the resulting equations, and reduce by means of the relations of the parts of the spherical triangle, we get NUMERICAL EXAMPLE. 159 cos d -= cos r cos 3 dA + sin r df. Differentiating the last of equations (80), and reducing, we find dv =- - sin r cos fp d) + cos r dfp. The equations thus derived give the values of the differential coefficients of 0 and V with respect to 2 and 9; and if the differences A/ and af/ are small, we shall have cos d A0 cos r cos aA -t silln A, A. =- - sin r cos P AA + cos r A. (8 The value of r required in the application of numbers to these equations may generally be derived with sufficient accuracy from (86), the algebraic sign of cos being indicated by the second of equations (81); and the values of j and 0 required in the calculation of the differential coefficients of these quantities with respect to the elements of the orbit, need not be determined with extreme accuracy. 58. EXAMPLE.-Since the spherical co-ordinates which are furnished directly by observation are the right ascension and declination, the formulae will be most frequently required in the form for finding V and 0 from o and 8. For this purpose, it is only necessary to write a and a in place of 2 and 9, respectively, and also 2', i', w', Zt, and u' in place of 2, i, w, Z, and u, in the equations which have been derived for the determination of f and 6, and for the differential coefficients of these quantities with respect to the elements of the orbit. To illustrate this clearly, let it be required to find the expressions for cos V aO and a^ in terms of the variations of the elements in the case of the example already given; for which we have 50~ 10' 7".29,' - 3530 45' 35".87, i — 190 26' 25".76. These are the elements which determine the position of the orbit of.Erynome @, referred to the mean equinox and equator of 1865.0. We have, further, logf= 0.62946, log g 0.34593, log h = 2.97759, F= 339~ 14' 0", G = 3500 11' 16", H-= 14~ 30' 48", u' — 1790 13' 58". In the first place, we compute (, 0, and r by means of the formulae 160 THEORETICAL ASTRONOMY. (83) and (85), or by means of (87), writing a, a, g', and i' instead of 2,, 2, and i, respectively. Hence we obtain 0 188~ 31'9", - - 1~ 59' 28", r- 19 17' 7". Since the equator is here considered as the fundamental plane, the longitude 0 is measured on the equator from the place of the ascending node of the orbit on this plane. The values of the differential coefficients are then found by means of the formulae COs7 O- - cos 0 sin'cos', dldb~' dg' do d_ r Cos di' -di' - cossin', dO r drj r csos dd- ft), sin sin (O - u'), cos 7 dO f cos (O - - F), d f sin sin ( -'- F), dcos d os ( t ) d - g sin sin (O -' - G), d M O J dM0 zi dO h dn h cos d cos (O - u'- H), d - sin sin (' - H), d~ A dtA. which give cos -dg, 0 d -+ 0.5072, C dos. di' -- - + 0.0204, cos d I, + 1.5051, d' =- + 0.0086, de' d,~ cos -r d - + 2.0978, d- + 0.0422, cos. d -4- 1.1922, d- - + 0.0143, dMp dyk do -1 cos d- + 538.00, dTherefore, the equations for cos ^A and ay become cos q =O + 1.5051 AX' + 2.0978 As + 1.1922 alM + 538.00 Alt, ^ =- + 0.0086 AX' + 0.0422 as + 0.0143 AMo - 1.71 ^A + 0.5072 A 2' + 0.0204 ai'. If we assign to the elements of the orbit the variations DIFFERENTIAL FORMULBE. 161 Aw= - 6".64, g - 14".12, i' -- -8".86, A -= + 10", A.0 = + 10", A = +r 0".01, we have A' = Aco' + cos i' A ~' - 19".96; and the preceding equations give cos hO + 8".24, = 6".96. With the same values of aw', A g', &c., we have already found cos 8 a c + 5".47, A a - 9".29, which, by means of the equations (88), writing a and a in place of A and I, give cos 0- + 8".23, n - 6".96. 59. In special cases, in which the differences between the calculated and the observed values of two spherical co-ordinates are given, and the corrections to be applied to the assumed elements are sought, it may become necessary, on account of difficulties to be encountered in the solution of the equations of condition, to introduce other elements of the orbit of the body. The relation of the elements chosen to those commonly used will serve, without presenting any difficulty, for the transformation of the equations into a form adapted to the special case. Thus, in the case of the elements which determine the form of the orbit, we may use a or loga instead of,/, and the equation kVl/ + m y - 3 /1- 0 a2 gives di Ada - = d log a, (89) in which 20 is the modulus of the system of logarithms. Therefore, the coefficient of AcZ is transformed into that of A log a by multiplying it by - -; and if the unit of the rnth decimal place of the logarithms is taken as the unit of A log a, the coefficient must be also multiplied by 10-m. The homogeneity of the equation is not disturbed, since, is here supposed to be expressed in seconds. If we introduce logp as one of the elements, from the equation p a cos' p 11 162 THEORETICAL ASTRONOMY. we get 2 d logp = d1. — 2Ao tan dsp, or do - I- 2 d logp - 3/ tan, d,. (90) 0 Hence it appears that the coefficients of A logp are the same as those of A log a, but since p is also a function of (p, the coefficients of o( are changed; and if we denote by cos 8 ( and ) the values of \ d(p \ the partial differential coefficients when the element / is used in connection with 5p, we shall have, for the case under consideration, d(d d fd s dA cos dco- 3-tan(p cos'.-y-, in which s - 206264".8. If the values of the differential coefficients with respect to / and yn have not already been found, it will be advantageous to compute the values of dr, d dr and dv by dip dp d logp d logp means of the expressions which may be derived by substituting in the equations (15) the value of d/u given by (90), and then we may __ dcd da ndd_ compute directly the values of cos 8, cos a -, and d (po d logp' d' d logp In place of M,, it is often convenient to introduce L,, the mean longitude for the epoch; and since Lo-Mo + 71 we have dL0 = dM3o +dr dM + dM aw dg, and, when X is used, dLo = dMo 4- d + (1 - cos i) dg2. Instead of the elements g and i which indicate the position of the plane of the orbit, we may use b = sini sin, C = sin i cos Q, and the expressions for the relations between the differentials of b and c and those of i and g are easily derived. The cosines of the angles which the line of apsides or any other line in the orbit makes with the three co-ordinate axes, may also be taken as elements of the DIFFERENTIAL FORMULAE. 163 orbit in the formation of the equations for the variation of the geocentric place. 60. The equations (48), by writing 1 and b in place of A and A, respectively, will give the values of the differential coefficients of the heliocentric longitude and latitude with respect to x, y, and z. Combining these with the expressions for the differential coefficients of the heliocentric co-ordinates with respect to the elements of the orbit, we obtain the values of cos b Al and Ab in terms of the variations of the elements. The equations for dx, dy, and dz in terms of du, d2, and di, may also be used to determine the corrections to be applied to the co-ordinates in order to reduce them from the ecliptic and mean equinox of one epoch to those of another, or to the apparent equinox of the date. In this case, we have du - dr - d g. When the auxiliary constants A, B, a, b, &c. are introduced, to find the variations of these arising from the variations assigned to the elements, we have, from the equations (99),, cot A -tan a cos i, cot B- cot a cos i -sini cosec a tane, cot C cot a2 cos i - sin i cosec g cot e, in which i may have any value from 0~ to 180~. If we differentiate these, regarding all the quantities involved as variable, and reduce by means of the values of sin a, sin b, and sin c, we get cos i sin A dA -in2 d - sin sin i di, sur2a sin a cos e dB = s (cos cos - sin i sin e cos g ) d sin b sin B sin i sin Q2 + s (cos sin i cos -cos i sin ) di d sin b sin' b sin e d C si2 c (cos i sin -- sin i cos e cos g ) dR ~~~~~sin c and these+ by means Sof (101)n reduce toSi d and these, by means of (101)i, reduce to 164 THEORETICAL ASTRONOMY. cos 7 dA- si d g — sin A cot a di, sina cos e Cos c cos a dB- =o b c 2 dg - sin B cotb di + ib de, (91) sin cos b cos a dC= dQ~.dg~ -~sin C cot c di+ de. sin c sin2 c Let us now differentiate the equations (101)1, using only the upper sign, and the result is da z - sin i sin A d -- cos A di, db - sin i sin B dQ + cos B di + cos ccosec b de, de -- - sin i sin C d -+ cos C di - co b cosec c de. If we multiply the first of these equations by cot a, the second by cot b, and the third by cot, and denote by 20 the modulus of the system of logarithms, we get d log sin a- -, sin i cot a sin A d + Ao cot a cos A di, cos b cos c d log sin b = - A, sin i cot b sin B dQ -+- A cot b cos B di + o b e de, sin b cos b cos c d log sin c — A sin i cot c sin C da + 0 cot c cos Cdi —a sin2 ds. (92) The equations (91) and (92) furnish the differential coefficients of A, B, C, log sin a, &c. with respect to 2, i, and e; and if the variations assigned to 2, i, and e are so small that their squares may be neglected, the same equations, writing AA, A, Ai, &c. instead of the differentials, give:'he variations of the auxiliary constants. In the case of equations (92), if the variations of, i, and e are' expressed in seconds, each term of the second member must be divided by 206264.8, and if the variations of log sin a, log sin b, and log sin are required in units of the rnth decimal place of the logarithms, each term of the second member must also be divided by 10. If we differentiate the equations (81),, and reduce by means of the same equations, we easily find cos b dl - cos i sec b du + cos b d - sin b cos (I- d) di, (93 db sin i cos( — ) du + sin(l ) di, which determine the relations between the variations of the elements of the orbit and those of the heliocentric longitude and latitude. By differentiating the equations (88), neglecting the latitude of DIFFERENTIAL FORMUL2E. 165 the sun, and considering 2, 9, J, and O as variables, we derive, after reduction, R cos f dA -= - cos (- ) do, 1? (94) df - a - sin f sin (A~- O) d0, which determine the variation of the geocentric latitude and longitude arising from an increment assigned to the longitude of the sun. It appears, therefore, that an error in the longitude of the sun will produce the greatest error in the computed geocentric longitude of a heavenly body when the body is in opposition. 166 THEORETICAL ASTRONOMY. CHAPTER III. INVESTIGATION OF FORMULE FOR COMPUTING THE ORBIT OF A COMET MOVING IN A PARABOLA, AND FOR CORRECTING APPROXIMATE ELEMENTS BY THE VARIATION OF THE GEOCENTRIC DISTANCE. 61. THE observed spherical co-ofdinates of the place of a heavenly body furnish each one equation of condition for the correction of the elements of its orbit approximately known, and similarly for the determination of the elements in the case of an orbit wholly unknown; and since there are six elements, neglecting the mass,-which must always be done in the first approximation, the perturbations not being considered,-three complete observations will furnish the six equations necessary for finding these unknown quantities. Hence, the data required for the determination of the orbit of a heavenly body are three complete observations, namely, three observed longitudes and the corresponding latitudes, or any other spherical coordinates which completely determine three places of the body as seen from the earth. Since these observations are given as made at some point or at different points on the earth's surface, it becomes necessary in the first place to apply the corrections for parallax. In the case of a body whose orbit is wholly unknown, it is impossible to apply the correction for parallax directly to the place of the body; but an equivalent correction may be applied to the places of the earth, according to the formule which will be given in the next chapter. However, in the first determination of approximate elements of the orbit of a comet, it will be sufficient to neglect entirely the correction for parallax. The uncertainty of the observed places of these bodies is so much greater than in the case of well-defined objects like the planets, and the intervals between the observations which will be generally employed in the first determination of the orbit will be so small, that an attempt to represent the observed places with extreme accuracy will be superfluous. When approximate elements have been derived, we may find the distances of the comet from the earth corresponding to the three observed places, and hence determine the parallax in right ascension DETERMINATION OF AN ORBIT. 167 and in declination for each observation by means of the usual formulae. Thus, we have r p cos o' sin ( -- 0) aJ cos tan so' tan C = ta r cos ( -0)' 7p sin (,' sin ( -- 8) A sin r in which a is the right ascension, 8 the declination, J the distance of the comet from the earth, (n' the geocentric latitude of the place of observation, 6 the sidereal time corresponding to the time of observation, p the radius of the earth expressed in parts of the equatorial radius, and 7r the equatorial horizontal parallax of the sun. In order to obtain the most accurate representation of the observed place by means of the elements computed, the correction for aberration must also be applied. When the distance a is known, the time of observation may be corrected for the time of aberration; but if A is not approximately known, this correction may be neglected in the first approximation. The transformation of the observed right ascension and declination into latitude and longitude is effected by means of the equations which may be derived from (92), by interchanging a and i, a and 9, and writing -e instead of s. Thus, we have tan a tanN sin a cos (N- e) ) tan A ~ - tan a, (1) cos tan, tan - =tan (N- e ) sin, and also cos (N —- ) cos f sin A cos N cos S sin a which will serve to check the numerical calculation of A and /. Since cos / and cos 8 are always positive, cos, and cos a must have the same sign, thus determining the quadrant in which i is to be taken. 62. As soon as these preliminary corrections and transformations have been effected, and the times of observation have been reduced to the same meridian, the longitudes having been reduced to the 168 THEORETICAL ASTRONOMY. same equinox, we are prepared to proceed with the determination of the elements of the orbit. For this purpose, let t, t', t" be the times of observation, r, r', r" the radii-vectores of the body, and u, u', u" the corresponding arguments of the latitude, R, R', R" the distances of the earth from the sun, and 0, 0', 0, the longitudes of the sun corresponding to these times. Let [rr'] denote double the area of the triangle formed between the radii-vectores r, r' and the chord of the orbit between the corresponding places of the body, and similarly for the other triangles thus formed. The angle at the sun in this triangle is the difference between the corresponding arguments of the latitude, and we shall have [rr'] - rr' sin (u' - u), [rr"] rr" sin (a"' - ), (2) [r'r"] r'r" sin (" - a ), If we designate by x) y, z, x', y', z', x/, y", z" the heliocentric coordinates of the body at the times t, t', and t", we shall have IXY = r sin a sin (A -+ u),' - r' sin a sin (A + iu'), x"I - r" sin a sin (A + u"), in which a and A are auxiliary constants which are functions of the elements a and i and these elements may refer to any fundamental plane whatever. If we multiply the first of these equations by sin (u" —'), the second by - sin (" —tu), and the third by sin (a' - u), and add the products, we find, after reduction, x x' x" sin (" - u') -, sin (u"- ) +,- sin (u'- u) = 0, r' r ri which, by introducing the values of [rr'], [rr"], and [r' r"], becomes [r'r"] x - [rr"] x' + [rr'] " 0. If we put [r'r"] [rr"]' [r' ] (3) we get nx - x' + n"x"" 0. (4) In precisely the same manner, we find ny- y' + n'l"y = 0, nz' + n"z" =. (5) nz -- z' + nZ"1- - O. DETERMINATION OF AN ORBIT. 169 Since the coefficients in these equations are independent of the positions of the co-ordinate planes, except that the origin is at the centre of the sun, it is evident that the three equations are identical, and express simply the condition that the plane of the orbit passes through the centre of the sun; and the last two might have been derived from the first by writing successively y and z in place of x. Let A, A', A" be the three observed longitudes, /, j', f" the corresponding latitudes, and J, A', d" the distances of the body from the earth; and let J cos -=p, z' cos' — p, O z" cos P" =p", which are called curtate distances. Then we shall have x - p cos os - R cos, x' p' cos A' - R cos', y =p sin A - R sin 0, y' p' sin' - R' sin (0, z - p tan P, z' p' tan 1', x" =p" cos " - R" cos (",. y" = p sin'" - R" sinl ", z -= p" tan i", in which the latitude of the sun is neglected. The data may be so transformed that the latitude of the sun becomes 0, as will be explained in the next chapter; but in the computation of the orbit of a comet, in which this preliminary reduction has not been made, it will be unnecessary to consider this latitude which never exceeds 1", while its introduction into the formula would unnecessarily complicate some of those which will be derived. If we substitute these values of x, x., &c. in the equations (4) and (5), they become =- n (p cos - R cos 0) - (p' cos A' - R' cos') + n" (p" cos " - R" cos O"), 0 -= n (p sin - R sin ) - (p' sin' - R' sin Q') (6) + n" (p" sin " - R" sin 0 "), - 0 = np tan t - p' tan f' + n"p" tan fi". These equations simply satisfy the condition that the plane of the orbit passes through the centre of the sun, and they only become distinct or independent of each other when n and n"q are expressed in functions of the time, so as to satisfy the conditions of undisturbed motion in accordance with the law of gravitation. Further, they involve five unknown quantities in the case of an orbit wholly unknown, namely, in, in", p, o', and p"; and if the values of n and n1" are first found, they will be sufficient to determine p, p', and p". 170 THEORETICAL ASTRONOMY. The determination, however, of n and n1^ to a sufficient degree of accuracy, by means of the intervals of time between the observations, requires that p' should be approximately known, and hence, in general, it will become necessary to derive first the values of ni, n", and p'; after which those of p and p/l may be found from equations (6) by elimination. But since the number of equations will then exceed the number of unknown quantities, we may combine them in such a manner as will diminish, in the greatest degree possible, the effect of the errors of the observations. In special cases in which the conditions of the problem are such that when the ratio of two curtate distances is known, the distances themselves may be determined, the elimination must be so performed as to give this ratio with the greatest accuracy practicable. 63. If, in the first and second of equations (6), we change the direction of the axis of x from the vernal equinox to the place of the sun at the time t', and again in the second, from the equinox to the second place of the body, we must diminish the longitudes in these equations by the angle through which the axis of x has been moved, and we shall have o - n,(p cos(A- 0') —R os(' — 0)) -(p' cos (' - 0') -') + 1n" (p" cos (A" - 0')- I" cos (() " —')), 0 — n (p sin (-') + R sin ('- )) p' sin (A'- 0') + " (p" sin (-') — R" sin(0" — 0')), (7) O n (psin (A'- A) + J sin(O —')) - R'sin(0' —') - n" (p" sin (A" - 1') - R" sin (0"- A')), 0 - np tan 1 - p' tan i' + n"p" tan /". If we multiply the second of these equations by tan', and the fourth by - sin (2'- 0'), and add the products, we get 0; n"p" (tan /' sin (" - 0') - tan A" sin (A'- 0')) -n"R" sin ( "-') tan I'+ np (tan /' sin ( - 0') - tan A sin (A'-O')) + nR sin ( (0'-' ) tan'. (8) -Let us now denote double the area of the triangle formed by the sun and two places of the earth corresponding to R and R' by [RR'], and we shall have [RR'] -= RR'sin (0 — ), and similarly [RR"] - RR" sin (0"- 0), [R'R"] ='R" sin (0" —'). ORBIT OF A HEAAENLY BODY. 1.71 Then, if we put - lRII1 C-11 N — t[R'R"] " [RR'(9 [RB"]' R ](9) [RRS] I[RIR"]' we obtain R" sin (O" —') R sin ((' — )) N. Substituting this in the equation (8), and dividing by the coefficient of p", the result is tan /' sin (- 0') -tan p' sin (" - o') \XC t 7i tan /" sin (A'- 0')- tan /i' sin (A"-(0') - N \ __ R sin (~'- ) tan fi' ( n tan " sin (A' - (') - tan i' sin (A" -') Let us also put tan 9i' sin ( -') — tan sin (A' — 0') tan /" sin (' - 0)') - tan i' sin (2" - (')' ( - tan _ sin (G' — 0) tan i' - taln " sin (A'- )') - tan i' sin (l"- (')' and the preceding equation reduces to P__f, ]JI-='p +Jr ( s- jlI"R. (11) We may transform the values of 11' id Md" so as to be better adapted to logarithmic calculation with the ordinary tables. Thus, if w' denotes the inclination to the ecliptic of a great circle passing through the second place of the comet and the second place of the sun, the longitude of its ascending node will be 0', and we shall have sin (A'- 0') tan w' tan /'. (12) Let j0,,o" be the latitudes of the points of this circle corresponding to the longitudes A and 2/', and we have, also, tan io =sin ( -')tan w', (13) tan A/i" sin (A" --') tan w'. Substituting these values for tan f', sin(2 - 0) and sin(A" - 0') in the expressions for il' and M/", and reducing, they become sin (/o' - fi) cos /" cos o sin (i" - lio") cos OS' ( 4) (14) 1== ^ tanw zsc(0- os/" cos sin(Q- 0)sin(/"/i" 172 THEORETICAL ASTRONOMY. When the value of -, has been found, equation (11) will give the relation between p and p" in terms of known quantities. It is evident, however, from equations (14), that when the apparent path of the comet is in a plane passing through the second place of the sun, since, in this case, 3-=0 and f"= /9o", we shall have M'= 0 and M"= oo. In this case, therefore, and also when - /9 and " —- P0' are very nearly 0, we must have recourse to some other equation which may be derived from the equations (7), and which does not involve this indetermination. >i; It will be observed, also, that if, at the time of the middle observation, the comet is in opposition or conjunction with the sun, the values of M' and M" as given by equation (14) will be indeterminate in form, but that the original equations (10) will give the values of these quantities provided that the apparent path of the comet is not in a great circle passing through the second place of the sun. These values are M'- = sin(A- 0') " sin(('- 0) sin (A" - 0')' sin (A"- 0') Hence it appears that whenever the apparent path of the body is nearly in a plane passing through the place of the sun at the time of the middle observation, the errors of observation will have great: influence in vitiating the resulting values of 31' and Ml'; and to obviate the difficulties thus encountered, we obtain from the third of equations (7) the following value of p":_n Osin ('- A) p n' sin (A" - A') - Bsin(0 A') —A sin (0' — A) -+ I sin (0" A') sin (A"- ) We may also eliminate p between the first and fourth of equations (7). If we multiply the first by tan', and the second by -cos (A- 0), and add the products, we obtain 0 = n"p" (tan f' cos (A" -') - tan A" cos (' - 0')) - n"R" tan /' cos (0"- 0') + np (tan g' cos (A -0') - tan ft cos (A' —')) - nR tan A' cos ((0' — () + R' tan 3', from which we derive ORBIT OF A HEAVENLY BODY. 173 n tan 9' cos ( - GO') -tan cos (A'-') PP'-' tan f" cos( (- 0')- tan' cos (Af- 0') (16) n,' R" tan f' cos ( "- Q') +, R tan' cos ('- 0)), R'tan' tan ri cos (' -') -tan f' cos (A" - 0') Let us now denote by I' the inclination to the ecliptic of a great circle passing through the second place of the comet and that point of the ecliptic whose longitude is 0'- 90~, which will therefore be the longitude of its ascending node, and we shall have cos ( - (0') tan I' - tan f'; (17) and, if we designate by 3, and 3,, the latitudes of the points of this circle corresponding to the longitudes A and A", we shall also have tan, = cos ( - (') tan I', 18 tan A,, cos (" - ()') tan I'. Introducing these values into equation (16), it reduces to n sin (C, - f) cos i' cos l,, P" P -n" I sin (/" - /,,) cos A cos /, (19) tanll' cos/ os ( f cos(O')+ n r (,, + C (,) G)) n - sin(/n-)i) from which it appears that this equation becomes indeterminate when the apparent path of the body is in a plane passings through that point of the ecliptic whose longitude is equal to the longitude of the second place of the sun diminished by 90~. In this case we may use equation (11) provided that the path of the comet is not nearly in the ecliptic. When the comet, at the time of the second observation, is in quadrature with the sun, equation (19) becomes indeterminate in form, and we must have recourse to the original equation (16), which does not necessarily fail in this case. When both equations (11) and (16) are simultaneously nearly indeterminate, so as to be greatly affected by errors of observation, the relation between p and p" must be determined by means of equation (15), which fails only when the motion of the comet in longitude is very small. It will rarely happen that all three equations, (14), (15), and (16), are inapplicable, and when such a case does occur it will indicate that the data are not sufficient for the determination of the elements of the orbit. In general, equation (16) or (19) is to be used when the motion of the comet in latitude is considerable, and equation (15) when the motion in longitude is greater than in latitude. 174 THEORETICAL ASTRONOMY. 64. The formulae already derived are sufficient to determine the relation between p" and p when the values of n and n" are known, and it remains, therefore, to derive the expressions for these quantities. If we put k (t'-) =, k(t)" -'), (20) k (t" — t) - r', and express the values of x, y, x, y", y z" in terms of x', y', z' by expansion into series, we have C _Zx' 1Z" dX t T2 1 dt k + 1.2'dt2 k2 1.2.3 dt' k3 +.'" (21) dx' ~r 1 d2x' T2 1 d3x' 9 ( ) " + d t. kd + J.2 2. a. + 1.2.3 dt3. +&c., and similar expressions for y, y", z, and z". We shall, however, take the plane of the orbit as the fundamental plane, in which case z,', and z" vanish. The fundamental equations for the motion of a heavenly body relative to the sun are, if we neglect its mass in comparison with that of the sun, d2x' kx2. t2 + 7-0~ d2' k~ dt2 r,3 If we differentiate the first of these equations, we get d3x' 3k2x' dr' k2 dx' dt3 r'* dt r3' dt Differentiating again, we find d4' r' 12k2 1 dr' 2 3k2 d2r' 6k2 dr' dx' W dt4 in6g r \ dt r''dt" * dto2 f'4 d dt' dy Writing y instead of x, we shall have the expressions for - and dY4 Substituting these values of the differential coefficients in equadtO tions (21), and the corresponding expressions for y and y", and putting ORBIT OF A HEAVENLY BODY. 175 r-1 I i- 2 3 d4. 12. dr2 3 d2r... 22 r", ~ _Yk t 4r6 5( dt) k2r'4' dt2 /) T- T,3 f"1 dr' b kr'3 4k24 dt (22.y — q -. dy, -, y" a, y + b". ^-^.2 3' dr2 a"a -b d+ 1 1 dr,,, 3 d dtx From these equations we easily derive y' x -- - - - - x'dy'-y'dx' Tt' y -' -' y' = b dy'- yd (23) y' -x"y t-(ab" - a"b) x dy'-y'dx' The first members of these equations are double the areas of the triangles formed by the radii-vectores and the chords of,the orbit between the places of the comet or planet. Thus, y'x - x'y = [rr'], y"x' - x"y' -[rr"], y"x - x"y = [rr"], (24) and x'dy' - y'dx' is double the area described by the radius-vector xf d~_- -l yd'x during the element of time dt, and, consequently, dy - y'dx' dt double the areal velocity. Therefore we shall have, neglecting the mass of the body, x'dy' - y' dx' - dt =2f =oVr, in which p is the semi-parameter of the orbit. The equations (23), therefore, become [rr'] = bk 1/, [r',] = b"k 7, [rr"] = (ab" + a"b) k 1'/ Substituting for a, b, at, b" their values from (22), we find, since 73:7 -- T t, 176 THEORETICAL ASTRONOMY. *. /2 l'S" dr'd d [rr'] — =" p (1- " kr' dt ) 6 f3 4 4 V1 r%~34 kr'4 dt. 1' [rr"] - - + r t )' (25) Er"] -=' 1 6, + 4 kr' 4 T.... [r'r"] [r'] From these equations the values of n = [rr" and n" [ r] may be derived; and the results are. 1 ( + (T4-7) 1T (7T + T..-I) dr ) n + 6 rf3 4 Ikr" dt',(1A-6 r'~ kr'4 (26) -7' ~~-~1 (r + -T) r( (7T + r~ - r2) dr' 26 1t + 1 3 kr'4 d' which values are exact to the third powers of the time, inclusive. In the case of the orbit of the earth, the term of the third order, dR' being multiplied by the very small quantity dt' is reduced to a superior order, and, therefore, it may be neglected, so that in this case we shall have, to the same degree of approximation as in (26), ~~~~~N" -~~~~~(27) N =- t ( +6 ( I+, ) *) [r'r"] From the equations (26) or from (25), since-= -'] we find =rl -, r' + 12 r3r"3 +" dr' of 4, in the case of an orbit wholly unknown, can be determined 9l only by successive approximations. In the first approximation to the elements of the orbit of a heavenly body, the intervals between the observations will usually be small, and the series of terms of (28) will converge rapidly, so that we may take n - T n" I' ORBIT OF A HEAVENLY BODY. 177 and similarly N 1 NV" -- ~' Hence the equation (11) reduces to pt M'p. (29) It will be observed, further, that if the intervals between the observations are equal, the term of the second order in equation (28) vanishes, and the supposition that -,,, is correct to terms of the third order. It will be advantageous, therefore, to select observations whose intervals approach nearest to equality. But if the observations available do not admit of the selection of those which give nearly equal intervals, and these intervals are necessarily very unequal, it will be more accurate to assume n N n"- N" and compute the values of N and N" by means of equations (9), since, according to (27) and (28), if r' does not differ much from R', the error of this assumption will only involve terms of the third order, even when the Values of r and r" differ very much. Whenever the values of p and p" can be found when that of their ratio is given, we may at once derive the corresponding values of r and r", as will be subsequently explained. The values of r and r" may also be expressed in terms of r' by means of series, and we have dr' T 1 d2r' "2 r 7=v-d-t-k+^. 2.; —&c., r " r' +dt k. + M dt2 + &c. from which we derive r _ T+ " dr' k dt' neglecting terms of the third order. Therefore,?': k (r" — r~ =di~ T; (30) 12 178 THEORETICAL ASTRONOMY. and when the intervals are equal, this value is exact to terms of the fourth order. We have, also, -- r dr' r + r" - 2r' + dt which gives r' -- (r + r") -- (r" - r)-,. ~(31) Therefore, when r and r" have been determined by a first approxidr' mation, the approximate values of r' and dt are obtained from these n equations, by means of which the value of -, may be recomputed from equation (28). We also compute N R'R"sin (" - (') (32) N" - RR' sin ((' - 0)' on N and substitute in equation (11) the values of -, and No thus found. If we designate by M the ratio of the curtate distances p and p", we have P n W P In the numerical application of this, the approximate value of p will be used in computing the last term of the second member. In the case of the determination of an orbit when the approximate elements are already known, the value of -, may be computed from n /r" sin (v"~-v') n- rr' sin (v'- v)' (34) N and that of,N from (32); and the value of M derived by means of these from (33) will not require any further correction. 65. When the apparent path of the body is such that the value of M', as derived from the first of equations (10), is either indeterminate or greatly affected by errors of observation, the equations (15) and (16) must be employed. The last terms of these equations may be changed to a form which is more convenient in the approximations to ithe value of the ratio of p" to p. Let Y, Y', Y" be the ordinates of the sun when the axis of ORBIT OF A HEAVENLY BODY. 179 abscissas is directed to that point in the ecliptic whose longitude is A', and we have Y =-R sin(O -A'), Y' R' sin (O' - ), Y- =R" sin (0"- I'). Now, in the last term of equation (15), it will be sufficient to put nN n N"' and, introducing Y, Y', Y", it becomes ( N Y- n- Y' + Y" cosee (A" — ). (35) It now remains to find the value of,' From the second of equations (26) we find, to terms of the second order inclusive, =- 1 6,3 We have, also, R'3 and hence 6 7T~ ( +,, ) ( R') n"~ N" G ~\I Therefore, the expression (35) becomes N sin (A"-A') j(NY —Y' + N" Y"+'T (T'+ r") - ) Y But, according to equations (5), NY- Y' + N" Y" -0, and the foregoing expression reduces to ~+ rr+,,) ( 1 1 ) R' sin ("' —') since Y' = R' sin (0'- A'). Hence the equation (15) becomes, n sin (A' - A) rt 1 1 \R'sin (A'- ) p P l ^'sin(A -A) 6T - (I +T" )(3 B" ssin (-A') (36) n" sin (A" —)!)` 180 THEORETICAL ASTRONOMY. If we put n sin (A'- ) nA" sin (A"- A')' n sin(A' - ) R 1 )' we have -M MoF. (37) Let us now consider the equation (16), and let us designate by X, X', X" the abscissas of the earth, the axis of abscissas being directed to that point of the ecliptic for which the longitude is 0', then X R Jcos( -'), X' = R', X" - R" cos (O" - 0'). It will be sufficient, in the last term of (16), to put n N n"' N' and for,- its value in terms of N" as already found. Then, since NX- X' + N"X" 0, this term reduces to,~' (i+ ~,,) ( 1 1) R' tan/~' - r )' - an cos 0-0 tan P" cos ('-')- tan' os (A"-' and if we put,n tan f' cos ( --') -tan P cos (- i- Q') IO ~n" tan fI" cos (' — 0') - tan P3' cos (t" - 0')' (38) W_ i-' (I1 1\ _ tan fi' ^' (T + )r'3 R'3tan cos" 0')-tan' cos ((a'-O')'' the equation (16) becomes M- Mo'F'. (39) P In the numerical application of these formulae, if the elements are not approximately known, we first assume n Tr when the intervals are nearly equal, and ORBIT OF A HEAVENLY BODY. 181 n N n~ N' as given by (32), when the intervals are very unequal, and neglect the factors F and F'. The values of p and p" which are thus obtained, enable us to find an approximate value of r', and with this a more exact value of ~7 may be found, and also the value of F or F,. n " Whenever equation (11) is not materially affected by errors of observation, it will furnish the value of I1 with more accuracy than the equations (37) and (39), since the neglected terms will not be so great as in the case of these equations. In general, therefore, it is to be preferred, and, in the case in which it fails, the very circumstance that the geocentric path of the body is nearly in a great circle, makes the values of F and F' differ but little from unity, since, in order: that the apparent path of the body may be nearly in a great circle, r' must differ very little from R'. 66. When the value of M has been found, we may proceed to determine, by means of other relations- between p and p", the values of the quantities themselves. The co-ordinates of the first place of the earth referred to the third, are X,-=R cos " - R cos 0, (40) y,= R" sin 0"- R sin 0. If we represent by g the chord of the earth's orbit between the places corresponding to the first and third observations, and by G the longitude of the first place of the earth as seen from the third, we shall have x, g cos G, y, = g sin G, and, consequently, R" cos ("- 0)-R g- coso(G- 0), (41 R" sin (0" -0 ) -g sin (G-0 ). If 4 represents the angle at the earth between the sun and comet at the first observation, and if we designate by w the inclination to the ecliptic of a plane passing through the places of the earth, sun, and comet or planet for the first observation, the longitude of the ascending node of this plane on the ecliptic will be 0, and we shall have, in accordance with equations (81),, cos + -- cos 9 cos (A - 0), sin. cos w - cos f3 sin ( - 0), sin 4 sin w - sin f, 182 THEORETICAL ASTRONOMY. from which tan w tan sin ( - 0)' ~~~tan w-~ =,~^,(42) n tan (A -)) tan. -- cos w Since cos / is always positive, cos b and cos(A- 0) must have the same sign; and, further, 4 cannot exceed 180~. In the same manner, if w" and 4lf represent analogous quantities for the time of the third observation, we obtain tan f" tan w" tan " sin (" - 0")' tan(" — 0" tan ~tn( (43) cos W" cos "' = cos ft" cos (" - 0"). We also have r2 2 +_ R2 - 2R cos,, which may be transformed into 2- (p sec -- R co 4)2 + R2 sin2; (44) and in a similar manner we find'=~- (p"o sec R" - -" cos t")2 + Rl2 sin~' ". (45) Let x designate the chord of the orbit of the body between the first and third places, and we have 2 = (X"- )2 + (" -- y y)2 + (Z- z)2 But x p cos - R cos 0, y- =p sin A - R sin 0, z =p tan f, and, since p" = 2p, x" = Mp cos A" - R" cos 0", y"-l Mp sin " - R" sin 0", " - Mp tan A" from which we derive, introducing g and G, x" - = Mp c os 2"- p cos - G, "- y = Mp sin A"- p sin A - g sin G, z" z - = Mp tan " — p tan P. Let us now put ORBIT OF A HEAVENLY BODY. 183 Mp cos A" - p cos A ph c'os C cos H, Mp sin A"- p sin A - ph cos C sin H, (46) Mp tan P"- p tan = ph sin C. Then we have x"- x - ph cos C cos H- g cos G, y"- y ph cos Csin H- g sin G, z" z -= ph sin C. Squaring these values, and adding, we get, by reduction, 2 = h2 -2gph cos cos(G -H) + g2; (47) and if we put cos C cos (G - H) = cos (, (48) we have 2 (ph -- g cos )2 + g2 sin. (49) If we multiply the first of equations (46) by cos A" and the second by sin A", and add the products; then multiply the first by sin lA, and the second by cos A", and subtract, we obtain h cos C cos (H - A") M- cos (" - ), h cos C sin (H — ") = sin (A" - i), (50) h sin - M tan " tan i, by means of which we may determine h, C, and H. Let us now put' g sin =- A, R sin - B h cos - b, R" sin 4" - B", h s b", (51) g cos - bR cos - c, g cos v - b"R" cos 4" — c", ph - g cos (p = d, and the equations (44), (45), and (49) become ^^T/d+, 2+' r- d ( + B2, 2 (52) r'= 4(l $- c )ii + B2 2 The equations thus derived are independent of the form of the orbit, and are applicable to the case of any heavenly body revolving around the sun. They will serve to determine r and r" in all cases in which the unknown quantity d can be determined. If p is known, 184 THEORETICAL ASTRONOMY. d becomes known directly; but in the case of an unknown orbit, these equations are applicable only when p or d may be determined directly or indirectly from the data furnished by observation. 67. Since the equations (52) involve two radii-vectores r and rl and the chord x joining their extremities, it is evident that an additional equation involving these and known quantities will enable us to derive d, if not directly, at least by successive approximations. There is, indeed, a remarkable relation existing between two radiivectores, the chord joining their extremities, and the time of describing the part of the orbit included by these radii-vectores. In general, the equation which expresses this relation involves also the semitransverse axis of the orbit; and hence, in the case of an unknown orbit, it will not be sufficient, in connection with the equations (52), for the determination of cl, unless some assumption is made in regard to the value of the semi-transverse axis. For the special case of parabolic motion, the semi-transverse axis is infinite, and the resulting equation involves only the time, the two radii-vectores, and the chord of the part of the orbit included by these. It is, therefore, adapted to the determination of the elements when the orbit is supposed to be a parabola, and, though it is transcendental in form, it may be easily solved by trial. To determine this expression, let us resume the equations k(t- T) - = tan.v + - tan3 - v and, for the time t", k (t" — T) v"- ( = ) tan Iv v+ 1 tan3 Iv". Subtracting the former from the latter, and reducing, we obtain 3k(t -t) sin (v" — v) I r" cos- (v"- v) l/2 q - cq cos cos, cos and, since r q sec2 Iv, this gives 3k (t" -t) sin I (v"- v) /r r" + cos (V"-v )/'. ( ~r~=~'-_=~,/~^-~ r+r +cos~-(v /- r/ ". (53) 1/2 1q / But we have, also, from the triangle formed by the chord x and the radii-vectores r and r", x2 - r+ - rT2 - 2rr" cos (v"- v) = (r + r")2 4rr" cos2 I (" - v). PARABOLIC ORBIT. 185 Therefore, os (r(- v) + r ) (r + r" — x) (-21/rr Let us now put r + r" -+ = n, r + r" — X - n m and n being positive quantities. Then we shall have r+ r' = (2 + n2), 2 cos ~ (v" - v) l/rr" = — mn; and, since m and n are always positive, it follows that the upper sign must be used when v"- v is less than 180~, and the lower sign when v"-v is greater than 180~. Combining the last equation with (53), the result is 3k (t" t) =- si ("v) +(m+ n2 + n mn). (55) 1/2q Now we have sin -I- (v" v) sin v" cos - sin v" cos.cos Squaring this, and reducing, we get sin2 ~ (v" - v) -- cos2 -lv + c vos 2' cos cos (v v), or, introducing r and q, q q mn sin (v"- v)- q- q r r Therefore, 1/2q sin 1 (v "-v)2' 7 (m +- n). Introducing this value into equation (55), we find 6k (t" - t) =- m3 n. Replacing in and n by their values expressed in terms of r, r", and n, this becomes 6k(t" ) ( + r" + ) - (r + r" - t)h (56) the upper sign being used when v"-v is less than 180~. This equation expresses the relation between the time of describing any parabolic arc and the rectilinear distances of its extremities from each other and from the sun, and enables us at once, when three of these quantities are given, to find the fourth, independent of either the 186 THEORETICAL ASTRONOMY. perihelion distance or the position of the perihelion with respect to the arc described. 68. The transcendental form of the equation (56) indicates that, when either of the quantities in the second member is to be found, it must be solved by successive trials; and, to facilitate these approximations, it may be transformed as follows:Since the chord z can never exceed r + r", we may put + sin r', (57) r --- s" and, since x, r, and r" are positive, sin' must always be positive. The value of'r must, therefore, be within the limits 0~ and 180~. From the last equation we obtain cos2r' (r + r/)2 _ 2 and substituting for x2 its value given by -- (r + r")2 - 4rr" cos2 (v"- v), this becomes co 4rr" cos2.(v" -v) cos - ~(r + r ~)" Therefore, we have 21/rr1 cos' = cos ( "- v) r+, (58) and also tan /' - = (59) 21/rr" cos (v" - -v) Hence it appears that when v"- v is less than 180~, r' belongs to the first quadrant, and that when v" - v is greater than 180~, cos r' is negative, and r' belongs to the second quadrant. If we introduce?' into the expressions for m2 and n2 they become ma (r + r") (1 + sin -'), n2 (r + r") (1 - sin), which give m=a (r + Y") (cos l' + sin 1,r)2 n2 (r + r") (~ cos r' sin r')2; and, since r' is greater than 90~ when v"'-v exceeds 180~, the equation (56) becomes t +' = (cos.' + sin Ir,)3 (cos /r' -- sin,-')3. (r + r")Is PARABOLIC ORBIT. 187 From this equation we get 6rt - = 6 cos + 2' sin 2 si, (r + r")2 2 or 61rf - 6 sin My -4 sin3'; (r + r/ )2 and this, again, may be transformed into 6r =3( sin s ) 4(in)3 (60) Let us now put si sin x — / (61) 1/2 or sin /' =V/2 sin x, and we have 3~J ~(~ - r3 sin x - 4 sin3 x = sin 3. (62) V/2(r+ r")-2 When v" v is less than 180', r' must be less than 90~, and hence, in this case, sin x cannot exceed the value a, or x must be within the limits 00 and 30~. When v" -v is greater than 180~, the angle r' is within the limits 90~ and 180~, and corresponding to these limits, the values of sin x are, respectively, I and V/2. Hence, in the case that v" - v exceeds 180~, it follows that x must be within the limits 30~ and 45~. The equation 3 sin 3x 1/2 (i +r")i is satisfied by the values 3x and 180~ - 3x; but when the first gives x less than 15~, there can be but one solution, the value 180" -3x being in this case excluded by the condition that 3x cannot exceed 135~. When x is greater than 150, the required condition will be satisfied by 3x or by 180~ - 3x, and there will be two solutions, corresponding respectively to the cases in which v" -v is less than 180~, and in which v" -v is greater than 180~. Consequently, when it is not known whether the heliocentric motion during the interval t - t is greater or less than 180~, and we find 3x greater than 450, the same data will be satisfied by these two different solutions. In practice, however, it is readily known which of the 188 THEORETICAL ASTRONOMY. two solutions must be adopted, since, when the interval t" - t is not very large, the heliocentric motion cannot exceed 180~, unless the perihelion distance is very small; and the known circumstances will generally show whether such an assumption is admissible. We shall now put 2vt 2~~' (63) (r -+ r") 2 and we obtain sin 3x - (64) 1/8 We have, also, sin / -/2 sin x, and hence cos' =- i/ - 2 sin2 x = I/cos 2x. Therefore sin r'- 2] sin x V cos 2x, and, since =- (r + r") sin r', we have X = 23 (r + r") sin x 1/cos 2x. If we put 3 sin x /cos 2x, (65) sin 3 the preceding equation reduces to --- 2 I. (66) /' (r + r") From equation (64) it appears that 7] must be within the limits 0 and 1X/8. We may, therefore, construct a table which, with V as the argument, will give the corresponding value of y, since, with a given value of (, 3x may be derived from equation (64), and then the value of p. from (65). Table XI. gives the values of C corresponding to values of a from 0.0 to 0.9. 69. In determining an orbit wholly unknown, it will be necessary to make some assumption in regard to the approximate distance of the comet from the sun. In this case the interval t" — t will generally be small, and, consequently, x will be small compared with r and r". As a first assumption we may take r = I, or r + r" = 2, and z =1 and then find x from the formula X' 1/2. PARABOLIC ORBIT. 189 With this value of x we compute d, r, and r" by means of the equations (52). Having thus found approximate values of r and r", we compute z by means, of (63), and with this value we enter Table XI. and take out the corresponding value of /i. A second value for z is then found from (66), with which we recompute r and r", and proceed as before, until the values of these quantities remain unchanged. The final values will exactly satisfy the equation (56), and will enable us to complete the determination of the orbit. After three trials the value of r + r" may be found very nearly correct from the numbers already derived. Thus, let y be the true value of log (r + r"), and let Ay be the difference between any assumed or approximate value of y and the true value, or yo - y + Ay. Then if we denote by yo' the value which results by direct calculation from the assumed value y,, we shall have Yo' -Yo -f (YO) f(Y + AY). Expanding this function, we have yo' - Yo f(y) + A ay + B Ay2 + &c. But, since the equations (52) and (66) will be exactly satisfied when the true value of y is used, it follows that f(Y)- o0, and hence, when Ay is very small, so that we may neglect terms of the second order, we shall have yo' - yo A ay = A (yo - y). Let us now denote three successive approximate values of log (r + s") by yo; Yo', yoI, and let Yo -o a, y= - yo - c a'; then we shall have a A (y -), a' = A (yo' - Y). Eliminating A from these equations, we get y (a' - a) = a'yO -ay, from which Y - a = y a'" (67) a-a 190 THEORETICAL ASTRONOMY. Unless the assumed values are considerably in error, the value of y or of log (r + r") thus found will be sufficiently exact; but should it be still in error, we may, from the three values which approximate nearest to the truth, derive y with still greater accuracy. In the numerical application of this equation, a and a' may be expressed in units of the last decimal place of the logarithms employed. The solution of equation (56), to find t" -t when x is known, is readily effected by means of Table VIII. Thus we have -- = sin 3x, V/2 (9' + r")P and, when r' is less than 90~, if we put sin 3x N —,sin' we get'- = 1/2 Nsinr' (r+ r") (68) or C —' /2 f N r + r". When r' exceeds 900, we put N' - sin 3x, and we have T- v'/2 N' (r + r")- (69) in which log i/2 9.6733937. With the argument r' we take from Table VIII. the corresponding value of N or N', and by means of these equations T =- k (t" - t) is at once derived. The inverse problem, in which r' is known and x is required, may also be solved by means of the same table. Thus, we may for a first approximation put X- =T 1/2, and with this value of x compute d, r, and r". The value of r' is then found from sin r' - rr + r"' and, the table gives the corresponding value of N or N'. A second approximation to x will be given by the equation 3' + " N-r PARABOLIC ORBIT. 191 or by 3' sill sn 3 in which log ~-=0.3266063. Then we recompute d, r, and r", and proceed as before until x remains unchanged. The approximations are facilitated by means of equation (67). It will be observed that d is computed from d-= ~V/2 A2, and it should be known whether the positive or negative sign must be used. It is evident from the equation d = ph - g cos (p, since p, h, and g are positive quantities, that so long as sp (which must be within the limits 0~ and 180~) exceeds 90~, the value of d must be positive; and therefore o( must be less than 90~, and g cos So greater than ph, in order that d may be negative. The equation (4;) shows that when x is greater than g, we have g cos 9 < lpph, and hence d must in this case be positive. But when X is less than g, either the positive or the negative value of d will answer to the given value of (o, and the sign to be adopted must be determined from the physical conditions of the problem. If we suppose the chords g and x to be proportional to the linear velocities of the earth and comet at the middle observation, we have, the eccentricity of the earth's orbit being neglected, 2 which shows that x is greater than g, and that d is positive, so long as r' is less than 2. The comets are rarely visible at a distance from the earth which much exceeds the distance of the earth from the sun, and a comet whose radius-vector is 2 must be nearly in opposition in order to satisfy this condition of visibility. Hence cases will rarely occur in which d can be negative, and for those which do occur it will generally be easy to determine which sign is to be used. However, if d is very small, it may be impossible to decide which of the two solutions is correct without comparing the resulting elements with other and more distant observations. 192 THEORETICAL ASTRONOMY. 70. When the values of r and r" have been finally determined, as just explained, the exact value of d may be computed, and then we have d + g cos ( h' (70) p" = Mp, from which to find p and p". According to the equations (90)1, we have r cos b cos ( (0) = p cos (A -0)-R, r cos b sin (I - 0) = p sin (A- 0), (71) r sin b = p tan A, and also r" cos b" cos (I" - 0") - p" cos (A" - ") - -", r" cos b" sin (I"- 0") -p" sin (A" - 0"), (72) r" sin b" " tan A", in which i and I" are the heliocentric longitudes and b, b" the corresponding heliocentric latitudes of the comet. From these equations we find r, r", 1, 1", b, and b"; and the values of r and r" thus found, should agree with the final values already obtained. When 1" is less than I, the motion of the comet is retrograde, or, rather, when the motion is such that the heliocentric longitude is diminishing instead of increasing. From the equations (82),, we have -- tan i sin (l - ~) - tan b, - tan i sin (/" — ) = tan b", which may be written 4- tan i (sin ( - x) cos (x - a) + sin (x - ) cos /- x)) = tan b, ~ tan i (sin (l"- x) cos (x -- g) + sin (x - 2) cos ("- x)) = tan b. Multiplying the first of these equations by sin (" - x), and the second by - sin (1- x), and adding the products, we get -+- tan i sin (x - 2) sin(l" - 1) = tan b sin (" - x) - tan b" sin (I - x); and in a similar manner we find - tan i cos (x - ) sin (I"- l) - tan b" cos ( — x) - tan b cos (l"- x). Now, since x is entirely arbitrary, we may put it equal to I, and we have PARABOLIC ORBIT. 193 tan i sill (I - ) = tan b, tan b" - tan b cos(l"- ) (74) tan i cos (l — ) + sin-(- 1) the lower sign being used when it is desired to introduce the distinction of retrograde motion. The formnule will be better adapted to logarithmic calculation if we put x =(l" + 1), whence l- x (l"- ) and 1-x -2(1 1"); and we obtain tan i sin (41 (P+ 1)- ) - ~ )" -- -b) tn' 2 cos b cos b" cos (l" -1)' ( sin (b"- b) ~tani cos(l(" + - 1 )) - ) 2 cos b cos b" sin 1 ( )' These equations may also be derived directly from (73) by addition and subtraction. Thus we have ~ tan i (sin (l"- 2 ) + sin (l- )) tan b" + tan b, ~ tanl (sin (- ) -sin (- ) tan b" - tan b; and, since sin(l"- Q) + sinll( - 2) 2 sin ~ ("+ 1- 22) cos ( — 1), sin ("- ) - sin (I - ) - 2 cos ("+- 1 2 ) sinl (1- 1), these become tan i sin (1 (1+ 1) 2 ) 4 - +- tan 6) COS. __, A~ ~(tan V - tan 5) tan cos( (1" + ) ) - - sin "-l) sin ( -l which may be readily transformed into (75). However, since b and b" will be found by means of their tangents in the numerical application of equations (7v) and (72), if addition and subtraction logarithms are used, the equations last derived will be more convenient than in the form (75). As soon as; and i have been computed from the preceding equations, we have, for the determination of the arguments of the latitude u and u", tan (1 — g6) tan (l"-' ) utan ~ = u M tan e'=: -- -(77) cos i cos Now we have =- v +-, in which (o =n — 7 in the case of direct motion, and w- = - 13 194 THEORETICAL ASTRONOMY. when the distinction of retrograde motion is adopted; and we shall have Ut" -- U = - v v, and, consequently, x2 r2 + r"2 - 2rr" cos (u" - u), (78) or 2 -= (r" - r cos ('t" -_ ))2 + r2 sin2 ("tt- u). (79) The value of x derived from this equation should agree with that already found from (66). We have, further, r q sec - ( — w), r" = q sec2- (" - w), or COS (, COS (- ) = By addition and subtraction, we get, from these equations, - (cos I (" — ) + cos A (-w)) — + v_ (COS, ) COS Y(, I 1 1 (c (os(' -- COS (' - - )) -,/ ~ / ~, from which we easily derive vq 4 2_ 1 1 2 O co ls (4('~ +a) 1-0) COS 4-v- ( -t6)= J + X-, V;+1' (80) 2 _s 1 sin A ( (tu"+ qu) - W) sin 4 (u"- ) —But 1 1 / r" 4 jr \ l/Tr 1" t/rrr\ r \'r J and if we put tan (45~ +') - since 4 I~ will not differ much from 1,' will be a small angle; and we shall have, since tan (450 + 0') - cot (45~ + 0') = 2 tan 20', 4r 4 I r;vr'r, -=2 tan 20', + r"= + 4- 2 sec 20', PARABOLIC ORBIT. 195 Therefore, the equations (80) become 1 i (1("+ ) —) tan 20' sin ( 1(i("+ u) -- w) s / ~ Vq s (" - u) Vrr" (81) 1 sec 20' -cos +) ) -~+ U) -) 1/ q - cos (Vt - u) CVrr from which the values of q and 0o may be found. Then we shall have, for the longitude of the perihelion T- -+ a, when the motion is direct, and 7r - 0, when i unrestricted exceeds 90~ and the distinction of retrograde motion is adopted. It remains now to find T, the time of perihelion passage. We have V =- 1 w, V" - t - o. With the resulting values of v and v" we may find, by means of Table VI., the corresponding values of M (which must be distinguished from the symbol M already used to denote the ratio of the curtate distances), and if these values are designated by 31 and M3", we shall have M M" t T —M g T -Mor T t M t" M" m t C in which m 3=, and log C0 9.9601277. When v is negative, the qg corresponding value of 1Mis negative. The agreement between the two values of T will be a final proof of the accuracy of the numerical calculation. The value of T when the true anomaly is small, is most readily and accurately found by means of Table VIII., from which we derive the two values of N and compute the corresponding values of T from the equation 2 T= t -3 Nr sin v, in which log -= 1.5883273. When v is greater than 900, we de 196 THEORETICAL ASTRONOMY. rive the values of N' from the table, and compute the corresponding values of T from 2 T=t - N'r 71. The elements q and T may be derived directly from the values of r, r", and x, as derived from the equations (52), without first finding the position of the plane of the orbit and the position of the orbit in its own plane. Thus, the equations (80), replacing u and,u" by their values v + w and v +'-'; become 2 1 sin -1 (v" + v) sin 4(V"-V) =C4- 4 2^~ ~~~~~~~ ^ ^ ~ (82) S os 4 (V' + V) Cs ( ) = + Adding together the squares of these, and reducing, we get 1 1 2 1 + -~ _= COS V (V~ V) 1 i' - cos// "q sin2 ~ (v" - v) or rr" sin2 1 (v" - v) r" + r - 2 V/rr" cos A (v" - v) Combining this equation with (59), the result is rr" sin2 (v" v) r r+ "-x cot " and hence, since x = (r + r") sin r', q =-sin (v" v) cot r. (83h We have, further, from (78), 2 (r" - r)2 + 4rr" sin2 (v"' ), from which, putting sinv --, (84) we derive 21/rr" cos v T eer sin q (v"' v). (85) Therefore, the equation (83) becomes PARABOLIC ORBIT. 197 q -- (r + r") cos2 Iry' cos2, (86) by means of which q is derived directly from r, r", and x, the value of v being found by means of the formula (84), so that cos is positive. When T' cannot be found with sufficient accuracy from the equation sin r - r + r"' we may use another form. Thus, we have. r + r+r + - r+ ~" ~ - 1 + sin — r r- 1 - sin y'- which give, by division, tan (450 + -r') r + r" + x (87) r+ r~In a similar manner, we derive tan (450 + I) - + (r_ -r) (88) In order to find the time of perihelion passage, it is necessary first to derive the values of v and v". The equations (59) and (85) give, by multiplication, tan (v" - v) == tan r' cos v, (89) from which v" - v may be computed. From (82) we get IrJ/ tan i (v" + v) tan v (v" - v) - r + 1 If we put tan —, t (90) r' this equation reduces to tan 4 (v" + v) - tan (' - 45~) cot 4 (v" - v), (91) and the equations (81) give, also, tan I (v" + v) = cot 4 (v" - v) sin 2', either of which may be used to find v" + v. 198 THEORETICAL ASTRONOMY. From the equations cosv _ 1 cos'v" _ /q. r' /q If- __ by multiplying the first by sin tv" and the second by - sin v, ad'ding the products and reducing, we easily find sin - (v"- v) sin v cos (v" -v) 1 Hence we have sI. v cot (v"- v) 1 /q 2 - r /r" sin. (v"' —v)'( (92) COSa iV oo o- =, which may be used to compute q, v, and v" when v" -v is known. When (v" - v) and - (v" + v), and hence v" and v, have been determined, the time of perihelion passage must be found, as already explained, by means of Table VI. or Table VIII. It is evident, therefore, that in the determination of an orbit, as soon as the numerical values of r, r", and x have been derived from the equations (52), instead of completing the calculation of the elements of the orbit, we may find q and T, and then, by means of these, the values of r' and v' may be computed directly. When this has been effected, the values of n and n" may be found from (3), or that of -, from (34). Then we compute p by means of the first of equations (70), and the corrected value of M froni (33), or, in the special cases already examined, from the equations (37) and (39). In this way, by successive approximations, the determination of parabolic elements from given data may be carried to the limit of accuracy which is consistent with the assumption of parabolic motion. In the case, however, of the equations (37) and (39), the neglected terms may be of the second order, and, consequently, for the final results it will be necessary, in order to attain the greatest possible accuracy, to derive P from (15) and (16). When the final value of M has been found, the determination of the elements is completed by means of the formulae already given. PARABOLIC ORBIT. 199 72. EXAMPLE.-TO illustrate the application of the formulae for the calculation of the parabolic elements of the orbit of a comet by a numerical example, let us take the following observations of the Fifth Comet of 1863, made at Ann Arbor:Ann Arbor M. T. d 1864 Jan. 10 6h 57m 208.5 19^ 14m 4S.92 + 340 6' 27".4, 13 6 11 54.7 19 25 2.84 36 36 52.8, 16 6 35 11.6 19 41 4.54 +39 41 26.9. These places are referred to the apparent equinox of the date and are already corrected for parallax and aberration by means of approximate values of the geocentric distances of the comet. But if approximate values of these distances are not already known, the corrections for parallax and aberration may be neglected in the first determination of the approximate elements of the unknown orbit of a comet. If we convert the observed right ascensions and declinations into the corresponding longitudes and latitudes by means of equations (1), and reduce the times of observation to the meridian of Washington, we get Washington M. T.,B 1864 Jan. 10 7h 24m 3s 297~ 53' 7".6 + 550 46' 58".4, 13 6 38 37 302 57 51.3 57 39 35.9, 16 7 1 54 310 31 52.3 + 59 38 18.7. Next, we reduce these places by applying the corrections for precession and nutation to the mean equinox of 1864.0, and reduce the times of observation to decimals of a day, and we have t 10.30837, A = 297~ 52' 51".1, f + 55~ 46' 58".4, t' - 13.27682, 2' = 302 57 34.4, P' - 57 39 35.9, "- 16.29299,."- 310 31 35.0, " — + 59 38 18.7. For the same times we find, from the American Nautical Almanac, - 290 6' 27".4, log R 9.992763, 0' 293 7 57.1, log R' 9.992830, 0" -296 12 15.7, log R" - 9.992916, which are referred to the mean equinox of 1864.0. It will generally be sufficient, in a first approximation, to use logarithms of five decimals; but, in order to exhibit the calculation in a more complete form, we shall retain six places of decimals. Since the intervals are very nearly equal, we may assume 200 THEORETICAL ASTRONOMY. n T N Then we have ta _\ tn t' tan' sin ( O0') tan3 sn (2'-') t - f t tan i" sin (2'- 0') - tan i' sin (" - 0')' and g sin (G -0) R" sin (" - 0), g cos (G - 0) - R" cos ( " -- 0) - R; h cos: cos (H — A") i- - cos (A" 2), h cos sin (H- )") sin (A"- ), h sin C _ M tan fl" tan f1; from which to find M1, G, g, g,, and h. Thus we obtain log M —- 9.829827, H — 94~ 24' 1".8, G - 22~ 58' 1".7, — 40 28 21.9, log g 9.019613, log h 9.^8532. d" cos fi Since ~ — i- 0.752, it appears that the comet, at the time cos of these observations, was rapidly approaching the earth. The quadrants in which G - G and II- " must be taken, are determined by the condition that g and h cos C must always be positive. The value of 2M should be checked by duplicate calculation, since an error in this will not be exhibited until the values of 2' and -' are computed from the resulting elements. Next, from cos - os 1 Cos ( - ), cos CoS' cos (" — "), cos s -c cos cos (G - H), we compute cos 4, cos )", and cos (; and then from gsin 5 =- A, h cos i = b, h Cos f" Ri sin, - B, hc b9 PR" sin -" =- B", g cos p - bR cos, - c, g cos ~ - b"R" cos "' = c!, we obtain A, B, B", &c. It will generally be sufficiently exact to find sin 4 and sin 4" from cos 4 and cos "; but if more accurate values of 4 and 4" are required, they may be obtained by means of the equations (42) and (43). Thus we derive log A = 9.006485, log B _ 9.912052,." 9.933366, log b-= 9.438524, log b" 62387, =- 0.125067, e..150561 NUMERICAL EXAMPLE. 201 Then we have 22 17 v-7'+ d=r A/(-r':' d/r + r" 2' d -+/ f A2 r - + B'2 r,) + Bus, - b" +'btf from which to find, by successive trials, the values of r, r", and x, that of u being found from Table XI. with the argument ^. First, we assume log x log T' /2 9.163132, and with this we obtain log r = 9.913895, log r" = 9.938040, log (r + r") = 0.227165. This value of log (r + r") gives = 0.094, and from Table XI. we find log, 0.000160. Hence we derive log x 9.200220, log r = 9.912097, log r"- 9.935187, log(r + r") = 0.224825. Repeating the operation, using the last value 6f log(r + r"), we get log x 9.201396, log r 9.912083, log r" 9.935117, log (r + r") = 0.224783. The correct value of log(r + r") may now be found by means of the equation (67). Thus, we have, in units of the sixth decimal place of the logarithms, a -224825 - 227165 - 2340, a' = 224783 - 224825 = - 42, and the correction to the last value of log(r + r") becomes a} a" P- _ - -- -- -- 0.8. Therefore, log (r + r") = 0.224782, and, recomputing I, #A, X, r, and r", we get, finally, log x - 9.201419, log r = 9.912083, log r" 9.935116, log (r + r") = 0.224782. The agreement of the last value of log (r + r") with the preceding one shows that the results are correct. Further, it appears from the 202 THEORETICAL ASTRONOMY. values of r and r" that the comet had passed its perihelion and was receding from the sun. By means of the values of r and r" we might compute approxicdr mate values of 7r' and t from the equations (30) and (31), and then n N a more approximate value of,, from (28), that of N. being found from (32). But, since r' differs but little from R', the difference bewen an N between -~ and y'' is very small, so that it is not necessary to consider the second term of the second member of the equation (33); and, since the intervals are very nearly equal, the error of the assumption nt T is of the third order. It should be observed, however, that an error in the value of M affects H, C, h, and hence also A, b, b", c, and c", aind the resulting value of p may be affected by an error which considerably exceeds that of JM. It is advantageous, therefore, to select observations which furnish intervals as nearly equal as possible in order that the error of JM may be small, otherwise it may become necessary to correct JM and to repeat the calculation of r, r", and x. We may also compute the perihelion distance and the time of perihelion passage from r, r", and x by means of the equations (86), (89), and (91) in connection with Tables VI. and VIII. Then r' and v' may be computed directly, and the complete expression for Mi may be employed. In the first determination of the elements, and especially when the corrections for parallax and aberration have been neglected, it is unnecessary to attempt to arrive at the limit of accuracy attainable, since, when approximate elements have been found, the observations may be more conveniently reduced, and those which include a longer interval nmiy be used in a more complete calculation. Hence, as soon as r, r, -and z have. been found, the curtate distances are next determined, and then the elements of the orbit. To find p and o", we have d + 0.122395, the positive sign being used since x is greater than g, and the formula d +- g cos,, h' give log p = 9.480952, log p" - 9.310779. NUMERICAL EXAMPLE. 203 From these values of p and,op, it appears that the comet was very near the earth at the time of the observations. The heliocentric places are then found by means of the equations (71) and (72). Thus we obtain I = 106~ 40' 50".5, b -+ 330 1' 10".6, logr -9.912082, 1" 112 31 9.9, b"= + 23 55 5.8, logr" 9.935116. The agreement of these values of r and r" with those previously found, checks the accuracy of the calculation. Further, since the heliocentric longitudes are increasing, the motion is direct. The longitude of the ascending node and the inclination of the orbit may now be found by means of the equations (74), (75), or (76); and we get a = 304~ 43' 11".5, i 64~ 31' 21".7. The values of ue and i," are given by the formulae tan(l —) t ) tan (l"- ) tanu ~ - —, tai u'" cos i cos u and I - S being in the same quadrant in the case of direct motion. Thus we obtain u 1420 52' 12'.4, t"- 153~ 18' 49".4. Then the equation 2 (r" — r cos ('" - t))2 + r sin" (u" - u) gives log x - 9.201423, and the agreement of this value of x with that previously found, proves the calculation of Q, i, A, and u". From the equations 4 / r" tan (45~ + — 0') 1-,/ -1 - si ( u+ )a -tan 20' -- sin ( 1 (,a" + u) - w) - ( V/q ~sin ({u" —' u)V;r' Ic (( + ) sec 20' cos I (u'"- ) V rr' we get 0' -0~ 22' 47".4, = 115~ 40' 6".3, log q = 9.887378. Hence we have X- w - =+ 60~ 23' 17".8, 204 THEORETICAL ASTRONOMY. and v -- u- o 270 12' 6".1, v'" =- - wz= 37 38' 43".1. Then we obtain logm =- 9.9601277 - log q 0.129061 and, corresponding to the values of v and v", Table VI. gives log M — 1.267163, log M" = 1.424152. Therefore, for the time of perihelion passage, we have M T-t - t - 13.74364, and T — t" -- ='- 19.72836. m The first value gives T 1863 Dec. 27.56473, and the second gives T- Dec. 27.56463. The agreement between these results is the final proof of the calculation of the elements from the adopted value of pl P If we find T by means of Table VIII., we have log N= 0.021616, log N" - 0.018210, and the equation 2 2 T — t- 3 N3 r sin v=t" - 3 N"r"2 sin v", in which log - 1.5883273, gives for T the values Dec. 27.56473 and Dec. 27.56469. Collecting together the several results obtained, we have the following elements: T - 1863 Dec. 27.56471 Washington mean time. 60' 23' 17".8 -- 6304 43 1.85 Ecliptic and Mean 7- 304 43 11.5 64 31 21.7 Equinox 1864.0, / - 64 31 21.7) log q 9.887378. 3Motion Direct. 73. The elements thus derived will, in all cases, exactly represent the extreme places of the comet, since these only have been used in finding the elements after p and p" have been found. If, by means NUMERICAL EXAMPLES. 205 of these elements, we compute n and n", and correct the value of 3JT the elements which will then be obtained will approximate nearer the true values; and each successive correction will furnish more accurate results. When the adopted value of 31 is exact, the resulting elements must by calculation reproduce this value, and since the computed values of 2, A", P, and /" will be the same as the observed values, the computed values of )/ and 9' must be such that when substituted in the equation for M, the same result will be obtained as when the observed values of') and f' are used. But, according to the equations (13) and (14), the value of M depends only on the inclination to the ecliptic of a great circle passing through the places of the sun and comet for the time t', and is independent of the angle at the earth between the sun and comet. Hence, the spherical coordinates of any point of the great circle joining these places of the sun and comet will, in connection with those of the extreme places, give the same value of M, and when the exact value of M has been used in deriving the elements, the computed values of 2' and f' must give the same value for w' as that which is obtained from observation. But if we represent by a' the angle at the earth between the sun and comet at the time tt, the values of it derived by observation and by computation from the elements will differ, unless the middle place is exactly represented. In general, this difference will be small, and since w' is constant, the equations cos'= cos /' cos (' - 0), sin' cosw' - cos i' sin (A' - 0'), (93) sin 4' sin w' - sin f', give, by differentiation, cos' d-' = cos w' sec f' d'., d3' = sin w' cos (I' - )') d4'. (94) From these we get cos fi' di' tan (2' - 0!) dfi' sin'' which expresses the ratio of the residual errors in longitude and latitudce, for the middle place, when the correct value of M has been used. Whenever these conditions are satisfied, the elements will be correct on the hypothesis of parabolic motion, and the magnitude of the final residuals in the middle place will depend on the deviation of the actual orbit of the comet from the parabolic form. Further, 206 THEORETICAL ASTRONOMY. when elements have been derived from a value of M which has not been finally corrected, if we compute A' and 9' by means of these elements, and then tan w'- tan f' sin s - (D')' (95) the comparison of this value of tan w' with that given by observation will show whether any further correction of M is necessary, and if the difference is not greater than what may be due to unavoidable errors of calculation, we may regard M as exact. To compare the elements obtained in the case of the example given with the middle place, we find v' 32 31' 13".5, ui' 148~ 11' 19".8, log r' 9.922836. Then from the equations tan (I' — ) = cos itan iu', tan b' tan i sin (t'- g), we derive 1'- 109~ 46' 48".3, b' - 28~ 24' 56".0. By means of these and the values of 0' and R', we obtain' = 302~ 57' 41".1,' = 57 39' 37".0; and, comparing these results with the observed values of A' and /', the residuals for the middle place are found to be Comp. - Obs. cos i AA' -= + 3".6, A - + 1".1. The ratio of these remaining errors, after making due allowance for unavoidable errors of calculation, shows that the adopted'value of M is not exact, since the error of the longitude should be less than that of the latitude. The value of w' given by observation is log tan w' 0.966314, and that given by the computed values of i' and i' is log tan w' = 0.966247. The difference being greater than what can be attributed to errors of calculation, it appears that! the value of M requires further cor NUMERICAL EXAMPLES. 207 rection. Since the difference is small, we may derive the correct value of M by using the same assumed value of -, and, instead of the value of tanw' derived from observation, a value differing as much from this in a contrary direction as the computed value differs. Thus, in the present example, the computed value of log tan w' is 0.000067 less than the observed value, and, in finding the new value of M, we must use log tan w'- 0.966381 in computing P0 and Pri involved in the first of equations (14). If the first of equations (10) is employed, we must use, instead of tan' as derived from observation, tan [' =tan w' sin (A' -'), or log tan A' - 0.966381 - log sin ('- 0') 0.198559, the observed value of A' being retained. Thus we derive log M- 9.829586, and if the elements of the orbit are computed by means of this value, they will represent the middle place in accordance with the condition that the difference between the computed and the observed value of tan w' shall be zero. A system of elements computed with the same data from log 1M- 9.822906 gives for the error of the middle place, C. -O. cos pi a' - - 1' 26".2, a'- -40".1. If we interpolate by means of the residuals thus found for two values of M, it appears that a system of elements computed from log M- 9.829586 will almost exactly represent the middle place, so that the data are completely satisfied by the hypothesis of parabolic motion. The equations (34) and (32) give log - = 0.006955, log N -0.006831, and from (10) we get log M' = 9.822906, log M" = 9.663729,. 208 THEORETICAL ASTRONOMY. Then by means of the equation (33) we derive, for the corrected value of M, log M= 9.829582, which differs only in the sixth decimal place from the result obtained by varying tan w and retaining the approximate values ~t!- Nr"' 74. When the approximate elements of the orbit of a comet are known, they may be corrected by using observations which include a longer interval of time. The most convenient method of effecting this correction is by the variation of the geocentric distance for the time of one of the extreme observations, and the formula which may be derived for this purpose are applicable, without modification, to any case in which it is possible to determine the elements of the orbit of a comet on the supposition of motion in a parabola. Since there are only five elements to be determined in the case of parabolic motion, if the distance of the comet from the earth corresponding to the time of one complete observation is known, one additional complete observation will enable us to find the elements of the orbit. Therefore, if the elements are computed which result from two or more assumed values of D differing but little from the correct value, by comparison of intermediate observations with these different systems of elements, we may derive that value of the geocentric distance of the comet for which the resulting elements will best represent the observations. In order that the formulae may be applicable to the case of any fundamental plane, let us consider the equator as this plane, and, supposing the data to be three complete observations, let A, A', A" be the right ascensions, and D, D', D" the declinations of the sun for the times t, t', t". The co-ordinates of the first place of the earth referred to the third are x = R" cos D" cos A" - R cos D cos A, y R" cos D" sin A" - R cos D sin A, z = " sin D" -- R sin D. If we represent by g the chord of the earth's orbit between the places for the first and third observations, and by G and IK, respectively, the right ascension and declination of the first place of the earth as seen from the third, we shall have x g cos K cos G, y = g cos Ksin G, z = g sin K, VARIATION OF THE GEOCENTRIC DISTANCE. 209 and, consequently, g cos K cos (G - A) = R" cos D" cos (A"- A) - R cos D, g cos K sin (G - A) R" cos D" sin (A" - A), (96) g sin K R" sin D" - R sin D, from which g, K, and G may be found. If we designate by x,, y,, z, the co-ordinates of the first place of the comet referred to the third place of the earth, we shall have x, = cos ao - g cos K cos G, y, ~= D os sin g cos K sin G, z, = si g sin sinK. Let us now put x, - h' cos." cos H',, = h' cos V' sin H', -,- h' sin C', and we get 7' cos V' cos (H' G) = cos s cos(- G) + g cos K, h' cos' sin (H' — G) - cos 8 sin (a -G), (97) h' sin! = d sin +- g sin K, from which to determine H', g', and h'. If we represent by p'1 the angle at the third place of the earth between the actual first and third places of the comet in space, we obtain cos'= cos' cos H' cos a" co o s cos sin H' cos a" sin a"+- sin C' sin a", or cos' =- cos V' cos " cos ("' - H') + sin V' sin a"; (98) and if we put e sinf - sin 8", e cosf cos s 8 cos (a"' -') this becomes cos' = e cos (' -f). (99) Then we shall have 2 = h, + D42 2h_'" cOs' or -2 (" — h' cos'g)2 + h12 sin2, (100) in which J" is the distance of the comet from the earth corresponding to the last observation. We have, also, from equations (44) and (45), r2 _ (J -R cos 4)2 + R2 sin',, rl2 = ('" -- /" cos,"')2 + R'"2 sin12 4~, 14 210 THEORETICAL ASTRONOMY. in which ~ is the angle at the earth between the sun and comet at the time t, and 4"' the same angle at the time t". To find their values, we have cos = cos D cos 8 cos (a -A) + sin D sin, (102) cos "-= cos D" cos a" cos (a"- A") + sin D" sin s", which may be still further reduced by the introduction of auxiliary angles as in the case of equation (98). Let us now put h' sin' - C, h' cos ~' - c, R sin B=, R cos - = b, (103) B" sin 4"= B", " rcos "'= b", and we shall have X /(" -- c) + C2 -r; V - ( - b) + B, (104) r"= l/("~ — b')' + B"2 These equations, together with (56), will enable us to determine J" by successive trials when J is given. We may, therefore, assume an approximate value of a" by means of the approximate elements known, and find r" from the last of these equations, the value of r having been already found from the assumed value of J. Then x is obtained from the equation 2vT, V/r + r" p/ being found by means of Table XI., and a second approximation to the value of /" from a" = G /X2 _ C2. (105) The approximate elements will give a" near enough to show whether the upper or lower sign must be used. With the value of J" thus found we recompute r" and z as before, and in a similar manner find a still closer approximation to the correct value of J". A few trials will generally give the correct result. When A" has thus been determined, the heliocentric places are found by means of the formulae r cos b cos (I - A) =- cos 8 cos (a - A) - R cos D, r'cos b sin ( - A) = A cos 8 sin (a -- A), (106) r sin b =- A sin - R sin D; VARIATION OF THE GEOCENTRIC DISTANCE. 211 r" cos b" cos (" - A") a- S cos " cos (al - A") - R cos D", r" cos b" sin (I" - A") A" cos A" sin (a" - A"), (107) r" sin b" = i" sin " - R" sin D", in which b, b", 1, I" are the heliocentric spherical co-ordinates referred to the equator as the fundamental plane. The values of r and r" found from these equations must agree with those obtained from (104). The elements of the orbit may now be determined by means of the equations (75), (77), and (81), in connection with Tables VI. and VIII., as already explained. The elements thus derived will be referred to the equator, or to a plane passing through the centre of the sun and parallel to the earth's equator, and they may be transformed into those for the ecliptic as the fundamental plane by means of the equations (109),. 75. With the resulting elements we compute the place of the comet for the time t' and compare it with the corresponding observed place, and if we denote the computed right ascension and declination by a0' and 8', respectively, we shall have a +t = a, + d', -, in which a' and d' denote the differences between computation and observation. Next we assume a second value of J, which we represent by a +8J, and compute the corresponding system of elements. Then we have a' + a" = C't' + d" - 0', a" and d" denoting the differences between computation and observation for the second system of elements. We also compute a third system of elements with the distance A - 8A, and denote the differences between computation and observation by a and d; then we shall have a =f(d - ), a =f(D), a" =f(d +- ), and similarly for d, d', and d". If these three numbers are exactly represented by the expression' w h i + x in which- a - x is the general value of the argument, since the values of a, a', and a" will be such that the third differences may be neglected, this formula may be assumed to express exactly any value of the function corresponding to a value of the argument not differing 212 THEORETICAL ASTRONOMY. much from d, or within the limits x = — A and x + -, the assumed values d - 8/, a, and J + 8A being so taken that the correct value of A shall be either within these limits or very nearly so. To find the coefficients mn, n, and o, we have m - n + o ~ a, n -a', om + n +- o - a", whence n = c-, n = (a"- -a), o - (a" + a) -a. Now, in order that the middle place may be exactly represented in right ascension, we must have ( ) +n( - ) -m -O, from which we find x _ _ -_ (n - (Ve- 4mo) =p, di 20 or x -p8a = 0. In the same manner, the condition that the middle place shall be exactly represented in declination, gives x — p'ad - 0. In order that the orbit shall exactly represent the middle place, both conditions must be satisfied simultaneously; but it will rarely happen that this can be effected, and the correct value of x must be found from those obtained by the separate conditions. The arithmetical mean of the two values of x will not make the sum of the squares of the residuals a minimum, and, therefore, give the most probable value, unless the variation of cos 8' tao', for a given increment assigned to A, is the same as that of A8'. But if we denote the value of x for which the error in a' is reduced to zero by x', and that for which A- =0, by x", the most probable value of x will be n2X + -nf 2X" x_ + 12, (108) in which n (= (a - a) and n' = (d" — d). It should be observed that, in order that the differences in right ascension and declination shall have equal influence in determining the value of x, the values of a, a', and a" must be multiplied by cos o'. The value of 8A is most conveniently expressed in units of the last decimal place of the logarithms employed. NUMERICAL EXAMPLE. 213 If the elements are already known so approximately that the first assumed value of J differs so little from the true value that the second differences of the residuals may be neglected, two assumptions in regard to the value of J will suffice. Then we shall have o = 0, and hence mnz- a, n - a" a' The condition that the middle place shall be exactly represented, gives the two equations (a"- a') x + a'd O0, (d"- d')x + d'cJ =- 0. The combination of these equations according to the method of least squares will give the most probable value of x, namely, that for which the sum of the squares of the residuals will be a minimum. Having thus determined the most probable value of x, a final system of elements computed with the geocentric distance a + x, corresponding to the time t, will represent the extreme places exactly, and will give the least residuals in the middle place consistent with the supposition of parabolic motion. It is further evident that we may use any number of intermediate places to correct the assumed value of A, each of which will furnish two equations of condition for the determination of x, and thus the elements may be found which will represent a series of observations. 76. EXAMPLE.-The formulae thus derived for the correction of approximate parabolic elements by varying the geocentric distance, are applicable to the case of any fundamental plane, provided that a,, A, D, &c. have the same signification with respect to this plane that they have in reference to the equator. To illustrate their numerical application, let us take the following normal places of the Great Comet of 1858, which were derived by comparing an ephemeris with several observations made during a few days before and after the date of each normal, and finding the mean difference between computation and observation: Washington M. T. a d 1858 June 11.0 141~ 18' 30".9 + 24~ 46' 25".4, July 13.0 144 32 49.7 27 48 0.8, Aug. 14.0 152 14 12.0 + 31 21 47.9, which are referred to the apparent equinox of the date. These places are free from aberration. 214 THEORETICAL ASTRONOMY. We shall take the ecliptic for the fundamental plane, and converting these right ascensions and declinations into longitudes and latitudes, and reducing to the ecliptic and mean equinox of 1858.0, the times of observation being expressed in days from the beginning of the year, we get t - 162.0, A = 1350 51' 44".2, =+ 90 6' 57".8, t' 194.0, )' -137 39 41.2,' = 12 55 9.0, t" 226.0, " — 142 51 31.8, P " + 18 36 28.7. From the American Nautical Almanac we obtain, for the true places of the sun, o = 80~ 24' 32".4, logR 0.006774, 0' -110 55 51.2, log R' 0.007101, Q" = 141 33 2.0, log R" 0.005405, the longitudes being referred to the mean equinox 1858.0. When the ecliptic is the fundamental plane, we have, neglecting the sun's latitude, D- 0, and we must write 2 and P in place of o and A, and 0 in place of A, in the equations which have been derived for the equator as the fundamental plane. Therefore, we have g cos (G - ) R" cos (0" - 0) -- R, g sin (G - ) = R" sin(0" - 0); cos 4 =C c os f (- c 4 " os (A " -- O") R cos - b, R" cos A"= b", R sin - B=, " sin " = B", from which to find G, g, b, B, b", and Bi, all of which remain unchanged in the successive trials with assumed values of A. Thus we obtain G 2010 7' 57".4, log B -9.925092, b + 0.568719, log g = 0.013500, log B" = 9.510309, b" = 0.959342. Then we assume, by means of approximate elements already known, log A == 0.397800, and' from h' cos V' cos (H' - G) - cos C cos (A - G) + g, h' cos C' sin (H' - G) = cost sin (A - G), h' sin' = A sin f, we find H', C', and'. These give H' = 153~ 46' 20".5, = +- 7~ 24' 16".4, log h' 0.487484. NUMERICAL EXAMPLE. 215 Next, from cos P'- cos' os s " cos (A" - H') + sin' sin f", h' cos ('- c, h'sin'=' C, we get log C= 9.912519, c =-+ 2.961673; and from r = ( -- b)2 + B2, we find log r - 0.323446. Then we have a" = ~ V X2 C2 r"- (" - b) + B"2, 2~ 2~ ar' - k (e'l —), (r + ")' X 1 (r + r")/ r -t- r'; from which to find z", r", and x. First, by means of the approximate elements, we assume log 4" -0.310000, which gives log r" = 0.053000, and hence we have - 0.3783, log p- = 0.002706, log x = 0.090511. With this value of x we obtain from the expression for a", the lower sign being used, since A" is less than c, log " - 0.309717. Repeating the calculation of r", J,6 and x, and then finding d" again, the result is log D" = 0.309647. Then, by means of the formula (67), we may find the correct value. Thus we have, in units of the sixth decimal place, a - 309717 - 310000 - - 283, a' = 309647 - 309717 — 70, and for the correction to the last result for log A" we have ~/2 aa a -23. Therefore, log 4" = 0.309624. By means of this value we get log r" = 0.052350, log x = 0.090628, 216 THEORETICAL ASTRONOMY. and this value of x gives, finally, log i" -- 0.309623, log r" - 0.052348. The heliocentric places of the comet are now found from the equations (71) and (72), writing J cosf and A" cos fi for p and p", respectively. Thus we obtain I = 159~ 43' 14".2, b -+ 10~ 50' 14".0, log r 0.323447, r" 144 17 47.8, b" + 35 14 28.7, log r" =0.052347. The agreement of these results for r and r" with those already obtained, proves the accuracy of the calculation. Since the heliocentric longitudes are diminishing, the motion is retrograde. Then from (74) we get a _ 165~ 17' 30".3, i 63~ 6' 32".5; and from tan (1 tan (- Q) tan u" - tan t(l- an cos i cos i we obtain u = 12~ 10' 12".6, u" =40~ 18' 51".2, the values of - t and I1- 2 being in the same quadrant when the motion is retrograde. The equation (79) gives log - 0.090630, which agrees with the value already found. The formulae (81) give w 129~ 6' 46".3, log q 9.760326, and hence we have - u - - -116~ 56' 33".7, v" = u" - - 88~ 47' 55".1, from which we get T= 1858 Sept. 29.4274. From these elements we find log r' = 0.212844, v' =- 107~ 7' 34".0, u' = 21~ 59' 12".3, and from tan (I' - ) - cos i tan u', tan b'= - tan i sin (I'- ), we get i' 154~ 56' 33".4, b' - + 19~ 30' 22".1. NUMERICAL EXAMPLE. 217 By means of these and the values of 0' and R', we obtain'- =137~ 39' 13".3, f' - + 12~ 54' 45".3, and comparing these results with observation, we have, for the error of the middle place, C. -O. cos i' A' — 27".2, fP' - 23".7. From the relative positions of the sun, earth, and comet at the time t" it is easily seen that, in order to diminish these residuals, the geocentric distance must be increased, and therefore we assume, for a second value of /, log D = 0.398500, from which we derive H' - 153~ 44' 57".6, V' + 7~ 24' 26".1, log h' = 0.488026, log C- 9.912587, log c 0.472115, logr -0.324207, log A" - 0.311054, log r" = 0.054824, log X - 0.089922. Then we find the heliocentric places I -159~ 40' 33".8, b += 10~ 50' 8".6, logr -0.324207, 1" 144 17 12.1, b" =-+ 35 8 37.8, log r" =- 0.054825, and from these, = 165~ 15' 41".1, i 63~ 2' 49".2, ut 12 10 30.8, " — 40 13 26.0, w 128 54 44.4, logq 9.763620, T=- 1858 Sept. 29.8245, log r'-0.214116, v' - 106~ 55' 43".8, u' 21~ 59' 0".6, l'f- 154 53 32.3, b' = 19 29 31.9, i'= 137 39 39.7, - =+12 55 2.9. Therefore, for the second assumed value of d, we have C. O. cos A'' — 1".5, A4' -6".1. Since these residuals are very small, it will not be necessary to make a third assumption in regard to A, but w3 may at once derive the correction to be applied to the last assumed value by means of the equations (109). Thus we have a' -1.5, a" = - 27.2, d' - 6.1, d"- 23.7, ~ log z = - 0.000700, 218 THEORETICAL ASTRONOMY. and, expressing ~ log A in units of the sixth decimal place, these equations give 25.7x - 1050 - 0. 17.6x - 4270 - 0. Combining these according to the method of least squares, we get 105 X 2.57 + 427 X 1.76 _ (2.57)2 + (1.76)2 + Hence the corrected value of log J is log J = 0.398500 + 0.000106 - 0.398606. With this value of log J the final elements are computed as already illustrated, and the following system is obtained:T- 1858 Sept. 29.88617 Washington mean time. -= 36~ 22' 36".9 ) o-165 15 24.8 a 165 15 24.8 A Mean Equinox 1858.0. i 63 2 14.2 log q = 9.764142 Motion Retrograde. If the distinction of retrograde motion is not adopted, and we regard i as susceptible of any value from 0~ to 1800, we shall have - 294~ 8' 12".7, i=116 57 45.8, the other elements remaining the same. The comparison of the middle place with these final elements gives the following residuals:C.-O. cos f AA = + 0".2, Ai - 4".3. These errors are so small that the orbit indicated by the observed places on which the elements are based differs very little from a parabola. When, instead of a single place, a series of intermediate places is employed to correct the assumed value of X, it is best to adopt the equator as the fundamental plane, since an error in oc or a will affect both A and 9; and, besides, incomplete observations may also be used NUMERICAL EXAMPLE. 219 when the fundamental plane is that to which the observations are directly referred. Further, the entire group of equations of condition for the determination of x, according to the formulse (109), must be combined by multiplying each equation by the coefficient of x in that equation and taking the sum of all the equations thus formed as the final equation from which to find x, the observations being supposed equally good. 220 THEORETICAL ASTRONOMY. CHAPTER IV. DETERMINATION, FROM THREE COMPLETE OBSERVATIONS, OF THE ELEMENTS OF THE ORBIT OF A HEAVENLY BODY, INCLUDING THE ECCENTRICITY OR FORM OF THE CONIC SECTION. 77. THE formule which have thus far been derived for the determination of the elements of the orbit of a heavenly body by means of observed places, do not suffice, in the form in which they have been given, to determine an orbit entirely unknown, except in the particular case of parabolic motion, for which one of the elements becomes known. In the general case, it is necessary to derive at least one of the curtate distances without making any assumption as to the form of the orbit, after which the others may be found. But, preliminary to a complete investigation of the elements of an unknown orbit by means of three, complete observations of the body, it is necessary to provide for the corrections due to parallax and aberration, so that they may be applied in as advantageous a manner as possible. When the elements are entirely unknown, we cannot correct the observed places directly for parallax and aberration, since both of these corrections require a knowledge of the distance of the body from the earth. But in the case of the aberration we may either correct the time of observation for the time in which the light from the body reaches the earth, or we may consider the observed place corrected for the actual aberration due to the combined motion of the earth and of light as the true place at the instant when the light left the planet or comet, but as seen from the place which the earth occupies at the time of the observation. When the distance is unknown, the latter method must evidently be adopted, according to which we apply to the observed apparent longitude and latitude the actual aberration of the fixed stars, and regard this place as corresponding to the time of observation corrected for the time of aberration, to be effected when the distances shall have been found, but using for the place of the earth that corresponding to the time of observation. It will appear, therefore, that only that part of the calculation of the DETERMINATION OF AN ORBIT. 221 elements which involves the times of observation will have to be repeated after the corresponding distances of the body from the earth have been found. First, then, by means of the apparent obliquity of the ecliptic, the observed apparent right ascension and declination must be converted into apparent longitude and latitude. Let'2 and p, respectively, denote the observed apparent longitude and latitude; and let () be the true longitude of the sun, 20 its latitude, and Ro its distance from the earth, corresponding to the time of observation. Then, if 2 and 13 denote the longitude and latitude of the planet or comet corrected for the actual aberration of the fixed stars, we shall have A - A - +- 20".445 cos (A - O) sec + - 0".343 cos (A - 281~) sec, (1) pf- -— = 20".445 sin (A - 0o) sin f - 0".343 sin (A - 281~) sin fl. In computing the numerical values of these corrections, it will be sufficiently accurate to use 20 and o instead of 2 and 9 in the second members of these equations, and the last terms may, in most cases, be neglected. The values of) and & thus derived give the true place of the body at the time t - 497s.784, but as seen from the place of the earth at the time t. When the distance of the planet or comet is unknown, it is impossible to reduce the observed place to the centre of the earth; but if we conceive a line to be drawn from the body through the true place of observation, it is evident that were an observer at the point of intersection of this line with the plane of the ecliptic, or at any point in the line, the body would be seen in the same direction as from the actual place of observation. Hence, instead of applying any correction for parallax directly to the observed apparent place, we may conceive the place of the observer to be changed from the actual place to this point of intersection with the ecliptic, and, therefore, it becomes necessary to determine the position of this point by means of the data furnished by observation. Let 00 be the sidereal time corresponding to the time to of observation, (p' the geocentric latitude of the place of observation, and po the radius of the earth at the place of observation, expressed in parts of the equatorial radius as unity. Then 00 is the right ascension and 5p' the declination of the zenith at the time t0. Let 1o and 0b denote these quantities converted into longitude and latitude, or the longitude and latitude of the geocentric zenith at the time to. The rectangular co-ordinates of the place of observation referred to the centre of the 222 THEORETICAL ASTRONOMY. earth and expressed in parts of the mean distance of the earth from the sun as the unit, will be -- po sin ro cos bo cos Io, yo =-Po sin r0 cos bo sin lo, o = Po sin r sin bo, in which ro 8t'.57116.1 Let Jo be the distance of the planet or comet from the true place of the observer, and J, its distance from the point in the ecliptic to which the observation is to be reduced. Then will the co-ordinates of the place of observation, referred to this point in the ecliptic, be x, - (A, - o) cos i cos ), y, = (d, - ) cos I sin A, (D, - do) sin, the axis of x being directed to the vernal equinox. Let us now designate by ( the longitude of the sun as seen from the point of reference in the ecliptic, and by R its distance from this point. Then will the heliocentric co-ordinates of this point be X — - R cos Q, Y —- R sin 0, Z=-0. The heliocentric co-ordinates of the centre of the earth are X-= - Ro cos 20 cos 00, Yo- - o cos 0 sin 0o, Zo = - Ro sin o0. But the heliocentric co-ordinates of the true place of observation will be X - x,, Y + y,, Z+ z,, or Xo +.0, YO + yO, Zo + zo, and, consequently, we shall have B cos 0 (, - Jo) cos f cos Rt = R cos 2O cos o - po sin r cos bo cos 1o, R sin ( - (, - o) cos sin A = Ro cos I, sin 30 - op sin r0 cos bo sin 1,, - (, - Jo) sin f = Ro sin A0 -po sin rO sin bo. If we suppose the axis of x to be directed to the point whose longitude is (O, these become DETERMINATION OF AN ORBIT. 223 R cos (0 - 0o) - (D, - o) cos A cos (s - o) - Ro cos Y - po sin cos cos (lo - o), R sin (( - 0() - (D, - 4) cos sin ( -- o) = (2) - p sin nr cos bo sin (1o - (o), - (D, - Do) sin 3 = R, sin -~ po sin r sin b,, from which R and D may be determined. Let us now put (J, -J) cos D; (3) then, since;, S, J and O - o 0 are small, these equations may be reduced to R = D cos (A - (o) -, p cos b, cos (lo - 0() + Ro, R (( - ) D sin ( - 0) — po cos bo sin (lo - 0), - D tan --, Po sin b, + Ro 0. Hence we shall have, if %O and -0 are expressed in seconds of arc,. o Posin bo, - Ro ocot, 206264.8 70p7 cos b cos (l —- O$) R = Ro + D cos (A - 0o)- "0 Po 0- (4) ~206264.8 ( 206264.8 D sin (A - (D) - o po cos bo sin (o — 0o) from which we may derive the values of ( and R which are to be used throughout the calculation of the elements as the longitude and distance of the sun, instead of the corresponding places referred to the centre of the earth. The point of reference being in the plane of the ecliptic, the latitude of the sun as seen from this point is zero, which simplifies some of the equations of the problem, since, if the observations had been reduced to the centre of the earth, the sun's latitude would be retained. We may remark that the body would not be seen, at the instant of observation, from the point of reference in the direction actually observed, but at a time different from to, to be determined by the interval which is required for the light to pass over the distance d, - Jo. Consequently we ought to add to the time of observation the quantity (J, - JO) 4978.78 = 4978.78 D sec f, (5) which is called the reduction of the time; but unless the latitude of the body should be very small, this correction will be insensible. The value of i derived from equations (1) and the longitude 0 224 THEORETICAL. ASTRONOMY. derived from (4) should be reduced by applying the correction for nutation to the mean equinox of the date, and then both these and the latitude j9 should be reduced by applying the correction for precession to the ecliptic and mean equinox of a fixed epoch, for which the beginning of the year is usually chosen. In this way each observed apparent longitude and latitude is to be corrected for the aberration of the fixed stars, and the corresponding places of the sun, referred to the point in which the line drawn from the body through the place of observation on the earth's surface intersects the plane of the ecliptic, are derived from the equations (4). Then the places of the sun and of the planet or comet are reduced to the ecliptic and mean equinox of a fixed date, and the results thus obtained, together with the times of observation, furnish the data for the determination of the elements of the orbit. When the distance of the body corresponding to each of the observations shall have been determined, the times of observation may be corrected for the time of aberration. This correction is necessary, since the adopted places of the body are the true places for the instant when the light was emitted, corresponding respectively to the times of observation diminished by the time of aberration, but as seen from the places of the earth at the actual times of observation, respectively. When 9 =- 0, the equations (4) cannot be applied, and when the latitude is so small that the reduction of the time and the correction to be applied to the place of the sun are of considerable magnitude, it will be advisable, if more suitable observations are not available, to neglect the correction for parallax and derive the elements, using the uncorrected places. The distances of the body from the earth which may then be derived, will enable us to apply the correction for parallax directly to the observed places of the body. When the approximate distances of the body from the earth are already known, and it is required to derive new elements of the orbit from given observed places or from normal places derived from many observations, the observations may be corrected directly for parallax, and the times corrected for the time of aberration. We shall then have the true places of the body as seen from the centre of the earth, and if these places are adopted, it will be necessary, for the most accurate solution possible, to retain the latitude of the sun in the formulae which may be required. But since some of these formulae acquire greater simplicity when the sun's latitude is not introduced, if, in this case, we reduce the geocentric places to the DETERMINATION OF AN ORBIT. 225 point in which a perpendicular let fall from the centre of the earth to the plane of the ecliptic cuts that plane, the longitude of the sun will remain unchanged, the latitude will be zero, and the distance P will also be unchanged, since the greatest geocentric latitude of the sun does not exceed 1"'. Then the longitude of the planet or comet as seen from this point in the ecliptic will be the same as seen from the centre of the earth, and if J, is the distance of the body front this point of reference, and, its latitude as seen from this point, we shall have A, cos fi, - d cos f3, A, sin 3, -- A sin / - ~O sin z, from which we easily derive the correction 3, - 3, or AS, to be applied to the geocentric latitude. Thus, we find i 0 c 2o, A?=- D~ cosA, (6) ~o being expressed in seconds. This correction having been applied to the geocentric latitude, the latitude of the sun becomes z 0.. The correction to be applied to the time of observation (already diminished by the time of aberration) due to the distance A,- i will be afbsoltely insensible, its maximum value not exceeding 0'.002. It should be remarked also that before applying the equation (6), the latitude'0 should be reduced to the fixed ecliptic which it is desired to adopt for the definition of the elements which determine the position of the plane of the orbit. 78. When these preliminary corrections have been applied to the data, we are prepared to proceed with the calculation of the elements of the orbit, the necessary formula for which we shall now investigate. For this purpose, let us resume the equations (6)3; and, if we multiply the first of these equations by tan /9 sin )" - tan f" sin 2, the second by tan 13" cosI - tan 9 cos 2", and the third by sin (2 - 2"), and add the products, we shall have 0 nR (tan p" sin ( - 0) - tan fi sin (" - 0)) - p' (tan f sin ("-,') - tan fi' sin (" -- ) + tan f" sin (A' - 2)) -R' (tan f" sin (A - 0') tan fi sin (" -- 0')).7) t n"R" (tan [," sin (2 - 0") - tan fi sin ("- - 0")). It should be observed that when the correction for parallax is applied 15 226 THEORETICAL ASTRONOMY. to the place of the sun, pt is the projection, on the plane of the ecliptic, of the distance of the body from the point of reference to which the observation has been reduced. Let us now designate by Kthe longitude of the ascending node, and by I the inclination to the ecliptic, of a great circle passing through the first and third observed places of the body, and we h:l-ve tan fi = sin ( - K) tan,. (8) tan i" - sin (" - K) tan:L.- Introducing these values of tan: and tan:/1 into the equation (7), since sill (A - 0) sin (A" - K) - sin (A" - 0) sin (A - K) sin (" - A) sin (0 -K), sin (R' - ) sin (A" - K) + sin (A" - A' sin ( -- K) = + sin (A" - 2) sin (A' -- ), sin (A -') sin (A"- K)-sin (2"-0') sin ( -K) - sin (A" - A) sin (0' - K), sin (A - 0") sin (A" - K) - sin (A"- 0") sin (2 - K) = - sin (A" - A) sin (0"- K), we obtain, by dividing through by sin (2" - 2) tan I, 0 -- nR sin (( - K) + p' (sin ('- K) - tan i' cot) I) -' sin (0'- K) + ""R" sin (0" —K). Let. /0 denote the latitude of that point of the great circle passing through the first and third places which corresponds to the longitude 2', then tan i = sin (A' - K) tan I, and the coefficient of p' in equation (9) becomes sin (1 -') ) cos rio cos' tan I Therefore, if we put _sin (f'- io) a- cos i tan I' (10) we shall have R' sin(' —) _) Rsin(s -K) / sec - -+ n ( ao ao,,ao a0 (11) + R" sin ("- K) + 9~L.,! ao This formula will give the value of p', or of A', when the values of nW and n" have been determined, since a0 and K are derived from the data furnished by observation. DETERMINATION OF AN ORBIT. (227 To find K and 1, we obtain from equations (8) by a transformation precisely similar to that by which the equations (75)3 were derived, sin (Qi"+ f3) tan I sin Ge (A" + 2) -- K)2 cos' cOs sec 2 - ), tan I cos (- (A" + A) - K) 2 cs c ose We may also compute K and I from the equations which may be derived from (74)3 and (76)3 by making the necessary changes in the notation, and using only the upper sign, since T is to be taken always less than 9a~. Before proceeding further with the discussion of equation (11), let us derive expressions for p and p" in terms of o', the signification of p and p", when the corrections for parallax are applied to the places of the sun, being as already noticed in the case of p'. 79. If we multiply the first of equations (6)3 by sin 0" tan i", the second by - cos 0" tan i/', and the third by sin (" — O"), and add the products, we get 0=np (tan A" sin (0"-) )-tanf sin (0"- ")) —Rtan " sin (0"- 0) — p (tan i" sin ( " —')-tan f' sin ( " —")~)+R' tan A" sin (0"O'), (13) which may be written Omnp (tan l sin (A"- ")-tan " sin (2-GO "))nR tan i" sin (0"-0) + p' (tan r" sil ('- - 0") tan 0o sil (A" O ")) p' (tan f' - tan Si) sill (,"- (") +- R' tan f" sin (O"- 0'). Introducing into this the values of tall p, tall /9", and tan 90 in terms of I and K, and reducing, the result is 0 - np sin (2"- A) sin ( 0 "- K) - nR sin (O "- 0) sin (A"- K) - p' sin ("-2A') sil (O "- K) - p'a0 sec i' sin (A"- 0 ") - R' sin (O " — O') sin (A"- K). Therefore we obtain K (llp' " ) ase sin (A" - ") \ -~ sin (2" -) sin (2" -) ) sin ()" - K) sin (A" —K) R' sill (0 "- O') —n sin (O "-) ) s' sin (" - A) sin (0"- K) But, by means of the equations (9)3, we derive R' sin (0"-') - nR sin((0"- 30) = (N- n) R sin (0"- 0(), 228 THEORETICAL ASTRONOMY. and the preceding equation reduces to p'sin("-A') aosec' sin(A"- ) \ ) P - ~ e sin ("\ - A) si n (iA" — A) sin ((" - K) )(14 + 1 X R sin ((" — ) sin (" --- K) "( sil (A" sin (( ) sin( - K) To obtain an expression for p" in terms of pto if we multiply the first of equations (6)3 by sin (0 tan i, the second by - cos o tan, and the third by sin (A- 0), and add the products, we shall have O=n"p" (tan ft sin (2" —- ) —tan fi" sin (2 —0)) — n"R" tan sin (0"- 0) -p'(tan Psin (' — ) —tan' sin (A — ))+R'tan ( sin ('- 0). (15) Introducing the values of tan 9, tan j', and tan j9" in terms of K and I; and reducing precisely as in the case of the formula already found for p, we obtain,,__' ( sin' -) acsec' sin (A ) n"~ sin (A" - ) sin (A"- ) sin (( - K) /+ (1 N" I R" sin (0 "-i- ) sin( (- K)'\ n" 1 / sill ("- A) sin ( -K) Let us now put, for brevity, R sin (- K) R' sin ('K) b, C -- aO ao d " sin (" - K) see p' _ R" sin (E"- 0) a0~ ^' ^ sin("- 2)' a0 sin - A2) sin(2"-') rfR sin(2"- 0") sin ("- 2) - (17) si (n' - A) -R sin (2 -0 i)' sin (A"l- 2) V +sf(^-^ sin ~Q) sin( ~ b h sil ( —K) 1_ sinin (2 - K) 2 d 2 b and the equations (11), (14), and (16) become p' see f' -- c +- nb + n"d, pM- = i+M 1- 9l (18) If n and n" are known, these equations will, in most cases, be sufficient to determine p, p', and p". DETERMINATION OF AN ORBIT. 229 80. It will be apparent, from a consideration of the equations which have been derived for p, p', and p", that under certain circumstances they are inapplicable in the form in which they have been given, and that in some cases they become indeterminate. When the great circle passing through the first and third observed places of the body passes also through the second place, we have a- 0, and equation (11) reduces to n"Rs" sin (" - K) + nR sin ( - K) R' sin (' K). If the ratio of n" to n is known, this equation will determine the quantities themselves, and from these the radius-vector r' for the middle place may be found. But if the great circle which thus passes through the three observed places passes also through the second place of the sun, we shall have K= (', or K= 180~+ 0', and hence n"R" sin (0"- 0) - niR sin (0' 0) = 0, or n" R sin ((D' — O) n - R" sin ("-') from which it appears that the solution of the problem is in this case impossible. If the first and third observed places coincide, we have A = A" and 9 = " -, and each term of equation (7) reduces to zero, so that the problem becomes absolutely indeterminate. Consequently, if the data are nearly such as to render the solution impossible, according to the conditions of these two cases of indetermination, the elements which may be derived will be greatly affected by errors of observation. If, however, i is equal to A" and i"1 differs from i9, it will be possible to derive p', and hence o and p"; but the formulae which have been given require some modification in this particular case. Thus, when A =", we have K= A"' 2, I-=90~, and /0 900, and hence a, as determined by equation (10), becomes -. Still, in this case it is not indeterminate, since, by recurring to the original equation (9), the coefficient of p', which is - a0 sec t', gives a0 sin:' cot I — cos j' sin (2' - K), (19) and when A = ", it becomes simply a0o= - cos i' sin ('- K). 230 THEORETICAL ASTRONOMY. Whenever, therefore, the difference 2" — is very small compared with the motion in latitude, a0 should be computed by means of the equation (19) or by means of the expression which is obtained directly from the coefficient of o' in equation (7). When 2A-" =-K, the values of JiJ, JillM, JA2, and.-12" cannot be found by means of the equations (17); but if we use the original form of the expressions for p and p" in terms of p', as given by equations (13) and (15), without introducing the auxiliary angles, we shall have p' tan i' sin ("- 0") - tan i" sin (2' - 0") an tan 3" sin( ") tan sin ( -- 0") N/ L 1\ _ R tan Fi" sin ((D"- )) n tan Af sin (" - 0") - tan fA" sin ( - 0")', p"' tan / sin ('- 0) - tan ft' sin ( - 0) I " tan f sin ("- 0) - tan A" sin ( - 0) / N"\ R" tan f sin ("- 0) - 1n" tan i sin (A" 0) - tanl sin (- 0)' Hence tan /' sin (A" - (") - tan i" sin (2'- 0") tan i sin (,"- 0") tan f" sin ( - 0")' tan: sin (' — 0) -taln' sin (2 - 0) tan f5 sin (" 0) -- taln sin (2 0)' (20 _M-2 R tan /" sin (0" - 0) 2 tan f sin (A" - ") -- tan fi" sin (A - 0")' tn _ __ R" tan sin (0)"- 0) 5 tan sin (2" --- 0) - tal f" sin ( - 0)' are the expressions for It, 1M,", M2, and 312"' which must be used when 2 =2" or when 2 is very nearly equal to 2"; and then p and o" will be obtained from equations (18). These expressions will also be used when A" — A 180~, this being an analogous case. When the great circle passing through the first and third observed places of the body also passes through the first or the third place of the sun, the last two of the equations (18) become indeterminate, and other formula must be derived. If we multiply the second of equations (7)3 by tan 9" and the fourth by -sin(/l"- 0'), and add the products, then multiply the second of these equations by tan 9 and the fourth by -sin (A- 0'), and add, and finally reduce by means of the relation NR sin (0' - 0) _ N"R" sin (0"- 0'), we get DETERMINATION OF AN ORBIT. 231 p' tan fi" sin (' - 0') - tan if' sin (A" ~-') (n" Nil\ iR" tan i" sin (Q" — 0')'+\ - N - tan f sin (A 0(D') - tan ft sill ( V" Y 0')' p tan' sin (A - O') - tan f sin (' - 0') (21 - pn"' tan A" sin (A -') - tan l3 sin (A" - 0') t N R\ R tanf sin ((D' (- ) a " sN"' talln sin ( -')-tan sin((A" —') These equations are convenient for determining p ando p from p'; but they become indeterminate when the great circle passing through the extreme places of the body also passes through the second place of the sun. Therefore they will generally be inapplicable for the eases in which the equations (18) fail. If we eliminate p" from the first and second of the equations (6)3 we get 0 - np sin (" - A) - nR sin (" - 0) p' sin (A" - A') + R' sin (A" - O')-n"R" sin (A" 0"), from which we derive p' sin (-') (22).Y sin(A" - A) +nI sin (A" 0) R' sin (A" - G') + n"R" sin (A" - (") n sin (" - A) Eliminating o between the same equations, the result is p' sin (A'-A) " sin A) (23) ~0l —, sin (I" — ~) nR sin (A - 0) - R' sin (A - 0') + — n"R" sin (A - 0") in" sin (A" - A) These formulae will enable us to determine p and ot" from p' in the special cases in which the equations (18) and (21) are inapplicable; but, since they do not involve the third of equations (6)3, they are not so well adapted to a complete solution of the problem as the formula previously given whenever these may be applied. If we eliminate successively p"t and p between the first and fourth of the equations (7)3, we get _p' tan f" cos (A'-') - tan i' cos (A" - 0') n' tan i" cos (A (') -tan f cos (A"-') tan 1" nR cos (0' — Q )- R' + n"-R" cos (("- 0') n+ tan f" cos (A -') -tan f cos (A" - 0')',, p' tan j,' cos (A -') - tan i cos (A'- Q') n' tan fi" cos (A -') - ta.n 1 cos (A" (0') tan nR cos (' - 0) - R' +- "R"cos (" - 0') - n"' tan f" cos ( -') - tan fi cos (A" - 0') 232 THEORETICAL ASTRONOMY. which may also be used to determine p and p" when the equations (18) and (21) cannot be applied. When the motion in latitude is greater than in longitude, these equations are to be preferred instead of (22) and (23.) 81. It would appear at first, without examining the quantities involved in the formula for p', that the equations (26)3 will enable us to find n and n" by successive approximations, assuming first that and from the resulting value of' determining r and then carrying and from the resulting value of,o determining rt, and then carrying the approximation to the values of n and n" one step farther, so as to include terms of the second order with reference to the intervals of time between the observations. But if we consider the equation (10), we observe that a0 is a very small quantity depending on the difference / ~ —3, and therefore on the deviation of the observed path of the body from the arc of a great circle, and, as this appears in the denominator of terms containing n and n" in the equation (11), it becomes necessary to determine to what degree of approximation these quantities must be known in order that the resulting value of p' may not be greatly in error. To determine the relation of a. to the intervals of time between the observations, we have, from the coefficient of o' in equation (7), ao see c' - tan f sin (A" -') - tan A' sin (A" - 2) + tan:" sin (A' - A). We may put tan f = tan'- Ar" + Br"2...., tan = tan + AT + B"...., and hence we have ao sec l3' (sin (i" - ( 2) - sin (A" -A) + sin (A' - )) tan r' + (r sin (2' —) - -' sin (2" —A')) A-+(-tsin (A'2)' sin ((" —.')) B+.., which is easily transformed into ao sec r'- 4 sin. (A'- ) sin 2 (A" - 2') sin A (A"- ) tan r' (25) + ( (r ) —sin ('-)-" sin (A"'))A+(ir sin (A'- )+-" sin ("2'))B+.. If we suppose the intervals to be small, we may also put sin ( ( - A), and sin ( A) ~ Af - A, sin (A' - ) = A' -. DETERMINATION OF AN ORBIT. 233 Further, we may put A =' -' A'rf + B'fr2...., A - A + A'r + B'r' +.... Substituting these values in the equation (25), neglecting terms of the fourth order with respect to r, and reducing, we get ao — Tr'r" (A'" tan f' + A'B - AB') cos pi It appears, therefore, that a0 is at least of the third order with reference to the intervals of time between the observations, and that an error of the second order in the assumed values of n and n" may produce an error of the order zero in the value of p' as derived from equation (11) even under the most favorable circumstances. Hence, in general, we cannot adopt the values Ir A~t! n = ~,) n' -,' omitting terms of the second order, without affecting the resulting value of p' to such an extent that it cannot be regarded even as an approximation to the true value; and terms of at least the second order must be included in the first assumed values of n and n". The equation (28)3 gives -- 22 ^2__.i2 -— 1 + 61 T23 (26) n 2 \ rJ3 dr' omitting the term multiplied by dt, which term is of the third order with respect to the times; and hence in this value of -i only terms 7f of at least the fourth order are neglected. Again, from the equations (26)3 we derive, since r' -r - r", n- +n" 1 + 2r (27) in which only terms of the fourth order have been neglected. Now the first of equations (18) may be written: 91// b+-d / see 13' = (0i + n") - c, (28) n11+t in which, if we introduce the values of ~- and n + -nZ as given by (26) and (27), onl term o the fourt rde ith respet to the (26) and (27), only terms of the fourth order with respect to the 234 THEORETICAL ASTRONOMY. times will be neglected, and consequently the resulting value of o' will be affected with only an error of the second order when a, is of the third order. Further, if the intervals between the observations are not very unequal, ~T2 _'l2 will be a quantity of an order superior to T2, and when these intervals are equal, we have, to terms of the fourth order,,n T, n 7 The equation (27) gives 2r'3 (n + n"- ) - rt". Hence, if we put P~ -- a~~sn~~, ~(29) Q 2r'3 (n + n -- 1), we may adopt, for a first approximation to the value of p', ITt P =- Q — e (30) and p' will be affected with an error of the first order when the i-ntervals are unequal; but of the second order only when the intervals are equal. It is evident, therefore, that, in the selection of the observations for the determination of an unknown orbit, the intervals should be as nearly equal as possible, since the nearer they approach to equality the nearer the truth will be the first assumed values of P and Q, thus facilitating the successive approximations; and when a0 is a very small quantity, the equality of the intervals is of the greatest importance. From the equations (29) we get +P( +2'), (31) n" - nP; and introducing P and Q in (28), there results _ r i Q 9 \b+Pd P see i= I + 2)- - (32) This equation involves both o' and r' as unknown quantities, but by means of another equation between these quantities p' may be eliminated, thus giving a single equation from which r' may be found, after which p' may also be determined. DETEIRMINATION OF AN ORBIT. 235 82. Let,t' represent the angle at the earth between the sun and planet or comet at the second observation, and we shall have, from the equations (93)3, tan i tan w' -s (tasin ('-')' tan - (- ) (33) COS WO~ cos - = cos f' cos (' - 0'), by means of which we may determine', which cannot exceed 180. Since cos y' is always positive, cos 4' and cos(,'- 0') must have the same sign. We also have r2._ J2 + R12 2J'R' cos 4', which may be put in the form 2 (p' sec i' - R' cos 4)2 + R"2 sin2 2', from which we get p' sec 1i'= R' cos' l/1'r'"- R' sin2 4'. (34) Substituting for p' sec j' its value given by equation (32), we have ( I + 2 ) bPd c-R cos 4' / r'2- R' sin2. 2-~'3 1 q- 1 P For brevity, let us put b +Pd co — c - k, (35) - oQ -- 10, and we shall have ko - 3 R' cos 4/ 1/ r2 - R2 sin2'. (36) When the values of P and Q have been found, this equation will give the value of r' in terms of quantities derived directly from the data furnished by observation. We shall now represent by z' the angle at the planet between the sun and earth at the time of the second observation, and we shall have R' sin (' sin z' (37) 236 THEORETICAL ASTRONOMY. Substituting this value of r', in the preceding equation, there results I, sin4 z' (ko -- R' cos,') sin z' - R' sin' cos z' in ^ (38)'aI sina x~" and if we put'02 sin r = R' sin,', o cos - k - R' cos,', (39) O R'" sin'," m0 -- V0, sina/ the condition being imposed that m0 shall always be positive, we have, finally, sin (z' - 0) -m, sin4 z'. (40) In order that mr may be positive, the quadrant in which: is taken must be such that ^ shall have the same sign as 10, since sin y' is always positive. From equation (37) it appears that sin z' must always be positive, or z'< 180~; and further, in the plane triangle formed by joining the actual places of the earth, sun, and planet or comet corresponding to the middle observation, we have r' sin (z' + 4') R' sin (z'+ 4') sin' sin z' Therefore, A, sin (z' + ) osi (41) P sin z and, since p' is always positive, it follows that sin (z + b') must be positive, or that z' cannot exceed 1800 -'. When the planet or comet at the time of the middle observation is both in the node and in opposition or conjunction with the sun, we shall have /'=- 0,'-= 180~ when the body is in opposition, and' - 0~ when it is in conjunction. Consequently, it becomes impossible to determine r' by means of the angle z'; but in this case the equation (36) gives o - - R' + r', when the body is in opposition, the lower sign being excluded by the condition that the value of the first member of the equation must be positive, and for' = 0, the upper sign being used when the sun is between the earth and the DETERMINATION OF AN ORBIT. 237 planet, and the lower sign when the planet is between the earth and the sun. It is hardly necessary to remark that the case of an observation at the superior conjunction when' =0, is physically impossible. The value of r' may be found from these equations by trial; and then we shall have I' - R' when the body is in opposition, and p' = R' -r when it is in inferior conjunction with the sun-! For the case in which the great circle passing through the extreme observed places of the body passes also through the middle place, which gives a0co 0, let us divide equation (32) through by c, andl we have b d / Q \-G+ P 1p'seci' - \ +2r l-P 31 se c The equations (17) give b R sin ( - K) d R_" sin ()" — K) c R'sin ()'-K)' c R'sin (~'-~ K)' and if we put b d 1 + P l+P we shall have (1+ 2r"3) c,- 1, since c = co when ct 0. Hence we derive rf - (42) But when the great circle passing through the three observed places passes also through the second place of the sun, both c and CO become indeterminate, and thus the solution of the problem, with the given data, becomes impossible. 83. The equation (40) must give four roots corresponding to each sign, respectively; but it may be shown that of these eight roots at least four will, in every case, be imaginary. Thus, the equation may be written mn sin4' - sill' cos -T cos z' sin C, 238 THEORETICAL ASTRONOOMY. and, by squaring and reducing, this becomes o2 sin8 z' - 2mn cos C sin5 -'- sin z' sin2 =_ 0. When C is within the limits - 90~ and + 90~, cos g will be positive, and, In, being always positive, it appears from the algebraic signs of the terms of the equation, according to the theory of equations, that in this case there cannot be more than four real roots, of which three will be positive and one negative. When C exceeds the limits - 90~ and + 90~, cos C will be negative, and hence, in this case also, there cannot be more than four real roots, of which one will be positive and three negative. Further, since sin2 g is real and positive, there must be at least two real roots-one positive and the other negative -whether cos h be negative or positive. We may also remark that, in finding the roots of the equation (40), it will only be necessary to solve the equation sin (z' - ) = % sin4 z', (43) since the lower sign in (40) follows directly from this by substituting 180~ - z' in place of't; and hence the roots derived from this will comprise all the real roots belonging to the general form of the equation. The observed places of the heavenly body,,onlly give the direction in space of right lines passing through the places of the earth and the corresponding places of the body, and any three points, one in each of these lines, which are situated in a plane passing through the centre of the sun, and which are at such distances as to fulfil the condition that the areal velocity shall be constant, according to the relation expressed by the equation (30),, must satisfy the analytical conditions of the problem. It is evident that the three places of the earth may satisfy these conditions; and hence there may be one root of equation (43) which will correspond to the orbit of the earth, or give p- 0. Further, it follows from the equation (37) that this root must be z -- 180~ -~,'; and such would be strictly the case if, instead of the assumed values of P and Q, their exact values for the orbit of the earth were adopted, and if the observations were referred directly to the centre of the earth, in the correction for parallax, neglecting also the perturbations in the motion of the earth. DETERMINATION OF AN ORBIT. 239 In the case of the earth, n Nit_ RR' sin (' - 0) RR" sin (0"- 0)' X _N R'R" sin(0- 0') RR"I sin(0" — 0)' and the complete values of P and Q become RR' sin ((' - ) R'R" (sin (" — 0')' rr( Ro RI sill (O- 0)+ ~ KR" sill(0 "- ) 1); 9 2' ~(BR"^f sin("-0) ~ and since the approximate values TIt P —, Q rTT differ but little from these, as will appear from the equations (27)3, there will be one root of equation (43) which gives z' nearly equal to 1800 -'. This root, however, cannot satisfy the physical conditions of the problem, which will require that the rays of light in coming from the planet or comet to the earth shall proceed from points which are at a considerable distance from the eye of the observer. Further, the negative values of sin z' are excluded by the nature of the problem, since r' must be positive, or z' < 180~; and of the three positive roots which may result from equation (43), that being excluded which gives z' very nearly equal to 1800 - /, there will remain two, of which one will be excluded if it gives z' greater than 180~ - 4', and the remaining one will be that which belongs to the orbit of the planet or comet. It may happen, however, that neither of these two roots is greater than 180 4- 1, in which case both will satisfy the physical conditions of the problem, and hence the observations will be satisfied by two wholly different systems of elements. It will then be necessary to compare the elements computed from each of the two values of z' with other observations in order to decide which actually belongs to the body observed. In the other case, in which cos. is negative, the negative roots being excluded by the condition that r' is positive, the positive root must in most cases belong to the orbit of the earth, and the three observations do not then belong to the same body. However, in the case of the orbit of a comet, when the eccentricity is large, and the intervals between the observations are of considerable magnitude, if 240 THEORETICAL ASTRONOMY. the approximate values of P and Q are computed directly, by means of approximate elements already known, from the equations rr' sin (Vu'- u) - r'" sin ('/ -4I')' Q 2r, (rr sin ('- u) +- rr" sin ( u') rr" sin ("-) - - ) - 1 it may occur that cos g is negative, and the positive root will actually belong to the orbit of the comet. The condition that one value of z' shall be very nearly equal to 180~ - -', requires that the adopted values of P and Q shall differ but little from those derived directly from the places of the earth; and in the case of orbits of small eccentricity this condition will always be fulfilled, unless the intervals between the observations and the distance of the planet from the sun are both very great. But if the eccentricity is large, the difference may be such that no root will correspond to the orbit of the earth. 84. We may find an expression for the limiting values of mn and C, within which equation (43) has four real roots, and beyond which there are only two, one positive and one negative. This change in the number of real roots will take place when there are two equal roots, and, consequently, if we proceed under the supposition that equation (43) has two equal roots, and find the values of mn and: which will accord with this supposition, we may determine the limits required. Differentiating equation (43) with respect to z', we get cos (z' -) 4mn sin z' cos z; and, in the case of equal roots, the value of z' as derived from this must alo satisfy the original equation sin (z' - C) =- nm sin 4z'. To find the values of m0 and _ which will fulfil this condition, if we eliminate 0Q between these equations, we have sin z' cos (z' - C) 4 cos z' sin (z' - C), from which we easily find sin (2z' - C) = sin C. (45) This gives the value of g in terms of z' for which equation (43) has DETERMINATION OF AN ORBIT. 241 equal roots, and at which it ceases to have four real roots. To find the corresponding expression for mn, we have sin (z' - _) cos (z' C) % -- sin. -z 4 sin 3z' cos z" in which we must use the value of g given by the preceding equation. Now, since sin (2z' - ) must be within the limits — 1 and + 1, the limiting values of sin g will be + ] and- 3, or C must be within the limits + 360 52'.2 and - 36~ 52'.2, or 143~ 7'.8 and 216~ 52'.2. If C is not contained within these limits, the equation cannot have equal roots, whatever may be the value of z0, and hence there can only be two real roots, of which one will be positive and one negative. If for a given value of C we compute z' from equation (45), and call this z0, or sin (2z - C) -- sin C, we may find the limits of the values of mn within which equation (43) has four real roots. The equation for z0' will be satisfied by the values 2zo- b, 180~- (2Z, - ); and hence there will be two values of mo, which we will denote by mn, and m2, for which, with a given value of C, equation (43) will have equal roots. Thus we shall have sin (z' - ) 1 sin. 4 -Z and, putting in this equation 180~ — (2z' - C) instead of 2zox -, or 90~ (z0'- C) in place of zo', cos z 2cos4 (' - ) It follows, therefore, that for any given value of C, if m0 is not within the limits assigned by the values of m, and m2, equation (43) will only have two real roots, one positive and one negative, of which the latter is excluded by the nature of the problem, and the former may belong to the orbit of the earth. But if P and Q differ so much from their values in the case of the orbit of the earth that z' is not very nearly equal to 180~ - 4/, the positive root, when C exceeds the limits + 36~ 52'.2 and - 36~ 52'.2, may actually satisfy the conditions of the problem, and belong to the orbit of the body observed. 16 242 THEORETICAI ASTRONOMY. When C is within the limits 143~ 7/.8 and 216~ 52'.2, there will be four real roots, one positive and three negative, if mo is within the limits mn and mn2; but, if mn surpasses these limits, there will be only two real roots. Table XII. contains for values of g from - 36~ 52'.2 to + 36~ 52'.2 the values of on1 and 2,, and also the values of the four real roots corresponding respectively to mn1 and mn2. In every case in which equation (43) has three positive roots and one negative root, the value of mo must be within the limits indicated by m, and n2,, and the values of z' will be within the limits indicated by the quantities corresponding to m1 and m2 for each root, which we designate respectively by z/1, z/, z3', and z/,. The table will show, from the given values of m0 and 180~ - I, whether the problem admits of two distinct solutions, since, excluding the value of z', which is nearly equal to 180~- ~', and corresponds to the orbit of the earth, and also that which exceeds 1800, it will appear at once whether one or both of the remaining two values of z' will satisfy the condition that z' shall be less than 180~ - B'. The table will also. indicate an approximate value of z', by means of which the equation (43) may be solved by a few trials. For the root of the equation (43) which corresponds to the orbit of the earth, we have p' 0, and hence from (36) we derive a= -~B" Substituting this value for k0 in the general equation (32), we have p'sec i=10( - l, ); and, since o' must be positive, the algebraic sign of the numerical value of 10 will indicate whether r' is greater or less than R'. It is easily seen, from the formulae for lo, b, d, &c., that in the actual application of these formulhe, the intervals between the observations not being very large, 1o will be positive when P'- i0 and sin ( 0' — K) have contrary signs, and negative when i' - go has the same sign as sin (0' - K). Hence, when 0' - K is less than 180~, r' must be less than R' if i' - -o is positive, but greater than R' if /' - / is negative. When )' - K exceeds 180~, r' will be greater than R' if i~' - o is positive, and less than R' if' - /9 is negative: We may, therefore, by means of a celestial globe, determine by inspection whether the distance of a comet from the sun is greater or less than DETERMINATION OF AN ORBIT. 243 that of the earth from the sun. Thus, if we pass a great circle through the two extreme observed places of the comet, r' must be greater than R' when the place of the comet for the middle observation is on the same side of this great circle as the point of the ecliptic which corresponds to the place of the sun. But when the middle place and the point of the ecliptic corresponding to the place of the sun are on opposite sides of the great circle passing through the first and third places of the comet, r' must be less than R'. 85. From the values of p' and r' derived from the assumed values Tt P - and Q =-T", we may evidently derive more approximate values of these quantities, and thus, by a repetition of the calculation, make a still closer approximation to the true value of p'. To derive other expressions for P and Q which are exact, provided that r' and pt are accurately known, let us denote by s" the ratio of the sector of the orbit included by r and r' to the triangle included by the same radii-vectores and the chord joining the first and second places; by s' the same ratio with respect to r and r", and by s this ratio with respect to r' and r". These ratios s, s', s must necessarily be greater than 1, since every part of the orbit is concave toward the sun. According to the equation (30), we have for the areas of the sectors, neglecting the mass of the body,.':T 1 PV p Il and therefore we obtain s" [rr']- = " itp, s' r [rr"] =T, s r] rp. (46) Then, since lerr'']' in rr"]' we shall have 7 8' 7Tt 8t n=,. —, (47) and, consequently, P. 8 (48) _ST sS -& = = ~'-7rr"7? -Q r" s'/ 8' 8- ", 2r' (48) Substituting for s, s', and s" their values from (46), we have Q=2^' ]+3 [r'r"] + [rr'] - E[r"] ]7" -p [rr]. [rr"]. [r'r"] s' (49) 244 THEORETICAL ASTRONOMY. The angular distance between the perihelion and node being denoted by w, the polar equation of the conic section gives 7) P = 1 + e cos (t' - ), P 1= + e cos (V'-w), (50) r =- 1 + e cos (t" - W). If we multiply the first of these equations by sin (a" — u'), the second by - sin ('/" - u), and the third by sin (' - u), add the products and reduce, we get P sin (u"'i')- Psin (" sin ) + - sin (' - ) = sin (t" -' u')'' r' -sin ('" -,) + sin (tt) - s ); and, since sin (i"' - u') = 2 sin I ('" -') cos 4 (u" -- ~''), sin (a" — t) - sin (U' — u) - 2 sin! ('"-'') cos (it" + a'- 2u), the second member reduces to 4 sin - (u" -t) sin ('" -u ) sin - (t' - u). Therefore, we shall have 4rr'r" sin (" -t-') sin 1 (" - t) sin ~ (u'- ~u) 2 r'r" sin ('"- u') - rr" sin (u'" — ) + rr' sin (' — )' If we multiply both numerator and denominator of this expression by by 2rr' cos ('a" - a') cos O (" -') cos - (a' -' ), it becomes, introducing [rr'], [rr"], and [r'r"], _[r"']. [rr"]. [rr'] I -- [r'r"] + [rr'] - [rr"] 2rr'r" cos 1 (t'"-1_') coS ('~" —) Cos I (a' —)' Substituting this value of p in equation (49), it reduces to q~ ss, rr" cos " ( " -t) C Cos cos (-co I )' (5) 86. If we compare the equations (47) with the formula (28)3, we derive S = -,7 2 _ (f 3 +,.r3) d' -_,1 +_.1 13 +4 kr'4 dt (52) DETERMINATION OF AN ORBIT. 245 Consequently, in the first approximation, we may take st 8-1. If the intervals of the times are not very unequal, this assumption will differ from the truth only in terms of the third order with respect to the time, and in terms of the fburth order if the intervals are equal, as has already been shown. Hence, we adopt for the first approximation, P Q = TT the values of r and zT" being computed from the uncorrected times of observation, which may be denoted by to, to', and to". With the values of P and Q thus found, we compute r', and from this p', p, and p", by means of the formulae already derived. The heliocentric places for the first and third observations may now be found from the formula (71)3 and (72)3, and then the angle u" — u between the radii-vectores r and r" may be obtained in various ways, precisely as the distance between two points on the celestial sphere is obtained from the spherical co-ordinates of these points. When u" - u has been found, we have sin ('" — u') -, sin (le" - ua), it' (53) n'r"' sin (u' - u) = ~ sin ('" - u), from which u"l - u' and u'u - may be computed. From these results the ratios s and s' may be computed, and then new and more approximate values of P and Q. The value of u" -, found by taking the sum of ut" - t' and a' - as derived from (53), should agree with that used in the second members of these equations, within the limits of the errors which may be attributed to the logarithmic tables. The most advantageous method of obtaining the angles between the radii-vectores is to find the position of the plane of the orbit directly from 1, 1", b, and b", and then compute'a, u', and u" directly from n2 and i, according to the first of equations (82),. It will be expedient also to compute r', 1' and b' from o', i', and i', and the agreement of the value of r', thus found, with that already obtained from equation (37), will check the accuracy of part of the numerical 246 THEORETICAL ASTRONOMY. calculation. Further, since the three places of the body must be in a plane passing through the centre of the sun, whether P and Q are exact or only approximate, we must also have tan b' = tan i sin (l' -- ), and the value of b' derived from this equation must agree with that computed directly from p', or at least the difference should not exceed what may be due to the unavoidable errors of logarithmic calculation. We may now compute n and n" directly from the equations' r" sin ('"- u'),, rr' sin (u'- ) (54) rr" sin ("- ua)' rr sin ('a- ua) but when the values of u, U', and u'a are those which result from the assumed values of P and Q, the resulting values of n and n"l will only satisfy the condition that the plane of the orbit passes through the centre of the sun. If substituted in the equations (29), they will only reproduce the assumed values of P and Q, from which they have been derived, and hence they cannot be used to correct them. If, therefore, the numerical calculation be correct, the values of ai and n" obtained from (54) must agree with those derived from equations (31), within the limits of accuracy admitted by the logarithmic tables. The differences u, - u' and u' - u will usually be small, and hence a small error in either of these quantities may considerably affect the resulting values of n and n"'. In order to determine whether the error of calculation is within the limits to be expected from the logarithmic tables used, if we take the logarithms of both members of the equations (54) and differentiate, supposing only n, n", and u' to vary, we get d loge = -c cot ('a" - u') du', d loge n"- + cot (u' - u) dt'. Multiplying these by 0.434294, the modulus of the common system of logarithms, and expressing du' in seconds of arc, we find, in units of the seventh decimal place of common logarithms, d logn = - 21.055 cot ('" -- u') du', d log n" = + 21.055 cot (' -- ) du'. If we substitute in these the differences between log and log n" as found from the equations (54), and the values already obtained by DETERMINATION OF AN ORBIT. 247 means of (31), the two resulting values of du' should agree, and the magnitude of dc' itself will show whether the error of calculation exceeds the unavoidable errors due to the limited extent of the logarithmic tables. When the agreement of the two results for n and I" is in accordance with these conditions, and no error has been made in computing n and n" from P and Q by means of the equations (31), the,accuracy of the entire calculation, both of the quantities which depend on the assumed values of P and Q, and of those which are obtained independently from the data furnished by observation, is completely proved. 87. Since the values of n and n' derived from equations (54) cannot be used to correct the assumed values of P and Q, from which r, 9r', ua, U', &c. have been computed, it is evidently necessary to compute the values for a second approximation by means of the series given by the equations (26)3, or by means of the ratios s and s". The expressions for n and n" arranged in a series with respect to the time involve the differential coefficients of r' with respect to t, and, since these are necessarily unknown, and cannot be conveniently determined, it is plain that if the ratios s and s" can be readily found from r r, r", u,, U', and T, T, z", so as to involve the relation between the times of observation and the places in the orbit, they may be used to obtain new values of P and Q by means of equations (48) and (51), to be used in a second approximation. Let us now resume the equation M1/ E — e sin E, or k(t-T) E- e sin E, az and also for the third place k (t"-T) - E" - e sin E". al Subtracting, we get E" - E- 2e sin (E" E) cos (E" + E). (55) a2 This equation contains three unknown quantities, a, e, and the difference E" —E. We can, however, by means of expressions involving r, r",'t, and u", eliminate a and e. Thus, since p -a(1 -e2), we have i/p -- a2l/ - e (E"- E - 2e sin (E" -- E) cos 1 (E"+ E)). (56) 248 THEORETICAL ASTRONOMY. From the equations /r sin -v -/ a (1 + e) sin -E, V/r" sin v" -/Va (1 + e) sin AE", l/r cos v Va (1 -e) cos!E,:: /r" cos " = (1 e) cos E", since v"-v - u- u, we easily derive r sin (u" — u) a/1 - el sin s (E"- E), (57) and also a cos (E" — E) - ae cos - (E" + E) l/rr cos ~ (uo " - ), or l/rr" cos (58) e cos (E + E) ~ cos (E" - ) - " c (58) Substituting this value of e cos -(E" + E) in equation (56), we get r'Vap a'/1 - e (E" - E- si (E" - E)) + 2a/ - e2 sin (E" - E) cos (u" — u) I/rr", and substituting, in the last term of this, for aVl/- e2, its value from (57), the result is'1p = a'/1 - e' (E" - E - sin (E"- E)) + rr" sin (u"- u). (59) From (57) we obtain aV ~1 -- -2 l (/rr' sin - (u" - u)) e p sin3 1 (E"- E) or ____ / -rr" sin (u"- i) V i 21/rr" cOs u" u) psin (.E" E) Therefore, the equation (59) becomes 1 E"- E- sin (E"- E) (rr] ~ sin A (E" E) 21/rr" cos (t + rr" (60) Let x' be the chord of the orbit between the first and third places, and we shall have X12 (r + r"))2'- 4rr" cos2 (It" - ut). Now, since the chord xl can never exceed r + r", we may put' - (r + r") sin y', (61) and from this, in combination with the preceding equation, we derive 21rr" cos (u"- u) -= (r + r") cos r. (62) DETERMINATION OF AN ORBIT. 249 Substituting this value, and [rr"] -- / p, in equation (60), it reduces to E"- E — Esin (E"- E) t, 2 1 1 sin3 (E" -E) (r + r")3 os3 y + - (63) The elements a and e are thus eliminated, but the resulting equation involves still the unknown quantities E"-E and s'. It is necessary, therefore, to derive an additional equation involving the same unknown quantities in order that E"-E may be eliminated, and that thus the ratio s', which is the quantity sought, may be found. From the equations r a- ae cosE, r" a - aecos E", we get r" r - 2a 2ae cos (E" + E) cos (E E). Substituting in this the value of e cos (El" - E) from (58), we have r" + r = 2a sin2 (E"- E) + 21/ r" ) c (r" ) cos (E" - E), and substituting for sin (E" -- E) its value from (57), there results, "+ r 2rr" sin' r (" ~) + 2V /r" cos- ("- u) (1 —2 sin2 (E" —E)). But, since 2rr" sin2 1 (u" - )__ ([rr"])2 2'7 - I 2 p 2prr" cos2' ('t"- ) s 21/rr" cos 2 (u"-u) we have 2r'2 1 r r(r -") coS r + (r + r") co (1 -2 sin ( E"-E)), from which we derive 1 f2" sin2 -r' sin2(E" E)- (r "cosr cos (64) which is the additional equation required, involving E"- E and s' as unknown quantities. Let us now put sin (E" - E) E"- - E sin(E"- E) Tt2 (r + r") cos3"., sin';-' x' sin' ("- E), CO y 250 THEORETICAL ASTRONOMY. and the equations (63) and (64) become 1m' 1 1.- -— 1 yt > st3;+ 8t t (66) 8~2 When the value of y' is known, the first of these equations will enable us to determine s', and hence the value of x', or sin2- (E"- E), from the last equation. The calculation of r' may be facilitated by the introduction of an additional auxiliary quantity. Thus, let,r tan' =, (67) and from (62) we find cos cos - (u" - ) -- 2 cos - (u" - u) coS2Z' tanl', or cos -- sin 2' cos ( - ). (68) We have, also, x'2 (r + r")2 - 4rr" cOS2' (in" — i), which gives X^2 = (r - I-)2 + 4rr" sin2 1 (i Z). -: Multiplying this equation by cos2(tt-u) and the preceding one by sin2 1 (al" - ), and adding, we get - (r + r")2 sin2 " u) +- (r -- r")2 COS2 r (n"- n). (69) From (67) we get r r^ cos' - r, sin X= + rl" and, therefore, Ir - r"''COS 2Z -- r __ r' so that equation (69) may be written ( = ) sin/2 r' sin2 1 (r- _ u) + cos2 2 cOS2 (u" - U). (r + r')sln We may, therefore, put sin r' cos G-' sin - (i" - u), sin r' sin G' = cos -. ('" - u) cos 2', (70) cosr cos' (t" - i ) sin 2X', DETERMINATION OF AN ORBIT. 251 from which r' may be derived by means of its tangent, so that sin /' shall be positive. The auxiliary angle G' will be of subsequent use in determining the elements of the orbit from the final hypothesis for P and Q. 88. We shall now consider the auxiliary quantity y' introduced into the first of equations (66). For brevity, let us put - (E"- E), and we shall have sin g Y 2g - sin 2g This gives, by differentiation, dy__if 4 sin2g dg = 3 cotdg d ~ —, y 3 cot 2g - sin 2g or cy' dy -= 3y' cot g - 4y' cosec g. dg The last of equations (65) gives xI'- sin2 gg, and hence dq d- - 2 cosec g. dx' Therefore we have dy' 6y' cos g - 8y 3(1 - 2x') y'- 4y'2 dx' sin2 g 2x' (1 - x') It is evident that we may expand y' into a series arranged in reference to the ascending powers of', so that we shall have' +-tx'- r'2+ x'3 +'4 + 4'5-+ &c. Differentiating, we get dy' dL = ft^+ 2 34x' + + 4 &c., and substituting for dy' the value already obtained, there results 2/x' + (4r - 23) x" + (6S - 4r) x'3 + (8s - 6) x'4+ (10: 8) x'+ &c. _ (3a - 42) (3 - 6a - 8a) x' + (3r - 6 - 4/2 8ar) x" + (3 6r -- 8r 8c) x' + (3S - 6 - 4 — 8 8a~8^) x'4 + (3 - 6s - 8r 8 - /89 - 8aC) Xt5 + &C. Since the coefficients of like powers of x' must be equal, we have 3a - 42 Q 0, 3f - 6 - 8I -7 2, 3r - 6f~ - 4 ~2 - 8ar = 2 (2r - fi), &c.; 252 THEORETICAL ASTRONOMY. and hence we derive -l 3- ^-1 26 6__ 228 _265896 _ 19139024 - 336875 T18965 ~-g SI 17l 875 Therefore we have y' 2.~ + J~-(-X' 6 22 8'"-265896 -34-190 +175 1 It + 82a X8 3 75 336875 X 2-189 6 875 X5 _~ 19139024 - + 4 + & + 9 1 3 9 6 &c. (71) If we multiply through by U, and put it 1 2 1 5 2 X3 1 3_84 It 1_ 5 9088 X15 3X T 5 75 667 3 7 5 I 4379d375 +T38278048 16 + - 41389356 - +&c (72) we obtain 10o y l + x- * (73) Combining this with the second of equations (66), the result is y' +- - If we put 5+-f, (74) we shall have 1 -a,1 2 But from the first of equations (66) we get -= 12 ('- 1); y and therefore we have (S'- 1) s,.+ ~. (75) As soon as /' is known, this equation will give the corresponding value of s'. Since $' is a quantity of the fourth order in reference to the difference 1 (E" - E), we may evidently, for a first approximation to the value of ^', take nm' and with this find s' from (75), and the corresponding value of x' from the last of equations (66). With this value of x' we find the corresponding value of ~', and recompute ~', s', and x'; and, if the DETERMINATION OF AN ORBIT. 253 value of E' derived from the last value of x' differs from that already used, the operation must be repeated. It will be observed that the series (72) for A' converges with great rapidity, and that for E" - E= 940 the term containing x'6 amounts to only one unit of the seventh decimal place in the value of $'. Table XIV. gives the values of $' corresponding to values of x' from 0.0 to 0.3, or from E"'- E — to E" -E - 132~ 50'.6. Should a case occur in which E"'- E exceeds this limit, the expression sin'll (E" - E) Y E"-E - sin (E"- E) may then be computed accurately by means of the logarithmic tables ordinarily in use. An approximate value of x' may be easily found with which y' may be computed from this equation, and then $' from (73). With the value of i' thus found, I' may be computed from (74), and thus a more approximate value of x' is immediately obtained. The equation (75) is of the third degree, and has, therefore, three roots. Since s' is always positive, and cannot be less than 1, it follows from this equation that o' is always a positive quantity. The equation may be written thus: t3 -t2 tSt 1- 0and there being only one variation of sign, there can be only one positive root, which is the one to be adopted, the negative roots being excluded by the nature of the problem. Table XIII. gives the values of log 8/2 corresponding to values of' from i'=- 0 to'= 0.6. When ^' exceeds the value 0.6, the value of s' must be found directly from the equation (75). 89. We are now enabled to determine whether the orbit is an ellipse, parabola, or hyperbola. In the ellipse x sin2 4 (E - E) is positive. In the parabola the eccentric anomaly is zero, and hence x- 0. In the hyperbola the angle which we call the eccentric anomaly, in the case of elliptic motion, becomes imaginary, and hence, since sin I (E" - E) will be imaginary, x' must be negative. It follows, therefore, that if the value of x' derived from the equation - t2 is positive, the orbit is an ellipse; if equal to zero, the orbit is a parabola; and if negative, it is a hyperbola. 254 THEORETICAL ASTRONOMY. For the case of parabolic motion we have x' - 0, and the second of equations (66) gives,-.n ( 76) If we eliminate s' by means of both equations, since, in this case, y' -, we get r,~ =j, + 4j'3 Substituting in this the values of m andjgiven by (65), we obtain - =3 sin z' cos r' + 4 sin3 r, (r + r") which gives ~-, - 6 sin o' cos- y' + 2 sin' r'^ (r +-{ r") or =- (sill 1 + cos C")3 +- (sin }r' -- cos r)3. (r + r")> This may evidently be written (r + " = (i + sin ) ( sin r, the upper sign being used when ry is less than 90~, and the lower sign when it exceeds 900. Multiplying through by (r + r")^, and replacing (r +- r") sin r by x, we obtain 6r'= (r +'r" + X) (r + r" — x), which is identical with the equation (56)3 for the special case of parabolic motion. Since x' is negative in the case of hyperbolic motion, the value of' determined by the series (72) will be different from that in the case of elliptic motion. Table XIV. gives the value of' corresponding to both forms; but when x' exceeds the limits of this table, it will be necessary, in the case of the hyperbola also, to compute the value of $' directly, using additional terms of the series, or we may modify the expression for y' in terms of E" and E so as to be applicable. If we compare equations (44), and (56),, we get tan E 1/- 1 tan!F; DETERMINATION OF AN ORBIT. 255 and hence, from (58),, tan E=- V-1. We have, also, by comparing (65), with (41),, since a is negative in the hyperbola, 62 +1 cos E- 2 which gives sin E - 1/ 1. 26 Now, since cos E - V-1 sin E= eE/-i in which e is the base of Naperian logarithms, we have E 1/ — 1 logo (cos E + /- 1 sin E), which reduces to E 1/~- = loge-, or E= V- 1 loge. By means of these relations between E and a, the expression for y' may be transformed so as not to involve imaginary quantities. Thus we have E" - E = (loge " - loge ) 1/- 1 - V-1 loge 11 62 sin (E" - E) - sin E" cos E - cos E" sin E =' - 1/ 1. 266"f From the value of cos E we easily derive sin E 6 - ~ 1coSE - +2 Va 21V/~ and hence sin 2 (E" -E)= -2 - 1. 2 1/66 Therefore the expression for y' becomes =Y (/ lg (6 - - (. (1/6a )3 loge — 4 V 66"a (6" 2 a62) 256 THEORETICAL ASTRONOMY. Since the auxiliary quantity a in the hyperbola is always positive) let us now put - A2 and we have (A- )3 y'= — ~-~ (77) A2 A- -2 4 logeA from which y' may be derived when A is known. We have, further, sin2 (E" - E) - (I - cos (E" - E)) - (- 2 V6 ) and therefore' ( — 4V _, (78) 4 V' sa" 40' or ~' — ^(^ -71)- ( 1(79) These expressions for y' and x' enable us to find $' when x' exceeds the limits of the table. Thus, we obtain an approximate value of xz by putting, rn' ^ ~5 —+j' from which we first find s' and then x' from the second of equations (66). Then we compute A from the formula (79), which gives A =-1 2x'+ 21/x'" -', (80) y' fiom (77), and d' from (73). A. repetition of the calculation, using the value of t' thus found, will give a still closer approximation to the correct values of x' and s'; and this process should be continued until.' remains unchanged. 90. The formule for the calculation of s' will evidently give the value of s if we use., r' rt" u'd and u", the necessary changes in the notation being indicated at once; and in the same manner using r", r, r', u, and u', we obtain s". From the values of s and s" thus found, more accurate values of P and Q may be computed by means of the equations (48) and (51). We may remark, however, that if the times of the observations have not been already corrected for the DETERMINATION OF AN ORBIT. 257 time of aberration, as in the case of the determination of an unknown orbit, this correction may now be applied as determined by means of the values of o, p', and p" already obtained. Thus, if to, to', and to" are the uncorrected times of observation, the corrected values will be t - to - Cp sec, t' = to - p' sec /', (81) tr" — tol- Cp" see fl", in which log C- 7.760523, expressed in parts of a day; and from these values of t, t', t" we recompute T, t', and r", which values will require no further correction, since p, p', and p", derived from the first approximation, are sufficient for this purpose. With the new values of P and Q we recompute r, r'', r", and u, ut, ut as before, and thence again P and Q, and if the last values differ from the preceding, we proceed in the same manner to a third approximation, which will usually be sufficient unless the interval of time between the extreme observations is considerable. If it be found necessary to proceed further with the approximations to P and Q after the calculation of these quantities in the third approximation has been effected, instead of employing these directly for the next trial, we may derive more accurate values from those already obtained. Thus, let x and y be the true values of P and Q respectively, with which, if the calculation be repeated, we should derive the same values again. Let Ax and Ay be the differences between any assumed values of x and y and the true values, or x0-x + Ax, Y-Y + AY. Then, if we denote by x,', yo' the values which result by direct calculation from the assumed values x0 and yo, we shall have xo - X f (Xo, Yo) = f ( + Ax, y -+ Al). Expanding this function, we get xo-.,r =f (x, y) + AAx + Bay + CAX2 + DAx Ay + EAy +.., and if Ax and Ay are very small, we may neglect terms of the second order. Further, since the employment of x and y will reproduce the same values, we have f(x, y) = 0, and hence, since Ax - x0 — x and Ay y - y, xo'- xo A (x0- x) + B (y- y). 17 258 THEORETICAL ASTRONOMY. In a similar manner, we obtain yo'- Yo- A' (x - x) + B' (Yo - y). Let us now denote the values resulting from the first assumption for P and Q by P1 and Q1, those resulting from P1, Q, by P2, Q2, and from P2, Q2 by P3, Q3; and, further, let Pi~- P = P PI a, P - P2= = Q1- Q-b, Q2 — Q1 b', Q3- Q2 b". Then, by means of the equations for x'J - x and yo'- Yo, we shall have a =A(P -x) + B(Q -y), b = A'(P -x) + B'(Q -y), a' - A (P, - x) + B (Q1 - y), b' A' (P- x) - B' (Q- y), a" A (P, - x) + B (Q- y), b"- A' (P2- x) + B' ( Q- y). If we eliminate A, B, A', and B' from these equations, the results are P(a'b" — a"b') + P, (a"b - ab") + P, (ab' - a'b) (a'b" - a"b') +- (a"b - ab") - (ab' - a'b) Q (a - a"b') + Q, (a"b - ab") + Q, (ab'- a'b) - (a'b" — a"b') - (a"b - ab") - (ab'- a'b) from which we get (a" +- a') (a'b"- aC"b') + a' (a"b - ab") (a'"- a"b') - (a"b - ab") + (ab' - a'b)' (b" + b') (C'b" - a"b') + b" (a"b - ab") Y Q~- ^(a'b" — c"b') + (a"b - ab") + (ab'- a'b)' In the numerical application of these formule it will be more convenient to use, instead of the numbers P, P, P, 2,, Q, &c., the logarithms of these quantities, so that a- log P1- log P, b log Q~ log Q, and similarly for a', b', a", b", —which may also be expressed in units of the last decimal place of the logarithms employed,-and we shall thus obtain the values of log x and logy. With these values of log x and log y for log P and log Q respectively, we proceed with the final calculation of r, r, r, and u6, u', ut. When the eccentricity is small and the intervals of time between the observations are not very great, it will not be necessary to employ the equations (82); but if the eccentricity is considerable, and if, in addition to this, the intervals are large, they will be required. It may also occur that the values of P and Q derived from the last hypothesis as corrected by means of these formula will differ so DETERMINATION OF AN ORBIT. 259 much from the values found for x and y, on account of the neglected terms of the second order, that it will be necessary to recompute these quantities, using these last values of P and Q in connection with the three preceding ones in the numerical solution of the equations (82). 91. It remains now to complete the determination of the elements of the orbit from these final values of P and Q. As soon as Q, i, and u, u', u" have been found, the remaining elements may be derived by means of r, r', and u' —, and also from r', r, and "t'-'; or, which is better, we will obtain them from the extreme places, and, if the approximation to P and Q is complete, the results thus found will agree with those resulting from the combination of the middle place with either extreme. We must, therefore, determine s' and x' from r, r", and t"~- Ua by means of the formulae already derived, and then, from the second of equations (46), we have ( strr" sin (-"- u)2 (83) ^^sin^-^(~ - (83) from which to obtain p. If we compute s and s" also, we shall have (/ sr" n sin ("t - ) ) 2 s"rr' sin ('tt _) ) 2 and the mean of the two values of p obtained from this expression should agree with that found from (83), thus checking the calculation and showing the degree of accuracy to which the approximation to P and Q has been carried. The last of equations (65) gives sin I (E" - E) =/Vx', (84) from which E'"- E may be computed. Then, from equation (57), since e - sin ~p, we have sin (- ('-a" a) &.& cos e- sin" 2 rr~" (85) sin(E"- E) for the calculation of a cos (. But p a (1 - e2) a cos2 s, whence cos =a CS (86) a cos p which may be used to determine p when e is very nearly equal to unity; and then e may be found from e = 1 - 2 sin2(450 -~ ). 260 THEORETICAL ASTRONOMY. The equations (50) give e cos (u -o) - -1, r e cos (U"- c) P= -, and from these, by addition and subtraction, we derive 2e cos - (u" ) cos ( ("' + u) ) —+) + _ 2, (87) 2e sin (u" - ) sin (t ("+ u) - w) -P- P by means of which e and co may be found. Since r-" 2V/r/ cos 2X' r + r" sin 2,' = — rwe have p?P 2p 4-+ - 2 v,'2 2r V7 l/rr sin 2/ P P 2p cot 2X r r7 r" and from equations (70), cot 2z sin 1 (u" — u) tan G' cos r' cot 2X/ = -/,?' sin 2;' = co (" — )' cos cos7('~u) Therefore the formulae (87) reduce to e sin (w - (u" + u)) --, P tan G', ch may be deri. from which also e and o) may be derived. Then sin =- e, and the agreement of cos op as derived from this value of (p with that given by (86) will serve as a further proof of the calculation. The longitude of the perihelion will be given by R - - v+ a, or, when i exceeds 90~, and the distinction of retrograde motion is adopted, by r = - w. DETERMINATION OF AN ORBIT. 261 To find a, we have p (a cos 0)2 cos2 p p or it may be computed directly from the equation Tt2 a 4s2 rr" cos2 (u" t u) in2 (E" -E)' (89) which results from the substitution, in the last term of the preceding equation, of the expressions for a cos o and p given by (83) and (85). Then for the mean daily motion we have k as We have now only to find the mean anomaly corresponding to any epoch, and the elements are completely determined. For the true anomalies we have v= u -, v' v' -, 1,' v"-,tt and if we compute r, r', r" from these by means of the polar equation of the conic section, the results should agree with the values of the same quantities previously obtained. According to the equation (45), we have tan ME =tan (45~ - ~) tan,v, tan E' tan (450 - 2o) tan v', (90) tan ME" = tan (450 -~ s) tan l", from which to find E, E', and E". The difference E" - E should agree with that derived from equation (84) within the limits of accuracy afforded by the logarithmic tables. Then, to find the mean anomalies, we have M -E -esinE, M' =E'- e sin E', (91) M" - E"- e sin E"; and, if -1u0 denotes the mean anomaly corresponding to any epoch T, we have, also, M,= - - t (t - T) M'- (e - T) = 1M"- (t"- T), in the application of which the values of t, t', and t" must be those which have been corrected for the time of aberration. The agree 262 THEORETICAL ASTRONOMY. ment of the three values of Mlf will be a final test of the accuracy of the entire calculation. If the final values of P and Q are exact, this proof will be complete within the limits of accuracy admitted by the logarithmic tables. When the eccentricity is such that the equations (91) cannot be solved with the requisite degree of accuracy, we must proceed according to the methods already given for finding the time from the perihelion in the case of orbits differing but little from the parabola. For this purpose, Tables IX. and X. will be employed. As soon as v, v', and v" have been determined, we may find the auxiliary angle V for each observation by means of Table IX.; and, with V as the argument, the quantities M, Mi', M" (which are not the mean anomalies) must be obtained from Table VI. Then, the perihelion distance having been computed from P + 1 -p e' we shall have T _- p 2 _ Mt1- 2 M -- 2 ~'C,\l~e~ Co t ^e Cm 1 j _e' (92) in which log Co - 9.96012771 for the determination of the time of perihelion passage. The times t, t, t" must be those which have been corrected for the time of aberration, and the agreement of the three values of T is a final proof of the numerical calculation. If Table X. is used, as soon as the true anomalies have been found, the corresponding values of log B and log C must be derived from the table. Then w is computed from tan tan |v 1 +- 9e tan - C 5(I + e)' and similarly for w' and w"; and, with these as arguments, we derive M, M', M' from Table VI. Finally, we have T - MBq3 __ M'B'q M]"B" qa CVo y (1 + 9e) Co0/L (1 + 9e) Col/To (1+ 9e) (93) for the time of perihelion passage, the value of Co being the same as in (92). When the orbit is a parabola, e 1= and p = 2q, and the elements q and wo can be derived from r, r", u, and ut' by means of the equa DETERMINATION OF AN ORBIT. 263 tions (76), (83), and (88), or by means of the formulae already given for the special case of parabolic motion. 92. Since certain quantities which are real in the ellipse and parabola become imaginary in the case of the hyperbola, the formula already given for determining the elements from r, r" t, and iu' require some modification when applied to a hyperbolic orbit. When s' and x' have been found, p, e., and co may be derived from equations (83) and (87) or (88) precisely as in the case of an elliptic orbit. Since x' = sin2 (E" - E), we easily find sin (E" - E) - 2 V'x' x and equation (85) becomes sin 1 (i;"- ui) 1/Tr' a cos p - (94) 2 i/x'~ x' But in the hyperbola x' is negative, and hence /x' t- x' will be imaginary; and, further, comparing the values of p in the ellipse and hyperbola, we have cos2 - tan2, or cos?p =- — 1 Itan x. Therefore the equation for a cos
4 sec.
The value of a may be found from
a =p cot2 4= (a tan )2(9
P
264 THEORETICAL ASTRONOMY.
or from
712
16s'2 rr" cos2 (tt" -- ) ('12- -')'
which is derived directly from (89), observing that the elliptic semitransverse axis becomes negative in the case of the hyperbola.
As soon as (o has been found, we derive from u, U', and u'" the
corresponding values of v, v', and v", and then compute the values
of F, F', and F" by means of the formula (57),; after which, by
means of the equation (69),, the corresponding values of V, N', and
N" will be obtained. Finally, the time of perihelion passage will
be given by
az.t a,
T=t-A N= - t' N'- t Y NN"
wherein log, k 7.87336575.
The cases of hyperbolic orbits are rare, and in most of those which
do occur the eccentricity will not differ much from that of the parabola, so that the most accurate determination of T will be effected by
means of Tables IX. and X. as already illustrated.
93. EXAMPLE.-TO illustrate the application of the principal formule which have been derived in this chapter, let us take the following observations of Eurynome ~:
Ann Arbor M.T. (79!) M
1863 Sept. 14 15h 53- 378.2 1h 0m 44".91 + 90 53' 30".8,
21 9 46 18.0 0 57 3.57 9 13 5.5,
28 8 49 29.2 0 52 18.90 +8 22 8.7.
The apparent obliquity of the ecliptic for these dates was, respectively, 23~ 27' 20".75, 23~ 27' 20".71, and 23~ 27' 20".65; and, by
means of these, converting the observed right ascensions and declinations into apparent longitudes and latitudes, we getAnn Arbor M. T. Longitude. Latitude.
1863 Sept. 14 15h 53- 37s.2 170 47' 37".60 + 30 8' 43".19,
21 9 46 18.0 16 41 36.20 2 52 27.46,
28 8 49 29.2 15 16 56.35 +2 32 42.98.
For the same dates we obtain from the American Nautical Almanac
the following places of the sun:
NUMERICAL EXAMPLE. 265
True Longitude. Latitude. log Ro.
172~ 1' 42".1 - 0.07 0.0022140,
178 37 17.2 + 0.77 0.0013857,
185 26 54.8 + 0.67 0.0005174.
Since the elements are supposed to be wholly unknown, the places
of the planet must be corrected for the aberration of the fixed stars
as given by equations (1). Thus we find for the corrections to be
applied to the longitudes, respectively,
-18".48, - 19".49, - 20".,
and for the latitudes,
+ 0".47, + 0".30, + 0.14.
When these corrections are applied, we obtain the true places of the
planet for the instants when the light was emitted, but as seen from
the places of the earth at the instants of observation.
Next, each place of the sun must be reduced from the centre of
the earth to the point in which a line drawn from the planet through
the place of the observer cuts the plane of the ecliptic. For this
purpose we have, for Ann Arbor,
/' = 42~ 5'.4,. log p 9.99935;
and the mean time of observation being converted into sidereal time
gives, for the three observations,
o0 3 29m 18, O' = 21" 48- 17J, 8 21P 18 55,
which are the right ascensions of the geocentric zenith, of which sp'
is in each case the declination. From these we derive the longitude
and latitude of the zenith for each observation, namely,
o0 60~ 33'.9,'0 - 347~ 0'.4, o1 —" 342~ 59'.2,
bo-+22 25.0, bo' -+50 -15-8, bo- + -53 41.6.
Then, by means of equations (4), we obtain
A o - 18".92, A' = 36".94, A" = - 25".76,
A log Ro - 0.0001084, A log Ro -0.0002201,
A log R" = - 0.0002796.
For the reduction of time, we have the values + 0S.15, + 0".28, and
+ 0".34, which are so small that they may be neglected.
266 THEORETICAL ASTRONOMY.
Finally, the longitudes of both the sun and planet are reduced to
the mean equinox of 1863.0 by applying the corrections
-50".95, - 51".52, - 52".14;
and the latitudes of the planet are reduced to the ecliptic of the same
date by applying the corrections - 0".15, - 0".14, and - 0".14,
respectively.
Collecting together and applying the several corrections thus obtained for the places of the sun and of the planet, reducing the uncorrected times of observation to the meridian of Washington, and
expressing them in days from the beginning of the year, we have the
following data:to - 257.68079, 17~ 46' 28".17, f + 30 8' 43".51,
to - 264.42570, )' 16 40 25.19, 13' - 2 52 27.62,
0" = 271.38625, A" - 15 15 44.03, i" +2 32 42.98,
o - 172~ 0' 32".23, log R 0.0021056,
0' - 178 35 48.74, logR' 0.0011656,
0" 185 25 36.90, log R" 0.0002378.
The numerical values of the several corrections to be applied to
the data furnished by observation and by the solar tables should be
checked by duplicate calculation, since an error in any of these reductions will not be indicated until after the entire calculation of the
elements has been effected.
By means of the equations
R'R" sin (0"- 0') N BR' sin(0' - 0)
RR" sin (0" — 0)' -RR" sin (0"- 0)'
tan t' tan (' O')
tan w s -- ~ tan Co ~,
sin (A'-( i)') cos Wwe obtain
log NV 9.7087449, log N" = 9.6950091,
~' -- 161~ 42' 13".16,
log (R' sin,') - 9.4980010, log (R' cos,') = 9.9786355S.
The quadrant in which 4/ must be taken is determined by the conditions that 4/ must be less than 180~, and that cos/' and cos ('- 0')
must have the same sign. Then from
NUMERICAL EXAMPLE. 267
sin (/s" + f )
tan Isin ((- (A" + A) - K)O- 2 (/ sec ('" - 2),
2 cos C3 cos pl
tan Icos ((" +- 2) - K) - 2 cos cosee (" 2);
2 cos IOos 13S'
sin (/-' - 1
tall o - sin (' - ) tan, a sin (Cos3 iotan
b Rsin (O K)' sin (0'- K)
in ( c
CIo ao
R" sin (e("-K) sec i' RR" sin (" — )
a csin ( )' ao sin (2" -')
we compute K, I,, a, b, c, d, fand h. The angle ITmust be less
than 90~, and the value of 0n must be determined with the greatest
possible accuracy, since on this the accuracy of the resulting elements
principally depends. Thus we obtain
K — 4 47' 29".48, log tan I=- 9.3884640,
30 20 52' 59"1 5 log ao 6.8013583~,
log b 2.5456342n, log c - 2.2328550,,
log d - 1.2437914, logf = 1.3587437,, logh - 3.9247691.
The formulae
sin (" - 2') B" sin (2" - 0")
-/11r sin (A"-2) d
iM sin (A' - A) R sin (A -0)
1 sin (A" - A) b
h sin (A"- K),,h sin ( - K)
- d' b~'
give
log M1 - 9.8946712, log M" -- 9.6690383,
log 31 = 1.9404111, log M" - 0.7306625,.
The quantities thus far obtained remain unchanged in the successive approximations to the values of P and Q.
For the first hypothesis, from
T - (to —, tot) " = (to' - to)
P, Q-",
b +Pd
bo P 1 + P o =o-c, lo 2o Q,
~0 sin 0 - R' sin',
mo cos o - R' cos
MO- R3 si 0
0 ~B ^^sin /
268 THEORETICAL ASTRON'OMY.
we obtain
log 9.0782249, log r" 9.0645575,
log P- 9.9863326, log Q - 8.1427824,
log cq = 2.2298567,, log o - 0.0704470,
log lo - 0.0716091, log % =0 0.3326925,
C 8~ 24' 49".74, logo, -- 1.2449136.
The quadrant in which C must be situated is determined by the condition that ^q shall have the same sign as 10.
The value of z' must now be found by trial from the equation
sin (z' - CQ - m sin4 z'.
Table XII. shows that of the four roots of this equation one exceeds
180~, and is therefore excluded by the condition that sin z' must be
positive, and that two of these roots give z' greater than 1800 --
and are excluded by the condition that z' must be less than 180~ —'.
The remaining root is that which belongs to the orbit of the planet,
and it is shown to be approximately 10~ 40'; but the correct value
is found from the last equation by a few trials to be
Z - 9~' 22".96.
The root which corresponds to the orbit of the earth is 18~ 20' 41",
and differs very little from 180~ - /.
Next, from
SR sin, R' sin (z'+ 4)
r' --, z' ~ cos ft,
n - -- 1 Q n nP,
f Nil
P M. +M ),
we derive
log r' - 0.3025672, log p' 0.0123991,
log n = 9.7061229, log in" - 9.6924555,
log p = 0.0254823, log p" =0.0028859.
The values of the curtate distances having thus been found, the
heliocentric places for the three observations are now computed from
NUMERICAL EXAMPLE. 269
r cos b cos (I - ) p cos ( - ) - R,
r cos b sin (I - ) - p sin (A - ),
r sinb -p tan f;
r' cos b' cos (I' - 0') p' cos ('-') - R',
r' cos b' sin (' 0')' sin (2'- O'),
9f sin b' = p' tan r';
r" cos b" cos (1" - 0") p" cos (" - Oi) - R",
r" cos b" sin (I" - 0") p" sin (" - 0"),
r" sin b" - p" tan f",
which give
I - 5~ 14' 39".53, log tan b 8.4615572, logr 0.3040994,
I' - 7 45 11.28, log tan b' 8.4107555, log r' 0.3025673,
1" 10 21 34.57, log tan b" = 8.3497911, log r" -0.3011010.
The agreement of the value of log r' thus obtained with that already
found, is a proof of part of the calculation. Then, from
tan b"- + tan b
tan i sin (I (1" + 1) - -2 cos (1"-.)'. -tanb" -tanb
tan i cos ( (l" + - ) ) -- 2.in b ( 1.~~IIID~VU C~O T Vj ~bl 2 sin (I" — )'
tan u — = tan = tan =
cost cosi cost
we get
= 2070 2' 38".16, i = 40 27' 23".84,
u 158~ 8' 25".78,' = 1600 39' 18".13, u" = 163~ 16' 4".42.
The equation
tan b' — tan i sin (I' — f )
gives log tan 6b' 8.4107514, which differs 0.0000041 from the value
already found directly from o'. This difference, however, amounts
to only 0".05 in the value of the heliocentric latitude, and is due to
errors of calculation. If we compute n and n" from the equations
r'r" sill (V'"-'),, rr' sin (u' - u)
rr" sn(- ) r" sin ('a"" - )r'
the results should agree with the values of these quantities previously
computed directly from P and Q. Using the values of u, u', and
It" just found, we obtain
log n = 9.7061158, log 2t" = 9.6924683,
270 THEORETICAL ASTRONOMY.
which differ in the last decimal places from the values used in finding
p and p". According to the equations
d log n - 21.055 cot (u"- u') du',
d log n" - + 21.055 cot (cu' - u) did',
the differences of log n and log yi" being expressed in units of the
seventh decimal place, the correction to u' necessary to make the two
values of loga agree is -0".15; but for the agreement of the two
values of log y", u' must be diminished by 0".26, so that it appears
that this proof is not complete, although near enough for the first
approxination. It should be observed, however, that a great circle
passing through the extreme observed places of the planet passes
very nearly through the third place of the sun, and hence the values
of p and pl as determined by means of the last two of equations (18)
are somewhat uncertain. In this case it would be advisable to compute p and p", as soon as pt has been found, by means of the equations (22) and (23). Thus, from these equations we obtain
log p 0.0254918, log p" = 0.0028874,
and hence
1 5~ 14' 40".05, log tan b - 8.4615619, log r = 0.3041042,
I"= 10 21 34.)1, log tan b" 8.3497919, log r" = 0.3011017,' =207~ 2' 32".97, i - 4~ 27' 25".13,
u = 158~ 8' 31".47, u'' 160~ 39' 23".31, at" 163~ 16' 9".22.
The value of log tan b4 derived from A' and these values of Q and i,
is 8.4107555, agreeing exactly with that derived from p' directly.
The values of n and nt' given by these last results for u,'t' and l",
are
log n = 9.7061144, log n" 9.6924640;
and this proof will be complete if we apply the correction de=d - 0".18
to the value of iu', so that we have't" u' -- 2~ 36' 46".09, t' - u - 2~ 30' 51".66.
The results which have thus been obtained enable us to proceed to
a second approximation to the correct values of P and Q, and we
may also correct the times of observation for the time of aberration
by means of the formulae
t = to Cp see f, t' - t' - Cp' see p', t" to — Cop" see P",
wherein log C= 7.760523, expressed in parts of a day. Thus we get
t =257.67467, t'- 264.41976, t" - 271.38044,
NUJMERICAL EXAMPLE. 271
and hence
log T = 9.0782331, log r' = 9.3724848, log " - 9.0645692.
Then, to find the ratios denoted by s and s", we have;Ia
tanX - r,
sin r cos G sin (" -'),
sin r sin G - cos (u" - u') cos 2Z,
cos r coo s (u" -') sin 2Z;
\r
tanX" -,
sin r" cos G" - sin - (t' -- ),
sin r" sin G" cos' (' - u) cos 2z",
cos r" cos i (' — ~) sin 2/;
= 2'. sin2'— (r'+r")3 cos' r cos r
~ ~2 2sin2 1,
~_ sin" ^^
m- (r + r')" cos3' os
from which we obtain
x - 440 57' 6".00, Z" 440 56' 57".50,
r= 1 18 35.90, /' 1 15 40.69,
log mn = 6.3482114, log mn" 6.3163548,
logj - 6.1163135, logj" = 6.0834230.
From these, by means of the equations
mn. m
7 _ 6+jl x _ ff 2!
using Tables XIII. and XIV., we compute s and s". First, in the
case of s, we assume
j += j -= 0.0002675,
I +K
and, with this as the argument, Table XIII. gives log s2 - 0.0002581.
Hence we obtain x' 0.000092, and, with this as the argument,
Table XIV. gives - = 0.00000001; and, therefore, it appears that a
repetition of the calculation is unnecessary. Thus we obtain
logs -0.0001290, logs"= 0.0001200.
When the intervals are small, it is not necessary to use the formulae
272 THEORETICAL ASTRONOMY.
in the complete form here given, since these ratios may then be found
by a simpler process, as will appear in the sequel. Then, from
8
P-= *, -'
77TT rt2
Q ss" rr cos (u" -') os (( "- - )) cos (u — u)'
we find
log P - 9.9863451, log Q 8.1431341,
with which the second approximation may be completed. We now
compute c, ko, lo, z', &c. precisely as in the first approximation; but
we shall prefer, for the reason already stated, the values of p and op"
computed by means of the equations (22) and (23) instead of those
obtained from the last two of the formula (18). The results thus
derived are as follows:log c, 2.2298499, log ko - 0.0714280,
log 1o 0.0719540, log 0 = 0.3332233,
C 8~ 24' 12".48, log m,=- 1.2447277,
z - 9 0' 30".84,
log r' 0.3032587, log p' - 0.0137621,
log n - 9.7061153, log n" —- 9.6924604,
log -p 0.0269143, log" — 0.0041748,
I = 5~ 15' 57".26, log tan b =8.4622524, logr =0.3048368,
1' - 7 46 2.76, log tan b' 8.4114276, log r' 0.3032587,
l" 10 22 0.91, log tan b" 8.3504332, log r"= 0.3017481,
a = 207~ 0' 0".72, i - 4~ 28' 35".20,
= 1580 12' 19".54, u'= 160~ 42' 45".82, u" 163~ 19' 7".14.
The agreement of the two values of log r' is complete, and the value
of log tan b' computed from
tan b' - tan i sin (T'- ),
is log tan b' 8.4114279, agreeing with the result derived directly
from p'. The values of n and nt obtained from the equations (54)
are
log n = 9.7061156, log n" = 9.6924603,
which agree with the values already used in computing p and p", and
the proof of the calculation is complete. We have, therefore,' -- u' = 2~ 36' 21".32, u't — u = 2~ 30' 26".28, u" — u - 5~ 6' 47".60.
From these values of Iul - u' and u' - u, we obtain
log s = 0.0001284, log s" - 0.0001193,
NUMERICAL EXAMPLE. 273
and, recomputing P and Q, we get
log P = 9.9863452, log Q = 8.1431359,
which differ so little from the preceding values of these quantities
that another approximation is unnecessary. We may, therefore, from
the results already derived, complete the determination of the elements
of the orbit.
The equations
tan X' 1-,
sin r' cos G' sin 1 (u" t- ),
sin' sin G' = cos - (a" - u) cos 2/,
cos = cos ^ ('t" - u) sin 2'/,'2 2 "' sin2 r'
m = (r +r" cos co
give
x' =440 53' 53".25, r' 20 33' 52".97, log tan' 8.9011435,
log m' 6.9332999, logj 6.7001345.
From these, by means of the formulae
and Tables XIII. and XIV., we obtain
log s'2 0.0009908, log x' = 6.5494116.
Then from
8'rr" sin (u"- U) )
we get
logp - 0.3691818.
The values of logp given by
= sr r" srr'sin (' sn ) )2
are 0.3691824 and 0.3691814, the mean of which agrees with the
result obtained from u" - -u, and the differences between the separate
results are so small that the approximation to P and Q is sufficient.
The equations
sin k (E"- E) - V',
sin-! (Ot" - /a)
cos p =_.- ~ -,
a cos
18
274 THEORETICAL ASTRONOMY.
give
] (E" - E) = 10 4' 42".903, log (a cos A)= 0.3770315,
log cos -p = 9.9921503.
Next, from
e sin (w - (u + Iu))= ~' - tan G',
cosr/rr
e cos ( ~ (le + ))=- sec= (e - ),
cos l/rr
we obtain
o = 190~ 15' 39".57, log e = log sin -= 9.2751434,
s= -10 51 39.62, - o + g -- 37~ 15' 40".29.
This value of p gives log cos S -9.9921501, agreeing with the result
already found.
To find a and u, we have
p k
a -- a y- 3X
COS2 g a3
cos? x
the value of k expressed in seconds of arc being log k = 3.5500066,
from which the results are
log a = 0.3848816, log4 -= 2.9726842.
The true anomalies are given by
V =- U w, V- =U' -w, v- - =u",_
according to which we have
v - 327~ 56' 39".97,'- = 330~ 27' 6".25, v" = 333~ 3' 27".57.
If we compute r, r', and r" from these values by means of the polar
equation of the ellipse, we get
logr - 0.3048367, log r' 0.3032586, log r" = 0.3017481,
and the agreement of these results with those derived directly from
p, p', and p" is a further proof of the calculation.
The equations
tan,E = tan (45~ - p) tan Iv,
tan lE' = tan (45~ - ) tan Iv',
tan E" = tan (45~- ~f) tan Iv"
give
E = 333~ 17' 28".18, E' = 335~ 24' 38".00, E" = 337~ 36' 19".78.
NUMERICAL EXAMPLE. 275
The value of (E" - E) thus obtained differs only 0".003 from that
computed directly from x'.
Finally, for the mean anomalies we have
M= E - e sin E, M' -E' e sin E', M" E" e sin E",
from which we get
M= 338~ 8' 36".71, M' - 339~ 54' 10".61, M" 341~ 43' 6".97;
and if M. denotes the mean anomaly for the date T= 1863 Sept. 21.5
Washington mean time, from the formulse
_A= d M, (t ~ T)
M' (t'- T)
M" - ( t" T),
we obtain the three values 339~ 55' 25".97, 3390 55' 25".96, and
339" 55' 25".96, the mean of which gives
M - 339~ 55' 25".96.
The agreement of the three results for M0 is a final proof of the
accuracy of the entire calculation of the elements.
Collecting together the separate results obtained, we have the following elements:
Epoch = 1863 Sept. 21.5 Washington mean time.
M- 339~ 55' 25".96
-r= 37 15 40.29)
= - 2097 0 0.72 Ecliptic and Mean
i 4 28 35.20 Equinox 1863.0.
i- 4 28 35.20
-=- 10 51 39.62
log a 0.3848816
log t= 2.9726842
- 939".04022.
If we compute the geocentric right ascension and declination of
the planet directly from these elements for the dates of the observations, as corrected for the time of aberration, and then reduce the
observations to the centre of the earth by applying the corrections
for parallax, the comparison of the results thus obtained will show
how closely the elements represent the places on which they are
based. Thus, we compute first the auxiliary constants for the equator,
using the mean obliquity of the ecliptic,
E = 23~ 27' 24".96,
276 THEORETICAL ASTRONOMY.
and the following expressions for the heliocentric co-ordinates of the
planet are obtained:
x -r [9.9997272] sin (296~ 55' 46".05 + u),
y = r [9.9744699] sin (206 12 42.79 + iu),
z =r [9.5249539] sin (212 39 14.62 + u).
The numbers enclosed in the brackets are the logarithms of sin a,
sin b, and sin c, respectively; and these equations give the co-ordinates
referred to the mean equinox and equator of 1863.0.
The places of the sun for the corrected times of observation, and
referred to the mean equinox of 1863.0, are
True Longitude. Latitude. Log R.
172~ 0' 29".5 - 0".07 0.0022146,
178 36 4.5 + 0.77 0.0013864,
185 25 42.0 0.67 0.0005182.
If we compute from these values, by means of the equations (104)1,
the co-ordinates of the sun, and combine them with the corresponding
heliocentric co-ordinates of the planet, we obtain the following geocentric places of the planet:
- =15~ 10' 29".06, = + 9~ 53' 16".72, log A = 0.02726,
a =14 15 0.22,' - 9 12 51.29, log J' = 0.01410,
a" -13 3 49.47, 3" -+ 8 21 54.46, log J" = 0.00433.
To reduce these places to the apparent equinox of the date of observation, the corrections
- 48".14, + 48".54, + 48".91,
must be applied to the right ascensions, respectively, and
+ 18".55, - 18".92, + 19".31,
to the declinations. Thus we obtain:
Washington M. T. Comp. a. Comp. d.
1863 Sept. 14.67467 1' 0" 45s.15 + 9~ 53' 35".3,
21.41976 0 57 3.25 9 13 10.2,
28.38044 0 52 18.56 -+8 22 13.8.
The corrections to be applied to the respective observations, in order
to reduce them to the centre of the earth, are +- 0.24, - 0.31, - 0.34
in right ascension, and + 4".5, + 4".8, + 5".1 in declination, so
that we have, for the same dates,
NUMERICAL EXAMPLE. 277
Observed ao. Observed J.
1 0' 45s.15 + 9~ 53' 35".3,
0 57 3.26 9 13 10.3,
0 52 18.56 +8 22 13.8.
The comparison of these with the computed values shows that the
extreme places are exactly represented, while the difference in the
middle place amounts to only 08.01 in right ascension, and to 0".1
in declination. It appears, therefore, that the observations are completely satisfied by the elements obtained, and that the preliminary
corrections for aberration and parallax, as determined by the equations (1) and (4), have been correctly computed.
It cannot be expected that a system of elements derived from observations including an interval of only fourteen days, will be so
exact as the results which are obtained from a series of observations
or from those including a much longer interval of time; and although
the elements which have been derived completely represent the data,
yet, on account of the smallness of 9' - -, this difference being only
31'".893, the slight errors of observation have considerable influence
in the final results.
When approximate elements are already known, so that the correction for parallax may be applied directly to the observations, in
order to take into account the latitude of the sun, the observed places
of the body must be reduced, by means of equation (6), to the point
in which a perpendicular let fall from the centre of the earth to the
plane of the ecliptic cuts that plane. The times of observation must
also be corrected for the time of aberration, and the corresponding
places of both the planet and the sun must be reduced to the ecliptic
and mean equinox of a fixed epoch; and further, the reduction to
the fixed ecliptic should precede the application of equation (6).
If the intervals between the times of observation are considerable,
it may become necessary to make three or more approximations to the
values of P and Q, and in this case the equations (82) may be applied.
But when approximate elements are already known, it will be advantageous to compute the first assumed values of P and Q directly
from these elements by means of the equations (44) or by means of
(48) and (51); and the ratios s and s" may be found directly from the
equations (46). In the case of very eccentric orbits this is indispensable, if it be desired to avoid prolixity in the numerical calculation,
since otherwise the successive approximations to P and Q will slowly
approach the limits required.
278 THEORETICAL ASTRONOMY.
The various modifications of the formulse for certain special cases,
as well as the formulae which must be used in the case of parabolic
and hyperbolic orbits, and of those differing but little from the
parabola, have been given in a form such that they require no further illustration.
94. In the determination of an unknown orbit, if the intervals are
considerably unequal, it will be advantageous to correct the first
assumed value of P before completing the first approximation in the
manner already illustrated. The assumption of
Q -t
is correct to terms of the fourth order with respect to the time, and
for the same degree of approximation to P we must, according to
equation (28)3, use the expression
Ti!+ 6T2- )
TfT
which becomes equal to only when the intervals are equal. The
first assumed values
P= Q=,
furnish, with very little labor, an approximate value of r'; and then,
with the values of P and Q, derived from
P r l+ ), Q=", (98)
the entire calculation should be completed precisely as in the example
given. Thus, in this example, the first assumed values give
log r' 0.30257,
and, recomputing P by means of the first of these equations, we get
log P = 9.9863404, log Q = 8.1427822,
with which, if the first approximation to the elements be completed,
the results will differ but little from those obtained, without this correction, from the second hypothesis. If the times had been already
corrected for the time of aberration, the agreement would be still
closer.
The comparison of equations (46) with (25)3 gives, to terms of the
fourth order,
NUMERICAL EXAMPLE. 279
T 2 Tt2
1 -1 r 1 r3l+ -
and, if the intervals are equal, this value of sa is correct to terms of
the fifth order. Since
loge s = loge (I + (s - 1)) = 8 - 1 - ( - 1)2 + &c.,
we have, neglecting terms of the fourth order,
10 r2: (99)
logs -. r (99)
in which log 20 —8.8596330. We have, also, to the same degree of
approximation,
log s' r3, log 5" = 6 03 (100)
For the values
log =- 9.0782331, log - =- 9.3724848, log r" = 9.0645692,
log r' - 0.3032587,
these formulse give
log s 0.0001277, log s' = 0.0004953, log s" = 0.0001199,
which differ but little from the correct values 0.0001284, 0.0004954,
and 0.0001193 previously obtained.
Since
sec3' 1 + 6 sinU2 r' + &c.,
the second of equations (65) gives
Tr2 6~ri
=n' = + ~-,~ + 6p~ sin2 lr' + &c.
+ (r - )+ (r + r")3 s
Substituting this value in the first of equations (66), we get
s'" (s' - 1 r + )- + y (r + rr)3 y (sin2 + &e.
If we neglect terms of the fourth order with respect to the time, it
will be sufficient in this equation to put y' -, according to (71), and
hence we have
T2
S'2 (S _ 1)4= __ ____
3 (r - r" )3
and, since s - 1 is of the second order with respect to r', we have,
to terms of the fourth order,
S2 (s'- 1) = logeo 8.
280 THEORETICAL ASTRONOMY.
Therefore,
T'2
log s'- 4 2o (101)
which, when the intervals are small, may be used to find s' from r
and r". In the same manner, we obtain
og22 T2
log S = 4 Ao (r,+ r,)3, log SI' 4 r*o (102)
For logarithmic calculation, when addition and subtraction logarithms are not used, it is more convenient to introduce the auxiliary
angles X, X', and X', by means of which these formulae become
z2osCs6 Z log s' 4Abl 2 Cos(V' lo eOTIN;s
log s 3= o~ cos logs ^ o~,.c3 logs" -c s6 x (103)
3 r,3 3g,
in which log2- = 9.7627230. For the first approximation these
equations will be sufficient, even when the intervals are considerable,
to determine the values of s and s" required in correcting P and Q.
The values of t, r', r", and r'X above given, in connection with
log r =0.3048368, log r" = 0.3017481,
give
log s 0.0001284, log s' 0.0004951, log s" = 0.0001193.
These results for log s and log s" are correct, and that for log s' differs
only 3 in the seventh decimal place from the correct value.
ORBIT FROM FOUR OBSERVATIONS. 281
CHAPTER V.
DETERMINATION OF THE ORBIT OF A HEAVENLY BODY FROM FOUR OBSERVATIONS,
OF WHICH THE SECOND AND THIRD MUST BE COMPLETE.
95. THE formulae given in the preceding chapter are not sufficient
to determine the elements of the orbit of a heavenly body when its
apparent path is in the plane of the ecliptic. In this case, however,
the position of the plane of the orbit being known, only four elements remain to be determined, and four observed longitudes will
furnish the necessary equations. There is no instance of an orbit
whose inclination is zero; but, although no such case may occur, it may
happen that the inclination is very small, and that the elements
derived from three observations will on this account be uncertain,
and especially so, if the observations are not very exact. The difficulty thus encountered may be remedied by using for the data in the
determination of the elements one or more additional observations,
and neglecting those latitudes which are regarded as most uncertain.
The formulae, however, are most convenient, and lead most expeditiously to a knowledge of the elements of an orbit wholly unknown,
when they are made to depend on four observations, the second and
third of which must be complete; but of the extreme observations
only the longitudes are absolutely required.
The preliminary reductions to be applied to the data are derived
precisely as explained in the preceding chapter, preparatory to a determination of the elements of the orbit from three observations.
Let t, t', t", t"' be the times of observation, r, r', r", r" the radiivectores of the body, tu, UC', ", Ui"' the corresponding arguments of
the latitude, R, Rn R", R"' the distances of the earth from the sun,
and 0', 0", 0"' the longitudes of the sun corresponding to
these times. Let us also put
[rVr"'] = rr"' sin ("' - u'),
an[r"r'] r"r"' sin ("' - u"),
and
n'- n" [rTr-"' ]
282 THEORETICAL ASTRONOMY.
Then, according to the equations (5)3, we shall have
nx - x' -+ n"i" - 0,
ny - + -"y" - O,
n'x' -'f x 0 (2)
n'y' - ~' + n"'y"' =- 0.
Let 2, A', A", 2"' be the observed longitudes, i, 9', P", a"' the observed latitudes corresponding to the times t, t', t", t"', respectively,
and d, Jz', a, A'mp the distances of the body from the earth. Further,
let
d"' cos A"' - p"',
and for the last place we have
x"' = C pw cos, cos ()'
y s"' - p"' sin R in "'.
Introducing these values of x"' and y"', and the corresponding values
of x, x', x", y, y', y" into the equations (2), they become
O = n (p cos A - R cos 0) - (p' cos A' -' cos 0')
+ ni" (p" cos A" R- " cos 0"),
O - n (p sin A -R sin 0) -(p' sin A' - R' sin 0')'
+ -- " (P" sin A" -'" sin 0"),
0 n' (p' cos A' -' cos 0') - (p" cos A" - R" cos 0") (3)
+ n"' (p"' cos A"' - R"' cos ( "'),
0' (p' sin A' -' sin 0') (p" sin A" - R" sin 0")
+- n"' (P"' sin "' - R"' sin G"').
If we multiply the first of these equations by sin 2, and the second
by - cos 2, and add the products, we get
0 = nR sin (A - 0) - (' sin (A' - A) + R' sin (A -'))
+'n (p" sin (A" - A) + " sin (A - ")); (4)
and in a similar manner, from the third and fourth equations, we
find
0 - n' (p' sil (A"' - A') - R' sin (A"' - 0')) (5)
- (p" sin (A"'- A") -- R" sin('- ")) -n'"R'" si ("' — 0"').
Whenever the values of n,', n", and n"' are known, or may be
determined in functions of the time so as to satisfy the conditions of
motion in a conic section, these equations become distinct or independent of each other; and, since only two unknown quantities p'
ORBIT FROMI FOUR OBSERVATIONS. 283
and p" are involved in them, they will enable us to determine these
curtate distances.
Let us now put
cos j' sin (' — ) =A, cos A sin (" - ) B,
cos P sin (A"'- ") - C, cos sin (A"'- 2') D- 1,
and the preceding equations give
Ap' sec'- Bn"p" sec " =nR sin (A -0) -' sin (- 0')
+ n"R" sin ( - 0"),
Dn'p' sec'- Cp" sec i"- n'R' sin ("' — 0') - R" sin (A"'- 0") (7)
+ n"'R"' sin (A"' - 0"').
If we assume for n and n" their values in the case of the orbit of
the earth, which is equivalent to neglecting terms of the second order
in the equations (26).3 the second member of the first of these equations reduces rigorously to zero; and in the same manner it can be
shown that when similar terms of the second order in the corresponding expressions for n' and n" are neglected, the second member
of the last equation reduces to zero. Hence the second member of
each of these equations will generally differ from zero by a quantity
which is of at least the second order with respect to the intervals of
time between the observations. The coefficients of p' and p" are of
the first order, and it is easily seen that if we eliminate p" from
these equations, the resulting equation for p' is such that an error of
the second order in the values of n and n" may produce an error of
the order zero in the result for p', so that it will not be even an
approximation to the correct value; and the same is true in the case
of p". It is necessary, therefore, to retain terms of the second order
in the first assumed values for n n, n", and n"/; and, since the
terms of the second order involve r' and r", we thus introduce two
additional unknown quantities. Hence two additional equations involving r', r", p', p" and quantities derived from observation, must
be obtained, so that by elimination the values of the quantities sought
may be found.
From equation (34)4 we have
p' sec i' =- R' cos' i: V r'2 - R2 sin2', (8)
which is one of the equations required; and similarly we find, for
the other equation,
p" sec i" = R" cos 4" + /'r"2 RB" sin2 4". (9)
284 TIEORETICAL ASTRONOMY.
Introducing these values into the equations (7), and putting
x' - =t 1/2"_ sin'2 4',
x2 == Vr"~R/12 sin'' (10)
x" — /r'a- R"I2 sin2 4",
we get
Ax' - Bn"x" = nR sin (2 - 0) - R' sin ( - 0')
+ nf"R" sin (2 -3 0") - AR' cos 4' + n"BR" cos 4",'x' - Cx" = l'R' sin ("' - 0(') - R sin (A"' - 0")
+- n"'R"' sin (A"' - 0"') - n'DR' cos 4' + CR" cos ".
Let us now put
B h' Dhi
A h', C -
or, cos i' sin (A" -) h" cos /' sin (A"' —')
cos /' sin (' - )' - os 3" sin (A"'- ")'
R R' sin(A -')
R' cos 4'+ ~A a,
cos, - R" sin ("' - 0"),,
hR" cos 4' R sin(2'
R" sin ( -(DI) _,
h'/~" cosa -" +- C-,
cos -' sin (' - 0')h"R' cos ~~ ) -- e.R sin (A - 0) _ d R"' sin (2"' - O"')
A C
and we have
x' = h'n"x" + nd'- a' + -n"e',
- h"n'x' + n"'d" - a" + n'. (12)
These equations will serve to determine x' and x", and hence r' and
r", as soon as the values of n, n', n", and n"' are known.
96. In order to include terms of the second order in the values of
n and "I, we have, from the equations (26)3,
"(+
T d= i t6 r 3 ) =- 1 + 3 (r' + )
and, putting
P' s q= (n + "- 1) r'3, (13)
these give
ORBIT FROM FOUR OBSERVATIONS. 285
= 7-2_1 2 _ l2
P) 7"\'-~Q6r" 13 (14)
^-^.
Let us now put
r"_-= k (t"' - "), r'- k (et — O, (15)
and, making the necessary changes in the notation in equations (26)3,
we obtain, r (+ ~ lr ( +~4 dt
7 - iI (1 + + r"' + 2 ( ~ + - L1 ) dr" 1
From these we get, including terms of the second order,
a"' = t ^1 + 6- r"+ ) ( I N'\ (, I l r II+I',
and hence, if we put
6 r,,1 - kr"'
P _ (_-.' T " = (n' + "T' - 1) r"3, (17)
we shall have, since r- = T + r'T",
Qua'" ^ ^~ 11) (18)'
When the intervals are equal, we have
p 7. a,,, P' T....
~{ =
and these eressins m these e eindin te e of an unknown orbit,
for the first approximation to th e quantities.
The equations (13) and (17) give
we shall have since' 1 + ( r3" )
1"' z1'P"
and, introducing these values, the equations (12) become
ond the inrod vian thee values, the euavetioecm
236 THEORETICAL ASTRONOMY.
+ p I + -r )'(h'' + Pd' +c') - a',
~~~~~~~~1 /'\ (20)
X 1 -+ P ( + rQ4 3 ) (h'. + P"d" + ")- a".
Let us now put
P'd' +- c' h'
l+"i
Uip,1~do -oi h+'i p(21)
"d" -+ — c", C
7Ipq- P" -- P"'
and we shall have
x-(1 + ) (f'x"+ c') - a,
(22)
~ -(i +;)(f I + Cot) " ~
We have, further, from equations (10),
13 (x,2 R 2 sin2')4 3
r (X 1 (23)- (x2, + R"2 sin2,")~,
If we substitute these values of r'3 and r"3 in equations (22), the two
resulting equations will contain only two unknown quantities x' and
x", when P', P", Q', and Q" are known, and hence they will be
sufficient to solve the problem. But if we effect the elimination of
either of the unknown quantities directly, the resulting equation
becomes of a high order. It is necessary, therefore, in the numerical
application, to solve the equations (22) by successive trials, which
may be readily effected.
If z' represents the angle at the planet between the sun and the
earth at the time of the second observation, and z" the same angle at
the time of the third observation, we shall have, R' sin {'
sin z
(24)
RB" sin 4" (24)
sin z
Substituting these values of r' and r" in equations (10), we get
(25)
x' =r cosz', (CO
and hence
ORBIT FROM FOUR OBSERVATIONS. 287, R' sin'
tan z'' sin
tan z -P
x
by means of which we may find z' and z" as soon as x' and x" shall
have been determined; and then r' and r" are obtained from (24) or
(25). The last equations show that when x' is negative, z' must be
greater than 90~, and hence that in this case r' is less than R'.
In the numerical application of equations (22), for a first approximation to the values of x' and x", since Q' and Q" are quantities of
the second order with respect to r or r"', we may generally put'- o, q"- o;
and we have
xfx = f"- +- co- a"',
or, by elimination,
coI fj —'c0" ffail a'
x' = - f'o~" — f" —a
1 -f'f"
e Co +i ff" Co' —f" a' — all
With the approximate values of x' and x" derived from these equations, we compute first r' and r" from the equations (26) and (24),
and then new values of x' and xa from (22), the operation being
repeated until the true values are obtained. To facilitate these approximations, the equations (22) give
if _X + a' Cot
-, f' (1 +')'
x" + aff c" (27)
fy T//3)
Let an approximate value of x' be designated by xo0, and let the
value of x" derived from this by means of the first of equations (27)
be designated by x0". With the value of xa" for x" we derive a
new value of x' from the second of these equations, which we denote
by x,'. Then, recomputing x" and x', we obtain a third approximate
value of the latter quantity, which may be designated by x2'; and,
if we put
x - xo' = ao, x2' 1 = ao'
288 THEORETICAL ASTRONOMY.
we shall have, according to the equation (67)3, the necessary changes
being made in the notation,
x =1' a — x 0 X2 a (28)
a,- a. o ao -- ao'
The value of x' thus obtained will give, by means of the first of
equations (27), a new value of x", and the substitution of this in the
last of these equations will show whether the correct result has been
found. If a repetition of the calculation be found necessary, the
three values of x' which approximate nearest to the true value will,
by means of (28), give the correct result. In the same manner, if
we assume for x" the value derived by putting Q' = 0 and Q"= 0,
and compute x', three successive approximate results for x" will
enable us to interpolate the correct value.
When the elements of the orbit are already approximately known,
the first assumed value of x' should be derived from
x='r" - r' R'2 sin2 2'
instead of by putting Q' and Q" equal to zero.
97. It should be observed that when A2 =- or 2"' = ", the equations (22) are inapplicable, but that the original equations (7) give,
in this case, either o" or p' directly in terms of n and n" or of n'
and n"' and the data furnished by observation. If we divide the
first of equations (22) by h', we have
h' r'3 )h'' h')
The equations (21) give
p,d'
f' 1 Co' P_ d + c
ah -I + P" h P' - i P'
and from (11) we get
a' R' cos /'' sin ( -- 0')' A' h' B
co R" sin(2 - 0~")
==- R cosB + R sin ( D" (29)
hB
d' R sin () - ~
h' B
Then, if we put
Cfr_ Pd'.'
Gd = pi h' 3- hr~~~~
ORBIT FROM FOUR OBSERVATIONS. 289
c' d'
its value may be found from the results for, and, derived by
means of these equations, and we shall have
a;' 1
XI72,'p~(Q- (x" at' --
(30)
ht 1 +r P' ( XI r, + Co )
When' =, we have h'= oo, and this formula becomes
0-(lZ 1+ r3, ) (X + c')- (1 + P'),
the value of t being given by the first of equations (29) This
equation and the second of equations (22) are sufficient to determine
x' and x" in the special case under consideration.
The second of equations (22) may be treated in precisely the same
manner, so that when )"' = -l, it becomes
0-( + Q (,+ C)- (i + ),
and this must be solved in connection with the first of these equations
in order to find x' and x".
98. As soon as the numerical values of x' and x"t have been
derived, those of r' and r" may be found by means of the equations
(26) and (24). Then, according to (41)4, we have,' sin (z' + c'),
siln z'
I, _ sin (z" + ) cos")
p" - - cos fi~.
sin z"
The heliocentric places are then found from p' and p" by means of
the equations (71),, and the values of r' and }r" thus obtained should
agree with those already derived. From these places we compute
the position of the plane of the orbit, and thence the arguments of
the latitude for the times t' and t".
The values of e', r", I', 6", n, n/, n', and n"'! enable us to determine r, r"', t, and t"'/. Thus, we have
[rr'r] -r r" sin (a~" -'),
and, from the equations (1) and (3),,
19
290 THEORETICAL ASTRONOMY,
nt
[ri"] - V- ['r"],
f/
[r"r"']-n'' [r'r"],
~'~:],,W [r'r"].
Therefore,
n"
r in ( ) sin (u', - s ),
r s sin'(i" si (' - ),
(32)
r" sin (u"'- 0') =? sin (,' ~ ),
r"' sin (') - - sin ( ).
From the first and second of these equations, by addition and subtraction, we get
r sin ((' - u) + -( (" ~-')) r + Fr sin (U"-'),
(33)
r cos (('' -u') + 1 (t" - u')) = -- cos 2 ('" - u'),
from which we may find r, ta' - u, and u =' - (u' - u).
In a similar manner, from the third and fourth of equations (32),
we obtain
9"' sin (('"' - u") + - (u"-')) " +' sinr (" ),
f r (34)
r' cos ((u"' - U") + (- t)) C,-'osr ( -l'),
from which to find r"' and u"'.
When the approximate values of r, rf, r", r"' and u, u', "/,'"
have been found, by means of the preceding equations, from the
assumed values of P', P", Q', and Q", the second approximation to
the elements may be commenced. But, in the case of an unknown
orbit, it will be expedient to derive, first, approximate values of r'
and r', using
r'g T~~T
and then recompute P' and P" by means of the equations (14) and
ORBIT FROM FOUR OBSERVATIONS. 291
(18), before finding u' and u". The terms of the second order will
thus be completely taken into account in the first approximation.
99. If the times of observation have not been corrected for the
time of aberration, as in the case of an orbit wholly unknown, this
correction may be applied before the second approximation to the
elements is effected, or at least before the final approximation is commenced. For this purpose, the distances of the body from the earth
for the four observations must be determined; and, since the curtate
distances p' and p" are already given, there remain only p and pi' to
be found. If we eliminate p' from the first two of equations (3), the
result is
"n" sin (" — 2')
P P sin (A ) (35)
n sin ()/- A)
nR sin (A'- 0) - R' sin (A'- 0') +- nm" R" sin (A' -0")
n sin (A'- A)
and, by eliminating p" from the last two of these equations, we also
obtain
P n'" sin (2"' —.")
n' R' sin (A" - (') - R" sin (2" - (") + n"' R"' sin (A" - ("')
"'1 sin (R"' — ")
by means of which p and p"' may be found. The combination of
the first and second of equations (3) gives
p, cos (A') p cos () "- ) (37)
nR cos (2 -- 0) -' cos (A- 0') + mR" " cos (A- 0")
+ n
and from the third and fourth we get
/ "' (z " (tIm'I-' A' ",
p"'- s ( Co "- A") P- cos (-"'-) (38)
(Alt- ~ co s --,).
n' R' cos ("'- (') -R" cos (A"' — ") + n"' R"' cos (A"'- "')
+ n"'
Further, instead of these, any of the various formulae which have
been given for finding the ratio of two curtate distances, may be
employed; but, if the latitudes /, i', &c. are very small, the values
of p and p"' which depend on the differences of the observed longitudes of the body must be preferred.
292 THEORETICAL ASTRONOMY.
The values of p'and p"' may also be derived by computing the
heliocentric places of the body for the times t and t"' by means of
the equations (82)i, and then finding the geocentric places, or those
which belong to the points to which the observations have been
reduced, by means of (90)1, writing p in place of J cos f. This
process affords a verification of the numerical calculation, namely,
the values of, and 2"' thus found should agree with those furnished
by observation, and the agreement of the computed latitudes d9 and
"' with those observed, in case the latter are given, will show how
nearly the position of the plane of the orbit as derived from the
second and third observations represents the extreme latitudes. If
it were not desirable to compute 2 and P'I in order to check the
calculation, even when: and 3"' are given by observation, we might
derive p and,o" from the equations
p r sinusin i cot,
p"' - r"' sin u"' sin i cot ", (
when the latitudes are not very small.
In the final approximation to the elements, and especially when
the position of the plane of the orbit cannot be obtained with the
required precision from the second and third observations, it will be
advantageous, provided that the data furnish the extreme latitudes
B9 and i9"', to compute p and p"' as soon as p' and,o" have been
found, and then find 1, 1"', b, and b"' directly from these by means
of the formule (71)3. The values of a and i may thus be obtained
from the extreme places, or, the heliocentric places for the times I'
and t"F being also computed directly from p' and p", from those
which are best suited to this purpose. But, since the data will be
more than sufficient for the solution of the problem, when the extreme
latitudes are used, if we compute the heliocentric latitudes b' and b-'from the equations
tan b' = tan i sin (1' - ),
tan b" = tan i sin (I"- g),
they will not agree exactly with the results obtained directly from p'
and p", unless the four observations are completely satisfied by the
elements obtained. The values of r' and r', however, computed
directly from p' and p" by means of (71)3, must agree with those
derived from x' and x".
The corrections to be applied to the times of observation on account
ORBIT FROM FOUR OBSERVATIONS. 293
of aberration may now be found. Thus, if to, to', t", and to"' are
the uncorrected times of observation, the corrected values will be
t - to -Cp sec f,
t' - to- Cp' see',
t" - to-t CP" secf", )
t"' = to"- Cp"' see i"',
wherein log C= 7.760523, and from these we derive the corrected
values of r, r', T", Z"', and zT'.
100. To find the values of P', P", Q', and Q", which will be
exact when r rrt, r"', and u,', Ull, u'll are accurately known, we
have, according to the equations (47)4 and (51)4, since Q' = Q,
P' —'
Tr S'
(41)
881' rr" cos t (t" - u'a) cos 2 (" -) cos, ('- )'
In a similar manner, if we designate by s"' the ratio of the sector
formed by the radii-vectores r" and r'" to the triangle formed by
the same radii-vectores and the chord joining their extremities, we
find
r s"'
T'
_ T < (42)
~ s "' r'r"' cos (u'"' - u") cos ('a"' - t) cos - (Z" -')'
The formule for finding the value of s"' are obtained from those for
s by writing ZX"f "', lG"', &c. in place of X,, G, &c., and using
r", r"', u"' - " instead of r', r", and u" - t', respectively.
By means of the results obtained from the first approximation to
the values of P', P", Q', and Q", we may, from equations (41) and
(42), derive new and more nearly accurate values of these quantities,
and, by repeating the calculation, the approximations to the exact
values may be carried to any extent which may be desirable. When
three approximate values of P' and Q', and of P" and Q", have
been derived, the next approximation will be facilitated by the use
of the formulae (82)4, as already explained.
When the values of P', P", Q', and Q" have been derived with
sufficient accuracy, we proceed from these to find the elements of the
orbit. After Q, i, r, r',r", r"',', U','a", and u'a" have been found,
the remaining elements may be derived from any two radii-vectores
294 THEORETICAL ASTRONOMY.
and the corresponding arguments of the latitude. It will be most
accurate, however, to derive the elements from r, r'", a, and a"'.
If the values of P', P", Q', and Q" have been obtained with'great
accuracy, the results derived from any two places will agree with
those obtained from the extreme places.
In the first place, from
tan o = I 7'
silno cos GO - sin ~ (u~"' - u), (43)
sin y sin GQ = cos 1 (t"' - t) cos 2o,
Cos y r cos - ("' - u) sin 2O
we find ro and G0. Then we have
ao — k (t'- t),
___________. sin (44)
MO (r +- r"')3 sro cos Cos (44)
o rn
n M o ^~9o
- +Jo + o 0 0 2,
from which, by means of Tables XIII. and XIV., to find so and x0.
We have, further,
So rr"' sin (t"' - ) )2
and the agreement of the value of p thus found with the separate
results for the same quantity obtained from the combination of any
two of the four places, will show the extent to which the approximation to P', P", Q', and Q" has been carried. The elements are now
to be computed from the extreme places precisely as explained in the
preceding chapter, using r"' in the place of r" in the formulse there
given and introducing the necessary modifications in the notation,
which have been already suggested and which will be indicated at
once.
101. EXAMPLE.-For the purpose of illustrating the application
of the formulne for the calculation of an orbit from four observations,
let us take the following normal places of Eurynome 0 derived by
comparing a series of observations with an ephemeris computed from
approximate elements.
Greenwich M. T. a d
1863 Sept. 20.0 140 30' 35".6 + 90 23' 49".7,
Dec. 9.0 9 54 17.0 2 53 41.8,
1864 Feb. 2.0 28 41 34.1 9 6 2.8,
April 30.0 74 29 58.9 + 19 35 41.5.
NUMERICAL EXAMPLE. 295
These normals give the geocentric places of the planet referred to the
mean equinox and equator of 1864.0, and free from aberration. For
the mean obliquity of the ecliptic of 1864.0, the American Nautical
Almanac gives
E-23~ 27' 24".49,
and, by means of this, converting the observed right ascensions and
declinations, as given by the normal places, into longitudes and latitudes, we get
Greenwich M. T. P
1863 Sept. 20.0 16~ 59' 9".42 + 2~ 56' 44".58,
Dec. 9.0 10 14 17.57 — 1 15 48.82,
1864 Feb. 2.0 29 53 21.99 2 29 57.38,
April 30.0 75 23 46.90 -3 4 44.49.
These places are referred to the ecliptic and mean equinox of 1864.0,
and, for the same dates, the geocentric latitudes of the sun referred.
also to the ecliptic of 1864.0 are
+ 0".60, +0".53, + 0".36, + 0".19.
For the reduction of the geocentric latitudes of the planet to the
point in which a perpendicular let fall from the centre of the earth
to the plane of the ecliptic cuts that plane, the equation (6)4 gives the
corrections - 0".57, - 0".38, - 0".18, and - 0".07 to be applied to
these latitudes respectively, the logarithms of the approximate distances of the planet from the earth being
0.02618, 0.13355, 0.29033, 0.44990.
Thus we obtain
t = 0.0, -16~ 59' 9".42, p ==+ 2~ 56' 44".01,
t' 80.0, A' -10 14 17.57, P' =-1 15 49.20,
t" = 135.0, A" =29 53 21.99, P" 2 29 57.56,
t'" 223.0, A"' 75 23 46.90, i"' = -3 4 44.56;
and, for the same times, the true places of the sun referred to the
mean equinox of 1864.0 are
O 177~ 0' 58".6, log R 0.0015899.
0' — 256 58 35.9, log R' 9.9932638,
0" - 312 57 49.8, log R" 9.9937748,
0"'- 40 21 26.8, log R"' " 0.0035149,
296 THEORETICAL ASTRONOMY.
From the equations
tan tan 3' tan (A' - )
t sin (A'-')' tan
tn w" (tan tan tan (A" - 0")
taln W' si. ~, tan " ~~
sin(,~ ) COS Wcos
we obtain
4' — 113~ 15' 20".10, log(R' cost') 9.5896777,
log(R' sin 4') = 9.9564624,
4"= 76 56 17.75, log (R" cos 4") 9.3478848,
log (R" sin 4") - 9.9823904.
The quadrant in which 4/ must be taken, is indicated by the condition that cos'5/ and cos(2'- 0') must have the same sign. The
same condition exists in the case of /". Then, the formulae
A -cos' sin (A' - ), B cos 1" sin (A"- ),
C cos 3" sin (A"'- A"), D - cos s (' sin (A"'-'),
B D
- h', - -h"
R' sin ( - 0')
a' =- R' cos -'+ -
a" " cos — B" sin (A"'- 0")
it - ^ f cos4- ^^
C
A
B" sin ( - 0")
c= h'R" cos 4~ + ~A ~ )
R' sin (A"' - (')
C" - h"R' cos -/'R sin (A- ) "' sin (A"' - I)
d — "A - C
give the following results:log A = 9.0699254X, log C = 9.8528803,
log B 9.3484939, log D 9.9577271,
log h' - 0.2785685, log h -= 0.1048468,
log a' 0.8834880,, log a" = 9.9752915,
log c' = 0.9012910,, log c" = 9.7267348n,
log d' = 0.4650841, log d" - 9.9096469.
We are now prepared to make the first hypothesis in regard to the
values of P', Q', P", and Q". If the elements were entirely unknown, it would be necessary, in the first instance, to assume for these
quantities the values given by the expressions
NUMERICAL EXAMPLE. 297
p,, rX QxA
P -- =-w~ -, Q,,==' -,";
then approximate values of r' and r" are readily obtained by means
of the equations (27), (26), and (24) or (25). The first assumed
value of x' to be used in the second member of the first of equations
(27), is obtained from the expression which results from (22) by
putting Q' = 0 and Q"- 0, namely,
c' - f'c'" — f'" - a'
1 -f'f"
after which the values of x' and x" will be obtained by trial from
(27). It should be remarked, further, that in the first determination
of an orbit entirely unknown, the intervals of time between the observations will generally be small, and hence the value of x' derived
from the assumption of Q' =0 and Q"- 0 will be sufficiently approximate to facilitate the solution of equations (27).
As soon as the approximate values of r' and r" have thus been
found, those of P' and P" must be recomputed from the expressions
- (1 ) (6 ~ ~- ~ 7- 6 3 )
~- 1 —~ r~, -- (1 ~ r" ~
With the results thus derived for P' and P", and with the values of
Q' and Q" already obtained, the first approximation to the elements
must be completed.
When the elements are already approximately known, the first
assumed values of P', P", Q', and Q" should be computed by means
of these elements. Thus, from
r'r" sin (v"-'),, r' sin (' - v)
r" sin (v" v)' rr sin (v — v)
r"r"' sin ("- "),,, rr" sin (v" - v')
rr"'r sin (v"' v')' r'r"' sin (v"- v')'
we find n, n', n", and n"'. The approximate elements of Eurynome
give
v -322~ 55' 9".3, logr -0.308327,
v' =353 19 26.3, log r' 0.294225,
v"= 14 45 8.5, log r" 0.296088,
v'- 47 23 32.8, log r"' - 0.317278,
298 THEORETICAL ASTRONOMY.
and hence we obtain
log n = 9.653052, log n" = 9.806836,
log n'= 9.825408, logrn"'= 9.633171.
Then, from
n
P' = TQ n"
P'=^' Q (n + " — 1) r",
Win
P" "" (n'+- n'"- 1)r
we get
log P' = 9.846216, log' = 9.840771,
log P"- 9.807763, log Q"- 9.882480.
The values of these quantities may also be computed by means of the
equations (41) and (42).
Next, from, P'd' + c' h
Co +h
% I- +P'' + P',, P"d"ff+ f ic"+ h"
C~ 1 +P"' - 1+ P',
we find
log Coe 0.541344n, logf' - 0.047658n,
log o" — 9.807665,, logf" 9.889385.
Then we have
St + C — -
XI' +- a Co"
f, + a'', sin a' R" sin 4"
tan z' -, tan z" -
R' sin 4' X',, R" sin 4" x"
sin z' cos Z sin z" Cos "'
from which to find r' and r". In the first place, from
x' = V'1/r2 R2 sin 4',
we obtain the approximate value
log x' 0.242737.
Then the first of the preceding equations gives
log x" 0.237687.
NUMERICAL EXAMPLE. 299
From this we get
" - 29~ 3' 11".7, log r" = 0.296092;
and then the equation for x' gives
log x' - 0.242768.
Hence we have
z' 270 20' 59".6, log r' = 0.294249;
and, repeating the operation, using these results for x' and r', we get
log x" - 0.237678, log x' = 0.242757.
The correct value of log xz may now be found by mieans of equation
(28). Thus, in units of the sixth decimal place, we have
a,= 242768 - 242737- + 31,' = 242757 - 242768- 11,
and for the correction to be applied to the last value of log', in
units of the sixth decimal place,
Alog x'=-, +3.
ac - a
Therefore, the corrected value is
log x'= 0.242760,
and from this we derive
log x" 70.237681.
These results satisfy the equations for x' and xa, and give
z' = 27~ 21' 1".2, log r' 0.294242,
z" =29 3 12.9, logr" -0.296087.
To find the curtate distances for the ist and second observations,
the formulae are,' sin (z' +') 1 " sin (z" + ") f
sin z' coS', p"' sin — " cos ",
which give
log p' = 0.133474, log p" = 0.289918.
Then, by means of the equations
300 THEORETICAL ASTRONOMY.
r' cos b' cos (I' 0') p' cos (' - 0') -R',
r cos b' sin (I' - 0') p' sin (' - 0'),
r' sin b' p' tan fi','r cos b cos (1" - 0") p" cos (A" ")- 0") - R",
r" cos b" sin (" -- 0") - p" sin (A" - 0"),
r" sin b" p" tan s",
we find the following heliocentric places:' = 370 35' 26".4, log tanb' = 8.182861n, log r' 0.294243,
" - 58 58 15.3, log tan b" = 8.6342091, log r" = 0.296087.
The agreement of these values of log r' and log r" with those obtained
directly from x' and x" is a partial proof of the numerical calculation.
From the equations
tan i sin ( " ( + -')- ) _ (tan b" tan b') sec 1 (" -'),
tan i cos Q- (l" + 1') - ) (tan b tan') cosec (1"'),
tan' -- tan (-') tan " tan (" - )
tan ttf tan u-
cos i cos i
we obtain
_ -206" 42' 24".0, i _ 40 36' 47".2,,' - 190 55 6.6 u" —212 20 53.5.
Then, from
1 1 + (~ ), ) a"' n'P",
q- P " 1t -- P
we get
log n" - 9.806832, log n 9.653048,
log i' = 9.825408, log n"' = 9.633171,
and the equations
r~n(t' 4)+1 (in" t6)) _ r+ si
r n11
sin ((u'" - ) + (u" )) - (') sin (u- t' ),
nr
r cos ((n' - u) + - (it" -')) cos (i"'),
r'" sin ((n"'" - u") + 1- (i" -')) - r" sin (" - in'),
in',
" Cos ((i"' - u") + I (i" - i')) = cos (i "- I),
NUMERICAL EXAMPLE. 301
give
logr 0.308379, u 160~ 30' 57".6,
log r"' = 0.317273, "'= 244 59 32.5.
Next, by means of the formule
tan (I - Q ) - cos i tan u, tan b - tan i sin (I -- ),
tan ("' - ) cos i tan "', tan b"' tan i sin (I"'- ),
p os- ) r cos b cos (I - (3) + R,
p sin (- 0) r cos b sin (I - 0),
p tan f =r sin b;
"' cos (A"' - 0"') -_ r" cos b"' cos (1"'- 0") + RI"',
p" sin ("' - o"') = r"' cos b"' sin (1"', 0"'),
p"' tan "' r"' sin b",
we obtain
I - 7 16' 51".8, "' =- 910 37' 40".0,
b-+ 1 32 14.4, b"'= — 4 10 47.4,
i - 16 59 9.0, i"' - 75 23 46.9,
P-+ 2 5640.1, fi"' — 3 4 43.4,
log p - 0.025707, log p"' 0.449258.
The value of 2"' thus obtained agrees exactly with that given by
observation, but A differs 0".4 from the observed value. This difference does not exceed what may be attributed to the unavoidable
errors of calculation with logarithms of six decimal places. The
differences between the computed and the observed values of 9 and
i'^ show that the position of the plane of the orbit, as determined
by means of the second and third places, will not completely satisfy
the extreme places.
The four curtate distances which are thus obtained enable us, in
the case of an orbit entirely unknown, to complete the correction for
aberration according to the equations (40).
The calculation of the quantities which are independent of P',
P", Q', and Q", and which are therefore the same in the successive
hypotheses, should be performed as accurately as possible. The
value of -,, required in finding x" from x', may be computed
directly from
o _ pd,
t a od'C lbd' Gc'
the values of hl and le being found by means of the equations (29);
the valueso~fad bein
302 THEORETICAL ASTRONOMY.
and a similar method may be adopted in the case of C' Further,
f'
in the computation of x' and x", it may in some cases be advisable
to employ one or both of the equations (22) for the final trial. Thus,
in the present case, x" is found from the first of equations (27) by
means of the difference of two larger numbers, and an error in the
last decimal place of the logarithm of either of these numbers affects
in a greater degree the result obtained. But as soon as.r" is known
Q"
so nearly that the logarithm of the factor 1 + — 3 remains unchanged,
the second of equations (22) gives the value of x" by means of the
sum of two smaller numbers. In general, when two or more formulae for finding the same quantity are given, of those which are
otherwise equally accurate and convenient for logarithmic calculation,
that in which the number sought is obtained from the sum of smaller
numbers should be preferred instead of that in which it is obtained
by taking the difference of larger numbers.
The values of r, r', r", r"', and u, t', u, ut"', which result from
the first hypothesis, suffice to correct the assumed values of P', P",', and Q". Thus, from
r-k ("- - t), - k- ( (t' — t), e- k (elOl- el),
tan% = z tan tan! i_
ta tnX rZI
sin r cos G sin (" — u'), sin y" cos G" _ sin 1 (L~'- i),
sisin G'sin G cos. (u" — u') cos 2y, sin r" sin G" -cos 1 (u'-'u) cos 2",
cos r cos - (t" — t') sin 2x, cos r" cos. (t'- u) sin 2/",
sinl "' cos G"'= sin I ("' - u"),
sin r"' sin G"' cos 2 (t"' -- Iu") cos 2/"'
cos "' c=os- (t"' - u") sin 2%"';
2 COS,/ c 7 Tcos, 1112 COS% 6 T
s c3 COSr y' Cos3 y r r cos yr,
sin it sinT" sind s"a
cos' cos cos /'
9Th 9j 5 fl~sll) 5 9 jflt!b 9 t!
_ _a __ _'
S) ~! " ~__ _ ___:t8
in connection with Tables XIII. and XIV. we find s, ", and s"'.
The results are
NUMERICAL EXAMPLE. 303
log = 9.9759441, log r"- 0.1386714, log T"' 0.1800641,
450 3' 39".1, /"' = 440 32' 1".4, X"'= 450 41' 55".2,
r 10 42 55.9, /'" 15 13 45.0, /"' 16 22 48.5,
log m 8.186217, log n"'= 8.516727, log m"'= 8.590596,
logj 7.948097, logj"- 8.260013, logj"'- 8.325365,
log s = 0.0085248, log s"-= 0.0174621, log s"'=- 0.0204063.
Then, by means of the formule
P' —~r' s'
r 2
TTQ t2
Q' ssO1 ~ rT" COS (a" -'a') COS - (,"'') cos s ('a- ~ )'
r'r' Cos (t'"' - a"') cos (It"' - l') cos ('a"t - t')'
we obtain
log P' = 9.8462100, log Q' 9.8407536,
log Pi" 9.8077615, log Q" 9.8824728,
with which the next approximation may be completed.
We now recompute co', c/", f', f", x', x", &c. precisely as already
illustrated; and the results are
log c0' = 0.5413485, log co = 9.8076649,,
logf' = 0.0476614., logf" 9.8893851,
log x' = 0.2427528, log x" 0.2376752,
z' 27~ 21' 2".71, z" 29~ 3' 14".09,
logr'= 0.2942369, logr" 0.2960826,
log p'= 0.1334635, logp" 0.2899124,
log n =9.6530445, log n" - 9.8068345,
log n' = 9.8254092, log n'"' = 9.6331707.
Then we obtain
l' = 370 35' 27".88, log tan' = 8.1828572l, log r' = 0.2942369,
1" = 58 58 16.48, log tan b" - 8.6342073., log r" = 0.2960827.
These results for log r' and log r" agree with those obtained directly
from zt and z", thus checking the calculation of ~' and 4" and of
the heliocentric places.
Next, we derive
-= 206~ 42' 25".89, i 4~ 36' 47".20,' = 190 55 6.27,'u"- 212 20 52.96,
304 THEORETICAL ASTRONOMY.
and fronm u"- u, r, ry, l, n"i, n' and n, we obtain
logr - 0.3083734, u = 160~ 30' 55".45,
log r"'- 0.3172674, It"' 244 59 31.98.
For the purpose of proving the accuracy of the numerical results,
we compute also, as in the first approximation,
l= 70 16' 51".54, l"'- 91~ 37' 41".20,
b=-+ 1 32 14.07, b"' —- 4 10 47.36,
A_ 16 59 9.38,'" - 75 23 46.99,
i -+ 2 56 39.54, P"'- 3 4 43.33,
log p -0.0256960, log p"' 0.4492539.
The values of 2 and 2"' thus found differ, respectively, only 0".04
and 0".09 from those given by the normal places, and hence the
accuracy of the entire calculation, both of the quantities which are
independent of P', P", Q', and Q", and of those which depend on
the successive hypotheses, is completely proved. This condition,
however, must always be satisfied whatever may be the assumed
values of P', P", Q, and Q".
From r, r', U%, u', &c., we derive
log s - 0.0085254, logs" = 0.0174637, log s"' = 0.0204076,
and hence the corrected values of P', P", Q', and Q" become
log P' - 9.8462110, log Q' = 9.8407524,
log P" =9.8077622, log Q" 9.8824726.
These values differ so little from those for the second approximation,
the intervals of time between the observations being very large, that
a further repetition of the calculation is unnecessary, since the results
which would thus be obtained can differ but slightly from those
which have been derived. We shall, therefore, complete the determination of the elements of the orbit, using the extreme places.
Thus, from
k (t"' t), tanXo 0 /,
sin r cos Go sin 1 ('/"' - ),
sin 7o sin G - cos - (u"'"- u) cos 2Z0,
cos ro = cos ("' - u) sin 2X0,
o _______ _o_,
T2 sin2 l'
~(r + r"')3 c' co s ro
_ mo + __o,m
o - -_ - _' x ) 02
NUMERICAL EXAMPLE. 305
we get
log r, = 0.5838863, log tan G = 8.0521953,
O = 420 14' 30".17, log m0 9.7179026,
log s2= 0.2917731, log x0= 8.9608397.
The formula
/ Sorr"' sin ("'-'t) )2
gives
logp — 0.3712401;
and if we compute the same quantity by means of
(8rI Isin (a sr'r" ) rr'sin (Vt'- _a)'2 ( s"'r"r"' sin (a"' u.") )
~ ~ ) = ~ T/ TJthe separate results are, respectively, 0.3712397, 0.3712418, and
0.3712414. The differences between these results are very small, and
arise both from the unavoidable errors of calculation and from the
deviation of the adopted values of P', P", Q', and Q" from the
limit of accuracy attainable with logarithms of seven decimal places.
A variation of only 0".2 in the values of u' - and "-' - u" will
produce an entire accordance of the particular results.
From the equations
sin j (E" E) (V O
a cos s ('-E ) rr
COS ~ P ~
a cosy
we obtain
(E"' E) _ 170 35' 42".12, log (a cos g) 0.3796883,
log cos = 9.9915518.
The formulae
e sin(o -- ( ( "' +- u)) - tan Go,
cos ro 1/rr'
e cos w -' (t"' + u')) = sec I (u"' - u),
cos ro 1/V "'
give
w = 197~ 38' 8".48, log e = log sin c - 9.2907881,
p- 11~ 15' 52".22,, _ w + = 44~ 20' 34".37.
This result for ( gives log cos (p= 9.9915521, which differs only 3
in, the last decimal place from the value found from p and a cos p.
Then, from
20
830 THEORETICAL ASTRONOMIY.
p- k
Cb == ~ ~p. - -- ~
0O 2 3 a3
COS a-?
the value of k being expressed in seconds of arc, or log k= 3.5500066,
we get
log a = 0.3881359, log s = 2.9678027.
For the eccentric anomalies we have
tan ME - tan 1 (u - o) tan (45~ ~- ),
tan -E' tan 1 (u' - w) tan (45~ -n),
tan 4E" - tan I (u" - w) tan (45~ -),
tan J-2E"' tan - (a"'- -W) tan (45~0- ),
from which the results are
E -329~ 11' 46".01, E" - 12~ 5' 33".63,
E — 354 29 11.84, E"' 39 34 34.65.
The value of (E"' - E) thus derived differs only 0".03 from that
obtained directly from x0.
For the mean anomalies, we have
M f E - e sin E, M1" = E" - e sin E",
M' - E' - e sin E', M"' - E"' - e sin E"',
which give
M = 334~ 55' 39".32, M" = 9~ 44' 52".82,
M'= 355 33 42.97, M'"' 32 26 44.74.
Finally, if JMo denotes the mean anomaly for the epoch T= 1864
Jan. 1.0 mean time at Greenwich, from
M- - M~ (t -- T) = M' — (t' - T)
= M" - /l (t" - T) M"' - t (t"' - T),
we obtain the four values
MO = 1~ 29' 39".40
39.49
39.40
39.40,
the agreement of which completely proves the entire calculation of
the elements from the data. Collecting together the several results,
we have the following elements:
NUMERICAL EXAMPLE. 307
Epoch =1864 Jan. 1.0 Greenwich mean time.
M - 1~ 29' 39".42
7 — 44 20 34.37 Ecliptic and Mean
Q_- 206 42 25 89 EP^
2 - 206 42 25 8 Equinox 1864.0.
i= 4 36 47.20
y I 11 15 52.22
log a - 0.3881359
log~ - 2.9678027
i - 928".54447.
102. The elements thus derived completely represent the four observed longitudes and the latitudes for the second and third places,
which are the actual data of the problem; but for the extreme latitudes the residuals are, computation minus observation,
a/ = ~- 4".47, A/' - + 1".23.
These remaining errors arise chiefly from the circumstance that the
position of the plane of the orbit cannot be determined from the
second and third places with the same degree of precision as from
the extreme places. It would be advisable, therefore, in the final
approximation, as soon as p', p, ", n1, it', and nt"' are obtained, to
compute from these and the data furnished directly by observation
the curtate distances for the extreme places. The corresponding
heliocentric places may then be found, and hence the position of the
plane of the orbit as determined by the first and fourth observations.
Thus, by means of the equations (37) and (38), we obtain
log p = 0.0256953, log p"' = 0.4492542.
With these values of p and p'", the following heliocentric places are
obtained:
I - 7~ 16' 51".54, log tan b 8.4289064, log r 0.3083732,
1"'- 91 37 40.96, log tan b"' 8.8638549,, logr"' 0.3172678.
Then from
tan sn sn ( (l"' + 1) - 2). 1 (tan b"' + tan b) see. (1"'- 1),
tan i c 1) - an i"' - ) ( t an tan b) cosec ("' - 1),
we get
= 206~ 42' 45".23, 4~ 36' 49".76.
For the arguments of the latitude the results are
u = 160~ 30' 35".99, u"' - 244~ 59' 12".53.
308 THEORETICAL ASTRONOMY.
The equations
tan b' tan i sin (tI - ),
tan b" = tan i sin (" - b ),
give
log tan b' = 8.1827129, log tan b"' 8.6342104,
and the comparison of these results with those derived directly from
p' and,op" exhibits a difference of + 1".04 in b', and of - 0".06 in
1". Hence, the position of the plane of the orbit as determined from
the extreme places very nearly satisfies the intermediate latitudes.
If we compute the remaining elements by means of these values
of r, r't/, and u, um', the separate results are:
log tan G = 8.0522282n, log on = 9.7179026,
log s2 =0.2917731, log x, = 8.9608397,
logp - 0.3712405, (E" - E) — 17~ 35' 42".12,
log (a cos p) = 0.3796884, log cos o - 9.9915521,
w = 197~ 37' 47".72, log e — 9.2907906,
-= 11 15 52.46, log cos o = 9.9915520,
log a -0.3881365, log,a 2.9678019,
E= 329~ 11' 47".24, E"' 39~ 34' 35".70,
11=334 55 40.46, f"'- 32 26 45.49,
i =o- 1 29 40.36, M 1 29 40.37.
Hence, the elements are as follows:
Epoch = 1864 Jan. 1.0 Greenwich mean time.
M= 1~ 29' 40".36
44 20 32.95 Ecliptic and Mean
d2 - 206 42 45.23
-206 42 45.23 Equinox 1864.0.
i- 4 36 49.76?= 11 15 52.46
log a = 0.3881365
i - 928".5427.
It appears, therefore, that the principal effect of neglecting the
extreme latitudes in the determination of an orbit from four observations is on the inclination of the orbit and on the longitude of the
ascending node, the other elements being very slightly changed. The
elements thus derived represent the extreme places exactly, and if
we compute the second and third places directly from these elements,
we obtain
JM' 355~ 33' 43".88, M" = 9~ 44' 53".73,
E' - 354 29 12.93, E" 12 5 34.81,
v' =353 16 59.07, v" -=14 42 45.96,
NUMERICAL EXAMPLE. 309
log r' = 0.2942366, log r" 0.2960828,
u' - 190~ 54' 46".79, u" 212~ 20' 33".68,
1' — 37 35 27.75, 1"- 58 58 16.50,
b' - 0 52 21.25, b --- 2 27 59.06,
A'-= 10 14 17.35, " — 29 53 21.99,
i'- -1 15 47.67, 13" — 2 29 57.62,
log p' 0.1334634, log p" = 0.2899122.
Hence, the residuals for the second and third places of the planet
areComp. - Obs.
A/' - 0".22, aF' + 1".53,
2A)." 0.00, Al" - -0.06;
and the elements very nearly represent the four normal places. Since
the interval between the extreme places is 223 cays, these elements
must represent, within the limits of the errors of observation, the
entire series of observations on which the normals are based. It
may be observed, also, that the successive approximations, in the
case of intervals which are very large, do not converge with the
same degree of rapidity as when the intervals are small, and that in
such cases the mnmerical calculation is very much abbreviated by the
determination, in the first instance, of the assumed values of P', P",
Q', and Q" by means of approximate elements already known. For
the first determination of an unknown orbit, the intervals will generally be so small that the first assumed values of these quantities, as
determined by the equations
fP' -. t (- Ci )_ \ i) = __i",
P i__ (1 _ I2 & r)
will not differ much from the correct values, and two or three
hypotheses, or even less, will be sufficient. But when the intervals
are large, and especially if the eccentricity is also considerable, several
hypotheses may be required, the last of which will be facilitated by
using the equations (82)4.
The application of the formule for the determination of an orbit
from four observations, is not confined to orbits whose inclination to
the ecliptic is very small, corresponding to the cases in which the
method of finding the elements by means of three observations fails,
310 THEORETICAL ASTRONOMY.
or at least becomes very uncertain. On the contrary, these formulae
apply equally well in the case of orbits of any inclination whatever,
and since the labor of computing an orbit from four observations
does not much exceed that required when only three observed places
are used, while the results must evidently be more approximate, it
will be expedient, in very many cases, to use the formula given in
this chapter both for the first approximation to an unknown orbit
and for the subsequent determination from more complete data.
CIRCULAR ORBIT. 311
CHAPTER VI.
INVESTIGATION OF VARIOUS FORBMITL FOR THE CORRECTION OF THE APPROXIMATE
ELEMENTS OF THE ORBIT OF A HEAVENLY BODY.
103. IN the case of the discovery of a planet, it is often convenient, before sufficient data have been obtained for the determination
of elliptic elements, to compute a system of circular elements, an
ephemeris computed from these being sufficient to follow the planet
for a brief period, and to identify the comparison stars used in differential observations. For this purpose, only two observed places
are required, there being but four elements to be determined, namely,
2, i, a, and, for any instant, the longitude in the orbit. As soon as
a has been found, the geocentric distances of the planet for the
instants of observation may be obtained by means of the formule
d - R cos ~ + q- 1 ~ RB sin" %,
A" = R" cos " + - /c'a R, 2 sin2 4",
the values of 4 and Q" being computed from the equations (42)3 and
(43)3. For convenient logarithmic calculation, we may first find z
and z" from
R sin, R" sin,"
sin z 7 --, sin 2 s n," (2)
a a
since the formule will generally be required for cases such that these
angles may be obtained with sufficient accuracy by means of their
sines. Then we have
R sin (z + 4) os R sin (Z" + os
sin -z"1 cos 3", (3)
sin z " sin z
from which to find p and p". These having been found, we have
tanQ-O^z psin(-~)
tan (I- 0) == \ -
p cos (A L 4- )-R
p tanj i
sin b t- ~,
for the determination of I and b, and similarly for I' and b". The
312 THEORETICAL ASTRONOMY.
inclination of the orbit and the longitude of the ascending node are
then found by means of the formulae (75)3, and the arguments of the
latitude by means of (77)3. Since tu"- u is the distance on the celestial sphere between two points of which the heliocentric spherical
co-ordinates are 1, b, and 1', b", we have, also, the equations
sin (u" - u) sin B -= cos b" sin (1"- 1),
sin (au" - u) cos B = cos b sin b" - sin b cos b" cos (" - 1),
cos (u" - u) - sin b sin b" -+ cos b cos b" cos (" - 1),
for the determination of't"- iA, the angle opposite the side 90~ - b"
of the spherical triangle being denoted by B. The solution of these
equations is facilitated by the introduction of auxiliary angles, as
already illustrated for similar cases.
In a circular orbit, the eccentricity being equal to zero, ua - u
expresses the mean motion of the planet during the interval t" — t,
and we must also have
t" —t -- (a" - u), (5)
the value of k being expressed in seconds of arc, or log k 3.5500066.
These formule will be applied only when the interval t" -- t is
small, and for the case of the asteroid planets we may first assume
a -2.7,
which is about the average mean distance of the group. With this
we compute p and p" by means of the equations (2) and (3), and the
corresponding heliocentric places by means of (4). If the inclination
is small, a u'- a will differ very little from 1"- 1. Therefore, in the
first approximation, when the heliocentric longitudes have been found,
the corresponding value of t"- t may be obtained from equation (5),
writing " - 1 in place of "l - u. If this comes out less than the
actual interval between the times of observation, we infer that the
assumed value of a is too small; but if it comes out greater, the
assumed value of a is too large. The value to be used in a repetition
of the calculation may be computed from the expression
log a - (log ( - t) + log k -- log (a - )),
the difference ut"-u being expressed in seconds of arc. With this
we recompute p, p", 1, and I", and find also b, b", 2, i,'t, and tL".
Then, if the value of a computed from the last result for u"s- u
differs from the last assumed value, a further repetition of the calcu
CIRCULAR ORBIT. 313
lation becomes necessary. But when three successive approximate
values of a have been found, the correct value may be readily interpolated according to the process already illustrated for similar cases.
As soon as the value of a has been obtained which completely
satisfies equation (5), this result and the corresponding values of 2,
i, and the argument of the latitude for a fixed epoch, complete the
system of circular elements which will exactly satisfy the two observed
places. If we denote by u0 the argument of the latitude for the epoch
T, we shall have, for any instant t,
- uo +- t (t -T),
L being the mean or actual daily motion computed from
k
- -3.
az
The value of u thus found, and r = a, substituted in the formulae for
computing the places of a heavenly body, will furnish the approximate ephemeris required.
The corrections for parallax and aberration are neglected in the
first determination of circular elements; but as soon as these approximate elements have been derived, the geocentric distances may be
computed to a degree of accuracy sufficient for applying these corrections directly to the observed places, preparatory to the determination of elliptic elements. The assumption of r'- a will also be
sufficient to take into account the term of the second order in the first
assumed value of P, according to- the first of equations (94.
104. When approximate elements of the orbit of a heavenly body
have been determined, and it is desired to correct them so as to satisfy
as nearly as possible a series of observations including a much longer
interval of time than in the case of the observations used in finding
these approximate elements, a variety of methods may be applied.
For a very long series of observations, the approximate elements
being such that the squares of the corrections which must be applied
to them may be neglected, the most complete method is to form the
equations for the variations of any two spherical co-ordinates which
fix the place of the body in terms of the variations of the six elements of the orbit; and the differences between the computed places
for different dates and the corresponding observed places thus furnish
equations of condition, the solution of which gives the corrections to
be applied to the elements. But when the observations do not in
314 THEORETICAL ASTRONOMY.
elude a very long interval of time, instead of forming the equations
for the variations of the geocentric places in terms of the variations
of the elements of the orbit, it will be more convenient to form the
equations for these variations in terms of quantities, less in number,
from which the elements themselves are readily obtained. If no assumption is made in regard to the form of the orbit, the quantities
which present the least difficulties in the numerical calculation are
the geocentric distances of the body for the dates of the extreme
observations, or at least for the dates of those which are best adapted
to the determination of the elements. As soon as these distances are
accurately known, the two corresponding complete observations are
sufficient to determine all the elements of the orbit.
The approximate elements enable us to assume, for the dates t and
t", the values of A and A"; and the elements computed from these
by means of the data furnished by observation, will exactly represent
the two observed places employed. Further, the elements may be
supposed to be already known to such a degree of approximation that
the squares and products of the corrections to be applied to the
assumed values of d and A" may be neglected, so that we shall have,
for any date,
coS A C os + cos d - A,- co
Ad8 d3 " (6)
doJ dJ"
If, therefore, we compare the elements computed from A and A" with
any number of additional or intermediate observed places, each observed spherical co-ordinate will furnish an equation of condition for
the correction of the assumed distances. But in-order that the equations (6) may be applied, the numerical values of the partial differential coefficients of a and 8 with respect to A and A" must be found.
Ordinarily, the best method of effecting the determination of these is
to compute three systems of elements, the first from A and A", the
second from A + D and A", and the third from A and A" + D", D
and D" being small increments assigned to A and A" respectively.
If now, for any date t', we compute a' and 8' from each system of
elements thus obtained, we may find the values of the differential
coefficients sought. Thus, let the spherical co-ordinates for the time
t' computed from the first system be denoted by a' and 8'; those
computed from the second system of elements, by a' + a sec 8' and
A' + d: and those from the third system, by a'+ a" sec a' and J'+ d".
Then we shall have
VARIATION OF TWO GEOCENTRIC DISTANCES. 315
da' a d
cos at' -, -a ---
cd D' dzJD'
(7)
dJ'~D`' dA" D"
and the equations (6) give
a a"I
C0os t, am A ^ A2,
cib~ ci" - ~(8)
d dl".
aB - ^AJ + - ^A2.
In the same manner, computing the places for various dates, for
which observed places are given, by means of each of the three systems
of elements, the equations for the correction of 4 and Ad", as determined by each of the additional observations employed, may be
f'ormed.
105. For the purpose of illustrating the application of this method,
let us suppose that three observed places are given, referred to the
ecliptic as the fundamental plane, and that the corrections for parallax,
aberration, precession, and nutation have all been duly applied. By
means of the approximate elements already known, we compute the
values of a and A" for the extreme places, and from these the heliocentric places are obtained by means of the equations (71)3 and (72)3,
writing A cos / and A" cos /i" in place of p and,op. The values of'-, i, u, and utt will be obtained by means of the formule (76)3 and
(77)3; and fiom r, r"t and u — u, the remaining elements of the
orbit are determined as already illustrated. The first system of elements is thus obtained. Then we assign an increment to 4, which
we denote by D, and with the geocentric distances a + D and A"
we compute in precisely the same manner a second system of elements. Next,- we assign to A" an increment D", and from a and
A"t -I D" a third system of elements is derived. Let the geocentric
longitude and latitude for the date of the middle observation computed from the first system of elements be designated, respectively,
by,'i and /9'; from the second system of elements, by 2.' and /^;
and from the third system, by ),3 and /9'. Then from
a = (' - 2') cos i,', d =P —,
a" -- (['- ],') COS/ " -, (9) -
we compute a, a", d, and dt, and by means of these and the values
of D and D" we form the equations
316 THEORETICAL ASTRONOMY.
Ad + a Ad" = cos i' a,',
d d"
D + D
for the determination of the corrections to be applied to the first
assumed values of d and D", by means of the differences between
observation and computation. The observed longitude and latitude
being denoted by 2' and 9', respectively, we shall have
COS P' A'- (' -,') cos 13',
for finding the values of the second members of the equations (10),
and then by elimination we obtain the values of the corrections Ad
and A/" to be applied to the assumed values of the distances.
Finally, we compute a fourth system of elements corresponding to
the geocentric distances A + ad and A" + ad" either directly from
these values, or by interpolation from the three systems of elements
already obtained; and, if the first assumption is not considerably in
error, these elements will exactly represent the middle place. It
should be observed, however, that if the second system of elements
represents the middle place better than the first system, 2,' and 32
should be used instead of,,' and,9/' in the equations (11), and, in
this case, the final system of elements must be computed with the
distances A + D + AJ and d" + a"/. Similarly, if the middle
place is best represented by the third system of elements, the corrections will be obtained for the distances used in the third hypothesis.
If the computation of the middle place by means of the final elements still exhibits residuals, on account of the neglected terms of
the second order, a repetition of the calculation of the corrections
Ad and Azd, using these residuals for the values of the second
members of the equations (10), will furnish the values of the distances for the extreme places with all the precision desired. The
increments D and D" to be assigned successively to the first assumed
values of A and A" may, without difficulty, be so taken that the
true elements shall differ but little from one of the three systems
computed; and in all the formulae it will be convenient to use, instead of the geocentric distances themselves, the logarithms of these
distances, and to express the variations of these quantities in units
of the last decimal place of the logarithms.
These formulae will generally be applied for the correction of
VARIATION OF TWO GEOCENTRIC DISTANCES. 317
approximate elements by means of several observed places, which
may be either single observations or normal places, each derived from
several observations, and the two places selected for the computation
of the elements from A and A" should not only be the most accurate
possible, but they should also be such that the resulting elements are
not too much affected by small errors in these geocentric places.
They should moreover be as distant from each other as possible, the
other considerations not being overlooked. When the three systems
of elements have been computed, each of the remaining observed
places will furnish two equations of condition, according to equations
(10), for the determination of the corrections to be applied to the
assumed values of the geocentric distances; and, since the number
of equations will thus exceed the number of unknown quantities,
the entire group must be combined according to the method of least
squares. Thus, we multiply each equation by the coefficient of Jd
in that equation, taken with its proper algebraic sign, and the sum
of all the equations thus formed gives one of the final equations
required. Then we multiply each equation by the coefficient of.A"
in that equation, taken also with its proper algebraic sign, and the
sum of all these gives the second equation required. From these
two final equations, by elimination, the most probable values of AJ
and AJ" will be obtained; and a system of elements computed with
the distances thus corrected will exactly represent the two fundamental places selected, while the sum of the squares of the residuals
for the other places will be a minimum. The observations are thus
supposed to be equally good; but if certain observed places are
entitled to greater influence than the others, the relative precision
of these places must be taken into account in the combination of the
equations of condition, the process for which will be fully explained
in the next chapter.
When a number of observed places are to be used for the correction
of the approximate elements of the orbit of a planet or comet, it will
be most convenient to adopt the equator as the fundamental plane.
In this case the heliocentric places will be computed from the assumed
values of d and J/, and the corresponding geocentric right ascensions
and declinations by means of the formule (106)3 and (107)3; and the
position of the plane of the orbit as determined from these by means
of the equations (76)3 will be referred to the equator as the fundamental plane. The formation of'the equations of condition for the
corrections ad and AI"1 to be applied to the assumed values of the
distances will then be effected precisely as in the case of A and fi, the
318 THEORETICAL ASTRONOMY.
necessary changes being made in the notation. In a similar manner,
the calculation may be effected for any other fundamental plane which
may be adopted.
It should be observed, further, that when the ecliptic is taken as
the fundamental plane, the geocentric latitudes should be corrected
by means of the equation (6)4, in order that the latitudes of the sun
shall vanish, otherwise, for strict accuracy, the heliocentric places
must be determined from A and A" in accordance with the equations
(89)v.
106. The partial differential coefficients of the two spherical coordinates with respect to A and A" may be computed directly by
means of differential formule; but, except for special cases, the
numerical calculation is less expeditious than in the case of the indirect method, while the liability of error is much greater. If we
adopt the plane of the orbit as determined by the approximate values
of A and /A" as the fundamental plane, and introduce Z as one of the
elements of the orbit, as in the equations (72)2, the variation of the
geocentric longitude 0 measured in this plane, neglecting terms of the
second order, depends on only four elements; and in this case the
differential formulae may be applied with facility. Thus, if we express r and v in terms of the elements yp, 1M0, and /a, we shall have
dr dr d d dr dM dr
da dcv d + dMc d d'
and
dv dv do' dv dMoi dv dc,
da dyd + dd -iMo da' J cd dd'
or
d (v-+ %) d% dv do dv dM, dv d,
dA dJ dp d + di d d + did d'
In like manner, we have
dr" dr" dco dr" dMo dr" de.
d- ~ d-' d- dMo da i df' d'd
d (v" +:X) dv" d, d+ " d.0o dv" d+, d%
D d d d dd d+ dol dJ'
dr d (v - ) dr" d (v" -- x)
As soon as the values of d — d~Z, dr- and d are
dA dA dJ dA
known, the equations necessary for finding the differential coefficients
of the elements Z p, 1 ~, and / with respect to A are thus provided.
In the case under consideration, when an increment is assigned to J,
VARIATION OF TWO GEOCENTRIC DISTANCES. 319
the value of A" remaining unchanged, r" and v" + X are not changed,
and hence
ddD dA ~~
dr" d (v") I
dJ -- O dJ -O
To find - and d( Zfrom the equations
J cos os s O= x + X,
J cos sin 0 = y + Y,
in which ^ is the geocentric latitude in reference to the plane of the
orbit computed from A and A" as the fundamental plane, and X, Y
the geocentric co-ordinates of the sun referred to the same plane, we
get
dx - cos. cos 0 dJ,
dy = cos sin 0 dd,
or, substituting for dx and dy their values given by (73),,
cos;- cos 0 dA cos u dr - r sin u d (v +- ),
cos v sin 0 dJ sinu dr + r cosu d (v + ).
Eliminating, successively, d (v -+ ) and dr, we get
dr
d cos - cos (0.- u),
(12)
d (v+) 1 (1
d+) _ - Gcos n sin (0 - u).
dJ r
Therefore, we shall have
dz dv dp cdv dMo dv d, 1
d.- d d -+ - +cos ~ sin (0 -- u),
d d d dj dd lo d1 d d r
^J' ^ ^J ^^ c dJ dp. dJ r
dZr d d dr dMo cdi dli
d- -' d A + - T' +- ~\ d' - -= c o s'c o s(0 u ),
d+ dv" d" dv" d il dv" d3)
_ _1. _ -.. _.
da de~d dr"F ddo
dr" d dr" dio dr" d,/
d' d d di i da da ~
and if we compute the numerical values of the differential coefficients
of r, r", v, and v" with respect to the elements po, 2Ml, and /z, these
equations will furnish, by elimination, the values of the four un-.. cl d- dill d a
known quantities d ) and diD
Id ddi dd di
In precisely the same manner we derive the following equations
320 TIHEORETICAL ASTRONOMY.
for the determination of the partial differential coefficients of these
elements with respect to d":dX dv dp dv cdlM. dv d,
d'- d' dA " dM, dJ" d,,' da
dr dp dr dMl dr dp,.
- -0- + 0+ -o
d li" d" d" d dlMo' d' d'(1
dc dv" dy dv" dMi dv" d,. 1..
d"+ dq dr " + dr" d" + dfji d- r" cos - in (O - n").
dr" dJ dr" dMl d " di'.
d' +' di" ~ dL" li+ d,.' d - cos cos — co "
Since the geocentric latitude V is affected chiefly by a change of the
position of the plane of the orbit, while the variation of the longitude
6 is independent of q and i when the squares and products of the
variations of the elements are neglected, if we determine the elements
which exactly represent the places to which J and d" belong, as well
as the longitudes for two additional places, or, if we determine those
which satisfy the two fundamental places and the longitudes for any
number of additional observed places, so that the sum of the squares
of their residuals shall be a minimum, the results thus obtained will
very nearly satisfy the several latitudes.
Let 0' denote the geocentric longitude of the body, referred to the
plane of the orbit computed from A and A" as the fundamental plane,
for the date t' of any one of the observed places to be used for correcting these assumed distances. Then, to find the partial differential
coefficients of 6' with respect to d and A", we have
d o',dO' dz,do' d, do' dhii
cos d - coos -. d — + cos -~-, + cos o' d
d d/- d- dc ld' d>, dd
do' d'
+- Cos f' T
d.' (15)
do',dO' do,dO' d', do' d11
dco' d* d'l" + os dp d'i" + co/ dJoi' dli"
dO' dp.
and by means of the results thus derived, we form the equation
cos7/ 9 0' = cos do' d dA cs' -d,-i" (16)
A fourth observed place will furnish, in the same manner, the additional equation required for finding AJ and Ld". If more than two
VARIATION OF TWO GEOCENTRIC DISTANCES. 321
observations are used in addition to the fundamental places on which
the assumed elements as derived from A and A" are based, the several
longitudes will furnish each an equation of condition, and the most
probable values of AJ and AzJ will be obtained by combining the
entire group of equations of condition according to the method of
least squares.
107. In the actual application of these formule to the correction
of the approximate elements, after all the preliminary corrections
have been applied to the data, we select the proper observed places
for determining the elements from the corresponding assumed distances d and 4", according to the conditions which have already been
stated, and fronm these we derive the six elements of the orbit. Since
the data furnished directly by observation are the right ascensions
and the declinations of the body, the elements will be derived in
reference to the equator as the plane to which the inclination and the
longitude of the ascending node belong. These elements will exactly
represent the two fundamental places, and, if the assumed distances
4 and Az are not much in error, they will also very nearly satisfy
the remaining places.
We now adopt as the fundamental plane the plane of the approximate orbit thus determined, and by means of the equations (83)2 and
(85)2, or by means of (87)2, writing a, 8, A', and i' in place of,, 39,
9g, and i, respectively, we compute the values of 0, /, and r for the
dates of the several places to be employed. Then the residuals for
each of the observed places are found from the formula
cos A SO sin r AS +- cosr cos a Aa,
A = Cosry -sin rcos ^Ao (1
the values of AC and a8 for each place being found by subtracting
from the observed right ascension and declination, respectively, the
right ascension and declination computed by means of the elements
derived from a and d". The values of 0, V, and r being required
only for finding cos'q 0, A^q, and the differential coefficients of 0 and
V, with respect to the elements of the orbit, need not be determined
with great accuracy.
dr d (v + z')
Next, we compute d and ( d from equations (12), and from
dr dr" dv dv" dr
(16)2 the values of dr dr" d' d &c., by means of which,
d-' d(p- d(p' d-p Mo,
using the value of u in reference to the equator, we form the equations (13). The accent is added to X to indicate that it refers to the
21
322 THEORETICAL ASTRONOMY.
equator as the plane for defining the elements. Thus we obtain four
equations, from which, by elimination, the values of the differential
coefficients of Z', (, Mo, and / with respect to J may be obtained.
In the numerical solution, by subtracting the third equation from
d"/
the first, the unknown quantity -d is immediately eliminated, so that
we have three equations to find the three unknown quantities do,
dMo di. dz
dj' and -Jd These having been found, d may be obtained from
the first or from the third equation.
In the same manner we form the equations (14), and thence derive
d-' dg dil d,
the values of dl' d' d' and d-' Then, by means of the formule (76)2, (78)2, and (79)2, we compute for the date of each place
to be employed in correcting the assumed distances the values of
do' do' do'
COS d,-' cos I' d-,, &c., and hence from (15) the values of cos B' do
d/f do I, d/
and cos r da' The results thus obtained, together with the residuals
dJl
computed by means of the equations (17), enable us to form, according to (16), the equations of condition for finding the values of the
corrections AJ and Ad". The solution of all the equations thus
formed, according to the method of least squares, will give the most
probable values of these quantities, and the system of elements which
corresponds to the distances thus corrected will very nearly satisfy
the entire series of observations. Since the values of cos' AO' are
expressed in seconds of arc, the resulting values of AJ and aA" will
also be expressed in seconds of arc in a circle whose radius is equal
to the mean distance of the earth from the sun. To express them in
parts of the unit of space, we must divide their values in seconds of
arc by 206264.8.
The corrections to be applied to the elements computed from d and
/", in order to satisfy the corrected values A + AJ and. A" -- A/",
may be computed by means of the partial differential coefficients
already derived. Thus, in the case of X', we have
dd dz,
Az'Z- -Ad + A ja",
from which to find AX'; and in a similar manner ay, A-1x, and Ad
d (v - 7') d (v". - /)
may be obtained. If, from the values of d ( ) and (- +,')
~~~w~~~~de compdute
we compute
VARIATION OF TWO GEOCENTRIC DISTANCES. 323
AeV" — (v" + z') Ad AX'
AV d A A
dA"
and apply these corrections to the values of v and v" found from J
and A", we obtain the true anomalies corresponding to the distances
A + AJ and A" -1- AZi". The corrections to be applied to the values
of r and r" derived from d and A" are given by
dr acr
Ar Ad, ar" ad"
If AJ and Ad" are expressed in seconds of arc, the corresponding
values of ar and ar" must be divided by 206264.8. The corrected
results thus obtained should agree with the values of r and r" computed directly from the corrected values of v, vi', p, and e by means
of the polar equation of the conic section. Finally, we have
d - sin 7 dD,
and similarly for dcz"; and the last of equations (73)2 gives
r sin u Ai' - r cos it sin i' A g' - sin,7 A l,
rf sin ai' - r" cos t sin i A' - sin ^^ Ai", (18)
from which to find hi' and A', u and u" being the arguments of
the latitude in reference to the equator. We have also, according to
(72)2, Aw A/ C A
am' - aC' -- cos i' a la I,
A-' 7;A/' + 2sin 2i' a',
from which to find the corrections to be applied to ow' and 7r'. The
elements which refer to the equator may then be converted into those
for the ecliptic by means of the formula which may be derived from
(109), by interchanging 2 and I and 180~ - i' and i.
The final residuals of the longitudes may be obtained by substituting the adopted values of AJ and adJ in the several equations of
condition, or, which affords a complete proof of the accuracy of the
entire calculation, by direct calculation from the corrected elements;
and the determination of the remaining errors in the values of' will
show how nearly the position of the plane of the orbit corresponding
to the corrected distances satisfies the intermediate latitudes.
Instead of po, M,, and p/, we may introduce any other elements
which determine the form and magnitude of the orbit, the necessary
324 THEORETICAL ASTRONOMY.
changes being made in the formule. Thus, if we use the elements
T, q, and e, these must be written in place of Mo, i, and (o, respectively, in the equations (13), (14), and (15), and the partial differential
coefficients of r, r", v, and v" with respect to these elements must be
computed by means of the various differential formulae which have
already been investigated. Further, in all these cases, the homogeneity of the formulae must be carefully attended to.
108. The approximate elements of the orbit of a heavenly body
may also be corrected by varying the elements which fix the position
of the plane of the orbit. Thus, if the observed longitude and latitude and the values of g and i are given, the three equations (91)i
will contain only three unknown quantities, namely, J, r, and u, and
the values of these may be found by elimination. When the observed
latitude F/ is corrected by means of the formula (6)4, the latitudes of
the sun disappear from these equations, and if we multiply the first
by sin (0 - ) sin /, the second (using, only the upper sign) by
- cos (0 - ~) sin f9, and the third by - sin (R - 0) cos /9, and add
the products, we get
sin sin (0 ~ 9 )
tanu — nn (19)
cos i sin fe cos ( - ) sin i cos f sin (A - )'
from which u may be found. If we multiply the second of these
equations by sin /9, and the third by - cos / sin (A - 2), and add the
products, we find
r -R sin ( ~ - ) (20)
sin u (sin i cot f sin (A - fi )- cos i(20)
The expression for r in terms of the known quantities may also be
found by combining the first and second, or by combining the first
and third, of equations (91)1. If we put
n cos N - sin A cos (O - ),
n sin N= cos sin (A - 0),
the formula for u becomes
cos Ntan u -=os (N tan (( - ). (21)
The last of equations (91)i shows that sin u and sin / must have the
same sign, and thus the quadrant in which u must be taken is determined. Putting, also,
m cos Mr sin u,
m sin M=- sin u cot f sin (A - ),
VARIATION OF THE NODE AND INCLINATION. 325
we have
-= -_ cos 1 R sin(O - )
cos (M + i) sin t(
When any other plane is taken as the fundamental plane, the
latitude of the sun (which will then refer to this plane) will be retained in the equations (91)1 and in the resulting expressions for u
and r.
The value of u may also be obtained by first computing w and 4
by means of the equations (42)3, and then, if z denotes the angle at
the planet or comet between the earth and sun, the values of u and
z, as may be readily seen, will be determined by means of the relations of the parts of a spherical triangle of which the sides are
1800 - (z + 4), 180~ + 0 - ~, and tu, the angle opposite to the
side u being that which we designate by w, and the side 180~ + (- 0 2
being included by this and the inclination i. Let S- 1800 -( + 4-),
and, according to Napier's analogies, this spherical triangle gives
tan - (S+ )- -) -i )cot( )
cos (i+,ow) 2 --,
(23)
sin. (i -- w')
from which S and u are readily found. Then we have
z - 180~ —, - S
R sin (24)
sin z
to find r.
If we assume approximate values of 2 and i, as given by a system
of elements already known, the equations here given enable us to find
r, u', ", and u" from 2,: and A", /, corresponding to the dates t
and t" of the fundamental places selected, and from these results for
two radii-vectores and arguments of the latitude, the remaining
elements may be derived. From these the geocentric place of the
body may be found for the date t' of any intermediate or additional
observed place, and the difference between the computed and the
observed place will indicate the degree of precision of the assumed
values of S2 and i. Then we assign to a the increment 8, i
remaining unchanged, and compute a second system of elements, and
from these the geocentric place for the time t'. We also compute a
third system from Q and i + Si, and by a process entirely analogous
to that already indicated in the case of the variation of two geocentric
326 THEORETICAL ASTRONOMY.
distances, we obtain the numerical values of the differential coefficients of At and /' with respect to 2 and i. Thus the equations
COS A cos/' da' d + Cos' dA2
dg' di
dg ^ + di
for finding the corrections Ax and ai to be applied to the assumed
values of these elements, will be formed; and each additional observation or normal place will furnish two equations of condition for
the determination of these corrections.
If the observed right ascensions and declinations are used directly
instead of the longitudes and latitudes, the elements 2 and i must
be referred to the equator as the fundamental plane, and the declinations of the sun will appear in the formula for u and r obtained from
the equations (91)1, thus rendering them more complex. Their derivation offers no difficulty, being similar in all respects to that of the
equations (19) and (20), and since they will be rarely, if ever, required, it is not necessary to give the process here in detail. In
general, the equations (23) and (24) will be most convenient for
finding r and u from the geocentric spherical co-ordinates and the
elements 2 and i, since w,', w", and 4"/ remain unchanged for the
three hypotheses.
When the equator is taken as the fundamental plane, ~ is the
distance between two points on the celestial sphere for which the
geocentric spherical co-ordinates are A, D and a,', those of the sun
being denoted by A and D. Hence we shall have
sin. sin B cos 8 sin (a - A),
sin. cos B = cos D sin - sin D cos a cos (a A), (26)
cos n -= sin D sin a cos cos os S cos (a - A),
from which to find ~ and B, the angle opposite to the side 90~ --
of the spherical triangle being denoted by B. Let K denote the
right ascension of the ascending node on the equator of a great circle
passing through the places of the sun and comet or planet for the
time t, and let w0 denote its inclination to the equator; then we shall
have
sin w0 cos (A - K) = cos B,
sin w sin (A - K) - sin B sin D, (27)
cos WU = sin B cos D,
from which to find w0 and K. In a similar manner, we may com
VARIATION OF THE NODE AND INCLINATION. 327
pute the values of v" —, Qg, and i from the heliocentric spherical
co-ordinates 1, b and 1", b".
From the equations
tan (& + ) COs i (i' —') cot K('- (2),
tan(o - sin ~ (i' -- T
sin (i' - w) o
the accents being added to distinguish the elements in reference to
the equator from those with respect to the ecliptic, the values of S,
and ut (in reference to the equator) may be found. Let so denote the
angular distance between the place of the sun and that point of the
equator for which the right ascension is K, and the equation
cot so cos wo cot (K- A) (29)
gives the value of so, the quadrant in which it is situated being determined by the condition that cos s and cos(K- A) shall have the
same sign. Then we have S = S0- s,, nd
z 1800 - - - So+- S (30)
R sin (30)
r -.
sin z
from which to find r.
109. In both the method of the variation of two geocentric distances and that of the variation of g and i, instead of using the
geocentric spherical co-ordinates given by an intermediate observation, in forming the equations for the corrections to be applied to the
assumed quantities, we may use any other two quantities which may
be readily found from the data furnished by observation. Thus, if
we compute r' and u' for the date of a third observation directly
from each of the three systems of elements, the differences between
the successive results will furnish the numerical values of the partial
differential coefficients of r' and u' with respect to J and d", or with
respect to Q and i, as the case may be. Then, computing the values
of r' and u' from the observed geocentric spherical co-ordinates by
means of the values of Q and i for the system of elements to be
corrected, the differences between the results thus derived and those
obtained directly from the elements enable us to form the equations
du' dau' (
d - + dz- -- lxu',
dJ dJ/,dr' dr' (31)
A zJ - WdJ A' Ar',
dJ Wj I
328 THEORETICAL ASTRONOMY.
or the corresponding expressions in the case of the variation of 2
and i, by means of which the corrections to be applied to the assumed values will be determined. In the numerical application of
these equations, Au' being expressed in seconds of arc, ^rt should also
be expressed in seconds, and the resulting values of AJ and Ad" will
be converted into those expressed in parts of the unit of space by
dividing them by 206264.8.
When only three observed places are to be used for correcting an
approximate orbit, from the values of r, r', r" and u,.', iu" obtained
by means of the formulae which have been given, we may find p and
a or - — the latter in the case of very eccentric orbits-from the first
a -
and second places, and also from the first and third places. If these
results agree, the elements do not require any correction; but if a
difference is found to exist, by computing the differences, in the case
of each of these two elements, for three hypotheses in regard to J
and S" or in regard to 2 and i, the equations may be formed by
means of which the corrections to be applied to the assumed values
of the two geocentric distances, or to those of a and i, will be
obtained.
110. The formule which have thus far been given for the correction of an approximate orbit by varying the geocentric distances,
depend on two of these distances when no assumption is made in
regard to the form of the orbit, and these formula apply with equal
facility whether three or more than three observed places are used.
But when a series of places can be made available, the problem may
be successfully treated in a manner such that it will only be necessary
to vary one geocentric distance. Thus, let x, y, z be the rectangular
heliocentric co-ordinates, and r the radius-vector of the body at the
time t, and let X, Y, Z be the geocentric co-ordinates of the sun at
the same instant. Let the geocentric co-ordinates of the body be
designated by x0, y,, ZQ, and let the plane of the equator be taken as
the fundamental plane, the positive axis of x being directed to the
vernal equinox. Further, let p denote the projection of the radiusvector of the body on the plane of the equator, or the curtate distance with respect to the equator; then we shall have
0o p cosa, yo P in, z0o- tan (32)
If we represent the right ascension of the sun by A, and its declination by D, we also have
VARIATION OF ONE GEOCENTRIC DISTANCE. 329
X =_ R cos D cos A, Y —- R cos D sin A, Z — R sin D. (33)
The fundamental equations for the undisturbed motion of the planet
or comet, neglecting its mass in comparison with that of the sun, are
d2x kx d2y kFy d z klz
+' O,- - "
dt2 r3 dt2 r3 dt + r 0;
but since
x- = = —yoSY Xo, O, - zZ,
and, neglecting also the mass of the earth,
d2X k2X d2 Y k2 y dZ k2Z
ct2 + + A' d t'' + -- -o
these become
dXo 2 o I7,2 1 ~04
dt2 r3 (R3 r3
d'x -o+~x(I__-) —o
dt2y + I Ok2RL 3 O,1=0 (34)
dt2 r3 " r3
dt2 + -3 + kZ ( -r3-'o
Substituting for x0, y,, and z0 their values in terms of a and 8, and
putting
(R3 r3)k ]Y(:3 r3)=)7 kZ(R3-r3)=' (33)
we get
d2x, +k2p
d- + — cos + -=-O0,
dlyo k'o
^+^ ^-+^0, (36)
d sin c- + w 0, (36)
d - - +p tan - +C 0.
Differentiating the equations (32) with respect to t, we find
dxo dp. dca
* cos c ~ - p sin cc,
cdt ct dt
d sin a t- + p cosa d (37)
dt= tan dt + P sec2 d
330 THEORETICAL ASTRONOMY.
Differentiating again with respect to t, and substituting in the equations (36) the values thus found, the results are
2kp d'p( da\ dp da\
by- cos a, and add the products, we obtain
dp_ _ - Coa - 2siP dat2 0(38)
rdt t d dt dt
k k;2p d2p d2CC 2,dd ddo
3 + tn tsil dl d-+ t + 2 se COS a (38)
t sin cos a 2 (- R ~ - ) R cos D )2 sin (a Jps ),
dt d da
and the fpreceding s equation becomes
by- cos c ), and a dd t he products, we obta in
d2o.
dp ( ( 2 -t + ddt2
- da (89)
The value of -d- thus found is independent of the diffeential coa sin (a p A). Then, adding the products, since sin A = e cos A,
the result is
~2 d( cot (a - A) d -- cot a sec- a - ]
at \ dt dt d
by sinA tan, the secnd by - cos A tan, and the third by
dt- sin cot (a - A)dding the produc ts, since sin A cos A,
from which we get
da2 d2a C 2ecau"( dJ2 d
- - cot (a - A) d + sec 2 + cot + ) cot
dp dt' dt2 dt' dt'
-T=p ca dd (40)
cot ( - A) - - cot 8-sec" a
dt dt
VARIATION OF ONE GEOCENTRIC DISTANCE. 331
When the ecliptic is taken as the fundamental plane, the last term
of the numerator of the second member of this equation vanishes,
and the epuation may be written
d, (41)
dt
the coefficient C being independent of p.
111. When the value of p is given, that of d will be determined
in terms of the data furnished directly by observation and of the
differential coefficients of a and ~ with respect to t from equation
(39), or from (40), the latter being preferred when the motion of the
body in right ascension is very slow. The value of d- having been
found, we may compute the velocities of the body in directions
parallel to the co-ordinate axes. Thus, since
Xo x + X, o -- y + Y, z= + Z,
the equations (37) give
dx dp. da dX
dtco dt sn -- ddt'
dy dp duc dY
dt -sin a + p cos dt dt (42)
clt dt dZ
dz _ dp dB dZ
-- tan - +- p sec' 8 -
dt dt dt dt
dx dy dz
by means of which d-x, d- and d may be determined.
-y means of which dt dt dt
dX dY dZ
To find the values of dt- d) and -, the equations
dt dt dt
X R cos 0,
Y- R sin 0 cos s,
Z - R sin ( sin s,
give, by differentiation,
dX dR d
-- cos 0 — ~ R sin 0 —,
cit dt dt
dY dR d (43)
- sin cos e -d +Rcos 0 cose (43
dZ dR dO(
dr= sin sin - + R cos 0 sins --
dt dt dt
332 THEORETICAL ASTRONOMY.
Now, according to equation (52),, we have
d k / (1 — e2) (1 + nl,)
dt 112 (44)
mn denoting the mass of the earth, and e0 the eccentricity of its orbit.
The polar equation of the conic section gives
dr r2e sin v dv
dt p dt
Let F denote the longitude of the sun's perigee, and this equation
gives
dR _2eosin (0-__d) d k l//1 m o
d?- 2e0sin( - ) d k — /i + 0 eo sinu( -F). (45)
dt 1 - eo2 dt V/1 -e2
If we neglect the square of the eccentricity of the earth's orbit, we
have simply
d~ kl- mo dR ~
dO 1+n ~ ~ - k 1/kV1 +- mo e sinll( - ). (46)
dt -- dt
aQ dR
The values of d and dt having been found by means of these.t t.t r dX dY
formulae, the equations (43) give the required results for -, -, and
dZ dt dt
d, and hence, by means of (42), we obtain the velocities of the
comet or planet in directions parallel to the co-ordinate axes.
112. The values of x, y, and z may be derived by means of the
equations
x - - a cos a cos a X,
y = cos 8 sin a - Y,
z -- sin - Z,
and from these, in connection with the corresponding velocities, the
elements of the orbit may be found. The equations (32)1 give immediately the values of the inclination, the semi-parameter, and the
right ascension of the ascending node on the equator. Then, the
position of the plane of the orbit being known, we may compute r
and u directly from the geocentric right ascension and declination by
means of the equations (28) and (30). But if we use the values of
the heliocentric co-ordinates directly, multiplying the first of equations (93), by cos 2, and the second by sin 2, and adding the products, we have
VARIATION OF ONE GEOCENTRIC DISTANCE. 333
r sin u = z cosec i,
r cos u x cos g +- y sin 2,
from which r and 16 may be found, the argument of the latitude u
being referred to the plane of xy as the fundamental plane. The
equation
r2 = x2 + y2 + z2
gives
dr x dx y dy z dz
dt r'dt r dt- r dt' (48
and, since
dr re sin v dv dv k l/p
dt p dt' dt r'
we shall have
Vp dr
e sin v —,,
esnv k dt (49)
e cosv — 1,
r
from which to find e and v. Then the distance between the perihelion and the ascending node is given by
t0 -- uv.
The semi-transverse axis is obtained from p and e by means of the
relation
P
a _ e2
Finally, from the value of v the eccentric anomaly and thence the
mean anomaly may be found, and the latter may then be referred to
any epoch by means of the mean motion determined from a.
In the case of very eccentric orbits, the perihelion distance will be
given by
P
q1 l+e;
and the time of perihelion passage may be found from v and e by
means of Table IX. or Table X., as already illustrated.
The equation (21)1 gives, if we substitute for f its value in terms
of p, denote by V the linear velocity of the planet or comet, and neglect the mass,
V2r -_ r' = 2'p.
dt'
Let,0 denote the angle which the tangent to the orbit at the extremity of the radius-vector makes with the prolongation of this
radius-vector, and we shall have
334 THEORETICAL ASTRONOMY.
dr dx dy dz
rVcos.0~o —r - x - y - y -Z,
^r^co^s -^ dt = dt dt+Zdt'
so that the preceding equation gives
k2p = V2r sin2 40.
Hence we derive the equations
Vr sin k0 kV/'p,
dx y dz (50)
Vr cos% —, x clt Y dt dt'
from which Vr and,0 may be found. Then, since
V2 k 2 ( 1.r a
we shall have
k2 2k2
r —V, (51)
ca r
by means of which a may be determined, and then e may be found
by means of this and the value of p.
The equations (49) and (50) give
V2
e cos (u w - ) -= r sin2 0,,
and, since
V2 2 1
2 r a7
these are easily transformed into
2ae sin (ut - ) = (2a r) sin 2o,,
2ae cos ( - w) =- (2a - r) cos 2o, - r.
If we multiply the first of these equations by - cos m and the second
by sin m, and add the products; then multiply the first by sin e and
the second by cos u, and add, we obtain
2ae sin w =- (2a - r) sin (2+ - + ) - r sin u, (52)
2ae cos w - (2a - r) cos (20, + u) - r cos t,
These equations give the values of o and e.
113. We have thus derived all the formulae necessary for finding
the elements of the orbit of a heavenly body from one geocentric
distance, provided that the first and second differential coefficients of
a and 8 with respect to the time are accurately known. It remains,
VARIATION OF ONE GEOCENTRIC DISTANCE. 335
therefore, to devise the means by which these differential coefficients
may be determined with accuracy from the data furnished by observation. The approximate elements derived from three or from a
small number of observations will enable us to correct the entire
series of observations for parallax and aberration, and to form the
normal places which shall represent the series of observed places.
We may now assume that the deviation of the spherical co-ordinates
computed by means of the approximate elements from those which
would be obtained if the true elements were used, may be exactly
represented by the formula
- A + Bh + Ch, (53)
h denoting the interval between the time at which the deviation is
expressed by A and the time for which this difference is AO. The
differences between the normal places and those computed with the
approximate elements to be corrected, will then suffice to form equations of condition by means of which the values of the coefficients
A, B, and C may be determined. The epoch for which h 0 may
be chosen arbitrarily, but it will generally be advantageous to fix it
at or near the date of the middle observed place. If three observed
places are given, the difference between the observed and the computed value of each right ascension will give an equation of condition,
according to (53), and the three equations thus formed will furnish
the numerical values of A, B, and C. These having been determined, the equation (53) will give the correction to be applied to the
computed right ascension for any date within the limits of the
extreme observations of the series. When more than three normal
places are determined, the resulting equations of condition may be
reduced by the method of least squares to three final equations, from
which, by elimination, the most probable values of A, B, and C will
be derived. In like manner, the corrections to be applied to the
computed latitudes may be determined. These corrections being
applied, the ephemeris thus obtained may be assumed to represent
the apparent path of the body with great precision, and may be employed as an auxiliary in determining the values of the differential
coefficients of a and 8 with respect to t.
Let f(a) denote the right ascension of the body at the middle
epoch or that for which h 0, and let f(a - no) denote the value of
a for any other date separated by the interval nwo, in which (o is the
interval between the successive dates of the ephemeris. Then, if we
put n successively equal to 1, 2, 3, &c., we shall have
336 THEORETICAL ASTRONOMY.
Function. I. Diff. II. Diff. III. Diff. IV. Diff: V. Diff.
f(a f, ( - ~ f ) f (a-I f-) ( )
(a) f'"_(a f(- ) ) f+) (a q,o) "'
fa (a ) f (a fv oa) f, (a )- w)
f (a -., ) f( (-f", (a - 2 ) fiv( + )
f(a +) 3 ( )
manner in either direction. If we exp an f + n) into a series,
the result is
/S f~(a + ld) a+ + + &c
f ( ) + - ++- + f++l- +
or, putting for brevity A dt w, B —2 (t c2, &c.,
f(a +- n - (a + A +B2+ C + D 4+ &c.
If we now put n successively equal to -4, -3, -2, -1, 0 +1,
&c., we obtain the values of f(a - 4co), f (cta 3),...... f (a + 4o)
in terms of A, B, C, &c. Then, taking the successive orders of
differences and symbolizing them as indicated above, we obtain a
e series of equations mans of which A, B, C, &.e will bte determined in terms of the successive orders of differences.I Finally, replacing A, B, C, &c. by the quantities which they represent, and
putting
f' + (a -.,w ) - lf' (a + ) - -f' (a),
f f'" ) ('a -,) + f + -,) "' (a), &c.,
we obtain ) + i f -f
c - f ( a) - o ()ft ( ta) - A/f" (a) + &c.),
d'a 1
dt - (- (a)- - (a) + () viii (a) + &c.),
ct3 = (fa (a) - f f(a) + Bfw" (a) - &.),
d4w -o w (p v (a - f Vc(a) + eqfuii (a) - &c.), (54)
d& e ( (a f (a - ).f vii (a) + &c(),
dt o 1 ( t ( )
dts6e - (fvi() - m Vio (a) + &c.),
dt - (fr(a) — ) )+&c.), (a)
- ( a) -If.. - (a) + V(a - v( ) - -&C.),
VARIATION OF ONE GEOCENTRIC DISTANCE. 337
by means of which the successive differential coefficients of a with
respect to t may be determined. The derivation of these coefficients
in the case of 8 is entirely analogous to the process here indicated for
a. Since the successive differences will be expressed in seconds of
arc, the resulting values of the differential coefficients of a and 8 withrespect to t will also be expressed in seconds, and must be divided by
206264.8 in order to express them abstractly.
du cld 2 d8 cd2
We may adopt directly the values of, dt-' - and ddetermined
by means of the corrected ephemeris, or, if the observed places do
not include a very long interval, we may determine only the values
dofa d. d d
of d-7' do' &c. by means of the ephemeris, and then find cl- and dt
directly from the normal places or observations. Thus, let A as, aft
be three observed right ascensions corresponding to the times t, t' t",
and we shall have
d' d'' (t, t)icit d3 d4 t) &
which give
d' (t' —t)'~'d" a- a ddl' d t4
T- dt -'-i (t' — t). +' -> - t) -t -4 &c.,
u~t"dd' a'!"- / d',, d "4-,'
+t l (t-, - - (') -~ -- ~'- t ~ 3~ — t ~') ~ -- c.
t ~ dtZ dt4
These equations, being solved numerically, will give the values of
cI2& dt
and, and we may thus by triple combinations of the observed
dt2
places, using always the same middle place, form equations of condition for the determination of the most probable values of these
differential coefficients by the solution of the equations according to
the method of least squares.
In a similar manner the values of t- and -d may be derived.
114. In applying these formula to the calculation of an orbit,
after the normal places have been derived, an ephemeris should be
computed at intervals of four or eight days, arranging it so that one
of the dates shall correspond to that of the middle observation or
normal place. This ephemeris should be computed with the utmost
22
338 THEORETICAL ASTRONOMY.
care, since it is to be employed as an auxiliary in determining quantities on which depends the accuracy of the final results. The comparison of the ephemeris with the observed places will furnish, by
means of equations of the form
A + Bh + C+h2 _ A,'
A' + B'h + C'h- - a',
h being the interval between the middle date t' and that of the place
used, the values of A, B, C, A', &c.; and the corrections to be
applied to the ephemeris will be determined by
A + Bnw + Cnw2( = a-,
A' -+ B'nw + C'n2o2-= A.
The unit of h may be ten days, or any other convenient interval,
observing, however, that c(o in the last equations must be expressed
in parts of the same unit. With the ephemeris thus corrected, we
dcu d2a d5 d8,
compute the values of dt, dt-, and d as already explained. These
differential coefficients should be determined with great care, since it
is on their accuracy that the subsequent calculation principally dedX ddY dZ
pends. We compute, also, the velocities dt - dt and dt by means
of the formulae (43), d and being computed from (46). The
dt dt
quantities thus far derived remain unchanged in the two hypotheses
with regard to A.
Then we assume an approximate value of A, and compute
p a cos;
and by means of the equation (40) or (39) we compute the value of
dp
dt It will be observed that if we put the equation (40) in the form
dp P
P + cotS,
the coefficient P remains the same in the two hypotheses. The three
dp
equations (38) may be so combined that the resulting value of
d'a dt
will not contain dt. This transformation is easily effected, and may
d2Ca
be advantageous in special cases for which the value of is very
uncertain.
The heliocentric spherical co-ordinates will be obtained from the
RELATION BETWEEN TWO PLACES IN THE ORBIT. 339
assumed value of a by means of the equations (106)3, and the rectangular co-ordinates from
x - r cos b cos I,
y r cos b sin 1,
z - r sin b.
dx dy dz
The velocities d-, d, and c will be given by (42), and from these
and the co-ordinates x, y, z the elements of the orbit will be computed by means of the equations (32)1, (47), (49), &c. With the
elements thus derived we compute the geocentric places for the dates
of the normals, and find the differences between computation and
observation. Then a second system of elements is computed from
a + AD, and compared with the observed places. Let the difference
between computation and observation for either of the two spherical
co-ordinates be denoted by n for the first system of elements, and by
n' for the second system. The final correction to be applied to a, in
order that the observed place may be exactly represented, will be
determined by
~ (n' — n) + n - O. (56)
Each observed right ascension and each observed declination will
thus furnish an equation of condition for the determination of AJ,
observing that the residuals in right ascension should in each case be
multiplied by cos A. Finally, the elements which correspond to the
geocentric distance A + AD will be determined either directly or by
interpolation, and these must represent the entire series of observed
places.
115. The equations (52)3 enable us to find two radii-vectores when
the ratio of the corresponding curtate distances is known, provided
that an additional equation involving r, r", x, and known quantities
is given. For the special case of parabolic motion, this additional
equation involves only the interval of time, the two radii-vectores,
and the chord joining their extremities. The corresponding equation
for the general conic section involves also the semi-transverse axis
of the orbit, and hence, if the ratio M of the curtate distances is
known, this equation will, in connection with the equations (52)3,
enable us to find the values of r and r" corresponding to a given
value of a. To derive this expression, let us resume the equations
340 THEORETICAL ASTRONOMY.
-E" - E- 2e sin (E" -E) cos (E"- E), (5
r + r" 2a - 2ae cos 4 (E" - E) cos (E" + E).
For the chord x we have
x (r + r")2 - 4rr" cos2 (" -r t)0,
which, by means of (58)4, gives
2 = (1' + r//)2
-4a2 (co2 (E/ -E)- 2ecos(- (E/+ E) +-e2 cos (Ez-+E));
and, substituting for r + r" its value given by the last of equations
(57), we get
X2 42 sin2 (E" - ) (I 2 cos2-(E" + E)). (58)
Let us now introduce an auxiliary angle h, such that
cos h = e cos 4 (E" - E),
the condition being imposed that h shall be less than 1800, and put
g (E" — E);
then the equations (57) and (58) become
-3 2g -2 sing cosh,
ay
r + r" -_ 2a (1 -cos g cos h), (59)
x 2a sin g sin h.
Further, let us put
h -g =, h -g= ~,
and the last two of equations (59) give
r + r" + - 4a sin' I,6
r + r" - 4a sin2 I. (60)
Introducing 8 and e into the first of equations (59), it becomes
( - sin e) - ( -sin ). (61)
a2
The formule (60) enable us to determine e and 8 from r + rl, x,
and a, and then the time' =k (t" - t) may be determined from
(61). Since, according to (58)4,'rr" cos (u"' u) -a (cos g - cos h) = 2Asin Z sin 8,
RELATION BETWEEN TWO PLACES IN THE ORBIT. 341
and since sin s is necessarily positive, it appears that when u" --
exceeds 180~, the value of sin 18 must be negative, and when
" — u- 180~, we have 8 -0; and thus the quadrant in which
8 must be taken is determined. It will be observed that the value
of E, as given by the first of equations (60), may be either in the
first or the second quadrant; but, in the actual application of the
formulae, the ambiguity is easily removed by means of the known
circumstances in regard to the motion of the body during the interval t~ - t.
In the application of the equations (52)3, by means of an approximate value of x we compute d, and thence r and r". Then we compute e and a corresponding to the given value of a, and from (61)
we derive the value of
tj~k'
If this agrees with the observed interval t - t, the assumed value
of z is correct; but if a difference exists, by varying x we may,readily find, by a few trials, the value which will exactly satisfy the
equations. The formulae (70)3 will then enable us to determine the
curtate distances p and p", and from these and the observed spherical
co-ordinates the elements of the orbit may be found.
As soon as the values of u and u" have been computed, since
e - = El- E, we have, according to equation (85)4,
sin - (u"- u)
cos ( P - ( ~I rr"
a sin - (~ - )
which may be used to determine 5p when the orbit is very eccentric.
To find p and q, we have
p = a cos'2, q - 2a sin2 (450 -_ );
and the value of wo may be found by means of the equations (87)4 or
(88),.
116. The process here indicated will be applied chiefly in the determination of the orbits of comets, and generally for cases in which
a is large. In such cases the angles e and a will be small, so that
the slightest errors will have considerable influence in vitiating the
value of t" - t as determined by equation (61); but if we transform
this equation so as to eliminate the divisor a' in the first member, the
uncertainty of the solution may be overcome. The difference e -sine
342 THEORETICAL ASTRONOMY.
may be expressed by a series which converges rapidly when z is small.
Thus, let us put
e - sin C = y sin sin, x =- sin2 I,
and we have
dy = 2 cosec -e — y cot -,
de:z 2yc9t',
de
d- = 4 cosec -e.
dx
Therefore
dy 8 -6y cos l 4- 3y ( -- 2x)
dx sin2 -e 2x(1 x)
If we suppose y to be expanded into a series of the form
y = a- fx + rx + - x3 - &c.,
we get, by differentiation,
dy
dx= + 2rx + 3x+2 + &c.,
and substituting for d- the value already obtained, the result is
2jx + (4r - 21) x2 + (68 - 4r) x + &c. 4 - 3a + (6 - 31) x
+ (6 - 3) X2 + (6r - 3a) x + &c.
Therefore we have
4 - 3a 0, 6 - 3f — 23,
63 - 3 -- 4r 23, 6r - 3 - 66 - 4r,
from which we get
-4, 46 4.6.8 4.6.8.10
"~ P-3.5' r"3.57' """7 579&c.
i-ay 4 -3.5.. o. 7.9 &
Hence we obtain
e-sin e =Si + 4 Snine+ s. 6.8.1sie + c.) (62)
and, in like manner,
6 8 4' 6.8.10
a-sin n'(1+sin + sin 4sin ( l ssin sin &c.), (63)
3 2 5 \ 5 T 7 4 5.7.9 4
which, for brevity, may be written
e- sin e _ 4 Q sin3, I,
sin Q'sin, z (64)
-- sin a -- ~ Q' sin3,,
RELATION BETWEEN TWO PLACES IN THE ORBIT. 343
Combining these expressions with (61), and substituting for sin 2s and
sin 2 their values given by the equations (60), there results
6 — Q (r + r" + x)2 T Q' (r + r" - ), (65)
the upper sign being used when the heliocentric motion of the body
is less than 180~, and the lower sign when it is greater than 180~.
The coefficients Q and Q' represent, respectively, the series of terms
enclosed in the parentheses in the second members of the equations
(62) and (63), and it is evident that their values may be tabulated
with the argument s or 8, as the case may be. It will be observed,
however, that the first two terms of the value of Q are identical with
the first two terms of the expansion of (cos4s)-2 into a series of
ascending powers of sin i, while the difference is very small between
the coefficients of the third terms. Thus, we have
(cos 2)- - - (1 - sin2 6) — 1 + sinC + s.11
sin 5= 215 1 sin ~
6. 11. 16, &
+ 5 10 15sin +
and if we put
Bo
-Q o (66)
(cos i-) 512
(COS 4l)e
we shall have
B - + + si4 T+1 Sin6 I- + &c. (67)
In a similar manner, if we put
(Q = B', 12 (68)
(cos Ia) 1
we find
Bo' = 1 + 9 sin4 1 + 2142 sin6 S + &c. (69)
Table XV. gives the values of Bo or Bo0 corresponding to s or a from
0~ to 60~.
For the case of parabolic motion we have
Q-=, Q' -,
and the equation (65) becomes identical with (56)3.
In the application of these formulhe, we first compute s and 8 by
means of the equations (60), and then, having found B0 and B0' by
means of Table XV., we compute the values of Q and Q' from (66)
and (68). Finally, the time r' k(t"- t) will be obtained from (65),
and the difference between this result and the observed interval will
344 THEORETICAL ASTRONOMY.
indicate whether the assumed value of X must be increased or diminished. A few trials will give the correct result.
117. Since the interval of time t — t cannot be determined with
sufficient accuracy from (65) when the chord z is very small, it
becomes necessary to effect a further transformation of this equation.
Thus, let us put
Q - Q = 6P, x - sin ]e, x'= sin2,
and we shall have
— P a -- x') ( 1 + ( + x') + 7.9 (2 + xx' + X) + &c. ).
Now, when X is very small, we may put
cos I -= cos 8,
and hence
sin2 l sin12
x x' - sin' l -~ sin2 la- si n ~- - 4 cos2'
which, by means of equations (60), becomes
x - X
8a cos2 t
Therefore we have, when x is very small,
40a cs ( + 1 sin2 2 + 8- sin4 + &c.) (70)
If we put
TO, (71)
the equation (65) becomes, using only the upper sign,
(r + r" + )- - (r + r" - ) - 6o, (72)
which is of the same form as (56)3. Hence, according to the equations (63)3 and (66),, we shall have
2r<'
x 2ro' /_, (73)
the value of, being found from Table XI. with the argument
2=o - (74)
yr -i — 3
RELATION BETWEEN TWO PLACES IN THE ORBIT. 345
It remains, therefore, simply to find a convenient expression for r0',
and the determination of x is effected by a process precisely the same
as in the special case of parabolic motion.
Let us now put
P c N
q 40a cos'4
and we shall have
C os, 1 2.8 3.8.10 4.8.10.12 6 &\
N_ 1+ 7sin"i+ -4sin4 +-7-_ Si4in
Q 7. 7.9 "9.11
or, substituting for Q its value in terms of sin es,
N- 1 + 3-a sin e + o 6sin 4 + 2o7 in6 e + &c. (75)
Therefore, if we put
a 40a r + -- cOS4lx, (76)
the expression for r,' becomes
T-0 Ar. (77)
Table XV. gives the value of log N corresponding to values of e
from -- 0 to - 60~.
If the chord z is given, and the interval of time t1 - t is required,
we compute Ar0' by means of (76), and, having found ro' from
x, ^r + r"
To - 2,
as in the case of parabolic motion, we have
t' t- Q ( -ls way we obtain normal places at convenient intervals]
throughout the entire period during' which the body was observed.
From three or more of these norlmal places, a new system of elements
should be conmpuited by means of soine one of the niethods which
have already been given; and these fundamental places beingl judiciouslly selected, the resultilni, elements will furnish a pretty cl1ose
approximatio t to thei truth o tllht te residuals A hich are found bycomparing them with all the direct, observed places may be regarded
as indicating very nearly the actual errors of those places. We 1ay
thel proceed to inlvestigate the character of the observations mlore
fully. But since the observations wxill have been made at iman!l diiferent places, bvy i-fferent observers,.with instrumlents of diiff1re!nt
sizes, atIdnd unde'la variety of dissimilat attendantl circumsta.nl esl it
may be easily.. understood that the investigation will involve m1uch
thalt is vague and uncertain.. In the theory of errors which has been
developed 1n fl s-chapter, it has been assumed that all consta nt
errors have been duly eliminated, and that the only errors w cllieh
remain are those accidental errors which must ever continue in a
greater or less degree undetermined. The greater the number and
404 THEORETICA.L ASTRONOMIY.
perfection of the observations employed, the more nearly will these
errors be determined, and the more nearly will the law of their distribution conform to that which has been assumed as the basis of
the method of least squares.
When all known errors have been eliminated, there may yet remain
constant errors, and also other errors whose law of distribution is
peculiar, such as may arise from the idiosyncrasies of the different
observers, from the systematic errors of the adopted star-places in
the case of differential observations, and from a variety of other
sources; and since the observations themselves furnish the only means
of arriving at a knowledge of these errors, it becomes important to
discuss them in such a manner that all errors which may be regarded,
in a sense more or less extended, as regulcr may be eliminated.
When this has been accomplished, the residuals which still remain
will enable us to form an estimate of the degree of accuracy which
may be attributed to the different series of observations, in order that
they may not only be combined in the most advantageous manner,
but that also no refinements of calculation may be introduced which
are not warranted by the quality of the material to be employed.
The necessity of a preliminary calculation in which a high degree
of accuracy is already obtained, is indicated by the fact that, however
conscientious the observer may be, his judgment is unconsciously
warped by an inherent desire to produce results harmonizing well
among themselves, so that a limited series of places may agree to
such an extent that the probable error of an observation as derived
from the relative discordances would assign a weight vastly in excess
of its true value. The combination, however, of a large number of
independent data, by exhibiting at least an approximation to the
absolute errors of the observations, will indicate nearly what the
measure of precision should be. As soon, therefore, as provisional
elements which nearly represent the entire series of observations have
been found, an attempt should be made to eliminate all errors which
may be accurately or approximately determined. The places of the
comparison-stars used in the observations should be determined with
care from the data available, and should be reduced, by means of the
proper systematic corrections, to some standard system. The reduction of the mean places of the stars to apparent places should also be
made by means of uniform constants of reduction. The observations
will thus be uniformly reduced. Then the perturbations arising from
the action of the planets should be computed by means of formula
which will be investigated in the next chapter, and the observed
COMBINATION OF OBSERVATIONS. 405
places should be freed from these perturbations so as to give the
places for a system of osculating elements for a given date.
147. The next step in the process will be to compare the provisional elements with the entire series of observed places thus corrected; and in the calculation of the ephemeris it will be advantageous to correct the places of the sun given by the tables whenever
observations are available for that purpose. Then, selecting one or
more epochs as the origin, if we compute the coefficients A, B, C in
the equation
A0 = A + Br + Cr2, (125)
in the case of each of the spherical co-ordinates, by means of equations of condition formed from all the observations, the standard
ephemeris may be corrected so that it may be regarded as representing
the actual path of the body during the period included by the observations. When the number of observations is considerable, it will be
more convenient to divide the observations into groups, and use the
differences between computation and observation for provisional
normal places in the formation of the equations of condition for the
determination of A, B, and C. It thus appears that the corrected
ephemeris which is so essential to a determination of the constant
errors peculiar to each series of observations, is obtained without first
having determined the most probable system of elements. The corrections computed by means of the equation (125) being applied to
the several residuals of each series, we obtain what may be regarded
as the actual errors of these observations. The arithmetical or probable mean of the corrected residuals for the series of observations
made by each observer may be regarded as the average error of observation for that series. The mean of the average errors of the several
series may be regarded as the actual constant error pertaining to all
the observations, and the comparison of this final mean with the
means found for the different series, respectively, furnishes the probable value of the constant errors due to the peculiarities of the
observers; and the constant correction thus found for each observer
should be applied to the corresponding residuals already obtained.
In this investigation, if the number of comparisons or the number
of wires taken is known, relative weights proportional to the number
of comparisons may be adopted for the combination of the residuals
for each series. In this manner, observations which, on account of
the peculiarities of the observers, are in a certain sense heterogeneous,
may be rendered homogeneous, being reduced to a. standard which
406 THEORETICAL ASTRONOMY.
approaches the absolute in proportion as the number and perfection
of the distinct series combined are increased. Wrhatever constant
error remains will be very small, and, besides, will affect all places
alike.
The residuals which now remain must be regarded as consisting
of the actual errors of observation and of the error of the adopted
place of the comparison-star. Hence they will not give the probable
error of observation, and will not serve directly for assigning the
measures of precision of the series of observations by each observer.
Let us, therefore, denote by E the mean error of the place of the
comparison-star, by e, the mean error of a single comparison; then
will -- be the mean error of In comparisons, and the mean error of
the resulting place of the body will, according to equation (35), be
given by
co C + es2. (126)
The value of %, in the case of each series, will be found by means of
the residuals finally corrected for the constant errors, and the value
of zs is supposed to be determined in the formation of the catalogue
of star-places adopted. Hence the actual mean error of an observation consisting of a single comparison will be
Vm,- ~/~(E2 _ E2). (127)
The value of e, for each observer having been found in accordance
with this equation, the mean error of an observation consisting of m
comparisons will be
The mean error of an observation whose weight is unity being denoted by e, the weight of an observation based on m comparisons will
be
p 2, (128)
The value of e may be arbitrarily assigned, and we may adopt for it
~ 10" or any other number of seconds for which the resulting values
of p will be convenient numbers.
When all the observations are differential observations, and the stars
of comparison are included in the fundamental list, if we do not take
into account the number of comparisons on which each observed
COMBINATION OF OBSERVATIONS. 407
place depends, it will not be necessary to consider e, and we may
then derive E, directly from the residuals corrected for constant errors.
Further, in the case of meridian observations, the error which corresponds to es will be extremely small, and hence it is only when these
are combined with equatorial observations, or when equatorial observations based on different numbers of comparisons are combined, that
the separation of the errors into the two component parts becomes
necessary for a proper determination of the relative weights.
According to the complete method here indicated, after having
eliminated as far as possible all constant errors, including the corrections assigned by equation (125) to be applied to the provisional
ephemeris, we find the value of a, given by the equation
n^,2 [nvv] - [Mn] s, (129)
in which in denotes the number of observations; i m, m n'n" &c. the
number of comparisons for the respective observations; and v, v', v",
&c. the corresponding residuals. Then, by means of equation (128),
assuming a convenient number for E, we compute the weight of each
observation. Thus, for example, let the residuals and corresponding
values of m be as follows:AO m AO m
+ 2".0 5, - ".0 7,
1.8 5, + 1.5 5,
- 0.4 10, + 4.1 8,
-5.5 5, 0.0 5.
Let the mean error of the place of a comparison-star be
e - 2".0;
then we have n - 8, and, according to (129),
8-,2 341.78 - 200.0,
which gives
e,- + 4".2.
Let us now adopt as the unit of weight that for which the mean erroi is
— 3".0;
then we obtain by means of equation (128), for the weights of the
observations,
2.5, 2.5, 5.1, 2.5, 3.6, 2.5, 4.1, 2.5,
respectively.
408 THEORETICAL ASTRONOMY.
In this manner the weights of the observations in the series made
by each observer must be determined, using throughout the same
value of s. Then the differences between the places computed from
the provisional elements to be corrected and the observed places corrected for the constant error of the observer, must be combined according to the equations (123) and (125), the adopted values of p, p',
p", &c. being those found from (128). Thus will be obtained the
final residuals for the formation of the equations of condition from
which to derive the most probable value of the corrections to be
applied to the elements. The relative weights of these normals will
be indicated by the sums formed by adding together the weights of
the observations combined in the formation of each normal, and the
unit of weight will depend on the adopted value of e. If it be desired to adopt a different unit of weight in the case of the solution
of the equations of condition, such, for example, that the weight of
an equation of average precision shall be unity, we may simply divide
the weights of the normals by any number p0 which will satisfy the
condition imposed. The mean error of an observation whose weight
is unity will then be given by
the value of being that used in the determination of the weights
the value of e being that used in the determination of the weights 1,
p'/ &c.
148. The observations of comets are liable to be affected by other
errors in addition to those which are common to these and to planetary observations. Different observers will fix upon different points
as the proper point to be observed, and all of these may differ from
the actual position of the centre of gravity of the comet; and further, on account of changes in the physical appearance of the comet,
the same observer may on different nights select different points.
These circumstances concur to vitiate the normal places, inasmuch as
the resulting errors, although in a certain sense fortuitous, are yet
such that the law of their distribution is evidently different from
that which is adopted as the basis of the method of least squares.
The impossibility of assigning the actual limits and the law of distribution of many errors of this class, renders it necessary to adopt
empirical methods, the success of which will depend on the discrimination of the computer.
If e0 denotes the mean error of an observation based on n corm
COMBINATION OF OBSERVATIONS. 409
parisons, and e, the mean error to be feared on account of the peculiarities of the physical appearance of the comet,
2 2+
~o + e2
will express the mean error of the residuals; and if n of these
residuals are combined in the formation of a normal place, the mean
error of the normal will be given by
-- [2 ] + 2. (130)
The value of es may be determined approximately from the data
furnished by the observations. Thus, if the mean error of a single
comparison, for the different observers, has been determined by means
of the differences between single comparisons and the arithmetical
mean of a considerable number of comparisons, and if the mean error
of the place of a comparison-star has also been determined, the
equation (126) will give the corresponding value of e02; then the
actual differences between computation and observation obtained by
eliminating the error of the ephemeris and such constant errors as
may be determined, will furnish an approximate value of es by means
of the formula
2_ _ _ C 2
in which n denotes the number of observations combined.
Sometimes, also, in the case of comets, in order to detect the operation of any abnormal force or circumstance producing different effects
in different parts of the orbit, it may be expedient to divide the
observations into two distinct groups, the first including the observations made before the time of perihelion passage, and the other
including those subsequent to that epoch.
149. The circumstances of the problem will often suggest appropriate modifications of the complete process of determining the relative weights of the observations to be combined, or indeed a relaxation from the requirements of the more rigorous method. Thus, if
on account of the number or quality of the data it is not considered
necessary to compute the relative weights with the greatest precision
attainable, it will suffice, when the discussion of the observations has
been carried to an extent sufficient to make an approximate estimate
of the relative weights, to assume, without considering the number
of comparisons, a weight 1 for the observations at one observatory, a
410 THEORETICAL ASTRONOMY.
weight I for another class of observations, 1 for a third class, and so
on. It should be observed, also, that when there are but few observations to be combined, the application of the formulae for the mean
or probable errors may be in a degree fallacious, the resulting values
of these errors being little more than rude approximations; still the
mean or probable errors as thus determined furnish the most reliable
means of estimating the relative weights of the observations made
by different observers, since otherwise the scale of weights would
depend on the arbitrary discretion of the computer. Further, in a
complete investigation, even when the very greatest care has been
taken in the theoretical discussion, on account of independent known
circumstances connected with some particular observation, it may be
expedient to change arbitrarily the weight assigned by theory to
certain of the normal places. It may also be advisable to reject
entirely those observations whose weight is less than a certain limit
which may be regarded as the standard of excellence below which
the observations should be rejected; and it will be proper to reject
observations which do not afford the data requisite for a homogeneous
combination with the others according to the principles already
explained. But in all cases the rejection of apparently doubtful
observations should not be carried to a.ny considerable extent unless
a very large number of good observations are available. The mere
apparent discrepancy between any residual and the others of a series,
is not in itself sufficient to warrant its rejection unless facts are
known which would independently assign to it a low degree of precision.
A doubtful observation will have the greatest influence in vitiating
the resulting normal place when but a small number of observed
places are combined; and hence, since we cannot assume that the law
of the distribution of errors, according to which the method of least
squares is derived, will be complied with in the case of only a few
observations, it will not in general be safe to reject an observation provided that it surpasses a limit which is fixed by the adopted theory
of errors. If the number of observations is so large that the distribution of the errors may be assumed to conform to the theory
adopted, it will be possible to assign a limit such that a residual
which surpasses it may be rejected. Thus, in a series of m observations, according to the expression (19), the number of errors greater
than nr will be
(1 2/ -t
Qm 1 -- I Jetdt)'
\ 1/7T ^0 /
COMBINATION OF OBSERVATIONS. 411
and when n has a value such that the value of this expression is less
than 0.5, the error nr will have a greater probability against it than
for it, and hence it may be rejected. The expression for finding the
limiting value of n therefore becomes
97 hr
- e-. l-dt 1 2 (131)
77 o 2m'
By means of this equation we derive for given values of m the corresponding values of nhr = 0.47694n, and hence the values of n.
For convenient application, it will be preferable to use e instead of r,
and if we put n' - 0.67449n, the limiting error will be n's, and the
values of n' corresponding to given values of m will be as exhibited
in the following table.
TABLE.
m it in 21? m m'
6 1.732 20 2.241 55 2.608 90 2.773
8 1.863 25 2.326 60 2.638 95 2.791
10 1.960 30 2.394 65 2.665 100 2.807
12 2.037 35 2.450 70 2.690 200 3.020
14 2.100 40 2.498 75 2.713 300 3.143
16 2.154 45 2.539 80 2.734 400 3.224
18 2.200 50 2.576 85 2.754 500 3.289
According to this method, we first find the mean error of an observation by means of all the residuals. Then, with the value of m as
the argument, we take from the table the corresponding value of a',
and if one of the residuals exceeds the value In'e it must be rejected.
Again, finding a new value of e from the remaining n - 1 residuals,
and repeating the operation, it will be seen whether another observation should be rejected; and the process may be continued until a
limit is reached which does not require the further rejection of observations. Thus, for example, in the case of 50 observations in
which the residuals -11".5 and +7".8 occur, let the sum of the
squares of the residuals be
[vv] - 320.4.
Then, according to equation (30), we shall have
- - 4- 2".56.
412 THEORETICAL ASTRONOMY.
Corresponding to the value m 50, the table gives n'- 2.576, and
the limiting value of the error becomes
n' = 6".6;
and hence the residuals -11".5 and +7".8 are rejected. Recomputing the mean error of an observation, we have
320.4 — 193.09 + "65.47 -~i- 1".65.
In the formation of a normal place, when the mean error of an
observation has been inferred from only a small number of observations, according to what has been stated, it will not be safe to rely
upon the equation (131) for the necessity of the rejection of a doubtful observation. But if any abnormal influence is suspected, or if
any antecedent discussion of observations by the same observer, made
under similar circumstances, seems to indicate that an error of a given
magnitude is highly improbable, the application of this formula will
serve to confirm or remove the doubt already created. Much will
therefore depend on the discrimination of the computer, and on his
knowledge of the various sources of error which may conspire continuously or discontinuously in the production of large apparent
errors. It is the business of the observer to indicate the circumstances peculiar to the phenomenon observed, the instruments employed, and the methods of observation; and the discussion of the
data thus furnished by different observers, as far as possible in accordance with the strict requirements of the adopted theory of errors,
will furnish results which must be regarded as the best which can be
derived from the evidence contributed by all the observations.
150. When the final normal places have been derived, the differences between these and the corresponding places computed from the
provisional elements to be corrected, taken in the sense computation
minus observation, give the values of n, n', n", &c. which are the
absolute terms of the equations of condition. By means of these
elements we compute also the values of the differential coefficients of
each of the spherical co-ordinates with respect to each of the elements
to be corrected. These differential coefficients give the values of the
coefficients a, b, c, a', b', &c. in the equations of condition. The
node of calculating these coefficients, for different systems of co-ordinates, and the mode of forming the equations of condition, have
been fully developed in the second chapter. It is of great import
CORRECTION OF THE ELEMENTS. 413
ance that the numerical values of these coefficients should be carefully checked by direct calculation, assigning variations to the elements, or by means of differences when this test can be successfully
applied. In assigning increments to the elements in order to check
the formation of the equations, they should not be so large that the
neglected terms of the second order become sensible, nor so small that
they do not afford the required certainty by means of the agreement
of the corresponding variations of the spherical co-ordinates as
obtained by substitution and by direct calculation.
As soon as the equations of condition have been thus formed, we
multiply each of them by the square root of its weight as given by
the adopted relative weights of the normal places; and these equations will thus be reduced to the same weight. In general, the
numerical values of the coefficients will be such that it will be convenient, although not essential, to adopt as the unit of weight that
which is the average of the weights of the normals, so that the
numbers by which most of the equations will be multiplied will not
differ much from unity. The reduction of the equations to a uniform
measure of precision having been effected, it remains to combine them
according to the method of least squares in order to derive the most
probable values of the unknown quantities, together with the relative
weights of these values. It should be observed, however, that the
numerical calculation in the combination and solution of these equations, and especially the required agreement of some of the checks of
the calculation, will be facilitated by having the numerical values of
the several coefficients not very unequal. If, therefore, the coefficient
a of any unknown quantity x is in each of the equations numerically
much greater or much less than in the case of the other unknown
quantities, we may adopt as the corresponding unknown quantity to
be determined, not x but,x, v being any entire or fractional number
such that the new coefficients, -, &c. shall be made to agree in
magnitude with the other coefficients. The unknown quantity whose
value will then be derived by the solution of the equations will be
vx, and the corresponding weight will be that of vx. To find the
weight of x from that of Zx, we have the equation
P-^ __ 2 (132)
In the same manner, the coefficient of any other unknown quantity
may be changed, and the coefficients of all the unknown quantities
may thus be made to agree in magnitude within moderate limits, the
414 THEORETICAL ASTRONOMY.
advantage of which, in the numerical solution of the equations, will
be apparent by a consideration of the mode of proving the calculation of the coefficients in the normal equations. It will be expedient,
also, to take for v some integral power of 10, or, when a fractional
value is required, the corresponding decimal. It may be remarked,
further, that the introduction of v is generally required only when
the coefficient of one of the unknown quanitiies is very large, as
frequently happens in the case of the variation of the mean daily
motion,.
When the coefficients of some of the unknown quantities are
extremely small in all the equations of condition to be combined, an
approximate solution, and often one which is sufficiently accurate for
the purposes required, may be obtained by first neglecting these
quantities entirely, and afterwards determining them separately. In
general, however, this can only be done when it is certainly known
that the influence of the neglected terms is not of sensible magnitude,
or when at least approximate values of these terms are already given.
When we adopt the approximate plane of the orbit as the fundamental plane, the equations for the longitude involve only four elements, and the coefficients of the variations of these elements in the
equations for the latitudes are always very small. Hence, for an
approximate solution, we may first solve the equations involving four
unknown quantities as furnished by the longitudes, and then, substituting the resulting values in the equations for the latitudes, they
will contain but two unknown quantities, namely, those which give
the corrections to be applied to 2 and i.
151. When the number of equations of condition is large, the
computation of the numerical values of the coefficients in the normal
equations will entail considerable labor; and hence it is desirable to
arrange the calculation in a convenient form, applying also the checks
which have been indicated. The most convenient arrangement will
be to write the logarithms of the absolute terms n, n', In", &c. in a
horizontal line, directly under these the logarithms of the coefficients
a, aC, a", &c., then the logarithms of b, b', b", &c., and so on. Then
writing, in a corresponding form, the values of log n, log n9', &c. on a
slip of paper, by bringing this successively over each line, the sums
[nn], [an], [bn], &c. will be readily formed. Again, writing on
another slip of paper the logarithms of a, a', a", &c., and placing
this slip successively over the lines containing the coefficients, we
derive the values [aa], [ab], [ac], &c. The multiplication by b, c, d,
CORRECTION OF THE ELEMENTS. 415
&c. successively is effected in a similar manner; and thus will be
derived [bb], [be], [bd], &c., and finally [ff] in the case of six unknown quantities. In forming these sums, in the cases of sums of
positive and negative quantities, it is convenient as well as conducive
to accuracy to write the positive values in one vertical column and
the negative values in a separate column, and take the difference of
the sums of the numbers in the respective columns. The proof of
the calculation of the coefficients of the normal equations is effected
by introducing s, s', s", &c., the algebraic sums of all the coefficients
in the respective equations of condition, and treating these as the
coefficients of an additional unknown quantity, thus forming directly
the sums [sn], [as], [bs], [cs], &c. Then, according to the equations
(76) and (77), the values thus found should agree with those obtained
by taking the corresponding sums of the coefficients in the normal
equations.
The normal equations being thus derived, the next step in the
process is the determination of the values of the auxiliary quantities
necessary for the formation of the equations (74). An examination
of the equations (54), (55), &c., by means of which these auxiliaries
are determined, will indicate at once a convenient and systematic
arrangement of the numerical calculation. Thus, we first write in a
horizontal line the values of [aca], [ab], [ac],... [as], [a], and directly under them the corresponding logarithms. Next, we write
under these, commencing with [ab], the values of [bb], [be], [bd],
F.. bs], [bn]; then, adding the logarithm of the factor ab] to the
L a[aa]
logarithms of [ab], [ac], &c. successively, we write the value of
[ab] [] under [bb], that of [a] ac] under [be], and so on. Sub[aa] [a]ct
tracting the numbers in this line from those in the line above, the
differences give the values of [bb.1], [bc.1],... [bs.1], [bn.], to be
written in the next line, and the logarithms of these we write directly
under them. Then we write in a horizontal line the values of [cc],
[cd],. [s], [en], placing [cc] under [bc.l], and, having added the
logarithm of [a] to the logarithms of [ac], [ad], &c. in succession,
[aca]
we derive, according to the equations (55) and (58), the values of
[CC.1], [cd.1],.. [cs.1], [cn.1], which are to be placed under the corresponding quantities [cc], [cd], &c. Next, we subtract from these,
respectively, the products
[be.ll] [bc.], [b.1] -[b.], [.1] [b],
[bb.1]^, [bbd. ] [bs.l], [bb.1],
[ub.] [bb.l]' [ bb.1 ] [ubS.
416 THEORETICAL ASTRONOMY.
and thus derive the values of [cc.2], [cd.2],.. [cs.2], [cn.2], which
are to be written in the next horizontal line and under them their
logarithms. Then we introduce, in a similar manner, the coefficients
[dd], [de],. [. dn], writing [dd] under [cd.2]; and from each of these
in succession we subtract the products
[ad] ], [ad] [a], [ad]
[aa] aa] [aa] [aa ],
thus finding the values of [dd.l], [de.1],.. [dn. 1]. From these we
subtract the products
[bd.1] [bd.1] [bd.i]
[bb.1] [bb.1] [b b],..b [b. 1],
respectively, which operation gives the values of [dd.2], [de.2],...
[dn.2]. From these results we subtract the products
[d.2] [ecd.2], [ed.2] [ [e.2], [ cn.2],
[cc.2] [cc.2] [cc.2]
and derive [dd.3], [de.3],.. [dn.3] under which we write the corresponding logarithms. Then we introduce [ee], f, ], [es], and [en],
writing [ee.] under [de.3]. First, subtracting ] [ae], [] [af],.
~~~~~[ae] l~[aa] [aa]
[a] [an], we get [ee.l], [ef.1], [es.1], and [en.1]; then subtracting
[aaJ
from these the products
[b5 e. 1] [b] [be],** [be.1 ][b ]
^^ be~~~[b. 1], r. 111],.
[bb.l] [bb.1] [bb.1] ^
we obtain the values of [ee.2], [ef.2], [es.2], and [en.2]. Again,
subtracting
[c.2] [e.2] [c2 f.2] ce. 2] [cn.2],
[ce.2] [cc.2] [ec.2],
we have the values of [ee.3], [ef.3], [es.3], [en.3]; and finally, subtracting from these the products
[ d. [de. [de.r] rd
[de.3ede [df.3],.. [d.3],
[dd.3] [dd. dd] [dd.3]
we derive the results for [ee.4], [ef.4], [es.4], and [en.4]; under which
the corresponding logarithms are to be written.
If there are six unknown quantities to be determined, we must
further write in a horizontal line the values of [ff], [fs], and [fin],
CORRECTION OF THE ELEMENTS. 417
placing [ff] under [ef.4], and by means of five successive subtractions entirely analogous to what precedes, and as indicated by the
remaining equations for the auxiliaries, we obtain the values of [ff.5],
[fs.5], and [fn.5].
The values of [bs.1], cs.l], [cs.2], &c. serve to check the calculation of the successive auxiliary coefficients. Thus we must have
[bb.1] + [be.1] + [bd.1] + [be.1] + [bf.1] = bs.1]
[be.1] + [c.1] + [cd.1] + [ce.1] + [cf.1] [cs.1], &c.,
[cc.2] + [cd.2] + [ce.2] + [cf.2] [s.2],
[ed.2] + [dd.2] + [de.2] + [df.2] - [d.2], &c.
Hence it appears that when the numerical calculation is arranged as
above suggested, the auxiliary containing s must, in each line, be
equal to the sum of all the terms to the left of it in the same line
and of those terms containing the same distinguishing numeral found
in a vertical column over the last quantity at the left of this line.
There will yet remain only the auxiliaries which are derived from
[sn] and [enn] to be determined. These additional auxiliaries will
be found by means of the formulae
[a][] [bn.1]
[sn.i1]- [as, [Si. [s2] -[ 1]- [bs.1],
[ac] [b.1]
[sn.3] [sn.2] [c2] [sn.4] [n.3] - [d [ds.3], (133)
Ice. 2] [dd.3]
[sn.5] = [sn.4] [4] sG] [6] -n [n.5] [n.5]
[ee.4] [es.4], [ff.5] [f.5],
and the equations (81) and (83). The arrangement of the numerical
process should be similar to that already explained.
The values of [sn.l], [sn.2], &c. check the accuracy of the results
for [bn.1], [cn.l], [cn.2], [dn.3], &c. by means of the equations
[bn.1] + [e.1] + [dn.1] + [en.1] -F [n.1.]- [sn.1],
[cn.2] + [dn.2] + [en.2] + [fit.2] -[sn.2],
[d2.3] + [en.3] + [f0.3] = [sn.3], (134)
[en.4] + [fn.4] [sn.4],
[0f.5] [-s.5].
It appears further, that, in the case of six unknown quantities, since
[fs.5] [ff.5], we have [sn.6] 0.
Having thus determined the numerical values of the auxiliaries
required, we are prepared to form at once the equations (74), by means
of which the values of the unknown quantities will be determined
27
418 THEORETICAL ASTRONOMY.
by successive substitution, first finding t from the last of these equations, then substituting this result in the equation next to the last
and thus deriving the value of w, and so on until all the unknown
quantities have been determined. It will be observed that the logarithms of the coefficients of the unknown quantities in these equations will have been already found in the computation of the auxiliaries.
If we add together the several equations of (74), first clearing them
of fractions, we get
0 = [a] x + ([ab] + [bb.1]) y + ([a] + [bc.1] + [cc.2])
+ ([ad] + [bd.1] + [cd.2] + [dd.3]) u
+ ([ae] + [be.1] + [ce.2] + [de.3] + [ee.4])w (135)
+ ([af] + [bf.1] + [cf.2] + [df.3] + [efj4] + [ff.5])t
+ [an] + [b6.1] + [c.2] + [dn.3] + [en.4] + [fi.5];
and this equation must be satisfied by the values of x, y, z, &c. found
from (74).
152. EXAMPLE.-The arrangement of the calculation in the case
of any other number of unknown quantities is precisely similar; and
to illustrate the entire process let us take the following equations,
each of which is already multiplied by the square root of its weight:0.707x + 2.052y - 2.372z - 0.221u + 6".58 0,
0.471x + 1.347y - 1.715z - 0.085u- + 1.63 - 0,
0.260x + 0.770y - 0.356z + 0.483 - 4.40 - 0,
0.092 +- 0.343y + 0.235z + 0.469 - 10.21 - 0,
0.414x + 1.204y - 1.506z - 0.205tu - 3.99 0,
0.040x + 0.150y + 0.104z + 0.206u - 4.34 = 0.
First, we derive
[nn] - 204.313,
[an] =+ 4.815, [aa] + 0.971,
[bn] = - 12.961, [ab] + 2.821, [bb] = +8.208,
[cn] - 25.697, [ac] - 3.175, [bc] - 9.168, [cc] = + 11.028,
[cn] — 10.218, [ad] - 0.104, [bd] - 0.251, [cd] -+ 0.938, [dd] + 0.594,
[si] - 18.139, [as] = 0.513, [bs] + 1.610, [cs] - 0.377, [ds] + 1.177.
The values of [sn], [as], [bs], [cs], and [ds], found by taking the
sums of the normal coefficients, agree exactly with the values computed directly, thus proving the calculation of these coefficients.
The normal equations are, therefore,
NUMERICAL EXAMPLE. 419
0.971x + 2.821y - 3.175z - 0.104- + 4.815 0,
2.821x + 8.208y - 9.168z - 0.251u + 12.961 0,
- 3.175x - 9.168y + 11.028z + 0.938u - 25.697 0,
-— 0.104x - 0.251y + 0.938z + 0.594u - 10.218 - 0.
It will be observed that the coefficients in these equations are nu-.merically greater than in the equations of condition; andc this will
generally be the case. Hence, if we use logarithms of five decimals
in forming the normal equations, it will be expedient to use tables
of six or seven decimals in the solution of these equations.
Arranging the process of elimination in the most convenient form,
the successive results are as follows:bb.1] = 0.0123, [bc.1] - 0.0562, [bd.1] = + 0.0511, [bs.1] = + 0.1196, [b?.l] =- 1.0278,
[cc.l] = + 0.6463, [cd.1] = + 0.5979, [cs.1] = + 1.3004, [cn.l] =- 9.9528,
[cc.2] = + 0.3895, [cd.2] = — 0.3644, [cs.2] = + 0.7539, [cn.2] = - 5.2567,
[dd.1] = + 0.5829, [ds.l] = + 1.2319, [dn.l] =- 9.7023,
[dd.2] = + 0.3706, [ds.2] = + 0.7350, [dn.2] =- 5.4323,
[dd.3] = + 0.0297, [ds.3] = + 0.0297 [dn.3] = — 0.5143,
[nn.l] = 180.436, [sn.1] = - 20.6828,
[nn.2]= 94.552, [sn.2] - 10.6889,
[nn.3]= 23.608, [sn.3]= - 0.5143,
[nn.4] = 14.698, [sn.4] = 0.
The several checks agree completely, and only the value of [mn.4]
remains to be proved. The equations (74) therefore give
x +- 2.9052y - 3.2698z- 0.1071u + 4.9588 = 0,
y + 4.5691z + 4.1545u - 83.5610 = 0,
z + 0.9356u - 13.4960 - 0,
u - 17.3165 0,
and from these we get
+ - 17".316, z_ - 2".705, y =+ 23".977, x = - 81".608.
Then the equation (135) becomes
0 + 0.9710x + 2.8333y - 2.7293z + 0.3412u - 1.9838,
which is satisfied by the preceding values of the unknown quantities.
If we substitute these values of x, y, z, and mt in the equations of
condition already reduced to the same weight by multiplication by
the square roots of their weights, we obtain the residuals
+ 0".67, - 1".34, + 2".17, - 2".01, - 0".40, - 0".72,
The sum of the squares of these gives
[vv] =-[nn.4] -~ 11.672,
and the difference between this result and the value 14.698 already
420 THEORETICAL ASTRONOMY.
found is due to the decimals neglected in the computation of the
numerical values of the several auxiliaries. The sum of all the
equations of condition gives generally
[a] x +- [b] y +- [l] z + [d] +... e + [$] = [v], (136)
which may be used to check the substitution of the numerical values
in the determination of v, v', &c. Thus, we have, for the values
here given,
1.984x + 5.866y - 5.610z + 0.647t - 6.75 - [v] 1."63.
It remains yet to determine the relative weights of the resulting
values of the unknown quantities. For this purpose we may apply
any of the various methods already given. The weights of u and z
may be found directly from the auxiliaries whose values have been
computed. Thus, we have
p = [dd.3] - 0.0297, pZ [dd.3] [c.2] -0.0312.
-[dd.2]
If we now completely reverse the order of elimination from the
normal equations, and determine x first, we obtain the values
[bb.2] ~ + 0.0425, [aa.2] + 0.0033,
[aa.3] - + 0.00056, [nn.4] 14.665,
and also
x ~ — 82."750, y -+ 24."365, z- 2."699, u + 17."272.
The small differences between these results and those obtained by the
first elimination arise from the decimals neglected. This second
elimination furnishes at once the weights of x and y, namely,
= [aa.3] 0.00056, p [aa.3] [bb.2] - 0.0072.
ac [ca.2]
We may also compute the weights by means of the equations (96).
Thus, to find the'weight of y, we have
[dd.2], - dd.1 - [.cd. [e] +- 0.02977,
and hence
[dd.3] [cc.2]
p [da. a ] [cI.2] [bb.1] - 0.0074.
[dd.21]b [cc.l1]
The equations (103) and (108) are convenient for the determination
of the values and weights of the unknown quantities separately.
CORRECTION OF THE ELEMENTS. 421
Thus, by means of the values of the auxiliaries obtained in the first
elimination, we find from the equations (100), (101), and (102),
A' - 2.9052, A" += 16.5442, A"'" - 3.3012,
B" -- 4.5691, B"' + 0.1202, C"' - 0.9356,
and then the equations (103) and (108) give
x - 81".609, y += 23".977, z - - 2".705, u -+ 17".316,
pZ 0.00057, py - 0.0074, pZ - 0.0312, p.- 0.0297,
agreeing with the results obtained by means of the other methods.
The weights are so small that it may be inferred at once that the
values of x, y, z, and u are very uncertain, although they are those
which best satisfy the given equations. It will be observed that if
we multiply the first normal equation by 2.9, the resulting equation
will differ very little from the second normal equation, and hence we
have nearly the case presented in which the number of independent
relations is one less than the number of unknown quantities.
The uncertainty of the solution will be further indicated by determining the probable errors of the results, although on account of the
small number of equations the probable or mean errors obtained may
be little more than rude approximations. Thus, adopting the value
of [vv] obtained by direct substitution, we have
[mm.4] F1.672
m -= -\A-/ ~ - ~ = = 2.416,
and hence
r -- 1".629,
which is the probable error of the absolute term of an equation of
condition whose weight is unity. Then the equations
r r?'
r= — r, r,
give
%r ~ 68".25, r = ~ 18".94, r = - 9".22, r =~ 9".45.
It thus appears that the probable error of z exceeds the value obtained
for the quantity itself, and that although the sum of the squares of
the residuals is reduced from 204.31 to 11.67, the results are still
quite uncertain.
153. The certainty of the solution will be greatest when the coefficients in the equations of condition and also in the normal equations
422 THEORETICAL ASTRONOMY.
differ very considerably both in magnitude and in sign. In the correction of the elements of the orbit of a planet when the observations extend only over a short interval of time, the coefficients will
generally change value so slowly that the equations for the direct
determination of the corrections to be applied to the elements will
not afford a satisfactory solution. In such cases it will be expedient
to form the equations for the determination of a less number of
quantities from which the corrected elements may be subsequently
derived. Thus we may determine the corrections to be applied to
two assumed geocentric distances or to any other quantities which
afford the required convenience in the solution of the problem,
various formulre for which have been given in the preceding chapter.
The quantities selected for correction should be known functions of
the elements, and such that the equations to be solved, in order to
combine all the observed places, shall not be subject to any uncertainty in the solution. But when the observations extend over a long
period, the most complete determination of the corrections to be
applied to the provisional elements will be obtained by forming the
equations for these variations directly, and combining them as already
explained. A complete proof of the accuracy of the entire calculation will be obtained by computing the normal places directly from
the elements as finally corrected, and comparing the residuals thus
derived with those given by the substitution of the adopted values
of the unknown quantities in the original equations of condition.
If the elements to be corrected differ so much from the true values
that the squares and products of the corrections are of sensible magnitude, so that the assumption of a linear form for the equations does
not afford the required accuracy, it will be necessary to solve the
equations first provisionally, and, having applied the resulting corrections to the elements, we compute the places of the body directly
from the corrected elements, and the differences between these and
the observed places furnish new values of n, n', n", &c., to be used
in a repetition of the solution. The corrections which result from
the second solution will be small, and, being applied to the elements
as corrected by the first solution, will furnish satisfactory results. In
this new solution it will not in general be necessary to recompute the
coefficients of the unknown quantities in the equations of condition,
since the variations of the elements will not be large enough to affect
sensibly the values of their differential coefficients with respect to
the observed spherical co-ordinates. Cases may occur, however, in
which it may become necessary to recompute the coefficients of one
CORRECTION OF THE ELEMENTS. 423
or more of the unknown quantities, but only when these coefficients
are very considerably changed by a small variation in the adopted
values of the elements employed in the calculation. In such cases
the residuals obtained by substitution in the equations of condition
will not agree with those obtained by direct calculation unless the
corrections applied to the corresponding elements are very small. It
may also be remarked that often, and especially in a repetition of the
solution so as to include terms of the second order, it will be sufficiently accurate to relax a little the rigorous requirements of a complete solution, and use, instead of the actual coefficients, equivalent
numbers which are more convenient in the numerical operations required. Although the greatest confidence should be placed in the
accuracy of the results obtained as far as possible in strict accordance
with the requirements of the theory, yet the uncertainty of the determination of the relative weights in the combination of a series of
observations, as well as the effect of uneliminated constant errors,
may at least warrant a little latitude in the numerical application,
provided that the weights of the results are not thereby much affected.
A constant error may in fact be regarded as an unknown quantity to
be determined, and since the effect of the omission of one of the
unknown quantities is to diminish the probable errors of the resulting
values of the others, it is evident that, on account of the existence of
constant errors not determined, the values of the variables obtained
by the method of least squares from different corresponding series of
observations may differ beyond the limits which the probable errors
of the different determinations have assigned. Further, it should be
observed that, on account of the unavoidable uncertainty in the estimation of the weights of the observations in the preliminary combination, the probable error of an observed place whose weight is
unity as determined by the final residuals given by the equations of
condition, may not agree exactly with that indicated by the prior
discussion of the observations.
154. In the case of very eccentric orbits in which the corrections
to be applied to certain elements are not indicated with certainty by
the observations, it will often become necessary to make that whose
weight is very small the last in the elimination, and determine the
other corrections as functions of this one; and whenever the coefficients of two of the unknown quantities are nearly equal or have
nearly the same ratio to each other in all the different equations of
condition, this method is indispensable unless the difficulty is reme
424 THEORETICAL ASTRONOMY.
died by other means, such as the introduction of different elements or
different combinations of the same elements. The equations (113)
furnish the values of the unknown quantities when we neglect that
which is to be determined independently; and then the equations
(114) give the required expressions for the complete values of these
quantities. Thus, when a comet has been observed only during a
brief period, the ellipticity of the orbit, however, being plainly indicated by the observations, the determination of the correction to be
applied to the mean daily motion as given by the provisional elements, in connection with the corrections of the other elements, will
necessarily be quite uncertain, and this uncertainty may very greatly
affect all the results. Hence the elimination will be so arranged that
ad shall be the last, and the other corrections will be determined as
functions of this quantity. The substitution of the results thus
derived in the equations of condition will give for each residual an
expression of the following form:A — o + /As..
Therefore we shall have
[vv] [v0v] + 2 [vr] aL + [rr] A/j2, (137)
which may be applied more conveniently in the equivalent form
[vv] [vo] [v ] [ + r] ( + [r ) (138)
The most probable value of Al/ will be that which renders [vv] a
minimum, or
A-i- [krr] (139)
and the corresponding value of the sum of the squares of the
residuals is
[vv] - [Vov] - [v~r]. (140)
[rr]
The correction given by equation (139) having been applied to t,
the result may be regarded as the most probable value of that element, and the corresponding values of the corrections of the other
elements as determined by the equations (114) having been also duly
applied, we obtain the most probable system of elements. These,
however, may still be expressed in the form
2 + Aoa i+, i + Boad, r + CGoa, &c.
CORRECTION OF THE ELEMENTS. 425
the coefficients Ao, B0, Co, &c. being those given by the equations
(114), and thus the elements may be derived which correspond to any
assumed value of i differing from its most probable value. The
unknown quantity A/l will also be retained in the values of the
residuals. Hence, if we assign small increments to /, it may easily
be seen how much this element may differ from its most probable
value without giving results for the residuals which are incompatible
with the evidence furnished by the observations.
If the dimensions of the orbit are expressed by means of the elements q and e, it may occur that the latter will not be determined
with certainty by the observations, and hence it should be treated as
suggested in the case of p; and we proceed in a similar manner when
the correction to be applied to a given value of the semi-transverse
axis a is one of the unknown quantities to be determined.
42(6 THEORETICAL ASTRONOMY.
CHAPTER VIII.
INVESTIGATION OF VARIOUS FORMULAE FOR THE DETERMINATION OF THE SPECIAL
PERTURBATIONS OF A HEAVENLY BODY.
155. WE have thus far considered the circumstances of the undisturbecl motion of the heavenly bodies in their orbits; but a complete
determination of the elements of the orbit of any body revolving
around the sun, requires that we should determine the alterations in
its motion due to the action of the other bodies of the system. For
this purpose, we shall resume the general equations (18),, namely,
d', x d2
2 +k2(1~ +n) — 2(1 + m) d+'
+ k2(+ m) 72 k(1 + 2) d(1)
dt2 r dy'
d2Iz V dO.
+ k2(1 +m) -ik2 (1 +n) dz
which determine the motion of a heavenly body relative to the sun
when subject to the action of the other bodies of the system. We
have, further,
o-Z I+yy'+ \ n'' 1 xx\+? ~ ~ ~+ zz'' ) &
which is called the iperturbinzg function, of which the partial differential coefficients, with respect to the co-ordinates, are
dQ m! I x' x x'\ m" Z<' x jx") \
/ +ep3 f3/ + + P/s-rt33+ &Cr,,
dx 1 1\ m \r'
d l+ m 3 p3y ) 1 +,I in 1-3 y )3 + &c., (2)
dx I p3 14I' Im\ p r"'
and in which iz', mn, &c. denote the ratios of the masses of the
several disturbing planets to the mass of the sun, and mi the ratio of
the mass of the disturbed planet to that of the sun. These partial
differential coefficients, when multiplied by k;(1 + qn), express the
PERTURBATTONS. 427
sum of the components of the disturbing force resolved in directions
parallel to the three rectangular axes respectively.
When we neglect the consideration of the perturbations, the general
equations of motion become
d2X0 + k (I + m)' 0,
dt2 r'03
d + I (1 +. )) YO 0, (3)
ci2 0
dt2 + k(1 +.) 0,
the complete integration of which furnishes as arbitrary constants of
integration the six elements which determine the orbitual motion of a
heavenly body. But if we regard these elements as representing the
actual orbit of the body for a given instant of time t, and conceive
of the effect of the disturbing forces due to the action of the other
bodies of the system, it is evident that, on account of the change
arising from the force thus introduced, the body at another instant
different from the first will be moving in an orbit for which the
elements are in some degree different from those which satisfy the
original equations. Although the action of the disturbing force is
continuous, we may yet regard the elements as unchanged during the
element of time dt, and as varying only after each interval dc. Let
us now designate by to the epoch to which the elements of the orbit
belong, and let these elements be designated by J, o, T 0,, i0, e0, and
aO; then will the equations (3) be exactly satisfied by means of the
expressions for the co-ordinates in terms of these rigorously-constant
elements. These elements will express the motion of the body subject to the action of the disturbing forces only during the infinitesima.l
interval dt, and at the time t + cdt it will commence to describe a
new orbit of which the elements will differ from these constant elements by increments which are called the perturbations.
According to the principle of the variation of parameters, or of
the constants of integration, the differential equations (1) will be
satisfied by integrals of the same form as those obtained when the
second members are put equal to zero, provided only that the arbitrary
constants of the latter integration are no longer regarded as pure
constants but as subject to variation. Consequently, if we denote the
variable elements by 31, w,, i, e, and a, they will be connected
with the constant elements, or those which determine the orbit at the
instant to, by the equations
428 THEORETICAL ASTRONOMY.'- 3 ( cdm -, r dQ dt,r 7
M o +lo f t',A -, + dt dt, a -- 20+q dt / 4
(4)
dci de da
i -= io J dt, e eo+ dt, a =a +J- dt
c dM1 dr
in which -, -, &c. denote the differential coefficients of the eledt dt'
ments depending on the disturbing forces. When these differential
coefficients are known, we may determine, by simple quadrature, the
perturbations JM, J~r, &c. to be added to the constant elements in
order to obtain those corresponding to any instant for which the
place of the body is required. These differential coefficients, however,
are functions of the partial differential coefficients of Q with respect
to the elements, and before the integration can be performed it
becomes necessary to find the expressions for these partial differential
coefficients. For this purpose we expand the function Q into a converging series and then differentiate each term of this series relatively
to the elements. This function is usually developed into a converging series arranged in reference to the ascending powers of the eccentricities and inclinations, and so as to include an indefinite number
of revolutions; and the final integration will then give what are
called the absolute or general perturbations. When the eccentricities
and inclinations are very great, as in the case of the comets, this
development and analytical integration, or quadrature, becomes no
longer possible, and even when it is possible it may, on account of
the magnitude of the eccentricity or inclination, become so difficult
that we are obliged to determine, instead of the absolute perturbations,
what are called the special perturbations, by methods of approximation known as mechanical quadrcatres, according to which we determine the variations of the elements from one epoch to to another
epoch t. This method is applicable to any case, and may be advantageously employed even when the determination of the absolute
perturbations is possible, and especially when a series of observations
extending through a period of many years is available and it is
desired to determine, for any instant to, a system of elements, usually
called osculating elements, on which the complete theory of the motion
may be based.
Instead of computing the variations of the elements of the orbit
directly, we may find the perturbations of any known functions of,
these elements; and the most direct and simple method is to determine the variations, due to the action of the disturbing forces, of
any system of three co-ordinates by means of which the position of
PERTURBATIONS. 429
the body or the elements themselves may be found. We shall, therefore, derive various formulae for this purpose before investigating the
formulae for the direct variation of the elements.
156. Let xz, y, zo be the rectangular co-ordinates of the body at
the time t computed by means of the osculating elements M0,, To g0,
&c., corresponding to the epoch to. Let x, y, z be the actual co-ordinates of the disturbed body at the time t; and we shall have
x - X o- x, y -= + Sy, z =- +z,
8x, oy, and oz being the perturbations of the rectangular co-ordinates
from the epoch to to the time t. If we substitute these values of x,
y, and z in the equations (1), and then subtract from each the corresponding one of equations (3), we get
d +xtk( +) ( -x~ )- k(l + ) d m)
dlt2 ~IT r dx'
dt2 + k ( + r3) ( ~o + k J ) 2 ( + m) d
d2t2 + 9 (1 + +) ( 5 )_ + ) d'(
dt - ( dyz
d,~t + r(l+I) 9"3 9os d+
Let us now put r - r' + r'; then to terms of the order Or2, which is
equivalent to considering only the first power of the disturbing force,
we have
-3 x 3 1 x-3x3,O
r' r \ ro
/YO + sy?/ 1( y -3Y~r
3 r 3 7'o
O3 o3 1'' 3 r o
z0 + az Z( _ \
r -, 3 Lo
and hence
d'2 2 d + (1 + )d+( +-)(
dt2 k I r+ r -
dt2 -k (1 + m)d ).
S~r ~= S~x + + _ Sz. (7)
Pt20 70 3 70
r._ x~0 + 7+ 0
430 THEORETICAL ASTRONOMY.
The integration of the equations (6) will give the perturbations 8x,
dy, and (z to be applied to the rectangular co-ordinates x0, Yo, zo computed by means of the osculating elements, in order to find the actual
co-ordinates of the body for the date to which the integration belongs.
But since the second members contain the quantities 8x, 8y, 8z which
are sought, the integration must be effected indirectly by successive
approximations; and from the manner in which these are involved
in the second members of the equations, it will appear that this integration is possible.
If we consider only a single disturbing planet, according to the
equations (2), we shall have
k2 (1 + m) d'P (8)
k2 (1 + mn) d~- ( - ) 2
^^ ^^-YP (8)
and these forces we will designate by X, Y, and Z respectively; then,
if in these expressions we neglect the terms of the order of the
square of the disturbing force, writing w, yo, zo in place of x, y, z,
the equations (6) become
d28x l' (1 -- m) +
dt2 XIo + 3 3~r - x,
d Jy 3 7-(1 + m)(3y 8 a (9)
dt2 -Y0+3r -, ( 9)
d2eZ k (1 + M1,) (3 2o 8r-
dt=2 - o 3-r- ro
which are the equations for computing the perturbations of the rectangular co-ordinates with reference only to the first power of the
masses or disturbing forces. We have, further,
p (' - x) + ( - y) + (z' - Z', (10)
in which, when terms of the second order are neglected, we use the
values xo, yo, z for x, y, and z respectively.
157. From the values of 8x, 8y, and 8z computed with regard to
the first power of the masses we may, by a repetition of part of the
calculation, take into account the squares and products and even the
higher powers of the disturbing forces. The equations (5) may be
written thus:
VARIATION OF CO-ORDINATES. 431
dt2 X+ 3 / )(( 1-3
d2y- - y k2(1 +'n)1 i \ ^\ y)
dt2 r03 ) r
dt2 -Z + P(i + (( n13) _Z S )
dt r703 r/
in which nothing is neglected. In the application of these formulae,
as soon as ax, 8y, and 6z have been found for a few successive intervals, we may readily derive approximate values of these quantities
for the date next following, and with these find
x = Xo + ax, y - yo + ay, z = z + z,
and hence the complete values of the forces X, Y, and Z, by means
of the equations (8). To find an expression for the factor
r3
r-3
which will be convenient in the numerical calculation, we have
r2 = (o X)2 + (yo + 8y)2 + + + ( + )2
-= _2 x + 2 ys8x + 2yo + 2z0&4 + dX2 + ay2 + 8Z2,
and therefore
r2:1 2 (xo+ 2 x) 8x + (yo + ay) ay + (zo + Gz) 8z
ro r2'0
Let us now put
X0 o + ax
q=i+ -- ~ + - 4- z (12)
702 r02'02
and
fq —1 - 1 - (1+2)-;
then we shall have
/_o(1 5 5.7 5.7.9,
- 3 -q+ 2 4q3 + &c.) (13)
and the values off may be tabulated with the argument q. The
equations (11) therefore become
-- z+k-( in (14)
d2 kT (f + az)
t2 +(fq -- da),
d2y k (1 + ) (fqy_ y)
d z k ( ) (fz- z).
432 THEORETICAL ASTRONOMY.
The coefficients of 8x, 8y, and 3z in equation (12) may be found at
once, with sufficient accuracy, by means of the approximate values
of these quantities; and having found the value of f corresponding
to the resulting value of q, the numerical values of dt,'xdt~ and
d 2Jz d'i d dt2 a
dt~ which include the squares and products of the masses, will be
obtained. The integration of these will give more exact values of
8x,'y, and Jz, and then, recomputing q and the other quantities which
require correction, a still closer approximation to the exact values of
the perturbations will result.
Table XVII. gives the values of logf for positive or negative
values of q at intervals of 0.0001 from q= 0 to q = 0.03. Unless
the perturbations are very large, q will be found within the limits of
this table; and in those cases in which it exceeds the limits of the
table, the value of
m3
fq -1may be computed directly, using the value of r in terms of r0 and
ox, eay, 8z.
In the application of the preceding formule, the positions of the
disturbed and disturbing bodies may be referred to any system of
rectangular co-ordinates. It will be advisable, however, to adopt
either the plane of the equator or that of the ecliptic as the fundamental plane, the positive axis of x being directed to the vernal
equinox. By choosing the plane of the elliptic orbit at the time to
as the plane of xy, the co-ordinate z will be of the order of the perturbations, and the calculation of this part of the action of the disturbing force will be very much abbreviated; but unless the inclination is very large there will be no actual advantage in this selection,
since the computation of the values of the components of the disturbing forces will require more labor than when either the equator
or the ecliptic is taken as the fundamental plane. The perturbations
computed for one fundamental plane may be converted into those
referred to another plane or to a different position of the axes in the
same plane by means of the formulae which give the transformation
of the co-ordinates directly.
158. We shall now investigate the formulae for the integration of
the linear differential equations of the second order which express the
variation of the co-ordinates, and generally the formulae for finding
the integrals of expressions of the form ff(x) dx and J jf(x) dx2
MECHANICAL QUADRATURE. 433
when the values of f(x) are computed for successive values of x increasing in arithmetical progression. First, therefore, we shall find
the integral of f(x) dx within given limits.
Within the limits for which x is continuous, we have
f(X) -=a + + rx + 8x + x +....; (15)
and if we consider only three terms of this series, the resulting equation
f(x) - a + fx + rx2
is that of the common parabola of which the abscissa is x and the
ordinate f(), and the integral of f(x) dx is the area included by the
abscissa, two ordinates, and the included arc of this curve. Generally, therefore, we may consider the more complete expression for
f(x) as the equation of a parabolic curve whose degree is one less
than the number of terms taken. Hence, if we take n terms of the
series as the value off(x), we shall derive the equation for a parabola
whose degree is n - 1, and which has n points in common with the
curve represented by the exact value of f(x).
If we, multiply equation (15) by dx and integrate between the
limits 0 and x', we get
X'
f(x)) dx = i +,"12 + rX3 + I Jt4 +. (16)
o
If now the values of f(x) for different values of x from 0 to x' are
known, each of these, by means of equation (15), will furnish an
equation for the determination of a, 9, r, &c.; and the number of
terms which may be taken will be equal to the number of different
known values of f(x). As soon as a, /, r, &c. have thus been found,
the equation (16) will give the integral required.
If the values of f(x) are computed for values of x at equal intervals and we integrate between the limits x 0, and x —=n, ax
being the constant interval between the successive values ofx) and
n the number of intervals from the beginning of the integration, we
obtain
92Ax
ff() dx an-x + _n2zx2 + l nrx3 + &c.
Let us now suppose a quadratic parabola to pass through the points
of the curve represented by f(x), corresponding to x = 0, x = Az
28
434 THEORETICAL ASTRONOMY.
and x - 2Ax; then will the area included by the arc of this parabola,
the extreme ordinates, and the axis of abscissas be
2Ax
ff(x) dx Ax (2a + 2A^x + — rAx2).
0
The equation of the curve gives, if we designate thle ordinates of the
three successive points by y, y,, and Y2,
Y~ f t ^ ~- = = (YY A -2y, ~t yO),
2A1 2AX2
and hence we derive
2Ax
ff(x) dx - 1 Ax (yo + 4Y1 + Y2)'
0
In a similar manner, the area included by the ordinates Y2 and y~,corresponding to x- = 2^x and x = 4ax,-the axis of abscissas, and
the parabola passing through the three points corresponding to Y2, Y3,
and y4, is found to be
4Ax
ff(x) dx = ^x (Y2 + 4Y3 + Y4);
2Ax
and hence we have, finally,
fnAx
ff(x) dx - 1 x (n -2 + 4yn-i + y?).
(n- 2) Ax
The sum of all these gives
n^x
ff(x) dx
0 (17)
A ((yo + qy) + 4 (y, + Yy + Y5 +... y- _ ) + 2 (y + Y4 +... Y_ -2)),
by means of which the approximate value of the integral within the
given limits may be found.
If we consider the curve which passes through four points corresponding to yo, y1 Y2, and y3, we have
y =f(x) -+ix + a X2 + rx 3 +
for the equation of the curve, and hence, giving to x the values 0,
AX, 2Ax, and 3Ax, successively, we easily find
MECHANICAL QUADRATURE. 435
Y = 0
1
/ - (2y3 - 9y2 + 18y1 - ly),
Y- 2^ (Y 3 + 4y - 5y1 + 2yo),
=, (Y3 - 3Y2 + 31- Yo)
Therefore we shall have
3Ax
ff(x) dZ- = ax (yo + 3y1 + 3Y2 + y). (18)
0o
In like manner, by taking successively an additional term of the
series, we may derive
4Az
f(x) dx- 2x (7yo + 32y1 + 12y2 + 32y3 + 7y,),
5A (19)
J'(x) d -288 (19yo + 75y1 +0y + + 50y + 75y + 19y).
0
This process may be continued so as to include the extreme values of
x for which f(x) is known; but in the calculation of perturbations it
will be more convenient to use the finite differences of the function
instead of the function itself directly. We may remark, further,
that the intervals of quadrature when the function itself is used,
may be so determined that the degree of approximation will be much
greater than when these intervals are uniform.
159. Let us put Ax- o, and let the value of x for which n — 0
be designated by a; then will the general value be
f(x) =f(a + nw),
o being the constant interval at which the values of f(x) are given.
Hence we shall have
dx - owdn,
f (x) dx -= f (a + no) dn.
If we expand the function f(a -t- nw), we have
f(a + nw) — f(a)) + n (a) + fd ) +'+ &c., (20)
da 1.2 da 2 1.2.3 da3
436 THEORETICAL ASTRONOMY.
and hence
df (a) 2ddf(a)
Jf(a + nw) dn - C + nf(a) + I 2, + I Wd2)
da da 2
(21)
C being the constant of integration. The equations (54)6 give'(22)
da - () -'(a) + f (a) - -f () +
a d2f(a)) (a)
dc f (a) 1f V (a) + ovi (a)..,
d ci~f(c4)
ca =fva (af a
de? i- -/ f(a) - fvIII(a) +...,
in which the functional symbols in the second members denote the
different orders of finite differences of the function. Hence we obtain
f f(a )+,,) dn- C + f(a)
+ b, (f' (a) - U1 f ( + _1- fv(), V (Vii(a) +..)
+'3(f" (a) - fiv~ (C) + of V ()- _-_ fh (a) F...)
+ 4' (a - I(a () +- 7Of" () -....)
+l5(fiV(a) _ iV (a) + if_ Vi(a) ) (23)
+ 14on (f (a) - (f" (a) +...)
+ - _,a (fvi (a) +- fv"' (a) +..)
+ 41-12 )8fvii ( + 4
If we take the integral between the limits — n and + ~', the terms
containing the even powers of n disappear. Further, since the values
of the function are supposed to be known for a series of values of r
at intervals of a unit, it will evidently be convenient to determine
the integral between the required limits by means of the sum of a
series of integrals whose limits are successively increased by a unit,
such that the difference between the superior and the inferior limit
of each integral shall be a unit. Hence we take the first integral
between the limits -- and +-2, and the equation (23) gives, after
reduction,
MECHANICAL QUADRATURE. 437
f(a + nu) dn =f(a) + -f" (a) - -l 7f (a) + g 67-f i (a,)
(24)
4 6 4 8 5 of iii(a) + &c.
It is evident that by writing, in succession, a + o, a + 2w,....
a + io in place of a, we simply add I to each limit successively, so
that we have
f(a + tnw) dn =fjf((a + iv) + (n - i) ) d(n -i)
417 6 i
-— f(a+i)+i2f. (a+i)a)- f ivf (a+iJ)+L 967/fvi (a+iw) —&C.
But since
f(a +- nw) dn = f (a + nw) dn +J (a + n) dn.....+f (a+n) dn,
- - - i —
if we give to i successively the values 0, 1, 2, 3, &c. in the preceding
equation, and add the results, we get
i +- n=i n=i
J a. + ^nw) dn = Vf(ac + mn) + 4 f (a + nU)
-~ ~==o n=o ^o-2 n= O n - (25)
z -i n=i
17 367 Vi
- 1i fiv(a + }) + U7-80aS( + (a ) -+ &C.
n==o0 n=o
Let us now consider the functions f(a), f (a + no), &c. as being
themselves the finite differences of other functions symbolized by'f,
the first of which is entirely arbitrary, so that we may put, in accordance with the adopted notation,
f(a)'f(a + w) - f(a- ),
f(a +,o) -'f(a + o) - f(a + 1)
f(a + n) - (a+ - (n + ) 0) - f(a + (n - )).
Therefore we shall have
n==i
f (a + n,) -'f(a + (i + -) ) - Y(a' — -),
n=O
and also
n=i
if (a + nw) =f (a + (i + )) -f' (a - w),
n==i
n =
if: (a + nw)=(a'" (a + + )-" - ), &c..6y=0f 2
n= = 6
438 THEORETICAL ASTRONOMY.
Further, since the quantity'(ac- wo) is entirely arbitrary, we may
assign to it a value such that the sum of all the terms of the equation
which have the argument a - co shall be zero, namely,'f(a- ) -- 2 4(a~ —)o + -- ~ 6- 8f (a- )+&c.
(26)
Substituting these values in (25), it reduces to
a + (i + ) i +
f (x) dx - f f(a + nw) dn
a-t -2 (27)
— (r,'ff(a + (i + 3)(t)) + J-3-f(a + (i + ))..-.
5 0f"' (a+(i_+ 1 ) - )+_ vof' (a+ (i+) )-&c. }
In the calculation of the perturbations of a heavenly body, the
dates for which the values of the function are computed may be so
arranged that for n - -, corresponding to the inferior limit, the
integral shall be equal to zero, the epoch of f(a- wo) being that of
the osculating elements. It will be observed that the equation (26)
expresses this condition, the constant of integration being included
in'f(ca - 1). If, instead of being equal to zero, the integral has a
given value when n -, it is evidently only necessary to add this
value to'f(a - -1o) as given by (26).
160. The interval o and the arguments of the function may always
be so taken that the equation (27) will furnish the required integral,
either directly or by interpolation; but it will often be convenient to
integrate for other limits directly, thus avoiding a subsequent interpolation. The derivation of the required formulae of integration
may be effected in a manner entirely analogous to that already indicated. Thus, let it be required to find the expression for the integral
taken between the limits — ~ and i.
The general formula (23) gives
aff + nw) dab If f(a) + If' (a) + a-jsf" (a) - 3S-f"' (a) - - 5 "fIf (a)
0o
+ 61 3f v () + f 3 vi(a) -&c.;
and since, according to the notation adopted,
f' (a) - 2 (f' (a - o) +f' ( + v))
-f' ( + 1) - If" (a), (28)
f"' (a) -f"' (a + o) - i (a)
f v (af) f v (a+ at) f vi (a), &c.,
MECHANICAL QUADRATURE. 439
this becomes
ff+(nw) dn=f (a)+if' (a+~w)- I4" (a)-+4.f"'. (a+2w) (29)
0
_ 11 ifiV (a)+(a + ) 19s fi(a) - &c.
Therefore we obtain
it'~~ (30)
11 v
7-4f... (a+(i+ 11 (+o)+ -If' (a+ito+) );4 _UfV(a7+(i) t
Of- 960_ J' (a [ tc) - c.
Now we have
i i +- i +1
f (a +- nw) dn ff(a - nw) dn (a + no) dn;
and if we substitute the values already found for the terms in the
second member, and also
fi(a + ito) f' (a) + (i + 1)w) -f" (a + (i - I)),
f'v(a+ - i) = f ('a + (i + )o) - f(a + (i- )), (31
f Vi (a + ito) _f v ( a + (3 + 1 ) cw) - f V (a + (3- ) ") &G.
we get
f( (x) dx= )ff (ca +?w1) dn
a -i) -_ i
- 5'f'( + (i + ) ) + (-Y (i 2) ) + (3 -2),f) -+ ( + )
- a ( +(~ t-o))+T 440f +(a+ ) t)+44f (a+3+ +-) o)
2- orV (a + + 2) t)w - T191/ ofv ( + ( -) ) + &c.
which is the required integral between the limits - 2 and i.
161. The methods of integration thus far considered apply to the
cases in which but a single integration is required, and when applied
to the integration of the differential equations for the variations of
the co-ordinates on account of the action of disturbing bodies, they
d~x d~y d~z
will only give the values of dt-, d- and dt- and another integration
becomes necessary in order to obtain the values of Ox, Sy, and Oz.
We will therefore proceed to derive formule for the determination
of the double integral directly.
440 THEORETICAL ASTRONOMY.
For the double integralfff(x) dx we have, since dx" = wdw2,
ffof( d2 =d ffr (a + 2W) dn2'
The value of the function designated by f(a) being so taken that
when n --,
ff(a + nw) dn 0,
the equation (23) gives
o
C=- ff(a + nw) dn.
Therefore, the general equation is
o
Sf(a + nw) dn =ff(a ~ nw)dn -+ nf(a)
-+ Pn + In- + r4n + l -n + &c.
the values of a, t,,... being given by the equations (22). Multiplying this by dn, and integrating, we get
o
fff(a + nw) dn2 = C' + nf (a + no) dn + In2f(a)
+, n- + pn4 + T -n + &c.,
C' being the new constant of integration. If we take the integral
between the limits - 2 and +, we find
+1 o
ff(a + no) dn2 =ff(a + n) d + -a + + go + &c.
From the equation (32) we get, for i 0,
o
f(a + n,) din - (a) -' (a) + f" (a) - f (a) + &c. (33)
Substituting this value, and also the values of a, re, &c.,-which
are given by the second niembers of the equations (22),-in the preceding equation, and reducing, we get
f(a+~to) dn2='f (a) — f' (a) + — f"' (a) - 8 of (a)+&c (34)
MECHANICAL QUADRATURE. 441
Hence
i-+
jf(a + nw) dn2'f(a + iw) ~- f' (a + 0i)
f"' (f + a W) - 367 (a + ) &C.
and
i + n-=z n =i
^fffa + n) dn2'f(a + n) - 4^ f' (a + mn)
_- n= 0 nn=O (35)
n = i n=i -
+ T -ZU s"' (a + n) - 36 6 f v (a+ 0) +.c.
n=o0 = =0
We may evidently consider'f(a -- ),'f(ca + -w), &c. as the differences of other functions, the first of which is arbitrary, so that we
have'f(a) -'f (a + -a,) + -- -(a - c) -- -"(a + w) -if"(a -'f(a + W,) ='f(a + 3w) + ~ff(a + -e) "- 7(a + 20) - f' (a)
('a + -no) (a + (n + + f( + ( ) ) r'f(a + (n+1+) w)
-'(a + (n - 1) w).
Therefore
n = i'f (a+no) — i (a+(it+1)))+V' (a+i)- "f (a) "f (Ca-c),
n==O
n =- i
n O
n=i
Vf
_.v (a+nto) —- vfv (a+(i+l)()o)+- iv(a+i )- if (a)-_fv(a-,),&c
n=O
(a) + f(a - ) -2f"(a -f"' (a - ), &c.,
and that, since "f(a )lu is arbitrary, we may put
"f - ) 4f(a) - a7 (2f" (a) + f" (a - ))
+ 963760 (3f'i (a) + 2f'v (a - c)) - &c., (
442 THEORETICAL ASTRONOMY.
the integral becomes
a +- (i ) u ir
f f(x) dx2 w2S fl(a + new) dn2
a- oa) "fc(a + (i + 1) )+ - f(a + io) - f(a + (i + 1) ) (37
4 8 (37)
31 I' 3 60 7!..
-4- +f(a-F)+w-8-F of (-+J(+1) ^+3-))- 08vf (a+o) -
which is the expression for the double integral between the limits
-2 and i+.
The value of'"f(a - ) given by equation (36) is in accordance
with the supposition that for n = - the double integral is equal to
zero, and this condition is fulfilled in the calculation of the perturbations when the argument a - Co corresponds to the date for which
the osculating elements are given. If, for in= — neither the single
nor the double integral is to be taken equal to zero, it is only necessary to add the given value of the single integral for this argument
to the value of'f(a - tw) given by equation (26), and to add the
given value of the double integral for the same argument to the value
of "f(t w) given by (36).
162. In a similar manner we may find the expressions for the
double integral between other limits. Thus, let it be required to
find the double integral between the limits -- and i.
Between the limits 0 and 2 we have
n- o
yf (a + nw) dn f f(a + noW) dn + f (a) + -41 a
1 1 1
+ 384P + 38-140r + 46008 + &C.
which gives
jff(a-+no) djn='2-,f(a) +f(a) — 1 f' (a)+, 4f" (a o () (a)(38)
0
1 i(a) 367 Fv (a)+3 78'7 -fvi (a)+&c.;
and this again, by means of (28), gives
f(a + nw) din = -fY(ac) + ( + ) f8(a + i) - -1-f (a + +) v)
~ 48 (i++a)})
+ -384f" (a+iw) (a-( 0) +3 ) j 8of i+ (a +iow)
3 68 77_f v(a+(i+I) aS)H + si15 46 7o'ofvi (a+-iw)+&c.
MECHANICAL QUADRATURE. 443
Therefore, since
i i +- i F+
ff(a + _f) dn2 Sf(a n) d f( + + ) dn2,
and'f(a + (i + ) W) -f(ca + (i + 1) w) - "f(a + iw),
f'(a + (( + ^-) oi) - f(a + (i + 1) V) - f(a + w),
f"' (a + (i + ) cl) -f" (a + (i + 1) w) -" (a + iw), &c.
we shall have
a + i i
d f (x) dX2 - " rjf(a + tu) dn2'
a-_ - (39)
_ —2 S+I)(a-+i)+-If (Ca+'i,) -4 off" (a+-)+6io)-8-f"v (a+iw) —&c. I,
which gives the required integral between the limits -- and i.
163. It will be observed that the coefficients of the several terms
of the formulae of integration converge rapidly, and hence, by a
proper selection of the interval at which the values of the function
are computed, it will not be necessary to consider the terms which
depend on the fourth and higher orders of differences, and rarely
those which depend on the second and third differences. The value
assigned to the interval w must be such that we may interpolate with
certainty, by means of the values computed directly, all values of the
function intermediate to the extreme limits of the integration; and
hence, if the fourth and higher orders of differences are sensible, it
will be necessary to extend the direct computation of the values of
the function beyond the limits which would otherwise be required,
in order to obtain correct values of the differences for the beginning
and end of the integration. It will be expedient, therefore, to take
w so small that the fourth and higher differences may be neglected,
but not smaller than is necessary to satisfy this condition, since otherwise an unnecessary amount of labor would be expended in the
direct computation of the values of the function. It is better, however, to have the interval o) smaller than what would appear to be
strictly required, in order that there may be no uncertainty with
respect to the accuracy of the integration. On account of the rapidity
with which the higher orders of differences decrease as we diminish
w, a limit for the magnitude of the adopted interval will speedily be
obtained. The magnitude of the interval will therefore be suggested
by the rapidity of the change of value of the function. In the comr
444 THEORETICAL ASTRONOMY.
putation of the perturbations of the group of small planets between
Mars and Jupiter we may adopt uniformly an interval of forty days;
but in the determination of the perturbations of comets it will evidently be necessary to adopt different intervals in different parts of
the orbit. When the comet is in the neighborhood of its perihelion,
and also when it is near a disturbing planet, the interval must necessarily be much smaller than when it is in more remote parts of its
orbit or farther from the disturbing body.
It will be observed, further, that since the double integral contains
the factor o2, if we multiply the computed values of the function by
W2, this factor will be included in all the differences and sums, and
hence it will not appear as a factor in the formule of integration.
If, however, the values of the function are already multiplied by o2,
and only the single integral is sought, the result obtained by the
formula of integration, neglecting the factor 02, will be (o times the
actual integral required, and it must be divided by ao in order to
obtain the final result.
164. In the computation of the perturbations of one of the asteroid
planets for a period of two or three years it will rarely be necessary
to take into account the effect of the terms of the second order with
respect to the disturbing force. In this case the numerical values of
the expressions for the forces will be computed by using the values
of the co-ordinates computed from the osculating elements for the
beginning of the integration, instead of the actual disturbed values
of these co-ordinates as required by the formulae (8). The values of
the second differential coefficients of Jx, oy, and 6z with respect to
the time, will be determined by means of the equations (9). If the
interval ce is such that the higher orders of differences may be neglected, the values of the forces must be computed for the successive
dates separated by the interval co, and commencing with the date
t0 - co corresponding to the argument a - c, t being the date to
which the osculating elements belong. Then, since the last terms
dp'x d'ay d_,z
of the formulae for -dt, dt2 and t involve 8x, 8y, and Jz, which
are the quantities sought, the subsequent determination of the differential coefficients must be performed by successive trials. Since the
integral must in each case be equal to zero for the date to, it will be
admissible to assume first, for the dates to - ~c and to + -o corresponding to the arguments a - o and a, that 8x 0, y = 0, and
8z- 0 and hence that the three differential coefficients, for each
VARIATION OF CO-ORDINATES. 445
date, are respectively equal to X0, Y0, and Z,. We may now by integration derive the actual or the very approximate values of the
variations of the co-ordinates for these two dates. Thus, in the case
of each co-ordinate, we compute the value of'f(a - 1w) by means
of the equation (26), using only the first term, and the value of
"f(a - w) from (36), using in this case also only the first term. The
value of the next function symbolized by "f will be given by
7"(a) = "7(a- ) +'f(a- w).
Then the formula (39), putting first i =-1 and then i = 0, and
neglecting second differences, will give the values of the variations
of the co-ordinates for the dates a - w and a. These operations will
be performed in the case of each of the three co-ordinates; and, by
means of the results, the corrected values of the differential coefficients will be obtained from the equations (9), the value of &r being
computed by means of (7). With the corrected values thus derived
a new table of integration will be commenced; and the values of'f(a- lo) and "f(a -- w) will also be recomputed. Then we obtain,
also, by adding f(a - o) to f(a), the value of'f(a -+ w), and, by
adding this to "f(a), the value of "f(a + w).
An approximate value of f(a + w) may now be readily estimated,
and two terms of the equation (39), putting i 1, will give an approximate value of the integral. This having been obtained for
each of the co-ordinates, the corresponding complete values of the
differential coefficients may be computed, and these having been
introduced into the table of integration, the process may, in a similar
manner, be carried one step farther, so as to determine first approximate values of Jx, 8y, and 8z for the date represented by the argument a + 2w, and then the corresponding values of the differential
coefficients. We may thus by successive partial integrations determine the values of the unknown quantities near enough for the calculation of the series of differential coefficients, even when the integrals are involved directly in the values of the differential coefficients.
If it be found that the assumed value of the function is, in any case,
much in error, a repetition of the calculation may become necessary;
but when a few values have been found, the course of the function
will indicate at once an approximation sufficiently close, since whatever error remains affects the approximate integral by only onetwelfth part of the amount of this error. Further, it. is evident
that, in cases of this kind, when the determination of the values of
the differential coefficients requires a preliminary approximate inte
446 THEORETICAL ASTRONOMY.
gration, it is necessary, in order to avoid the effect of the errors in
the values of the higher orders of differences, that the interval m
should be smaller than when the successive values of the function to
be integrated are already known. In the case of the small planets
an interval of 40 days will afford the required facility in the approximations; but in the case of the comets it may often be necessary to
adopt an interval of only a few days. The necessity of a change in
the adopted value of (w will be indicated, in the numerical application of the formul, by the manner in which the successive assumptions in regard to the value of the function are found to agree with
the corrected results.
The values of the differential coefficients, and hence those of the
integrals, are conveniently expressed by adopting for unity the unit
of the seventh decimal place of their values in terms of the unit of
space.
165. Whenever it is considered necessary to commence to take into
account the perturbations due to the second and higher powers of the
disturbing force, the complete equations (14) must be employed. In
this case the forces X, Y, and Z should not be computed at once for
the entire period during which the perturbations are to be determined.
The values computed by means of the osculating elements will be
employed only so long as simply the first power of the disturbing
force is considered, and by means of the approximate values of 8x,
Jy, and 8z which would be employed in computing, for the next place,
the last terms of the equations (9), we must compute also the corrected values of X, Y and Z. These will be given by the second
members of (8), using the values of x, y, and z obtained from
x -x= + 8x, - Yo + y, z -= zo + sz.
We compute also q from (12), and then from Table XVII. find the
d'8x d28y
corresponding value of f. The corrected values of dta, dt~ y and
d'6z dt' dt2'a
dt2 will be given by the equations (14), and these being introduced,
in the continuation of the table of integration, we obtain new values
of Ax, By, and 8z for the date under consideration. If these differ
much from those previously assumed, a repetition of the calculation
will be necessary in order to secure extreme accuracy. In this repetition, however, it will not be necessary to recompute the coefficients
of ax, ay, and Oz in the formula for q, their values being given with
sufficient accuracy by means of the previous assumption; and gene
VARIATION OF CO-ORDINATES. 447
rally a repetition of the calculation of X, Y, and Z will not be
required.
Next, the values of 8x, 8y, and 8z may be determined approximately, as already explained, for the following date, and by means
of these the corresponding values of the forces X, Y. and Z will be
found, and also f and the remaining terms of (14), after which the
integration will be completed and a new trial made, if it be considered necessary. In the final integration, all the terms of the formulse of integration which sensibly affect the result may be taken
into account. By thus performing the complete calculation of each
successive place separately, the determination of the perturbations in
the values of the co-ordinates may be effected in reference to all
powers of the masses, provided that we regard the masses and co-ordinates of the disturbing bodies as being accurately known; and it is
apparent that this complete solution of the problem requires very
little more labor than the determination of the perturbations when
only the first power of the disturbing force is considered. But
although the places of the disturbing bodies as given by the tables
of their motion may be regarded as accurately known, there are yet
the errors of the adopted osculating elements of the disturbed body
to detract from the absolute accuracy of the computed perturbations;
and hence the probable errors of these elements should be constantly
kept in view, to the end that no useless extension of the calculation
may be undertaken. When the osculating elements have been corrected by means of a very extended series of observations, it will be
expedient to determine the perturbations with all possible rigor.
When there are several disturbing planets, the forces for all of
these may be computed simultaneously and united in a single sum,
so that in the equations (14) we shall have IX, 2 Y, and.SZ instead
of X, Y, and Z respectively; and the integration of the expressions
d2ax d12y d 2z
for dt2, ~2 and dt will then give the perturbations due to the
action of all the disturbing bodies considered. However, when the
interval eo for the different disturbing planets may be taken differently,
it may be considered expedient to compute the perturbations separately, and especially if the adopted values of the masses of some of
the disturbing bodies are regarded as uncertain, and it is desired to
separate their action in order to determine the probable corrections
to be applied to the values of m, mn, &c., or to determine the effect
of any subsequent change in these values without repeating the calculation of the perturbations.
448 THEORETICAL ASTRONOMY.
166. EXAMPLE.-TO illustrate the numerical application of the
formulae for the computation of the perturbations of the rectangular
co-ordinates, let it be required to compute the perturbations of
Eurynome @ arising from the action of Jupiter' from 1864 Jan. 1.0
Berlin mean time to 1865 Jan. 15.0 Berlin mean time, assuming the
osculating elements to be the following:Epoch 1864 Jan. 1.0 Berlin mean time.
Mo= 10 29' 5".65
" 206_ 439 5.69 Ecliptic and Mean
0 24O-0 36 52 11 Equinox 1860.0
o - 4 36 52.11
o -= 11 15 51.02
log c0 0.3881319
o = 928".55745.
From these elements we derive the following values:Berlin Mean Time. o yo zo log r
1863 Dec. 12.0 +'l.53616 -t 1.23012 -0.03312 0.294084,
1864 Jan. 21.0 1.15097 1.59918 0.07369 0.294837,
March 1.0 0.69518 1.87033 0.10978 0.300674,
April 10.0 +- 0.19817 2.03141 0.13936 0.310864,
May 20.0 -0.31012 2.08092 0.16134 0.324298,
June 29.0 0.80326 2.02602 0.17523 0.339745,
Aug. 8.0 1.26055 1.87959 0.18122 0.356101,
Sept. 17.0 1.66729 1.65711 0.17990 0.372469,
Oct. 27.0. 2.01414 1.37473 0.17209 0.388214,
Dec. 6.0 2.29597 1.04766 0.15870 0.402894,
1865 Jan. 15.0 -2.51077 + 0.68978 - 0.14066 0.416240.
The adopted interval is w - 40 days, and the co-ordinates are referred to the ecliptic and mean equinox of 1860.0. The first date,
it will be observed, corresponds to to - (o, and the integration is to
commence at 1864 Jan. 1.0.
The places of.Jupiter derived from the tables give the following
values of the co-ordinates of that planet, with which we write also
the distances of Eurynome from Jtupiter computed by means of the
formula
= (X - )2 + (yl- y) + (z'- z)
Berlin Mean Time. X' y'' log r' log p
1863 Dec. 12.0 -4.09683 -3.55184 +0.10533 0.73425 0.86866,
1864 Jan. 21.0 3.89630 3.76053 0.10152 0.73368 0.86713,
March 1.0 3.68416 3.95803 0.09744 0.73305 0.86292,
April 10.0 — 3.46098 -4.14366 +0.09304 0.73237 0.85622,
NUMERICAL EXAMPLE. 449
Berlin Mean Time. x/ z/ log log p
1864 May 20.0 -3.22739 -4.31684 +0.08839 0.73164 0.84732,
June 29.0 2.98405 4.47693 0.08346 0.73086 0.83656,
Aug. 8.0 2.73162 4.62343 0.07827 0.73003 0.82428,
Sept. 17.0 2.47085 4.75576 0.07284 0.72915 0.81077,
Oct. 27.0 2.20247 4.87345 0.06720 0.72823 0.79628,
Dec. 6.0 1.92728 4.97606 6.06134 0.72726 0.78098,
1865 Jan. 15.0 -1.64600 -5.06301 0.05531 0.72625 0.76498.
These co-ordinates are also referred to the ecliptic and mean equinox
of 1860.0.
If we neglect the mass of Eurynome and'adopt for the mass of
Jupiter
1047.819'
we obtain, in units of the seventh decimal place,
2,'k" =- 4518.27,
and the equations (9) become
2'26 4l8 27(xf1xo X )i 0.47346 (3o r Ox)
W2 dy = 4518.27( p -, 0.4 73,3 -3 ( r (40)
dt\' ro " \ 0 ro
W2 -' _ 4518.27( )+ 0.47346 r -
ty92 dl83 4518-27 (- 3 ~r ) +(
dt P3\ r1'3 \03 7
Substituting for the quantities in the first term of the second member
of each of these equations the values already found, we obtain
Argument. Date. w2XQ Y2 Yo 2Z
a - w 1863 Dec.. 12.0 + 53.00 + 47.09 - 1.43,
a 1864 Jan. 21.0 53.71 46.31 0.91,
a + w March 1.0 54.23 45.18 - 0.37,
a + 2w April 10.0 54.69 43.59. + 0.22,
a + 3wo May 20.0 55.23 4151- 0.70,
a + 40 June 29.0 56.06 -8.96 1.19,
a + 5o Aug. 8.0 57.30 35.92 1.66,
a + 6w Sept. 17.0 59.0O 32.47 2.08,
a + 7w Oct. 27.0 61.55 28.60 2.43,
a + 8w Dec. 6.0 64.85 24.34 2.69,
a + 9w 1865 Jan. 15.0 + 69.09 + 19.78 + 2.83,
which are expressed in units (' the seventh decimal place.
We now, for a first approx:imation, regard the perturbations as
29
45I0 THEORETICAL ASTRONOMY.
being equal to zero for the dates Dec. 12.0 and Jan. 21.0, and, in the
case of the variation of x, we compute first'f(a _) w - - f' (a - )- - (53.71 - 53.00) - 0.03,
53.71
f(a- ) -f (a) - + + 2.24,
and the approximate table of integration becomes
f (a - ) + 53.00 f(a- ) — _ 0.03,f(a-w) + 2.24)
f(a) + 53.71. " f() + 2.21.
Then the formula (39), putting first i =- 1, and then i = 0, gives
53.00
Dec. 12.0 x = + 2.24 + 125 + 6.66,
53.71
Ja,. 21.0 x = + 2.21 + - + 6.69.
In a similar manner, we find
Dec. 12.0 + —. 5.85 z- 0.16,
Jan. 21.0 + 5.82 = 0.14.
By means of these results we compute the complete values of the
second members of equations (40), Jr being found from'- Yo x + r Y + ZO:70 "o To
and thus we obtain
dax,d__y d2d
Date. 02 dd2 ddy
dt2 cdt dt'
Dec. 12.0 + 53.86 -t 47.76 1.45 + 8.85,
Jan. 21.0 -+ 54.23 + 47.25 -0.96 + 8.63.
We now commence anew the table of integration, namely,
2x y.
f'f f f f f f' "f
+53.86_ 0.02 + 2.2C. +47.76 + 002+ 1.97, -1.45 -0 -0.04,
+54.23 542 + 2.24, 4-47.25 +4727 + 1.99, -0.96 0.98 -0.06,
+56.45, f49.26, -1o.04,
the formation of which is made evilenti by what precedes.
We may next assume for approximate values of the differential
coefficients, for the date March:0, + 54.6, + 46.7, and — 0.5,
respectively; and these give, for this date,
NUMERICAL EXAMPLE. 451
54.6
x= + 56.45 -+- 5 = 4 + 61.00,
1-2i
y + 49.26 + 46 + 53.15,
12 = + 53.15,
a —- 1.04- 0.5- 1.08.
12
By means of these approximate values we obtain the following
results:1864 March 1.0 dt - 55.01, - - 53.86, - -1.00,
dtt2 dt 2 dt2
r -+- +71.03.
Introducing these into the table of integration, we find, for the corresponding values of the integrals,
=-1-t 61.03; y - +53.75, Oz= - 1.12.
These results differ so little from those already derived from the
assumed values of the function that a repetition of the calculation is
unnecessary. This repetition, however, gives
d' 3x d2sy d23Z
d, -- + 55.04, + - 53.91, -- 1.00.
d dtd d
Assuming, again, approximate values of the differential coefficients
for April 10.0, and computing the corresponding values of Ox, oy,
and Jz, we derive, for this date,
d~d~ dd3 d~z
- 2 d = + 48.06, o2 -- 63.19, 2 2
dt2 dt2 dt2
Introducing these into the table of integration, and thus deriving
approximate values of Sx, jy, and 8z for May 20, we carry the process one step further. In this manner, by successive approximations,
we obtain the following results:dx d2jy 426
Date. 2 d 2 dy 2 d
_dt2 dt2 dCt2
1863 Dec. 12.0 + 53.86 + 47.76 - 1.45,
1864 Jan. 21.0 54.23 47.25 0.96,
March 1.0 55.04 53.91 1.00,
April 10.0 48.06 63.19 1.54,
May 20.0 32.85 65.40 2.07,
June 29.0 16.74 54.48 1.75,
Aug. 8.0 8.62 31.39 - 0.36,
Sept. 17.0 -- 14.20 + 2.09 + 1.86,
452 THEORETICAL ASTRONOMY.
Date. ~2 dld 02 ddY d2
ct2 dt2 dt2
1864 Oct. 27.0 + 34.84 - 26.32 + 4.44,
Dec. 6.0 68.79 47.87 6.86,
1865 Jan. 15.0 + 112.64 - 58.39 + 8.68.
The complete integration may now be effected, and we may use both
equation (37) and equation (39), the former giving the integral for
the dates Jan. 1.0, Feb. 10.0, March 21.0, &c., and the latter the
integrals for the dates in the foregoing table of values of the function.
The final results for the perturbations of the rectangular co-ordinates,
expressed in units of the seventh decimal place, are thus found to be
the following:Berlin Mean Time. 6x 6y 6z
1863 Dec. 12.0 + 6.7 + 5.9 -0.2,
1864 Jan. 1.0 0.0 0.0 0.0,
21.0 + 6.8 5.9 0.1,
Feb. 10.0 27.1 23.5 0.5,
March 1.0 61.0 53.7 1.1,
21.0 108.9 97.4 2.0,
April 10.0 169.7 155.7 3.1,
30.0 242.7 229.9 4.7,
May 20.0 325.7 320.3 6.7,
Junel 9.0 417.1 427.2 9.3,
29.0 514.6 549.1 12.3,
July 19.0 616.1 684.9 15.7,
Aug. 8.0 720.8 831.4 19.5,
28.0 827.4 986.0 23.4,
Sept. 17.0 936.8 1144.6 27.0,
Oct. 7.0 1049.4 1303.8 30.2,
27.0 1168.2 1460.0 32.6,
Nov. 16.0 1295.4 1609.4 33.9,
Dec. 6.0 1435.6 1749.6 33.8,
26.0 1592.8 1877.6 32.0,
1865 Jan. 15.0 + 1772.6 + 1992.3 -28.2.
During the interval included by these perturbations, the terms of
the second order of the disturbing forces will have no sensible effect;
but to illustrate the application of the rigorous formulae, let us commence at the date 1864 Sept. 17.0 to consider the perturbations of
the second order.
In the first place, the components of the disturbing force must be
computed by means of the equations
NUMERICAL EXAMPLE. 453
oJ2X_ oJ2'k2 __- oT( y) w2 Yw2m'k2 y y ),
w3Z==m^^ k(-z - z)
The approximate values of 8x, 3y, and 8z for Sept. 17.0 given immediately by the table of integration extended to this date, will suffice
to furnish the required values of the disturbed co-ordinates by means
of
x =- o + 6x, y = + Yo - +Sy, z =z z - z;
and to find p - Po + kp, we have
ap —~ ~x- ~s ~ Y- y z 8_P P P
or
J log p - - ((.' -.) x + (y' - y) y + (z'- z) z),
in which 2o is the modulus of the system of logarithms. Thus we
obtain, for Sept. 17.0,
~ log p = +- 0.0000084,
W2X= + 59.09, w2Y= - 32.48, w2Z — + 2.08,
which rerequire no further correction.
Next, we compute the values of
o +- YO + a2 y Zo + ~ az
po2'0r2
which also will not require any further correction, and thus we form,
according to (12), the equation
q = - - 0.29996ax + 0.29815ay - 0.03237z.
The approximate values of tx, gy, and z being substituted in this
equation, we obtain
q -= + 0.0000061,
corresponding to which Table XVII. gives
Hence we derive
(fqx - ax) - 44.87, - f (fy - y) - 30.40,'0 2'r0
ua2,
(fqz-z) —0.21,
whchalo il nt euie nyfuthr oretinan tusPefom
454 THEORETICAL ASTRONOMY.
and the equations (14) give
d"axod 2'ay dd2~x
+ 14.22, + 2.08, d 1.87.
dt2 dt' +' dt'
These values being introduced into the table of integration, the
resulting values of the integrals are changed so little that a repetition
of the calculation is not required.
We now derive approximate values of Ax, 8y, and Jz for Oct. 27.0,
and in a similar manner we obtain the corrected values of the differential coefficients for this date; and thus by computing the forces for
each place in succession from approximate values of the perturbations,
and repeating the calculation whenever it may appear necessary, we
may determine the perturbations rigorously for all powers of the
masses. The results in the case under consideration are the following:Date. 2 ddx 02 dy 2d2
dt2 dt2 dt2
1864 Sept. 17.0 + 14.22 + 2.08 + 1.87,
Oct. 27.0 34.84 — 26.31 4.44,
Dec. 6.0 68.77 47.86 6.86,
1865 Jan. 15.0 + 112.60 - 58.39 + 8.68.
Introducing these results into the table of integration, the integrals
for Jan. 15.0 are found to be
x= + 1772.6, 8y + 1992.3, 8z - 28.2,
agreeing exactly with those obtained when terms of the order of the
square of the disturbing forces are neglected.
If the perturbations of the rectangular co-ordinates referred to the
equator are required, we have, whatever may be the magnitude of the
perturbations,
(x, = x,
y, cos e ay - sin e sz, (41)
Jz, = sin e 8y + cos e z,
x,, y,, z, being the co-ordinates in reference to the equator as the fundamental plane. Thus we obtain, for 1865 Jan. 15.0,
ax, = + 1772.6, ay, = + 1838.9,, =z + 767.2.
These values, expressed in seconds of arc of a circle whose radius is
the unit of space, are
xr, - + 36".562, dy, = + 37".930, az, = + 15".825.
VARIATION OF CO-ORDINATES. 455
The approximate geocentric place of the planet for the same date is
a = 183~ 28', = - 5~ 39', log a = 0.3229,
and hence, neglecting terms of the second order, we derive, by means
of the equations (3)2, for the perturbations of the geocentric right
ascension and declination,
aa ~- 17".03, -+ =- 5".67.
167. The values of 8x, By, and Jz, computed by means of the coordinates referred to the ecliptic and mean equinox of the date t, must
be added to the co-ordinates given by the undisturbed elements and
referred to the same mean equinox. The co-ordinates referred to the
ecliptic and mean equinox of t may be readily transformed into those
referred to the ecliptic and mean equinox of another date t'. Thus,
let0 denote the longitude of the descending node of the ecliptic of t'
on that of t, measured from the mean equinox of t, and let' be the
mutual inclination of these planes; then, if we denote by x', y', z'
the co-ordinates referred to the ecliptic of t as the fundamental plane,
the positive axis of x, however, being directed to the point whose
longitude is 0, we shall have
x' = x cos 8 + y sin 8,
y' = —x sin + y cos 0, (42)
2f -2.
Let us now denote by x", y", z" the co-ordinates when the ecliptic
of i is the plane of xy, the axis of x remaining the same as in the
system of x', y', z'. Then we shall have
F,
y" =y' cos - z' sin a, (43)
z' - y' sin - + z' cos a.
Finally, transforming these so that the axis of z remains unchanged.
while the positive axis of x is directed to the mean equinox of t and
denoting the new co-ordinates by x,, y,, z,, we get
x, - x" cos (0 + p) ~ Y" sin (0 + p),
y, x"" sin (o + p) + y" cos (0 + p), (44)
Z, =- Z,
in which p denotes the precession during the interval t'- t. Eliminating x", y", and z" from these equations by means of (43) and (42),
observing that, since - is very small, we may put cos' = 1, we get
456 THEORETICAL ASTRONOIY.
x, cosp - y sinCS + _ 1 sin ( -+ p),
8
y, x sinp + y cosp - z cos (~ + p), (45)
z, —z - x sin 0 - + y os,
8s
in which s =206264.8, V being supposed to be expressed in seconds
of arc. If we neglect terms of the order p3, these equations become
x,= x-x P x_ yP + (sin + p cos 0) z,
P8 S S27
y, _y i P- y - (cos -_ sin ), (46)
z, = z - xa sin 0+ -ycos0.
S S
These formulae give the co-ordinates referred to the ecliptic and mean
equinox of one epoch when those referred to the ecliptic and mean
equinox of another date are known. For the values of p, r, and 0,
we have
p (50".21129 + 0".0002442966r) (t'- ),
= ( 0".48892 - 0".000006143T) (t'- t),
o 351~ 36' 10" + 39".79 (t - 1750) - 5".21 (t' -— ),
in which = -(t' - t) - 1750, t and t' being expressed in years from
the beginning of the era. If we add the nutation to the value of p,
the co-ordinates will be derived for the true equinox of t'.
The equations (45) and (46) serve also to convert the values of 0x,
Sy, and 8z belonging to the co-ordinates referred to the ecliptic and
mean equinox of t into those to be applied to the co-ordinates referred to the ecliptic and mean equinox of t'. For this purpose it
is only necessary to write 8x, 8y, and 8z in place of x, y, and z respectively, and similarly for x,, y,, z,.
In the computation of the perturbations of a heavenly body during
a period of several years, it will be convenient to adopt a fixed equinox and ecliptic throughout the calculation; but when the perturbations are to be applied to the co-ordinates, in the calculation of an
ephemeris of the body taking into account the perturbations, it will
be convenient to compute the co-ordinates directly for the ecliptic
and mean equinox of the beginning of the year for which the
ephemeris is required, and the values of 8x, Jy, and 8z must be
reduced, by means of the equations (45), as already explained, from
the ecliptic and mean equinox to which they belong, to the ecliptic
and mean equinox adopted in the case of the co-ordinates required.
VARIATION OF CO-ORDINATES. 457
In a similar manner we may derive formulae for the transformation
of the co-ordinates or of their variations referred to the mean equinox
and equator of one date into those referred to the mean equinox
and equator of another date; but a transformation of this kind will
rarely be required, and, whenever required, it may be effected by first
converting the co-ordinates referred to the equator into those referred
to the ecliptic, reducing these to the equinox of t' by means of (45)
or (46), and finally converting them into the values referred to the
equator of t'.' Since, in the computation of an ephemeris for the
comparison of observations, the co-ordinates are generally required
in reference to the equator as the fundamental plane, it would appear
preferable to adopt this plane as the plane of xy in the computation
of the perturbations, and in some cases this method is most advanta.geous. But, generally, since the elements of the orbit of the disturbed planet as well as the elements of the orbits of the disturbing
bodies are referred to the ecliptic, the calculation of the perturbations
will be most conveniently performed by adopting the ecliptic as the
fundamental plane. The consideration of the change of the position
of the fundamental plane from one epoch to another is thus also rendered more simple. Whenever an ephemeris giving the geocentric
right ascension and declination is required, the heliocentric co-ordinates of the body referred to the mean equinox and equator of the
beginning of the year will be computed by means of the osculating
elements corrected for precession to that epoch, and the perturbations
of the co-ordinates referred to the ecliptic and mean equinox of any
other date will be first corrected according to the equations (46), and
then converted into those to be applied to the co-ordinates referred to
the mean equinox and equator. If the perturbations are not of considerable magnitude and the interval t' - t is also not very large, the
correction of Ox, By, and az on account of the change of the position
of the ecliptic and of the equinox will be insignificant; and the
conversion of the values of these quantities referred to the ecliptic
into the corresponding values for the equator, is effected with great
facility.
In the determination of the perturbations of comets, ephemerides
being required only during the time of describing a small portion of
their orbits, it will sometimes be convenient to adopt the plane of the
undisturbed orbit as the fundamental plane. In this case the positive axis of x should be directed to the ascending node of this plane
on the ecliptic, and the subsequent change to the ecliptic and equinox,
whenever it may be required, will be readily effected.
458 THEORETICAL ASTRONOMY.
168. The perturbations of a heavenly body may thus be determined rigorously for a long period of time, provided that the osculating elements may be regarded as accurately known. The peculiar
object, however, of such calculations is to facilitate the correction of
the assumed elements of the orbit by means of additional observations according to the methods which have already been explained;
and when the osculating elements have, by successive corrections,
been determined with great precision, a repetition of the calculation
of the perturbations may become necessary, since changes of the elements which do not sensibly affect the residuals for the given differential equations in the determination of the most probable corrections,
may have a much greater influence on the accuracy of the resulting
values of the perturbations.
When the calculation of the perturbations is carried forward for a
long period, using constantly the same osculating elements,-and
those which are supposed to require no correction,-the secular perturbations of the co-ordinates arising from the secular variation of
the elements, and the perturbations of long period, will constantly
affect the magnitude of the resulting values, so that Ax, cy, and 8z
will not again become simultaneously equal to zero. Hence it
appears that even when the adopted elements do not differ much
from their mean values, the numerical amount of the perturbations
may be very greatly increased by the secular perturbations and by
the large perturbations of long period. But when the perturbations
d28x: d2oy
are large, the calculation of the complete values of dt2' ~-' and
d2sz
dt2 (which is effected indirectly) cannot be performed with facility,
dGb
requiring often several repetitions in order to obtain the required
accuracy, since any error in the value of the second differential coefficient produces, by the double integration, an error increasing proportionally to the time in the values of the integral. Errors, therefore,
in the values of the second differential coefficients which for a moderate period would have no sensible effect, may in the course of a long
period produce large errors in the values of the perturbations, and it
is evident that, both for convenience in the numerical calculation and
for avoiding the accumulation of error, it will be necessary from time
to time to apply the perturbations to the elements in order that the
integrals may, in the case of each of the co-ordinates, be again equal
to zero. The calculation will then be continued until another change
of the elements is required.
CHANGE OF THE OSCULATING ELEMENTS. a59
The transformation from a system of osculating elements for one
epoch to that for another epoch is very easily effected by means of
the values of the perturbations of the co-ordinates in connection
with the corresponding values of the variations of the velocities
dx dy dz
dt, - d- and -d. The latter will be obtained from the values of the
second differential coefficients by means of a single integration according to the equations (27) and (32). Thus, in the case of the
example given, we obtain for the date 1865 Jan. 15.0, by means of
(32), in units of the seventh decimal place,
dx day dz
40 dt + 385.9, 40 -- + 214., 40 9.7.
-dt d4 — 2,t dThe velocities in the case of the disturbed orbit will be given by the
formulae
dx dxo dRx dy dyo day d d dz dzo dSz 47
dt dt + dt' dt dt + dt' dt = dt + dt -
dt tdt
To obtain the expressions for the components of the velocity
resolved parallel to the co-ordinates, we have, according to the equations (6)2,
dx dr dv
= in a sin- r sinsi +) a cos (A + ) dt'
at dt at
dy., /-, dr dv
dyt sin b sin (B +- u) d- + r sin b cos (B - u) —,
dt t at
dz dv
= sin c sin (C + u) r + r sin c cos (C - ) u.
These equations are applicable in the case of any fundamental plane,
if fhe auxiliaries sin a, sin b, sin c, A, B, and C are determined in
reference to that plane. To transform them still further, we have
dr kl/ + m.
-=, -- me sin (U -- ),
dt l -p
dv kl/p(l+m) k1/1 ~m (
rd_ p+)(1 _ e cos (u - )),
dt r - 1/p
in which w denotes the angular distance of the perihelion from the
ascending node. Substituting.these values, we obtain, by reduction,
460 THEORETICAL ASTRONOMY.
d- = i/ ((e cos w cos uo) cos A - (esin ~ si sin A)sin) ) i a,
dy k1/1 +
d —,_ ~ -p ((e cos -- cos u) cosB - (e sin o + sin u) sin B) sin b,
dt v/p
dz kl-+m
dt- /- ((e cosw - + cos ic) cos C - (e sin - + siSn u) sin C) sin c.
Let us now put
1/p + m (e sin (o + sin u) - Vsin U,
/p;(48)
~k/1 +~m (e cos +- cos u) = Vcos U,
and we have
dx
dt = Vsin a cos (A + U),
dy
dy Vsin b cos (B + U), (49)
dt =Vsin c cos (C U).
These equations determine the components of the velocity of a heavenly body resolved in directions parallel to the co-ordinate axes,
and for any fundamental plane to which the auxiliaries A, B, &c.
belong. When the ecliptic is the fundamental plane, we have
sin c = sin i, C- 0.
The sum of the squares of the equations (48) gives
= 2k2(1m)(1 + e2+ cos( )) k2(+m) ( - I)
and hence it appears that V is the linear velocity of the body.
The determination of the osculating elements corresponding to any
date for which the perturbations of the co-ordinates and of the velocities have been found, is therefore effected in the following manner:First, by means of the osculating elements to which the perturbations belong, we compute accurate values of r0, x0, y, z0, and by
dx
means of the equations (48) and (49) we compute the values of -dt
dy, dz
d and -d. Then we apply to these the values of the perturbations, and thus nd, dy d These having been
tions, and thus find x, y, z, - -i, and d These having been
CHANGE OF THE OSCULATING ELEiMENTS. 461
found, the equations (32), will furnish the values of 2, i, and p;
and the remaining elements may be determined as explained in Art.
112. Thus, from
Vr sin 4- ~kp (1 +- m),
dx dy dz
=Vr cos ^x dt + dtZ dt'
we obtain Vr and 0o, and from
r sin u - (- x sin + y cos 2) sec i,
r cos u -- x cos +- y sin 2,
we derive r and u; and hence Vfrom the value of Vr. When i is
not very small, we may use, instead of the preceding expression for
r sin u,
r sin iu z z cosec i.
Next, we compute a from
r
2a-r=2 k'ld n)1
2 k 2 (Iq-)'
r V2
and from
2ae sin ow - (2a - r) sin (2% -+ u) - r sin u,
2ae cos w - (2c -~ r) cos (2+ + t- ) - r cos u,
we find co and e. The mean daily motion and the mean anomaly or
the mean longitude for the epoch will then be determined by means
of the usual formule.
In the case of a very eccentric orbit, after r and ut have been found,
dr
dt will be given by equations (48)6, and the values of e and v will
be given by the equations (49)6. Then the perihelion distance will
be found from
p
q-1!+ e'
and the time of perihelion passage will be found from v and e by
means of Table IX. or Table X.
dx dy
In the numerical values of the velocities dt' d-, &c., more decimals
t dt'
must be retained than in the values of the co-ordinates, and enough
must be retained to secure the required accuracy of the solution. If
it be considered necessary, the different parts of the calculation may
be checked by means of various formulae which have already been
given. Thus, the values of 2g and i must satisfy the equation
462 THEORETICAL ASTRONOMY.
z cos i -- y si i cos + + sin i sin -- 0.
We have, also,
_ dx 2 /d y \ \2 d 2
V2 (dt)+( dYt ) ( dt
r2 2 + I-2 + 2-,
z - r sin t sin i,
which must be satisfied by the resulting values of V, r, and u; and
the values of a and e must satisfy the equation
p - a (1 -e2) a cos2 o.
169. When the plane of the undisturbed orbit is adopted as the
fundamental plane, we obtain at once the perturbations
S (r cos i), 8 (r sin u), az,
and from these the perturbations of the polar co-ordinates are easily
derived. There are, however, advantages which may be secured by
employing formulae which give the perturbations of the polar co-ordinates directly, retaining the plane of the orbit for the date to as the
fundamental plane.
Let w denote the angle which the projection of the disturbed
radius-vector on the plane of xy makes with the axis of x, and / the
latitude of the body with respect to the plane of xy; then we shall
have
X = r cos t' cos w,
y = r cos f sin w, (50)
z - r sin f.
Let us. now denote by X, Y, and Z, respectively, the forces which are
expressed by the second members of the equations (1), and the first
two of these equations give
xd -d dy d ( Yx — y) dt + C,
C being the constant of integration. The equations (50) give
dx d(rcos f). d
d- COS W r cos f sin wd- rco
dt dt dt
dy d (r cos f) dw
dt= sin w dt - r cs cos w cwdt dt dt
and hence
dy dx 2 2 dw
-dt -Y dt-r Cos dt
VARIATION OF POLAR CO-ORDINATES. 463
Therefore we have
r2 cos2 dj( Yx - Xy) dt + C.
If we denote by S0 the component of the disturbing force in a direction perpendicular to the disturbed radius-vector and parallel with
the plane of xy, we shall have
X- -- SO sin w, Y= So cos w,
and
Yx - Xy = S r cos /f.
Therefore
r cos t 3 SO r cos f dt + C.
In the undisturbed orbit we have 9 0, and
dwo -kV/po (1 + n);
and thus the preceding equation becomes
r2 cosZ dwt — So r cos l dt + k o (-+ -n). (51)
The equations (1) also give
1 xd2A + yd2 d2y' k(1 + nm) _ Z- (52)
r d r2 r r
If we denote by R the component of the disturbing force in the
direction of the disturbed radius-vector, we have
R= XY + Yy+Z (53)
r r r
We have, also,
xd2x + yd2y + zd2z d (xdz + ydy + zdz) -- (dx2 + dy2 + dz2)
= d (rdr) - (dr2 + r2dv2) - rd2r - rdv2,
v denoting the true anomaly in the disturbed orbit, or, since
dv2 = cos2 9 dW2 + d)2,
xd2x + yd2y + zd2z = rd2r - r2 cos2 f dw2 - r2di2.
Hence the equation (52) becomes
d2r dw2 df2 k2 (1 ) + )
-- -r cos d - r -- ( = R. (54)
d dt2 " dt2 r2
464 THEORETICAL ASTRONOMY.
170. The equations (51) and (54), in connection with the last of
equations (1), completely represent the motion of a heavenly body
about the sun when acted upon by disturbing forces, and, when completely integrated, they will give the values of w, r, and z for any
point of the orbit; but, since they cannot be integrated directly, we
must, as in the case of the rectangular co-ordinates, find the equations
which give by integration the values of 8w, Jr, and z. In the case
of the undisturbed orbit, we have
dwo
r I o (l+
(55)
d2ro dWo2 k2(1+ m)_ (55)
dt2 o dt2 + o2
If we denote by 8w the variation of w arising from the action of the
disturbing force, we have aw= wo + -w; and hence we easily find,
from (51),
dsw 1 f' I 2\ p (1+ )
dt -r'cos2 pJ,Srcos dt-l- 2 4 (56)\
dt rc Cos fi dt 1 C — r. cos2 ~
We have, further,
r2 - ro2 + 2rr + 8r2,
which gives
r _1+2r &
r' ro2
Let us now put
q c ( -o siOS2 i,' -- ai1 cos' (57)
o2 c 8s2 —Co 2 2
and. we have
1' - (58)
The equation (56), therefore, becomes
dt r cos' j So r cosp f dt -g (59)
in which we put
dwo kVpo (1 +n) (60)
go dt ro2
If we substitute r + 8r for r in equation (54), and combine the
result with the second of equations (55), we get
d2r dw22 dt2 + k2 ( r
VARIATION OF POLAR CO-ORDINATES. 465
and if we put
"ro + l r, r
_q" - Or, fT- 1-__ (61)
we have
1 + 2q" (62)
and hence
d'Jr k' (1 +- n) dbw
dt - R + )+ f,,q, + go2fr + 2g dto
%2 ftq dt
(e 910 dt (63)
-r sin 2i (go+dt ) + t r( dt
Finally, we have, from the last of equations (1),
d2z Ik2(1 + n)
=Z- Z, (64)
by means of which the value of z may be found, since, in the case of
the undisturbed motion, we have z0 = 0.
The values off' corresponding to different values of q' may be
tabulated with the argument q', and, since the equation (62) is of the
same form as (58), the same table will give the value of f" when q"
is used as the argument. Table XVII. gives the values of logf or
logfI corresponding to values of q' or q" from - 0.03 to + 0.03.
Beyond the limits of this table the required quantities may be computed directly.
171. When we consider only terms of the first order with respect
to the disturbing force, we have
flqY 2ar
ro,
and the equations become
d(t o -t rJ 0
d"6o
d2 —- -R+ - fsorOdt + (2k. (1 + -n)) 3)r, (65)
d2_z k2 (1 + nm)
dt2 ro
In determining the perturbations of a heavenly body, we first consider only the terms depending on the first power of the disturbing
force, for which these equations will be applied. The value of dr
30
466 THEORETICAL ASTRONOMY.
will be obtained from the second equation by an indirect process, as
already illustrated for the case of the variation of the rectangular
co-ordinates. Then aw will be obtained directly from the first
equation, and, finally, z indirectly from the last equation. Each of
the integrals is equal to zero for the date t,, to which the osculating
elements belong.
When the magnitude of the perturbations is such that the terms
depending on the squares and products of the masses must be considered, the general equations (59), (63), and (64) will be applied.
The values of the perturbations for the dates preceding that for
which the complete expressions are to be used, will at once indicate
approximate values of 8?w, dr, and z; and with the values
r r -- 9 r d r, w = W -- + w, sin I -_,
the components of the disturbing force will be computed. We compute
also q' from the first of equations (57), and q" from the first of (61);
then, by means of Table XVII., we derive the corresponding values
of log f and logf". The coefficients of 8r in the expressions for
q and q" will be given with sufficient accuracy by means of the
approximate values of or and sin /9, and will not require any further
correction. Then we compute S0r cos, and find the integral
fSor cos H dt;
dsw
and the complete value of dt will be given by (59). The value
of -ct2 will then be given by equation (63). The term r t ( ) will
always be small, and, unless the inclination of the orbit of the disturbed body is large, it may generally be neglected. Whenever it shall
be required, we may put it equal to ( - )' The corrected values
of the differential coefficients being introduced into the table of integration, the exact or very approximate values of 8w, 8r, and z will
be obtained. Should these results, however, differ much from the
corresponding values already assumed, a repetition of the calculation
may become necessary. In this manner, by computing each place
separately, the terms depending on the squares, products, and higher
powers of the disturbing forces may be included in the results. It
will, however, be generally possible to estimate the values of Jw, &r,
VARIATION OF POLAR CO-ORDINATES. 467
and z for two or three intervals in advance to a degree of approximation sufficient for the computation of the forces for these dates.
In order that the quantity o, representing the interval adopted in
the calculation of the perturbations, may not appear in the integration, we should introduce it into the equations as in the case of the
variation of the rectangular co-ordinates. Thus, in the determinadcw
tion of 8w we compute the values of o ot- and since the second
member of the equation contains the integralfS0rcos dt, if we
introduce the factor w2 under the sign of integration, this integral,
omitting the factor w in the formule of integration, will become
wofsor cos 9 dt, as required. The last term of the equation ~i-il be
multiplied by co.
In the case of Or, each term of the equation for must contain
the factor 02. If the second of equations (65) is employed, the first
and third terms of the second member will be multiplied by c'2; but
since the value of S0 is supposed to be already multiplied by o2, the
second term will only be multiplied by o.
The perturbations may be conveniently determined either in units
of the seventh decimal place, or expressed in seconds of arc of a
circle whose radius is unity. If they are to be expressed in seconds,
the factor s- 206264.8 must be introduced so as to preserve the
homogeneity of the several terms, and finally 8r and 4z must be converted into their values in terms of the unit of space.
172. It remains yet to derive convenient formule for the determination of the forces So, R, and Z. For this purpose, it first becomes
necessary to determine the position of the orbit of the disturbing
planet in reference to the fundameital plane adopted, namely, the
plane defined by the osculating elements of the disturbed orbit at the
instant to. Let i' and 2' denote the inclination and the longitude of
the ascending node of the disturbing body with respect to the ecliptic,
and let I denote the inclination of the orbit of the disturbing body
with respect to the fundamental plane. Further, let N denote the
longitude of its ascending node on the same plane measured from the
ascending node of this plane on the ecliptic or from the point whose
longitude is ~0,, and let N' be the angular distance between the ascending node of the orbit of the disturbing body on the ecliptic and
the ascending node on the fundamental plane adopted. Then, from
the spherical triangle formed by the intersection of the plane of the
468 THEORETICAL ASTRONOMY.
ecliptic, the fundamental plane, and the plane of the orbit of the disturbing body with the celestial vault, we have
sin - I sin' (N + N') sin - (' a sin ( +
sin 1Icos (N + N) cos (' - o) sin ( -i),
cos I sin - (N N') sin (' — 0) co (' + i0), (. cos cos(- N') I- cos' (2' ~ o) cos I (i'- o),
from which to find 2V, N', and I.
Let f' denote the heliocentric latitude of the disturbing planet
with respect to the fundamental plane, w' its longitude in this plane
measured from the axis of x, as in the case of w, and ut' the argument of the latitude with respect to this plane. Then, according to
the equations (82),, we have
tan (w' - N) = tan uo' cos I,
tan' = tan Isin (w' - N).
If t' denotes the argument of the latitude of the disturbing planet
with respect to the ecliptic, we have
-' -' U - N'. (68)
This formula will give the value of u,', and then wt and' will be
found from (67). We have, also,
cos Ut' = cos /' cos (w' - N),
which will serve to indicate the quadrant in which w' - N must be
taken.
The relations here derived are evidently applicable to the case in
which the elements of the orbits of the disturbed and disturbing
planets are referred to the equator, the signification of the quantities
involved being properly considered.
The co-ordinates of the disturbing planet in reference to the plane
of the disturbed orbit at the instant t0 as the fundamental plane will
be given by' - r' cos 1 cos w',
y' r' cos 3' sin w', (69)
z' -r' sin f'.
To find the force R, we have
R=X-+ Y + Zr r r
VARIATION OF POLAR CO-ORDINATES. 469
and
( p3n r3 3 )3
( P3(3)
P3 r 3
we get
Rn'l (hr' cos'cosC ('- ) + hr' sinin sin' — ). (71)
The equation
S, r cos= Y-. XXy
gives
X- nSubs2 h r' cose' sin (w' - y w), (72)
from which to find S. Finally, we have
Z=n'k(/ r -- s )ri (73)
from which to find Z.
When we determine the perturbations only with respect to the
first power of the disturbing force, the expressions for R, S0, and Z
become
R = n' (' hr' cos t cos (2- w) ~' ),, - m'k2 h r' cos f' sin (20' - wQ),
Z -m'l2 h r' sin f'.
To compute the distance p, we have
2 = (' - )2 + (y' y)2 + (Z'- Z)2
which gives
p2 r= - _r_ 2r r' cos cos f' cos (w' - w) - 2r r' sin 3 sin', (75)
and, if we neglect terms of the second order, we have
po2 - r2 + r02 - 2ro r' cos i' cos (2' w o). (76)
If we put
cos y — cos. os s f cos (w' - w) + sin f sin i', (77)
we have
P2 r2 +- r2 - 2rr' cos y
- r2 sin2r + (r - r' cos Y);
470 THEORETICAL ASTRONOMY.
and hence we may readily find p from
p sin n - r' sin r,
p co n r - r' cos ry,
the exact value of the angle n, however, not being required.
Introducing r into the expression for R, it becomes
R = m'k' h r' cos r- ) (79)
by means of which 1R may be conveniently determined.
173. When we neglect the terms depending on the squares and
higher powers of the masses in the computation of the perturbations,
the forces R, S, and Z will be computed by means of the equations
(74), po being found from (76) or from (78), when we put
cos r cos f' cos (,w - WV).
But when the terms of the order of the square of the disturbing
force are to be taken into account, the complete equations must be
used. Thus, we find p from (78), S0 from (72), Z from (73), and R
from (71) or (79). The values of Jw, 8r, and z, computed to the
point at which it becomes necessary to consider the terms of the
second order, will enable us at once to estimate the values of the
perturbations for two or three intervals in advance to a degree of
approximation sufficient for the calculation of the forces; and the
values of R, S0, and Z thus found will not require any further correction.
When the places of the disturbing planet are to be derived from
an ephemeris giving the heliocentric longitudes and latitudes, the
values of a' and i' will be obtained from two places separated by a
considerable interval, and then the values of u' will be determined
by means of the first of equations (82), or by means of (85),. When
the inclination i' is very small, it will be sufficient to take
t' -'-' +- s tani2 i' sin 2 (1' - ),
in which s — 206264.8. But when the tables give directly the longitude in the orbit, u' +- 2', by subtracting 2' from each of these
longitudes we obtain the required values of iu.
It should be observed, also, that the exact determination of the
values of the forces requires that the actual disturbed values of r',
w', and &' should be used. The disturbed radius-vector r' will be
VARIATION OF POLAR CO-ORDINATES. 471
given immediately by the tables of the motion of the disturbing
body, but the determination of the actual values of w' and /' requires that we should use the actual values of N', N, and I in the
solution of the equations (68) and (67). Hence the disturbed values
of g' and i' should be used in the determination of these quantities
for each date by means of (66). It will, however, generally be the
case that for a moderate period the variation of 2' and i' may be
neglected; and whenever the variation of either of these has a sensible effect, we may compute new values of N, N', and I from time to
time, by means of which the true values may be readily interpolated
for each date. We may also determine the variations of V, N', and
I arising from the variation of Q' and i', by means of differential
formule. Thus the relations will be similar to those given by the
equations (71)2, so that we have
sin N' siln N'
8 sNr-' i —2g cos N' g - I cos Ii',
sin(' - ) si O I
sin N sin N'
N - sin( )cos N'' — s-i', (80)
sin(2'~-b) sQis
8I = sin N' sin i' s 2' + cosN' ai"',
from which to find c81T, XN, and ~1.
When the perturbations are computed only in reference to the first
power of the mass, the change of 2' and i' may be entirely neglected; but when the perturbations are to be computed for a long
period of time, and the terms depending on the squares and products
of the disturbing forces are to be included, it will be advisable to
take into account the values of 8NV, 8N', and 81, and, using also the
value of u' in the actual orbit of the disturbing body, compute the
actual values of w' and f'.
In the case of several disturbing bodies, the forces will be determined for each of these, and then, instead of R, S0, and Z, in the
formule for the differential coefficients, CR, 2'S, and 2Zwill be used.
174. By means of the values of 8wo, ar, and z, the heliocentric or
the geocentric place of the disturbed planet may be readily found.
Thus, let the positive axis of x be directed to the ascending node of
the osculating orbit at the instant t0 on the plane of the ecliptic;
then, in the undisturbed orbit, we shall have
o o
u denoting the argument of the latitude. Let x,, y,, z, be the co-or
472 THEORETICAL ASTRONOMY.
dinates of the body referred to a system of rectangular co-ordinates
in which the ecliptic is the plane of xy, and in which the positive
axis of x is directed to the vernal equinox. Then we shall have
x, = x cos g o -- y cos i sin g + z- sin i, sin go,
y, x sin g o + y cos i cos ag0 - z sin i, cos a,
z, - y sin i, + z cos,
or, introducing the values of x and y given by (50),
x, =- r cos f cos w cos 0 - r cos / sin w cos i sin + z si i sin, sin
y, - r cos if cos w sin 0o +- r cos i sin w cos i cos -, z sin i, cos a o, (81)
z, - r cos A sin w sin i + z cos i.
Introducing also the auxiliary constants for the ecliptic according to
the equations (94), and (96)i, we obtain
x, -- r cos f sin a sin (A -- w) + z cos a,
y, = r cos i sin b sin (B + w) + z cos b, (82)
z, = r cos i sin i sin w + z cos i,
by means of which the heliocentric co-ordinates in reference to the
ecliptic may be determined.
If the place of the disturbed body is required in reference to the
equator, denoting the heliocentric co-ordinates by x,,, y,, z,,, and the
obliquity of the ecliptic by s, we have
XI,, - X
=y, cos -s z, sin e,
z,, y, sin e + z, cos e.
Substituting for x,, y,, z, their values given by (81), and introducing
the auxiliary constants for the equator, according to the equations
(99), and (101), we get
x,,= r cos f sin a sin (A + w) -]- z cos a,
y,, = r cos f sin b sin (B + w) + z cos b, (83)
z, - r cos f sin c sin ( C+ w) - z cos c.
The combination of the values derived from these equations with the
corresponding values of the co-ordinates of the sun, will give the
required geocentric places of the disturbed body. These equations
are applicable to the case of any fundamental plane, provided that
the auxiliary constants a, A, b, B, &c. are determined with respect
to that plane. In the numerical application of the formulae, the
value of w will be found from
w- = U -+- aw,
VARIATION OF POLAR CO-ORDINATES. 473
Uo being the argument of the latitude for the fundamental osculating
elements, and care must be taken that the proper algebraic sign is
assigned to cos a, cos b, and cos c.
If the values of wr, 20, and io used in the calculation of the perturbations are referred to the ecliptic and mean equinox of the date
t0, and the rectangular co-ordinates of the disturbed body are required
in reference to the ecliptic and mean equinox of the date t0or, the
value of w must be found from
w-o + V0 + a v,
the value of oo referred to the ecliptic of to' being reduced to that of
to", by means of the first of equations (115),. Then 0o and i0 should
be r dfrm e reduced from the ecliptic and mean equinox of t to the ecliptic
and mean equinox of to0 by means of the second and third of the
equations (115),, and, using the values thus found in the calculation
of the auxiliary constants for the ecliptic, the equations.(82) will
give the required values of the heliocentric co-ordinates. If the coordinates referred to the mean equinox and equator of the date to"
are to be determined, the proper corrections having been applied to
o0 and i, the mean obliquity of the ecliptic for this date will be
employed in the determination of the auxiliary constants a, A, &c.
with respect to the equator, and the equations (83) will then give
the required values of the co-ordinates.
If we differentiate the equations (83), we obtain, by reduction,
dxl, dw dr
dx= r cos f sin a cos (A + w) d+ sec f sin a sin (A + w) d
dt dt dt
dz
+ (cos a - tan f sin a sin (A + w)) dt-,
dy,, d d
dyf i= r cos f sin b cos (B + w) d + see f sin b sin (B + w) dr
dt d(t t 84
+ (cos b - tan sin b sin (B +w)) -t,
dz,, dw dr
d = r cos f sin c cos ( C + w) t- + sec f sin c sin ( +- w)
dt (t dt
dz
+(cos c - tan f sin c sin (C +w)) di'
by means of which the components of the velocity of the disturbed
body in directions parallel to the co-ordinate axes may be determined.
d/6x do d23r d'z
The values of -B and -d will be obtained from d2J and d2 by a
single integration, and then we have
single integration, and then we have
474 THEORETICAL ASTRONOMY.
ldw kl/p'(l+rm) daw dr kl/1 - m dar (
dt ro2 + d' _- dt_ eo sinV0 + -t (85)
dw dr
from which to find -,t and d
dt dt
175. EXAMPLE.-In order to illustrate the calculation of the perturbations of r, w, and z, let us take the data given in Art. 166, and
determine these perturbations instead of those of the rectangular coordinates.
In the first place, we derive from the tables of the motion of
Jupiter the values' = 98~ 58' 22".7, i' = 1 8' 40".5,
which refer to the ecliptic and mean equinox of 1860.0. We find,
also, from the data given by the tables the values of u' measured
from the ecliptic of 1860.0. Then, by means of the formulke (66),
using the values of g0 and i0 given in Art. 166, we derive
N- 194~ 0' 49".9, N' 301~ 38' 31".7, I- 5~ 9' 56".4.
The value of u0' is given by equation (68), and then w' and fi are
found from the equations (67). Thus we have
Berlin Mean Time. log10^w0 WO= t0 log wo'
1863 Dec. 12.0, 0.294084 192~ 4/ 24.5 0.73425 140 18'54".6 -0~ 1'38/.1
1864 Jan. 21.0, 0.294837 207 40 52.2 0.73368 17 21 44.2 0'8 9.1
March 1.0, 0.300674 223 3 5.9 0.73305 20 25 5.2 0 34 39.9
April 10.0, 0.310864 237 51 38.3 0.73237 23 28 59.8 0 51 7.6
May 20.0, 0.324298 251 52 47.9 0.73164 26 33 32.1 1 7 29.7
June 29.0, 0.339745 264 59 30.0 0.73086 29 38 44.8 1 23 43.5
Aug. 8.0, 0.356101 277 10 24.6 0.73003 32 44 41.2 1 39 46.3
Sept. 17.0, 0.372469 288 28 4.1 0.72915 35 51 24.6 1 55 35.2
Oct. 27.0, 0.388214 298 57 16.3 0.72823 38" 58 57.5 2 11 7.5
Dec. 6.0, 0.402894 308 43 48.7 0.72726 42 7 23.3 2 26 20.3
1865 Jan. 15.0, 0.416240 317 53 39.1 0.72625 45 16 43.9 -2 41 10.6
The values of po may be found from (76) or (78)as already given in
Art. 166.
The forces R, So, and Z may now be determined by means of the
equations (74), h being found from (70), and if we introduce the
factor 02 for convenience in the integration, as already explained, we
obtain the following results:
Date. o2R (S0 2Z wSodt
1863 Dec. 12.0, + 1".4608 + 0".1476 + 0".0009 + 0".0282
1864 Jan. 21.0, + 1.4223 - 0.6757 -+0.0101 - 0.2361
NUMERICAL EXAMPLE. 475
Date. So2R 21.0'o (2 S8dt
1864 March 1.0, + 1".2616 - 1".4512 + 0".0190 - 1".3060
April 10.0, 1.0018 2.1226 0.0273 3.1035
May 20.0, 0.6760 2.6473 0.0347 5.5020
June 29.0, + 0.3179 2.9988 0.0406 8.3402
Aug. 8.0, -0.0452 3.1650 0.0449 11.4378
Sept. 17.0, 6.3944 3.1437 0.0470 14.6076
Oct. 27.0, 0.7180 2.9392 0.0466 17.6640
Dec. 6.0, 1.0097 2.5586 0.0432 20.4273
1865 Jan. 15.0, — 1.2674 - 2.0081 + 0.0362 - 22.7245
The integral oiSodrodt is obtained from the successive values of o2Soo0
by means of the formula (32).
Next we compute the values of the differential coefficients by
means of the formult (65). For the dates 1863 Dec. 12.0 and 1864
Jan. 21.0 we may first assume r — 0, and, by a preliminary integration, having thus derived very approximate values of 8r for these
dates, the values of d~2 will be recomputed. Then, commencing
anew the table of integration, we may at once derive an approximate
value of Jr for the date March 1.0 with which the last term of the
expression for ~-dt may be computed. Continuing this indirect process, as already illustrated in the case of the perturbations of the rectangular co-ordinates, we obtain the required values of the second
d'2
differential coefficient. In a similar manner, the values of ct will
dsw
be obtained. The values of -d will then be given directly by means
of the first of equations (65); and the final integration will furnish
the perturbations required. Thus we derive the following results:dsw _d___ d,
Date. 2dd d2 O2 d~z w o z
dct dt2 dt"
1863 Dec. 12.0, -0".0423 +1".4509 +0".0009 -0".00 +0".18 +0".00
1864 Jan. 21.0, 0.1086 1.3405 0.0101 0.02 0.17 0.00
Mar. 1.0, 0.7162 +0.7829 0.0183 0.40 1.47 0.01
Apr. 10.0, 1.6114-0.0455 0.0251 1.55 3.53 0.04
May 20.0, 2.4795 0.9344 0.0300 3.61 5.54 0.09
June 29.0, 3.0807 1.7333 0.0326 6.42 6.62 0.18
Aug. 8.0, 3.2971 2.3752 0.0331 9.64 5.98 0.29
Sept. 17.0, 3.1080 2.8533 0.0311 12.88 +2.98 0.44
Oct. 27.0, —2.5425-3.1872+0.0265-15.73 -2.86 +0.62
476 THEORETICAL ASTRONOMY.
Date. do 2 d2r 2 d w
dt dt2 dt2
1864 Dec. 6.0, — 1".6443 — 3".4009 -0".0190 — 17".85 — 11".88 +0".83
1865 Jan. 15.0,-0.4511 -3.5334 +0.0079 -18.92-24.29 +1.05
It has already been found that, during the period included by these
results, the perturbations arising from the squares and products of
the disturbing forces are insensible, and hence the application of the
complete equations for the forces and for the differential coefficients
is not required. The equations (83) will give, by means of the
results for w u0 + -w, r =- r + Jr, and z, the values of the heliocentric co-ordinates of the disturbed body, and the combination of
these with the co-ordinates of the sun will give the geocentric place.
When we neglect terms of the second order, we have, according to
the equations (84),
ax,, x= cot (A + w) aw + - ar + z cos a,
ro
aY,, y- cot (B ~ w) aw + Jr + z cos b, (86)
r0
8z,, = z0 cot (C + w) 1W +-~r 0 + cos,
the heliocentric co-ordinates x0, Y0, z being referred to the same fundamental plane as the auxiliary constants, a, b, A, &c. Thus, in the
case of Eurynome, to find the perturbations of the rectangular co-ordinates, referred to the ecliptic and mean equinox of 1860.0, from
1864 Jan. 1.0 to 1865 Jan. 15.0, we have
A 2960 34/ 37/.5, B -= 2060 43/ 34/.4, C= 0,
log cos a 8.557354,, log cos b =8.856746, log cos log cos o = 9.998590,
log x0 0.399807, log o 9.838709, log z == 9.148170,,
w == wo + wt =- 317~ 53/ 20".2,
and hence, by means of (86), we derive
ax, + 36".559, ay, = + 41".083, az, - 0".588.
If we express these in parts of the unit of space, and in units of the
seventh decimal place, we obtain
ax, + 1772.4, ay, = + 1991.8, 2, - 28.5,
agreeing with the results already obtained by the method of the variation of rectangular co-ordinates, namely,
3ax2, - 1772.6, ay, + 1992.3, z,=- 28.2.
CHANGE OF THE OSCULATING ELEMENTS. 477
176. By using the complete formule, the perturbations of r, w,
and z may be computed with respect to all powers of the disturbing
force, and for a long series of years, using constantly the same fundamental osculating elements. But even when these elements are so
accurate as not to require correction, on account of the effect of the
large perturbations of long period upon the values of 8w and or, the
numerical values of the perturbations will at length be such that a
change of the osculating elements becomes desirable, so that the
integration may again commence with the value zero for the variation
of each of the co-ordinates. This change from one system of elements to another system may be readily effected when the values of
the perturbations are known. Thus, having found the disturbed
values of r, w, and z, we have
dv' dw2 di3 lhkp (1 + m)
-r —- cos' +
dt2 dt'2 +dt2 r4
p being the semi-parameter of the instantaneous orbit of the disturbed
body. In the undisturbed orbit we have
dv, kvpo (1 + m)
o- dt 2
and hence we derive
P ~2oro4 dt,
dv
- go 214dt
Substituting for dt the value above given, there results
/ 21 1 dco \ 1 d1 —
P -Po-,4 Cos + * dtj ). 2 d' (87),N — o dt rok2d
"Q \ \ 90 d Io d'
by means of which p may be determined. To find d-, we have
dfi 1 dz tan i dr
dt r cos dt r dt(88)
We have, also,
dr kl/l +m. kl/l -m. d8r
d- -- _ - e sin v - /- e in o- d at
and if we put
- I+ a, r= ~p dar (89)
0o Vl+w dt
478 THEORETICAL ASTRONOMY.
this equation becomes
e sin v eo sin vo -- aeo sin o v- r. (90)
We have, further,
e cos v - - 1,
and, putting
P ro 1 -, (91)
Po r
we obtain
e cos v - eo cos v + f.
This equation, combined with (90), gives
e sin (v - vo) - ae, sin v0 cos vo + r cos vO-0 f- sin vo
~ (92)
e cos (v- v,) eo + ece sin2 v + r sin vo + Po cos v,,
by means of which the values of e and v may be found, those of the
auxiliaries o, 9, r, being found from (89) and (91). Then we have
e - sin m, a -=p se2 s,
p. =- +, tan E=- tan (45~ -') tan v,
az
M-1 E-e sin E,
by means of which %, a, j,, and l1 may be determined. In the case
of orbits of great eccentricity, we find the perihelion distance from
P
q — -b+e'
and the time of perihelion passage will be derived from e and v by
means of Table IX. or Table X.
It remains yet to determine the values of g, i, and co or r. Let
00 denote the longitude of the ascending node of the instantaneous
orbit on the plane of the osculating orbit, defined by 92 and i0, measured from the origin of w, and let ^o denote its inclination to this
plane. Then we have
tan %r sin (w - 0o) = tan f,
di c^od~ (93)
tan, cos (w - O) d sec2 ) (
and hence
CHANGE OF THE OSCULATING ELEMENTS. 479
gd- dt
dt
by means of which 00 may be found. The quadrant in which o0 is
situated is determined by the condition that sin(w - 00) and tan k
must have the same sign. The value of ^ will be found from the
first or the second of equations (93).
If we denote by: the argument of the latitude of the disturbed
body with respect to the adopted fundamental plane, we have
tan tan (w -- (95)
cOS )o
and the angle ( must be taken in the same quadrant as w - 00.
Then, from the spherical triangle formed by the intersection of the
planes of the ecliptic and instantaneous orbit of the disturbed body,
and the fundamental plane, with the celestial vault, we derive
cos 1 i sin (u- ( ) + 1 (- 2o)) Sill O cos (i -0),
cos M i cos( (- M0 + (-n- ~o)) - cos: co
(96)
sin i sin (i C)-. ) ) sill sin o) -
sin - i cos ( (ut ) - 1 (2 - go)) cos 0so sin j(i + o).
These equations will furnish the values of i, u -, and 2 - 0, and
hence, since, and Q2 are given, those of a and a. The value of v
having been already fonnd, we have, finally,
V) - u
7 ==It- V + a,
and the elements are completely determined. These elements will
be referred to the ecliptic and mean equinox to which g20 and i0 are
referred, and they may be reduced to the equinox and ecliptic of any
other date by means of the formule which have already been given.
The elements of the instantaneous orbit of the disturbed body may
also be determined by first computing the values of x,,, y,, z,,, in
reference to the fundamental plane to which a and i are to be referred, by means of the equations (83), and also those of dt', dt' d
by means of (85) and (84), and then determining the elements from
the co-ordinates and velocities, as already explained.
It should be observed that when the factor (02, or the square of the
480 THEORETICAL ASTRONOMY.
adopted interval, is introduced into the expressions for the forces and
differential coefficients, the first integrals will be
dsr dcw dz
dt' dt' dt'
and that when these quantities are expressed in seconds of arc, they
must be converted into their values in parts of the unit of space
whenever they are to be combined with quantities which are not expressed in seconds. In other words, the homogeneity of the several
terms must be carefully attended to in the actual application of the
formule.
When the elements which correspond to given values of the perturbations have been determined, if we compute the heliocentric
longitude and latitude of the body for the instant to which the elements belong, the results should agree with those obtained by computing the heliocentric place from the fundamental.osculating elements and adding the perturbations.
177. The computation of the indirect terms when the perturbations of the co-ordinates r, w, and z are determined, is effected with
greater facility than in the case of the rectangular co-ordinates,
although the final results are not so convenient for the calculation of
an ephemeris for the comparison of observations. This indirect calculation, which, when the perturbations of any system of three coordinates are to be computed, cannot in any case be avoided without
imlpairing the accuracy of the results, may be further simplified by
determining, in a peculiar form, the perturbations of the mean
anomaly, the radius-vector, and the co-ordinate z perpendicular to the
fundamental plane adopted.
Let the motion of the disturbed body be, at each instant, referred
to the plane of its instantaneous orbit; then we shall have j= 0,
and the equations (51) and (54) become
r2 = r dt + ol + mk o + ),
(97)
dUr dW2+ k' (1 + (97)
dt2 dt2 r2
in which R denotes the component of the disturbing force in the
direction of the disturbed radius-vector, and S the component in the
plane of the disturbed orbit and perpendicular to the disturbed radiusvector, being positive in the direction of the motion. The effect of
VARIATION OF POLAR CO-ORDINATES. 481
the components R and S is to vary the form of the orbit and the
angular distance of the perihelion from the node. If we denote by
Z the component of the disturbing force perpendicular to the plane
of the instantaneous orbit, the effect of this will be to change the
position of the plane of the orbit, and hence to vary the elements
which depend on the position of this plane.
Let us take a fixed line in the plane of the instantaneous orbit,
and suppose it to be directed from the centre of the sun to a point
whose angular distance back from the place of the ascending node is
a, and let the value of a be so taken that, so long as the position of
the plane of the orbit is unchanged, we shall have
The line thus taken in the plane of the orbit may be regarded as
fixed during all changes in the position of this plane. Let Z denote
the angle between this fixed line and the semi-transverse axis; then
will
X -- + - a, (98)
and when the position of the plane of the orbit is unchanged, we have
X -- 7r.
But if, on account of the action of the component Z, the position of
the plane of the orbit is changed, we have, according to the equations
(72)2, the relations
da -cosi d2,
d dz - cosi dg, (99)
dn = dX + (1 -cos i) d g =- d 2 + 2 sin'2 i d.
We have, further,,I = v + X -', (100)
v being the true anomaly in the instantaneous orbit.
The two components of the disturbing force which act in the plane
of the disturbed orbit will only vary X and the elements which determine the dimensions of the conic section. We have, therefore, in the
case of the osculating elements, for the instant t0,
Xo -- o + -o - o.
Let us now suppose 2 to denote the true longitude in the orbit, so
that we have
=v + v7C +V + +2,
31
482 THEORETICAL ASTRONOMY.
or
= -v +X -- ( -- g); (101)
then, since X is equal to t when the position of the plane of the orbit
is unchanged, it follows that a- a represents the variation of the
true longitude in the orbit arising from the action of the component
Z of the disturbing force. The elements may refer to the ecliptic or
the equator, or to any other fundamental plane which may be adopted.
178. For the instant t we have, in the case of the disturbed motion,
the following relations:E- e sin E= M-~ ~ (t - to),
r cos v - a cos E- ae,
r sin v - at/l - e2 sin E,
Let us first consider only the perturbations arising from the action of
the two components of the disturbing force in the plane of the disturbed orbit, and let us put, =v +. (103)
Further, let MH +, (t - t) + AM be the mean anomaly which, by
means of a system of equations identical in form with the preceding,
but in which the values of a., e, X0 are used instead of the instantaneous values a, e, and Z, gives the same longitude A,, so that we have
E, - e, sin E, = Mo + L-o (t-to) -+ SM,
r, cos, ao cos E, -- a(e,
(104)
r, sin v, - ao/1 - eo sin E,
A, ==, + Xo V, + - o.
If, therefore, we determine the value of 8M so as to satisfy the condition that i, =v + X, the disturbed value of the true longitude in
the orbit, neglecting the effect of the component Z of the disturbing
force, will be known. The value of r, will generally differ from that
of the disturbed radius-vector r, and hence it becomes necessary to
introduce another variable in order to consider completely the effect
of the components R and S. Thus, we may put
r r, (1 + (), (105)
and v will always be a very small quantity. When IM and v have
been found, the effect of the disturbing force perpendicular to the
plane of the instantaneous orbit may be considered, and thus the
complete perturbations will be obtained.
VARIATION OF CO-ORDINATES. 483
dw
In the equations (97), r2~ dt expresses the areal velocity in the instantaneous orbit, and it is evident that, since the true anomaly is not
affected by the force Z perpendicular to the plane of the actual orbit,
dr2 must also represent this areal velocity, and hence the equations
idt
become
bdedt -Sfr dt + kl/pO ( + M)
(10o6)
d2r ( dv,\2 k2(i +M) ()1
df r -dt / + -- R.
179. If we differentiate each of the equations (104), we get
dE d__
(1- e Cos E,) t,
dr,. dv, dE,
cos v, -- r, sin v, d- - a, sin E, dE
dt d t (107)
dr, vE,, / cs dE,
sin - + r, cos vI -b a 1Ie
dt, dv,
dt dt
From the second and the third of these equations we easily derive
drv, edE,
r, = (a/ - eo r, sin v, cos E,- ar, cos v, sin E,) d.
dE,
Substituting in this the values of r, sin,, r, cos v,, and dt"- and reducing, we get
dr,.- sin L +- -i
or
dr, l/l + m., doM\
dt /- eosi n v, 1 dt
dr,
From the same equations, eliminating -t-, we get
dv, __ 77dE
rd -- (al/ - eo2 r, cos v, cos, + aor, sin v, sin E,) dE,
which reduces to
r -t o (1 + ) a (109)
---- P
484 THEORETICAL ASTRONOMY.
or
2 dvf + /~^~~^ ^, ^~/^, I d8M\
r k Po m (1 + M) (m + V2\ ( 0 dt )
Combining this with the first of equations (106), we get
dt (1 +) 1 + 1((~ ( )S/rd (1(10)
from which i3f may be found as soon as v is known.
The equation (105) gives
dr dr, dv
(I + V) +
dt dtc
d2r drr cr dv d'V (111)
dt2 = (1 +.) 2r, +r 2 dr r
dt2 dt9,- dt dt r,-dt.
Differentiating equation (108) and substituting for dt its value
already found, we obtain
d2r, k2 (1+-m) e, cos, 1 daM\ k/l1 +me sinv, d2aM
dt2 r,2 \' dt d l J /, dt2
and the last of the preceding equations becomes
d2r d2V k+ (1 + m) eo cosV ( 1 d+M\ 2
^ dt-_'dt2 r,2 o\+ dt~)
kl/+m. (1 +v d2aM dv 2 dv d&M
+ d 2d
+ /- e0 in O dt2, dt
The equation (110) gives
1 d2sM 2 dv 2 dy 1 r
o' dc + (1 + v)' dt + ( + ) dt k' V1 (1 +) J
1 Sr
(1 +)* kV/(1+m)'
which is easily reduced to
1 + d2d M dv 2 dv d&M 1 Sr
o dt dt+ 2T dt+ dt 1+ v k /p;o( + nt)
and hence we derive
d2r d2v k2 (1 + M) ecos (1+ 1\ dMt ) pesinv,
The equation (109) gives
The equation (109) gives
VARIATION OF CO-ORDINATES. 485
dv,) 2 k2p0(1+m) (1~1 2^
d ~t r?~3 dt
and, since
r,- 1 - ecosv,,
r,
this becomes
( dv, )2 2 (1 + m) (+)(1+ di
\ ( + ) r (1 + (113)
+ +r,2 [cosL+(+ -+. dt
Combining equations (112) and (113) with the second of equations
(106), we get
d2i 1 + k 2(1 ( + m) ( + 1 d_ _ i 2
eo sin v, k2 (1 + m) (I + ) (1
Po r
From (110) we derive
(1+)4(1+1. d) t+_ 2 fSrdt
(I+~)'j\ d+ dt
Po (1 + m )
+ ( m srdt):
and the preceding equation becomes
dt2 = By + 2' + 1 fSr dteo sin v, S
w _h i t 3 e comp(le te) e r t I d m
which is the complete expression for the determination of v.
180. It remains now to consider the effect of the component of the
disturbing force which is perpendicular to the plane of the disturbed
orbit. Let x,, y, z, denote the co-ordinates of the body referred to
the fundamental plane to which the elements belong, and x, y the
co-ordinates in the plane of the instantaneous orbit. Further, let a
denote the cosine of the angle which the axis of x makes with that
of %, and G the cosine of the angle which the axis of y makes with
that of y, and we shall have
z, -= ax +-y (116)
If the position of the plane of the orbit remained unchanged, these
486 THEORETICAL ASTRONOMY.
cosines a and: would be constant; but on account of the action of
the force perpendicular to the plane of the orbit, these quantities are
functions of the time. Now, the co-ordinate z, is subject to two distinct variations: if the elements remain constant, it varies with the
time; and, in the case of the disturbed orbit, it is also subject to a
variation arising from the change of the elements themselves. We
shall, therefore, have
indd _w I d \p "l
dt dt )' + [ dt
in which ( dt ) expresses the velocity resulting from the constant
elements, and d j that part of the actual velocity which is due
to the change of the elements by the action of the disturbing force.
But during the element of time dt the elements may be regarded as
dz,
constant, and hence the velocity dt in a direction parallel to the
axis of z, may be regarded as constant during the same time, and as
receiving an increment only at the end of this instant. Hence we
shall have
dz, \ r_ -dF 0]
dt dt ) L dtJ
Differentiating equation (116), regarding a and. as constant, we
get
dz, dz dxdv dy
(dIz) dId d +iy (117)
{^k~~dt dt dt
and differentiating the same equation, regarding x and y as constant,
we get
-- x d- y - 0o. (118)
LdtJ dt Y-dt (8)
Differentiating equation (117), regarding all the quantities involved
as variable, the result is
d2z, du dx d1i dy d Qx+ d2y
dt2 dt' dt +dt' dt dt + dt2 ( 9)'
Now, we have
Z,. =- X+ ftY- Zcosi, (120)
in which Z, denotes the component of the disturbing force parallel
to the axis of z,, and i the inclination of the instantaneous orbit to
VARIATION OF CO-ORDINATES. 487
the fundamental plane. Substituting for X and Y their values given
by the equations (1), and reducing by means of (116), we obtain
d X d~ 2
z.-~ y+ 2V+ (1 n+m)~ +Zcosi,
Z c dt2 dt2,3
or
d d x, Ad d2y
dt2 dt2 + dt + Zco
Comparing this with (119), there results
da dx df Z dy_ os (1
_dt-' -d t- + =it-.dt Z COS i. (121)
cidtct di dt
181. The equation (120) gives
dlz, k'(1 + m)
d2Z - k2 ( + M)' z,+Zcosi + ax+ fXy. (122)
The component of the disturbing force perpendicular to the plane of
the disturbed orbit does not affect the radius-vector r; and hence,
when we neglect the effect of this component, and consider only the
components R and S which act in the plane of the orbit, we have
d -o, k(1+ + ) ~ +Pop Y1 (123)
in which z0 denotes the value of z, obtained when we put Z= 0.
Let us now denote by Jz, that part of the change in the value of z,
which arises from the action of the force perpendicular to the plane
of the disturbed orbit, so that we shall have
zr-zo + azt, a = U0 + 8a,? - 0 -+- S.
Substituting these in equation (122) and then subtracting equation
(123) from the result, we get
d28 - - r( + ) az + Zcos i + Xa + YAf (124)
The equations (116) and (117) give
dci, dx dy
a%, = -s7$au + Ydfl am +.
dt rdt -
If we eliminate 89 between these equations, there results
\ dt dt -- dt, dt'
488 THEORETICAL ASTRONOMY.
and since the flctor of &a in this equation is double the areal velocity
in the disturbed orbit, we have
1 (dy d~z
l/p (1- ) - ( dt z Y t (125)
k I-i +(l m)\dt I dt)
Eliminating &oa from the same equations, we obtain, in a similar
manner,.I (/ dsd,d dx d (126)
k/p (-m) dt dt (126)
Substituting these values in equation (124), it becomes
d2__ k (I + m
d?_2, 3 k8 1q-m z', + Z cos:/
dt2 r8
d;l~.. — ( ( X d _ )dr (Y(127)
" kp(l1+) \m\i dt dt )' - dt )
If we introduce the components R and S of the disturbing force, we
have
X -R - s Y RY +S,
r r r r
and hence
dy dx R kl/ ) dr
d: d- r dt
Yx Xy - Sr.
Therefore the equation (127) becomes
dz;~ — k2 (1 -), -Z +Zcosi
dX~~t'2~ ~~- ~ r( ), O128)
R S dr Sr dz ( )
Jk,+ kl/p(1~m)d
(r kl/Vv (I +) rn )dt /p (1 +2 ) dt
We have, further,
dr = ( + + r, d
dt dt dt
which, by means of the equations (108) and (109), gives
dr e( sin, -r2d + r, d p ) esin v, r,
Substituting this value in the equation (128), we obtain
VARIATION OF CO-ORDINATES. 489
d2_z, k2 (1 I + e) *_ IR eo sin v, \A,
dt2 r3 Po / + v
(129)
Sr / daz, e z, dv \
ki /p(1 + ) \dt 1 + d'
which is the complete expression for the determination of Jz,.
182. The equations (110), (115), and (129) determine the complete
perturbations of the disturbed body. The value of v must first be
obtained by an indirect process from the equation (115), and then JlIf
is given directly by means of (110). The value of 8z will also be
determined by an indirect process by means of (129).
In order to obtain the expressions for the forces R, S, and Z, let w
denote the longitude of the disturbed body measured in the plane of
the instantaneous orbit from its ascending node on the fundamental
plane to which g and i are referred, it being the argument of the
latitude in the case of the disturbed motion. Let w' denote the longitude of the disturbing body measured from the same origin and in
the plane of the orbit of the disturbed body, and let 9' denote its
latitude in reference to this plane. Finally, let N, N', I, and 0'
have the same signification in reference to the plane of the instantaneous orbit that they have in reference to the plane of the undisturbed
orbit in the case of the equations (66). Then we shall have
sin Isin - (N + N') sin 4 (' - g ) sin (i' + i),
sin -1Icos (N — N') - (' - sin (i' -i),
cos I sin (N- N') -sin.2 (2'-2 (130)
cos —cos-cos( N')cos) cos ~ (it- ),
from which to determine N, N', and I. We have, also,
aO' i' - N',
tan (w' - N) t tan I' cos 1, (131)
tan 13' - tan Isin (w' - N),
from which to find w' and i', u' being the argument of the latitude
of the disturbing body in reference to the plane to which g and i
are referred.
Since, when the fYotion of the disturbed body is referred to the
plane of its instantaneous orbit, ji- 0, the equations (71), (72), and
(73) become
R m' k2( h r' cos' cos (w' w)- ),
(132)
=- m'k2h r' cos i' sin (w' - w),
Z = n'k2 h r' sin',
490 THEORETICAL ASTRONOMY.
by means of which the required components of the disturbing force
may be found, the value of h being given by
1 1
P3 3'
To find p, we have
p2 _ r2 + r2 _ 2rr' cos' cos (w' - w), (133)
or, putting
cos ry = cos b' cos (w'- w),
the equations
p sin n r' sill r,
pcos n r — r' cos y(. 1
The values of r' and u' for the actual places of the disturbing
body will be given by the tables of its motion, and the actual values
of' and i' will also be obtained by means of the tables. The determination of the actual values of r and w requires that the perturbations shall be known. Thus, when 8IM and v have been found,
we compute, by means of the mean anomaly MH + - (t - to) -4 8Ji
and the elements a, e0, the values of v, and r,. Then, since
v +-X =v, + T, we have, according to (100),
w -= v, + -0 o-. (135)
We have, also,
r=(1 +) r,.
In the case of the fundamental osculating elements, we have
^o =_ S o)y
which may be used as an approximate value of a; but the complete
determination of w requires that a 20 + -8a shall also be determined. The exact determination of the forces also requires that the
actual values of 2 and i as well as those of a' and i', shall be used
in the determination of N, N', and I for each instant. When these
have been found, it will be sufficient to compute the actual values of
N, N', and I at intervals during the entire period for which the perturbations are required, and to interpolate their values for the intermediate dates. The variations of these quantities arising from the
variations of S, i, g2', and i' may also be determined by means of
differential formulae. Thus, from the differential relations of the
parts of the spherical triangle from which the equations (130) are
derived, we easily find
VARIATION OF CO-ORDINATES. 491
sini sinn' sin N
dN' c s osN d(P2' - 2) - cos Idi' + sTdi,
sin I sin I sin I
d cos N' di'- cos NdI + sin - sin iNd ((' - ).
When i and I are very small, it will be better to use
sin i sin N' sin i' sin N
sin sin(2' - g )' sin I sin(g'- g )' 137)
in finding the numerical values of these coefficients. By means of
these formulhe we may derive the values of AN, AN', and I1 corresponding to given values of OQ, bi, 2b', and ji'. The formulae
by means of which ~(, 8n2, and Si may be obtained directly, will be
presently considered.
The results for ON, 6N', and AI being applied to the quantities to
which they belong, we may compute the actual values of w' and I'.
The value of r will be found from the given value of v, and that of
w will be given by means of equation (135). Then, by means of
the formulie (132), the forces R, S, and Zwill be obtained. The
perturbations will first be computed in reference only to terms depending on the first power of the disturbing force, and, whenever it
becomes necessary to consider the terms of the second order, the
results already obtained will* enable us to estimate the values of the
perturbations for two or more intervals in advance with sufficient
accuracy for the determination of the three required components of
the disturbing force; and when there are two or more disturbing
bodies to be considered, the forces for each of these may be computed
at once, and the values of each component for the several disturbing
bodies may be united into a single sum, thus using 2R, IS, and 2Z
in place of?, X, and Z respectively. The approximate values of the
perturbations will also facilitate the indirect calculation in the determination of the complete values of the required differential coefficients.
183. When only the perturbations due to the first power of the
disturbing force are required, the osculating elements 20 and i0 will
be used in finding N, N', and I, and r0, w0 will be used instead of r
and w in the calculation of the values of.R,, and Z. The equations
for. the determination of the perturbations 8M, v, and Sz,, neglecting
terms of the second order, are, according to the equations (110),
(115), and (129), the following: —
492 THEORETICAL ASTRONOMY.
d M 1 f dt-2
dt lo k/po (1+ ) fSro t- 2 (138)
d2P R1 2k2 (1 + m) 1 eo sin v k2 (1 -)
dt2 rO ro3 kl/0o (l + m) eS Po _r2
d2_, k2 (1+ m)
d Z cos it -- s,.
The value of v is first found by integration from the results given
by the second of these equations, and then lM is found from the first
equation. Finally, 8z, is found by means of the last equation. The
integrals are in each case equal to zero for the dates to which the
fundamental osculating elements belong, and the process of integration is analogous, in all respects, to that already illustrated in the
case of the variation of the rectangular co-ordinates. It will be obd2y
served, however, that the expression for - involves only one indirect term, the coefficient of which is small, and the same is true in
d2J, d1IM
the case of dt', while d is given directly. When the perturbations have been found for a few dates, the values for the following
date can be estimated so closely that a repetition of the calculation
will rarely or never be required; and the actual value of r may be
used instead of the approximate value r0 in these expressions for the
differential coefficients. Neglecting terms of the second order, we
have
log r - log r, + 2ov,
wherein 20 denotes the modulus of the system of logarithms. We
may also use v, instead of v.; but in this case, since r, and v, depend
on AM, only the quantities required for two or three places may be
computed in advance of the integration.
A comparison of the equations (138) with the complete equations
(110), (115), and (129) shows that, if the values of i' and w' are
known to a sufficient degree of approximation, we may, with very
little additional labor, consider the terms depending on the squares
and higher powers of the masses. It will, however, appear from
what follows, that when we consider the perturbations due to the
higher powers of the disturbing forces, the consideration of the effect
of the variation of z, in the determination of the heliocentric place
of the disturbed body, becomes much more difficult than when the
terms of the second order are neglected; and hence it will be found
advisable to determine new osculating elements whenever the consideration of these terms becomes troublesome.
VARIATION, OF CO-ORDINATES. 493
The results may be conveniently expressed in seconds of arc, and
afterwards v and Az, may be converted into their values expressed in
units of the seventh decimal place, or, giving proper attention to the
homogeneity of the several terms of the equations, in the numerical
operations, 86Al may be expressed in seconds of arc, while v and 8z,
are obtained directly in units of the seventh decimal place. It will
be advisable, also, to introduce the interval io into the formulae in
such a manner that this quantity may be omitted in the case of the
formulae of integration.
184. In the case of orbits of great eccentricity, the mean anomaly
and the mean daily motion cannot be conveniently used in the numerical application of the formule. Instead of these we must
employ the time of perihelion passage and the elements q and e.
Thus, let T0 be the time of perihelion passage for the osculating elements for the date to, and let To -+- T be the time of perihelion passage to be used in the formula in the place of To and in connection
with the elements q, and e0 in the determination of the values of r,
and v,, so that we have
V + X -, + -.
In the case of parabolic motion we have, neglecting the mass of the
disturbed body,
(t - (To + T)) = tan -v, + I tan' v,, (139)
the solution of which to find v, is effected by means of Table VI. as
already explained. To find r,, we have
r, q0 see2 "V,.
For the other cases in which the elements M. and cannot be employed, the solution must be effected by means of Table IX. or Table
X. Thus, when Table IX. is used, we compute M from
M= (t-(To +T)) 1 -4 e
o 2
wherein log Co = 9.9601277, and with this as the argument we derive
from Table VI. the corresponding value of V. Then, having found
i 1-+ e' by means of Table IX. we derive the coefficients required
in the equation
v, = V + A (1OOi) + B (100i)2 + C(100i)3, (140)
494 THEORETICAL ASTRONOMY.
from which v, will be determined. Finally, r, will be found from
r -(1 + eo) (141)
1 +1 60 COS V,
When Table X. is used, we proceed as explained in Art. 41, using
the elements T — To + T, q0, and eo, and thus we obtain the required
values of v, and r,.
It is evident, therefore, that, for the determination of the perturbations, only the formula for finding the value of 8M requires modification in the case of orbits of great eccentricity, and this modification is easily effected. The expression
Mo + l, (t-to) + aM- M,
gives
_o (to - To) + Io (t - to) + aM= [o (t - ( + S T)),
or, simply,
sMl== - T,
and the equation (110) becomes
d 1~ 1 1 1 (Sr dt, (142)
dt- (1 + v)2 (1+2 k'po (l+-m)
by means of which the value 8T required in the solution of the equations for r, and v, may be found.
If we denote by t, the time for which the true anomaly and the
radius-vector computed by means of the fundamental osculating elements have the values which have been designated byv, and r,, respectively, we have
+1 dM dt,
1 o C -* dt
and the equation (110) becomes
it, 1 2+ 1 1 S dtl (143)
dt - (1 +v) (1 + v)2 k/i'o (1-+m)
or, putting t, = t + At,
d rt I 1 - f r
dt (l+ 1 (+ (1+) (Sr ldt. (144)
dt' (1 - +) 2 (1 - )2 kV'po ( 1f- d)
If we determine 8t by means of this equation, the values of the
radius-vector and true anomaly will be found for the time t+ 8t
instead of t, according to the methods for the different conic sections,
VARIATION OF CO-ORDINATES. 495
using the fundamental osculating elements. The results thus obtained
are the required values of r, and v, respectively.
185. When the values of the perturbations v, az,, and OM, JT, or
8t have been determined, it remains to find the place of the disturbed
body. The heliocentric longitude and latitude will be given by
cos b cos (I - 2) cos ( - g),
cos b sin (I - ) -sin ( - g) cos i,
sin b - sin (A -- ) sin i,
or, since -=, - + g,
cos b cos (I - ) - cos (, -a),
cos b sin (I - ) = sin (A, - ) cos i, (145)
sin b sin (A, - a) sin i,
in which,= v, -+ r. If we multiply the first of these equations
by cos ( - h), and the second by -sin ( - h), in which h may
have any value whatever, and add the results; then multiply the first
by'sin(~ - h), and the second by cos (I - h), and add, we get
cos b cos (l-h)-=cos (A, —) cos ( 2 -h)-sin (A,-A) sin ( -h) cos i,
cos b sin ( —h)-cos (A, —) sin ( -h) +sin (A, —o) cos (~ -h) cos i
sill b =in (2, —r) sin i.
But, since 2,- a (2, - 0) - ( - o), these equations may be
written
cos b oss (- h)
-cos ( ) os s( (a- 20 o)os s ( — h) +sin (6 — o) sin ( — h) cos i)
+sin (A,- ) (sin ( — o)cos( 2 -h)-cos ( — o) sin (Q -h) cos ),
cos b sin ( — ) (146)
=cos (A,-s o) (cos (a- Ao) sin (2 -h)-sin ((- oC) cos ( 2 -h) cosi)
+sin (,- o) (sin (a-, o) sin (Q2 -h)+cos (o-, o) cos (g -h) cos i),
sinb —sin(A, —,) cos (a — ) sin i-cos (,,-' ) sin (6-g o) sin i.
Let us now conceive a spherical triangle to be formed, of which two
of the sides are --- o and a - h, respectively, and let the angle
included by these sides be i. Since h is entirely arbitrary, we may
assign to it a value such that the other angle adjacent to the side
a - 2, will be equal to io. Let the third side be designated by
ho- 20, and the angle opposite to a -- 2 by /'. The auxiliary
triangle thus formed gives the following relations: —
496 THEORETICAL ASTRONOMY.
cos (h, — g o)=cos (a- g o) cos (2 -h)-l-sin (a- o) sin ( s -h) cos i,
sin (ho- g) sin i, -sin (g -h) sin i, (147)
sin (h0 — o) cos i-sin (- ~, o) cos ( g -h)-cos (aT- o) sin ( -h) cos,
sin (1ho — 0 ) cos v'-cos (o- g o) sin ( — h) —sin ( — ) cos (2 -h) cos i.
Combining these with the preceding equations, we easily derive
cos b cos ( —h) =cos (A,- g ) cos (h,- o) +sin (A, — ) sin (ho-,- o) cos i,
cos b sin (1 —h) sin (A,- o) cos (ho- ) o) i,-coso (c, — o) sin (ho- o)
+cos (, — a) sin (ho~ —2o) (1+cos V') (148)
+sin (,- ) ((cos -cos i) cos (h — )+Ssin (- ) sin (2 -h) sin i),
sin b -sin sin (A,- o) +-(cos (- g o) sin i-sin in) sin (A, - g )
cos (A,- g ) sin ( — g0o) sin i.
Since the action of the component of the disturbing force perpendicular to the plane of the disturbed orbit does not change the radiusvector, we have
r sin b - r sin io sin (A, —~o0) + 8z,,
and hence the last of these equations gives
=sin (A, - o) (cos (T - - o) s -s i - sin )(
-cos (A, - o) sin (, - %o) sin i.
From the relation of the parts of the auxiliary spherical triangle, we
have
sin i sin (a -- o) = sin -r sin (ho - 0o),
sin i cos (o - no) - sin v' cos (ho - 2o) cos i + cos j' sin i0.
Therefore,
= sin(A, - o) (cos os io (ho - g2,) sin'-sin i(1 - os)), (150)
- cos (A, - 2 o) sin (ho - o) sin n',
and
z, sin, = sin (A,- 0) (cos io cos (h,- 0o) (1+cos V')-sin i sin 7')
- I cos'
r^ ~~ I~~~os^~~ -qC'~(151)
- cos (A,- 2o) sin (ho — o) (1 + cos ).
We have, further, from the auxiliary spherical triangle,
cos i = sin o sin V' cos (ho - o) - cos cos',
from which we get
cos i -cos io = sin i cos (ho- go) sin' - os io (1 + cos ).
We have, also,
sin (a - go) sin i sin -' sin (ho- o),
sin ( - h) sin i sini, sin (ho — o),
VARIATION OF CO-ORDINATES. 497
or
sin ( - P o) sin (g - h) sin2 i sin2' (io - Po) sin io sin'.
Hence we derive
(cos i-cos i) cos (ho-. o) +sin (~ —g o) sin (2 -h) sin2 i-sin iosin 7'
- (1+cos r') cos io cos (ho — o ).
Combining this and the equation (151) with the equations (148), we
obtain
cob cos ( —h)=cos(2,,- ~o) cos (h- 0) +sin(2, — o)sinl (ho — o) cos i,
cos b sin ( —h) sin (,- o)cos (ho — os) icoso (A,- g o) sin (ho- o)
sin /' _z,
- -cos' r
sin b =sin (A,- o) sin i0+ --
If we multiply the first of these equations by cos (ho - Q2), and the
second by -sin (h/0- 2), and add the results; then multiply the
first by sin (ho - 2 0), and the second by cos (ho - a ), and add, we get
sin -' 8z,
cos b cos (1-2o-(h-ho))-cos (,- Qo)+sin (ho- o) l cs l, —,
sin'2' (z,
cos b sin (1- o-(h —ho))=sin (A,- So)cos-cos(ho — o) 1 -cos
sin b =-sin (2, — o) sin i0o-+'- (152)
Let us now put
p = sin (e - 6'T) sin i,' = cos ( - go) sin i -sin, ( 3)
and there results, from (149),
- q' sin (, - o) -p' cos (A,- o). (154)
Comparing this with equation (150), we observe that
p' = sin' sin (^ho- o),
q' = sin l' cos (ho - ao) cos i - sin i, (1 - cos').
Therefore, we have
sin 7' _ p'
-~ cost sin (ho -- 2o) = -cos ~"
sin'2' __ q_
n,cos (^o - 0) =tan + (1 - c )
I - cos? ~/tl os (I~cos)32
32
498 THEORETICAL ASTRONOMY.
and, if we put F= h- h,, the equations (152) become
cosbcos(l — S r) —cos(A, — o)+ I -- cos.
1. -- eOS r'l r' (155)
cosbs (inn ~o~Z')-sin (i,-~- o) cos io( tan i,+cos io ( ) C r,
sin b =sin (A,- o) sin oi,+-.
As soon as r, p', qt, and v' are known, these equations will furnish
the exact values of I and b, those of i, and r being found by means
of the perturbations v and lfM.
186. The value of F may be expressed in terms of p' and q'
Thus, if we differentiate the first of equations (147) and reduce by
means of the remaining equations of the same group, we get
d (ho - a o) - cos' d ( - ) + cos i, da + sin i, sin ( - o0) di,
and if we interchange a - h and h -- 20 in this equation, we must
also interchange i and io, which are the angles opposite to these sides,
respectively, in the auxiliary spherical triangle, so that we shall have
d ( -- h) = cos' d (o - g2) + cos i do,
io being constant. Adding these equations, observing that 20 is also
constant, we get
(1-cos -') d ( g -h+-o)=sin i sin (T — ) di+(cos i+cos io) d; (156)
and since da = cos i d g, this becomes
(1 - cos 7') d (h - ho) — sin i sin ( ~- Qgo)di
+ (sin2i - cos' -_ cosi cos i) d-,
cos l'
which, since
cos 7' = sin i sin i, cos (a - P o) cos i cos io, (157)
may be written
(1-cos I') dF= - sin i sin ( — 2 o) di+tan i (sin i-sin io cos (o — g o)) da.
(158)
The differentiation of the equations (153) gives
dp' sin (a - 2,o) cos i di + sin i cos (a - go0) d,
dq' cos ( - go) cos i di - sin i sin (o - Qo) d<,
from which we derive
VARIATION OF CO-ORDINATES. 499
q'dp' —p'dq' = sin2 i da - sin i0 dp'
- cos i (-sin io sin (~- g o) di+-tani (sin i-sin i cos ( — o)) d6).
Combining this with equation (158), we get
cos i (1 - cos?') dF == q'dp' — pdq',
and hence
r= - _.dtt dt t, (159)
cos i (1- cos')
the integral being equal to zero for the instant to which the fundamental osculating elements belong. It is evident from the equations
(153) that p' and q' are of the order of the first power of the disturbing forces, and hence, since i^ differs but little from 180~-(i+io),
it follows that, so long as i is not very large, ris at least of the
second order.
The last of equations (145) gives, - r sin i sin A, cos a - r sin i cos A, sin a,
and since
x z r cos,, y r sin X,,
this becomes, -- x sin i sin C -- y sin i cos a.
Comparing this with equation (116), it appears that
-a =- sin i sin, - sin i cos a, (160)
and hence, by means of (153), we derive
p'=- a coso -- f sin o0,
q' - - sin g o + f cos o ~- sin i,
and also
dp - d. d.
== - COS g d - sin g0 ~
dt -dt dt
- da d (161)
d - sin g' dt + dCt
dt t d- -
From the equations (118) and (121), observing that
dy dx
d-Y dt kl/p (1 + m),
we derive, by elimination,
d r sin, cos i df _ c rs cs, os' 7. - - /_ (l.')'.' l+
500 THEORETICAL ASTRONOMY.
Therefore we shall have
dp' r cos i sin (A, go)
dt k (+m)
(162)
dq' r cos i cos (A, - Q o) Z
dt kl (l+ m)
by means of which p' and q'may be found by integration, the integral in each case being zero for the date to at which the determination of the perturbations begins.
When the value of Oz, has already been found by means of the
equation (129), if we compute the value of q', that of p' will be
given by means of (154), or
p = q' tan (~, 0) cs (, —0)
and if p' is determined, q' will be given by
q'- =p' cot (2,- d2) +
- t I cot (A,-2 + r sin (i,- go)
If both p' and q' are found from the equations (162), 8z, may be determined directly from (154); but the value thus obtained will be
less accurate than that derived by means of equation (129).
Since the formula for d, completely determines the perturbations
due to the action of the component Z perpendicular to the plane of the
instantaneous orbit, instead of determining p' and qt by an independent
integration by means of the results given by the equations (162), it
dsz,
will be preferable to derive them directly from 8z, and - The
equations (161) give
= -- cos o 0 -- sin 0 i, q' -- sin Qo a +- cos o 3a.
Substituting for ao and A3 their values given by (125) and (126),
and putting
x".x cos 20 + y sin o0, y"- - x sin 0 + y cos b0,
we obtain
I y,, diz,
kV'p (1+ + y" ~'- ~) -\'
1 Il +dz, dx" (163)
ki/p(1l + q) ( dt f dt )
VARIATION OF CO-ORDINATES. 501
Substituting further the values
x" - r cos (, - o2), y" r sin (A — ),
and also
dA, kl/p (I -- n)
dt r2
dr kV11 + n. ki/ (l + ) e sin v
~ —=^ ~-~ es~ sin v --
dt - s/p r -+- e cos v
we easily find, since A,- v =,'=-(cos (A,-XO)+Hecos(x-G )z r sin/ (<, - (~) (d,6d
~' — (cos(0- g~o) + Gecos(- P/o)) — + -~ r
ki//(l+ in) dt
- r cos (, - ) dz, (164)
- q=+ (sin (A,- Q0) + esin (X - /2o)) p + d )' -
which may be used for the determination ofJp' andiq'. These equations require, for their exact solution, that the disturbed values e, X
and p shall be known, but it is evident that the error will be slight,
especially when e is small, if we use the undisturbed values e,, p,,
and Xo ro. The actual values of A, and r are obtained directly from
the values of the perturbations.
When p' and q' have been found, it remains only to find cos i, and
1 - cos v', in order to be able to obtain Fby means of the equation
(159). From (153) we get
p' - q' in2 i -+ 2q' sin i0,
and hence
cos i 1/1 p'2 (q' ~ sin io,), (165)
from which cos i may be found. The equation (157) gives
1 - cos' = cos i, (cos i, + cos i) - q' sin i,, (166)
by means of which the value of 1 - cos v' will be obtained.
If we substitute the values of pt, q, dt, and dt given by the
equations (153) and (162) in (159), it is easily reduced to
Zdt, (167)
(1 - cos I') k/p (1 + m)
which may be used for the determination of r. When we neglect
terms of the order of the cube of the disturbing force, in finding 1I
we may use po in place of p and put 1 - cos' =2 cos io, so that the
formula becomes
502 THEORETICAL ASTRONOMY.
~r — f~ 1 - z, Zdt. (168)
2 cos2 i kl/o1 (1+r m)
187. By means of the formulae which have thus been derived, we
may find the values of all the quantities required in the solution of
the equations (155), in order to obtain the values of I and b for the
disturbed motion. From r, 1, and b the corresponding geocentric
place may be found. The heliocentric longitude and latitude may
also be determined directly by means of the equations (145), provided
that 2, a, and i are known; and the required formulae for the determination of these elements may be readily derived. Thus, the equations (160) give, by differentiation,
da.. di do
-d sin f cos dt sill i Cos dtdt at dt
d3. di... da
d-t = cos a cosi -- sin i sin ddt c t dt'
whence. da d. d
sin -cos -sin -
dt dt dt'
di da dfi
cos d - — sin a - - cos -
dt dt dt
du du
Introducing the values of dt and - already found into these equations, and putting
a T=Of f==,- 8+ i+ 8= i - - a8 + 8,
we obtain
ci - m) cot i sin (A, -- ) rZ,
di 1 (169)
t kp(1 +~ cos (2,- ~) rZ,
dt kl/p(l1 +- m)
and also, since da - cos i d2,
d _ 1 sin (A,- 6) rZ, (170)
dt kVlp(1 -+) sini
by means of which the variations of a, i, and a due to the action
of the disturbing forces, may be determined. The integral is in each
case equal to zero at the initial date to to which the fundamental osculating elements belong and at which the integration is to commence.
VARIATION OF CO-ORDINATES. 503
If we find i, and then a- a from
6 _ ff - ~Sil (,i a) rZdt, (171)
the true longitude in the orbit will be obtained from
R-Ax,+ ~ 6 —.
d~i d8a d 2g
It is evident that since the expressions for d — dt and -t require, for an accurate solution, that the disturbed values i, a, and p
shall be known, and require, besides, that three separate integrations
shall be performed, unless the perturbations are computed only in
reference to the first power of the disturbing force, in which case we
use io, Po, and g20 in place of i, p, and a, respectively, in the equations
(169) and (170), the action of the component Zcan be considered in
the most advantageous manner by means of the variation of z, arising
from this component alone; and even when only the perturbations
of the first order are to be determined it will still be preferable to
derive 8z, by the indirect process from the expression for d-F', and to
determine the heliocentric place by means of the equations (155).
When we neglect the terms of the second order, these equations
become
cosb cos ( - g20) - cos (A,- g0),.,Z
cos b sin (I - Po) = sin (A, g~0) cos i, - tan 0' —, (172)
s sin b = sin (,- g 0) sin i0+ -
by means of which I and b are determined immediately from the perturbations 8s, v, and az,. The peculiar advantage of determining
the effect of the action of the component Z by means of the partial
variation of z, is apparent when we observe that the expressions for
d~6 deg~
d and. -a involve sin i as a divisor; and in the case of orbits whose
dt dc
inclination is small, this divisor may be the source of a considerable
amount of error.
188. The determination of the perturbations so as to include the
higher powers of the masses is readily effected by means of the comd8M d'v dz,f
plete expressions for t, and —, when the correct values of
R, X, Z, i, and p are known. The corrected values of i and p
504 THEORETICAL ASTRONOMY.
which are required only in the case of az,-may be easily estimated
with sufficient accuracy, since we require only cosi, while V/p appears as the divisor of a term whose numerical value is generally
insignificant. To obtain the actual values of R, S, and Z, the corrections to be applied to N, N', and I must first be determined by
means of the formulae (136). The values of 8i' and 82' will be
found by means of the data furnished by the tables of the motion of
the disturbing body, and the corresponding corrections for N, N',
and I having been found by means of the terms of (136) involving
di' and dQg, there remain the corrections due to 6i and 802 to be
applied. These may be found in terms of the quantities p' and q'
already introduced. Thus, the equations
dp' -- cos i sin (6 - o) di + sin i cos (6 - o) do,
dq' =- cosi cos (6 -,) di - sin i sin (6 - d) da,
give
cos i di = sin (a - 0 o) dp' + cos ( - Q O) dlq',
sin i d =- cos (6 - Q ) dp' - sin (a — 2 o) dq'.
The equations (136) give, observing that da cos i d 2,
d -- cos N di - tan i sin N d,
sin NV tan i
dN' -- + dit- cos Nda,
sin I sin I
and, substituting the preceding values of di and da, these become
sin (N + a- Q) dp' cos (N + 6-,)d
cos i cos d
d NV'-_ cos (N+ 6- ao)d + sin (N + 6 ) -
sin I cos i sin os dq.
If we neglect the perturbations of the third order, these equations
give
1 - sin N - cos N -,
cos cos
-cosi0 2 COSt0 (173)
(N'= cosec I cos N - sin N2V,
\ cos i cos to
by means of which 8I and 8N may be determined, p' and q' being
found by means of the equations (164), using e, 70, and p, in place
of e, x, and p. The results for 8I and AN' obtained from (173)
being applied to the values of I' and N' as already corrected on
account of Ji' and 8 j', give the required values of these quantities.
NUMERICAL EXAMPLE. 505
iWhen we consider only di and dQ, since
sin i' cos' - cos i sin I sin i cIos N c,
we easily find
N= cos I aN' 8- a, (174)
and if we add the quantity cos I N' to the value of N already corrected on account of 3i' and 820', and denote the result by N,, the
required value of N will be N, - 8a. Then, according to (131), we
may compute w' + Ja and d' by means of the formulae
i' -t' N',
tan ((w' + a,)- N,) tan u,' cos, (175)
tan i' tan Isin ((w' +- ) - N,),
using the values of N' and I as finally corrected. We have, further,
according to (135),
w - + = v, + =o - go,
by means of which we may compute the value of w + Oa; then the
value of w' - w required in the equations (132), and also in finding
the value of p, will be given by
w' -- = (w' + a) - (0 + 8a),
and the forces R, S, and Z may be accurately determined.
By thus determining the correct values of R, S, and Z from date
to date, the perturbations AM, v, and 8z, may be determined in reference to the higher powers of the disturbing forces according to the
process already explained. The only difficulty to be encountered is
that which arises from the quantities r, p', and q', required in the
determination of the heliocentric place of the disturbed body by
means of the equations (155). If an exact ephemeris for a short
period is required, by means of the complete perturbations we may
determine new osculating elements, and by means of these the required
heliocentric or geocentric places.
189. EXAMPLE.-We will now illustrate the application of the
formulae for the determination of the perturbations JM, v, and 8z, by
a numerical example; and for this purpose let it be required to
determine the perturbations of Eurynomne @ arising from the action
of Jupiter from 1864 Jan. 1.0 to 1865 Jan. 15.0, Berlin mean
506 THEORETICAL ASTRONOMY.
time, the fundamental osculating elements being those given in
Art. 166.
In the first place, by means of the formula (130), using the values
a - 2060 39' 5".7, i 4- 36' 52".1,
dg'- 98 58 22.7, i'1= 18 40.5,
which refer to the ecliptic and mean equinox of 1860.0, we obtain
N= 194~ 0' 49".9, N'- 301~ 38' 31".7, - 5~ 9' 56".4.
Then, by means of the data furnished by the Tables of Jupiter, we
find the values of u', the argument of the latitude of Jupiter in reference to the ecliptic of 1860.0, and from the equations (131) we derive
w' and i'. The values of r' are given by the Tables of Jupiter, and
the values of r0 and %v are found from the elements given in Art.
166. The results thus obtained are the following:Berlin Mean Time. log o0 v0 log r w' W
1863 Dec. 12.0, 0.294084 354~ 26/ 18/.0 0.73425 14~18/ 54".6 -0~ 1 38/.1
1864 Jan. 21.0, 0.294837 10 2 45.7 0.73368 17 21 44.2 0 18 9.1
March 1.0, 0.300674 25 24 59.4 0.73305 20 25 5.2 0 34 39.9
April 10.0, 0.310864 40 13 31.8 0.73237 23 28 59.8 0 51 7.6
May 20.0, 0.324298 54 14 41.4 0.73164 26 33 32.1 1 7 29.7
June 29.0, 0.339745 67 21 23.5 0.73086 29 38 44.8 1 23 43.5
Aug. 8.0, 0.356101 79 32 18.1 0.73003 32 44 41.2 1 39 46.3
Sept. 17.0, 0.372469 90 49 57.6 0.72915 35 51 24.6 1 55 35.2
Oct. 27.0, 0.388214 101 19 9.8 0.72823 38 58 57.5 2 11 7.5
Dec. 6.0, 0.402894 111 5 42.2 0.72726 42 7 23.3 2 26 20.3
1865 Jan. 15.0, 0.416240 120 15 32.6 0.72625 45 16 43.9 -2 41 10.6
The value of w for each date is now found from
wU —Vo + r- 0 -Qo + 197~ 38' 6".5,
and the components of the disturbing force are determined by means
of the formule (132), p being found from (133) or (134), and h from
(70). The adopted value of the mass of Jupiter is
1047.879'
and the results for the components R, S, and Z are expressed in units
of the seventh decimal place. The factor (2 is introduced for convenience in the integration, (o being the interval in days between the
successive dates for which the forces are to be determined. Thus we
obtain the following results:
NUMERICAL EXAMPLE. 507
Date. 2/R 2r5o o2Z coSo of Srodt
1863 Dec. 12.0, + 70.82 + 7.16 + 0.04 + 1.37
1864 Jan. 21.0, 68.95 - 32.76 0.49 - 11.45
March 1.0, 61.16 70.38 0.92 63.32
April 10.0, 48.57 102.91 1.32 150.48
May 20.0, 32.77 128.34 1.68 266.75
June 29.0, + 15.41 145.39 1.96 404.35
Aug. 8.0, - 2.19 153.44 2.17 554.54
Sept. 17.0, 19.12 152.41 2.29 708.21
Oct. 27.0, 34.81 142.50 2.25 856.39
Dec. 6.0, 48.95 124.04 2.09 990.36
1865 Jan. 15.0, - 61.45 - 97.36 + 1.75 - 1101.73
The single integration to find (wSrodt is effected by means of the
formula (32).
The equations for the determination of the required differential
coefficients are
W d8IM I d -- 2w
w t ~^- -==-^ -,~=- ^ I S3t ~1 2
zd2~~X, w2k2
d 2 2 R 2P I Sil 2 2 k2
2 r, + k\,/p J P po <
~ clP",.
-2 d ~ 2Zcos -- 9 z.
Substituting in these the results already obtained, and also
log Jo - 2.967809, logpo 0.371237, log e, - 9.290776,
we obtain first, by an indirect process, as illustrated in the case of
the direct determination of the perturbations of the rectangular cod'v d2Sff
ordinates, the values of W2 dt' and 2 and having found
dt2 then, having found v,
w dt is given directly by the first of these equations. The integration of the results thus derived, by the formulae for mechanical quadrature, furnishes the required values of v, 8M, and cz,. The calculation of the indirect terms in the determination of v and Jz,, there
being but one such term in each case, is, on account of the smallness
of the coefficient, effected with very great facility.
The final results are the following:
508 THEORETICAL ASTRONOMY.
dl2 ~ d2v d2sz,
Date. 2 d —- t2 2 d M v z,
dt dt 2 dt
1863 Dec. 12.0, 0".028 + 36.16 + 0.04 + 0".01 + 4.41 + 0.02
1864 Jan. 21.0, 0.072 33.61 0.49 -0.01 4.31 0.04
March 1.0, 0.499 22.55 0.89 0.27 37.11 0.54
April 10.0, 1.213 + 5.58 1.21 1.11 91.96 1.93
May 20.0, 2.070 - 13.52 1.45 2.75 152.22 4.52
June 29.0, 2.902 31.59 1.53 5.24 199.05 8.54
Aug. 8.0, 3.546 46.65 1.60 8.49 214.54 14.10
Sept. 17.0, 3.858 57.88 1.52 12.22 183.69 21.24
Oct. 27.0, 3.723 65.19 1.28 16.05 + 95.29 29.90
Dec. 6.0, 3.056 68.83 0.92 19.49 - 58.00 39.82
1865 Jan. 15.0, - 1.800 - 69.19 + 0.40 -21.97 -279.84 +50.64
Since, during the period included by these results, the perturbations
of the second order are insensible, we have, for the perturbations of
Eurynome arising from the action of Jupiter from 1864 Jan. 1.0 to
1865 Jan. 15.0,
sM-= - 21".97, - 0.00002798, z, 0.00000506.
It is to be observed that Jz, is not the complete variation of the coordinate z, perpendicular to the ecliptic, but only that part of this
variation which is due to the action of the component Z alone; and
hence the results for 8z, differ from the complete values obtained
when we compute directly the variations of the rectangular coordinates.
Let us now determine the heliocentric longitude and latitude for
1865 Jan. 15.0, Berlin mean time, including the perturbations thus
derived. From the equations
M, - Mo + P0 (t- to) + 3M,
E, e0 sin E, - M,,
r, =a 0(1 - e cos E,),
sin (v,- E,) -sin sin E, ao,
r,, = v, -+, r - r, (1 + v),
we obtain
M, 99~ 29' 35".51, E,- 110~ 0'33".75,
log r, 0.4162304, v, - 120 15 13.80,
log r = 0.4162183, A, - 164 32 25.97.
The calculation of the values of r, and v, from the values of M,, a0,
and e,, may be effected by means of the various formulae for the
NUMERICAL EXAMPLE. 509
determination of the radius-vector and true anomaly from given
elements. If we substitute these results for ),, r, and 8z, in the equations (172), we get
I 1640 37' 59".05, b - 3 5' 32".54,
which are referred to the ecliptic and mean equinox of 1860.0, and
from these we may derive the geocentric place of the disturbed body.
If the place of the body is required in reference to the equinox and
ecliptic of any other date, it is only necessary to reduce the elements
%T, Qo, and i0 to the equinox and ecliptic of that date; and then,
having computed i, and r, we obtain by means of the equations (172)
the required values of I and b. In the determination of the perturbations it will be convenient to adopt a fixed equinox and ecliptic
throughout the calculation; and afterwards, when the heliocentric or
geocentric places are determined, the proper corrections for precession
and nutation may be applied.
In order to compare the results obtained from the perturbations
M,; v, and 8z, with those derived by the method of the variation of
rectangular co-ordinates, we have, for the date 1865 Jan. 15.0,
o= - 2.5107584, -- + 0.6897713, zo - 0.1406590;
and for the perturbations of these co-ordinates we have found'x - + 0.0001773, y -+ 0.0001992, az -0.0000028.
Hence we derive
z - 2.5105811, y - + 0.6899705, z — 0.1406618,
and from these the corresponding polar co-ordinates, namely,
log r 0.4162182, 1 164~ 37' 59".05, b - 30 5' 32".54,
from which it appears that the agreement of the results obtained by
the two methods is complete.
190. When the perturbations become so large that the terms of the
second order must be retained, the approximate values which may be
obtained for several intervals in advance by extending the columns
of differences, will serve to enable us to consider the neglected terms
partially or even completely, and thus derive the complete perturbations for a very long period. But on account of the increasing difficulties which present themselves, arising both from the consideration
510 THEORETICAL ASTRONOMY.
of the perturbations due to the action of the component Z in computing the place of the body, and from the magnitude of the numerical values of the perturbations, it will be advantageous to determine,
from time to time, new osculating elements corresponding to the
values of the perturbations for any particular epoch, and thus commencing the integrals again with the value zero, only the terms of
the first order will at first be considered, and the indirect part of the
calculation will, on account of the smallness of the terms, be effected
with great facility. The mode of effecting the calculation when the
higher powers of the masses are taken into account has already been
explained, and it will present no difficulty beyond that which is inseparably connected with the problem. The determination of F, p',
rl' dap' dq'
and q' may be effected from the results for -t dt, and by means
of the formulae for integration by mechanical quadrature, as already
illustrated, or we may find F by a direct integration, and the values
of 1p and q' by means of the equations (164), dt being found from
dt' by a single integration. The other quantities required for the
complete solution of the equations for the perturbations will be
obtained according to the directions which have been given; and in
the numerical application of the formulae, particular attention should
be given to the homogeneity of the several terms, especially since, for
convenience, we express some of the quantities in units of the seventh
decimal place, and others in seconds of arc.
The magnitude of the perturbations will at length be such that,
however completely the terms due to the squares and higher powers
of the disturbing forces may be considered, the requirements of the
numerical process will render it necessary to determine new osculating
elements; and we therefore proceed to develop the formulhe for this
purpose.
Cd22v d'WZ
191. The single integration of the values of wt2 and W2 wdill
th vade d& o6, dv d^z,
give the values of ~-~d and o d-, and hence those of d and d-,
dJ'M
which, in connection with dt, are required in the determination of
the new system. of osculating elements. Since r2 dv rersents double
the areal velocity i the disturbed oents i we havedouble
the areal velocity in the disturbed orbit, we have
CHANGE OF THE OSCULATING ELEMENTS. 511
dv, kl/p (1 + mn)
dt r2
The equation (109) gives
dt, do' dt
do, _ EVI (qI- mn)(1 -~ + dsM )
Hence, since r - r, (1 + v), we obtain
p p(o l+ o1. dM ) (1+y), (176)
by means of which we may derive p. This formula will furnish at
once the value of p, which appears in the complete equation for
dt2, and also in the equations (164); and the value of cos i may be
determined by means of (165).
In the disturbed orbit we have
dr kVl + mn
- - - e sin v,
dt l/p
and the equations (108) and (111)'give
dr kil-. I d8M\ dv.d -.. e0sin v, 1+ - (1 + -) - + r,
d d't d)'
Therefore we obtain
/o e sin v - 1/ e sin v, (1 )+ / +. dv
/Po dt k/ l — m
which, by means of (176), becomes.,I dM,1 +,) rj/p d (7
esin v = e sin v, 1 ( + dt + (177)
p, Po f dt V(- I /lim dtThe relation between r and r, gives
P Po
1 + e cos v 1 + eo cos v,
and, substituting in this the value of p already found, we get
e cos v = (1 +eo cosv,) (1 -1 d ) (1 + )31. (178)
\ Po dt / Y)
512 THEORETICAL ASTRONOMY.
Let us now put
o dt
r,Vp d' (179)
k-/l + m dt
a and f9 being small quantities of the order of the disturbing force,
and the equations (177) and (178) become
e sin v= e0 sin v, + aeo sin v, +-,
e cos V = eo cos V, + ace COS V, +- a.
These equations give, observing that r, (cos v, + eo) =po cos,,
e sin (v,- v) =c a sin v, - f cos v,,
a0o (180)
e cos (v,- v) = e- P cos E, + ~ sin v,,
from which e, v,- v, and v may be found; and thus, since
- l0o + -(v - ), (181)
we obtain the values of the only remaining unknown quantities in
the second members of the equations (164). The determination of
p' and q' may now be rigorously effected, and the corresponding
value of cos i being found from (165), dt and dt will be given by
(162). Then, having found also 1 cos i' by means of (166), r may
be determined rigorously by the equation (159), and not only the
complete values of the perturbations in reference to all powers of the
masses, but also the corresponding heliocentric or geocentric places
of the body, may be found.
If we put
- =ca sin v, - f cos v,,,' aPo cos E, + f sin v,, (182)
and neglect terms of the third order, the equations (180) give
t2
e - eo + + 2e-'
r' r'a' (183)
VV 2 - — 8 - s,
e0 eo
in which s= -206264"1.8. These equations are convenient for the
CHANGE OF THE OSCULATING ELEMENTS. 513
determination of e and v, -v, and hence X by means of (181), when
the neglected terms are insensible.
The values of p, e, and v having been found, we have
k/1 +- m
sin -= e, a =p sec2, 3 -- (184)
tan -1 E - tan (45~ - ) tan, v, M — E — e sin E,
from which to find the elements p, a,,e, and M. The mean anomaly
thus found belongs to the date t, and it may be reduced to any other
epoch denoted by t, by adding to it the quantity / (t0 - t). When we
neglect the terms of the -f4 order, we have
sin vp - sin'o
-— o - cos.o - ( -'o) sin go
and if we substitute for sin c - sin o, - e - e the value given by
the first of equations (183), the result is
2' sin o +- r'2'D 9~ 2 sin %oP cos g o-' sin go tan go
from which we get
a, 812 sin'o s 12' +- +. + s- + ~s - s s, (185)
~ cos o 2 cos 23so 2 sin cos'o
by means of which (p may be found directly, terms of the third order
being neglected.
In the case of the orbits of comets for which e differs but little
from unity, instead of OMf we compute by means of the formula
(142) the value of AT, and since we have
d8T 1 d8M
dt /to dt
the equation for p becomes
P-p(l- daT) (1 + v)4; (186)
and for a we have
(- dt) (1 + ). (187)
Then e, v, and q will be found by means of the equations
33
514 THEORETICAL ASTRONOMY.
e sin (v, - v) = a sin v, - cos v,,
e cos (v, - v) = e + a (cos v, + e) + 13 sin v,, (188)
p
l+e
and the time of perihelion passage will be derived from e and v by
means of Table IX. or Table X.
There remain yet to be found the elements a, Qg, and i, which determine the position of the plane of the disturbed orbit in space.
The values of p' and q' will be found from the equations (164), and
F, whenever it may be required, will be determined as already
explained. Then we shall have
sin i sin (a - go) =p', (189)
sin i cos (a - 20) q' +- sin i0,
from which to find i and a. When we neglect the terms of the third
order, these equations give
sin i -sin io-q'
si smn
and hence
p' p'q'
a 0 + sin i, sin2 oi
q' q" sin io P' +
i -- io - c-o s si q- 32 c (190)
cos i + 2 cos i, 2 sin i cos (g
in which s -206264".8. The auxiliary spherical triangle which we
have employed in the derivation of the equations (155) gives directly
cos 1 (i + io) tan 2 (6 o)
cos (i - i) tan ( -- h+ho go)
and since h - h= r, we have
t Cos (i-i)
tan t- ( ~ — )- COS (i- o) tan (- go), (191)
by means of which the value of 2 may be found. This equation
gives, when we neglect terms of the third order,
I — __o sin io
g~ =- do + I + cos I co ( - io) (+ -- Po). (192)
0 00 0 cos ig'L 2 cos2 j
Substituting in this the values of a- 20 and i —io given by (190),
we get
a = 1 - 3 sin io,
-ini -- iio - ~- -pqs - /r, (193)
sin i cos i sin.2 cos i
CHANGE OF THE OSCULATING ELEMENTS. 515
r being expressed in seconds of arc. Finally, for the longitude of
the perihelion, we have
7 X+ ~- a, (194)
and the elements of the instantaneous orbit are completely determined. When we neglect terms of the third order, this equation,
substituting the values given by (190) and (192), becomes
= tan. is + tan' io (1 + 2 cos +. (195)
cos 2c 2 cos' i,
It should also be observed that the inclination i which appears in
these formulae is supposed to be susceptible of any value from 0~ to
180~, and hence when i exceeds 90~ and the elements are given in
accordance with the distinction of retrograde motion, they are to be
changed to the general form by using 180 - i instead of i, and
2 -- instead of wr.
The accuracy of the numerical process may be checked by computing the heliocentric place of the body for the date to which the
new elements belong by means of these elements, and comparing the
results with those obtained directly by means of the equations (155).
We may remark, also, that when the inclination does not differ much
from 90~, the reduction of the longitudes to the fundamental plane
becomes uncertain, and r may be very large, and hence, instead of
the ecliptic, the equator must be taken as the fundamental plane to
which the elements and the longitudes are referred.
192. Although, by means of the formulae which have been given,
the complete perturbations may be determined for a very long period
of time, using constantly the same osculating elements, yet, on
account of the ease with which new elements may be found from 8J1M,
dJM d, d_,
v, z,, d- - ~t and dt' and on account of the facility afforded in
the calculation of the indirect terms in the equations for the differential coefficients so long as the values of the perturbations are small,
it is evident that the most advantageous process will be to compute
MM, v, and 8z, only with respect to the first power of the disturbing
force, and determine new osculating elements whenever the terms of
the second order must be considered. Then the integration will
again commence with zero, and will be continued until, on account
of the terms of the second order, another change of the elements is
required. The frequency of this transformation will necessarily de
516 THEORETICAL ASTRONOMY.
pend on the magnitude of the disturbing force; and if the disturbed
body is so near the disturbing body that a very frequent change of
the elements becomes necessary, it may be more convenient either to
include the terms of the second order directly in the computation
of the values of 8M, r, and 8z,, or to adopt one of the other methods
which have been given for the determination of the perturbations of
a heavenly body. In the case of the asteroid planets, the consideration of the terms of the second order in this manner will only
require a change of the osculating elements after an interval of several years, and whenever this transformation shall be required, the
equations for Ap, i, Q, and 7r, in which the terms of the third order
are neglected, may be employed. It should be observed, however,
that the perturbations of some of the elements are much greater than
the perturbations of the co-ordinates, and hence when terms depending on the squares and higher powers of the masses have been
neglected in the computation of these perturbations, it may still be
necessary to include the values of the terms of the second order in
the incomplete equations referred to. No general criterion can be
given as to the time at which a change of the osculating elements
will be required; but when, on account of the magnitude of the
values of UM, v, and 8z,, it appears probable that the perturbations
of the second order ought to be included in the results, by computing
a single place, taking into account the neglected terms, we may at
once determine whether such is the case and whether new elements
are required.
193. We have already found the expressions for the variations of
a and i due to the action of the disturbing forces, and we shall now
consider those for the variation of the other'elements of the orbit
directly. Let x, y, z be the co-ordinates of the body at any given
time referred to any fixed system of co-ordinates. These will be
known functions of the six elements of the orbit and of the time.
If the body were not subject to the action of the disturbing forces,
these six elements would be rigorously constant, and the co-ordinates
would vary only with the time; but on account of the action of these
forces the elements must be regarded as continuously varying in order
that the relation between the elements and the co-ordinates at any
instant shall be expressed by equations of the same form as in the
case of the undisturbed motion. The co-ordinates will, therefore, in
the disturbed motion, be subject to two distinct variations: that
which results from considering the time alone to vary, and that which
VARIATION OF CONSTANTS. 517
results from the variation of the elements themselves. Let these two
kinds of partial variations be symbolized respectively by ( d) and
dt, 2and similarly in the case of the other co-ordinates; then will
the total variations be given by
dx _ dx_ \ d rx dy_ dy \ _,
d ~ dt ( dt dt ~d+ l _dt'
dz _ dz \+ ~dz (196)
dt \dt Ldt
But if we differentiate twice in succession the equations which express the values of x, y, and z as functions of the elements and of
the time, regarding both the elements and the time as variable, the
substitution of the results in the general equations for the motion of
the disturbed body will furnish three equations for the determination
of the variations of the elements. There are, however, six unknown
quantities to be determined; and hence we may assign arbitrarily
three other equations of condition. The supposition which affords
the required facility in the solution of the problem is that
[do' [fd] 2 []-~' (197)
and hence that
dx dx \ dy dy\ dz dz \
dt dt }' dt \dt' dt \dt
It thus appears that in order that the integrals of the equations (1)
shall be of the same form as those of the equations (3),-the arbitrary constants of integration which result from the integration of
the latter being regarded as variable when the disturbing forces are
considered,-the first differential coefficients of the co-ordinates with
respect to the time have the same form in the disturbed and undisdx dy dA
turbed orbits. But since -, -d and are the velocities of the
dt dt dt
disturbed body in directions parallel to the co-ordinate axes respectively, it follows that during the element of time dt the velocity of
the body must be regarded as constant, and as receiving an increment
only at the end of this instant. The equations (197) show also that
if we differentiate any co-ordinate, rectangular or polar, referred to a
518 THEORETICAL ASTRONOMY.
fixed plane and measured from a fixed origin, with respect to the elements alone considered as variable, the first differential coefficient
must be put equal to zero, and this enables us at once to effect the
solution of the problem under consideration. It is to be observed,
further, that the functions whose first differential coefficients with
respect to the time when only the elements are regarded as variable
are thus put equal to zero, must not involve directly the motion of
the disturbed body, since the second differential coefficients of the coordinates have not the same form in the case of the disturbed motion
as in that of the undisturbed motion.
194. If we suppose the disturbing force to be resolved into three
components, namely, R in the direction of the disturbed radiusvector, S in a direction perpendicular to the radius-vector and in the
plane of disturbed orbit, positive in the direction of the motion, and
Z perpendicular to the plane of the instantaneous orbit, the latter
will only vary 2 and i and the longitude of the perihelion so far as
it is affected by the change of the place of the node, while the forces
1 and S will cause the elements I1, 7r, e, and a to vary without affecting 2 and i.
Let us now differentiate the equation
V2- k2 (1+ ) ( 2 ),
regarding the elements as variable, and we get
2 rdr] - 1 da 2V dV
~2 L dt J - -2 dt k (I + m) dt'
or
da 2a V dV
dt k2 (1 + m) dt
dV
The differential coefficient dt is here the increment of the accelerating force, in the direction of the tangent to the orbit at the given
point, due to the action of the disturbing force; and if we designate
the angle which the tangent makes with the prolongation of the
radius-vector by O0, we shall have
dV
/ -=R cos bS + sin 0.
Substituting this value in the preceding equation, we obtain
VARIATION OF CONSTANTS. 519
da 2a2
da = k2 (1 m) (R V cos (P0 + SV sin sb).
dt k2 (1 — M)
But we have, according to the equations (50)6,
Vcos (d o - dt- H esnv,
c /dr \ k~l + — m)
Vsin -- =r( d~ t (I + )-'
in which v denotes the true anomaly in the instantaneous orbit; and
hence there results
da 2a2
cdt kV Z (l (e sin v t P), (198)
dt /p ( r
by means of which the variation of a may be found.
If we introduce the mean daily motion p, we shall have
dg a / da
dt~,t (199)
dt a dt
and hence
d/2 3a/~
dl a - s ~((e sin vR + S), (200)
dct k/p (1 + - m) r
for the determination of 8u.
The first of the equations (97) gives
d / dv\
dt 2 -dt- Sr;
and hence we obtain
d (l/p) Sr
dt kl/1 + m
or
-p -2pr S. (201)
dt kl/1 + m
The equation p = a (1 - e2) gives
dp p da de
dt a dt t2
Equating these values of d- and introducing the value of dalready found, we get
de kpl/psin vR+ - e - (202)a
dt k p(1+? e r a
520 THEORETICAL ASTRONOMY.
and since
P - 1 e cos v, r 1 e cos E,
r a
E being the eccentric anomaly in the instantaneous orbit, this becomes
de 1
dt -- /1 (p sin vR + p (cos v cos E) S), (203)
dt - k I/j (I T +m)
which will give the variation of e. If we introduce the angle of
eccentricity %, we shall have
de dcs
~ - ==cos