Vad JOWY SUGIS GIANT DIAG FDPALFIOWMO Ds) EI ALEU YY @ waAOG Analytical Geometry of Three Dimensions WILLIAM H. McCreEADOVER BOOKS ON MATHEMATICS HANDBOOK OF MATHEMATICAL Functions, Milton Abramowitz and Irene A. Stegun. (0-486-61272-4) ‘TeNsOR ANALYSIS ON ManiFoins, Richard L, Bishop and Samuel I. Goldberg. (0-486-64039-6) VECTOR AND TENSOR ANaLysis WITH AppLicaTions, A. I, Borisenko and I. E. Tarapov. (0-486-63833-2) THe History OF THE CaLcuLUS AND Its ConcerTUAL DEVELOPMENT, Carl B. Boyer. (0-486-60509-4) THE QUALITATIVE THEORY OF ORDINARY DIFFERENTIAL EQUATIONS: AN Introvuction, Fred Brauer and John A. Nohel. (0-486-65846-5) PrincipLes oF Stanistics, M. G. Bulmer. (0-486-63760-3) ‘Tue, THEORY oF Spivors, Elie Cartan, (0-486-64070-1) Apvancep Numer THeory, Harvey Cohn, (0-486-64023-X) Staristics Manuat, Edwin L. Crow, Francis Davis, and Margaret Maxfield. (0-486-60599-X) FourIER SERIES AND ORTHOGONAL Functions, Harry F. Davis. (0-486-65973-9) ComPUTABILITY AND Unsotvasiuity, Martin Davis. (0-486-61471-9) AsyMPToric METHODS IN ANALYSIS, N. G. de Bruijn. (0-486-64221-6) ‘Tue Matuemanics oF GAMES OF STRATEGY, Melvin Dresher. (0-486-64216-X) Appuitp PARTIAL DIFFERENTIAL Equations, Paul DuChateau and David Zachman. (0-486-41976-2) Asymptoric Expansions, A. Erdélyi, (0-486-60318-0) ‘CoMPLEx VaRUAnLES: HARMONIC AND AxatyTiC FuNcTIONS, Francis J. Flanigan. (0-486-61388-7) DiFreRENTIAL. TopoLocy, David B. Gauld. (0-486-45021-X) ON FORMALLY UNDECIDABLE PROPOSITIONS OF PRINCIPIA. MATHEMATICA. AND Retaren Systems, Kurt Gédel, (0-486-66980-7) A History oF Greek Maruemarics, Sir Thomas Heath. (0-486-24073-8, 0-486-24074-6) Two-volume set PRooABILTY: ELEMENTS OF THE MATHEMATICAL TueoRY, C. R. Heathcote. (0-486-41149-4) LvtRopucTION To NUMERICAL ANALYSIS, Francis B. Hildebrand. (0-486-65263-3) METHODS oF AppuieD MaTueMatics, Francis B, Hildebrand. (0-486-67002-3) Torowosy, John G, Hocking and Gall S. Young. (0-486-65676-4) MATHEMATICS AND Loaic, Mark Kac and Stanislaw M. Ulam. (0-486-67085-6) MATHEMATICAL FOUNDATIONS OF INFORMATION Theory, A. I. Khinchin. (0486-60434-9) Axrrumetic REFRESHER, A. Albert Klaf. (0-486-21241-6) CALCULUS REFRESHER, A. Albert Klaf. (0-486-20370-0) (continued on back flap)(continued from front flap) Intropuctory Reat Anatysis, A. N. Kolmogorov and S. V. Fomin. (0-486-61226-0) SPECIAL FUNCTIONS AND THEIR AppuicaTioNs, N, N. Lebedev. (0-486-60624-4) Chance, Luck aNp Stansmics, Horace C. Levinson. (0-486-41997-5) ‘TensoRS, DIFFERENTIAL FORMS, AND VARIATIONAL PrincipLes, David Lovelock and Hanno Rund. (0-486-65840-6) ‘Survey oF Matrix Theory AND Mareix Ivequatities, Marvin Marcus and Henryk Minc. (0-486-67102-X) ABSTRACT ALGEBRA AND SowuTION BY Rapicats, John E. and Margaret W. Maxfield, (0-486-67121-6) FUNDAMENTAL Concerts OF ALGesra, Bruce E. Meserve. (0-486-61470-0) FunpaMentat Concepts of Geometry, Bruce E. Meserve. (0-486-63415-9) Firty CHALLENGING PROBLEMS IN PROBABILITY WITH SOLUTIONS, Frederick Mosteller. (0-486-65355-2) NUMBER THEORY AND ITs History, Oystein Ore. (0-486-65620-9) ‘MATRICES AND TRANSFORMATIONS, Anthony J. Pettofrezzo. (0-486-63634-8) ‘Tue Umprat, Carcutus, Steven Roman. (0-486-44139-3) PROBABILITY THEORY: A Concise. Course, Y. A. Rozanov, (0-486-63544-9) Linear ALcenra, Georgi E. Shilov. (0-486-63518-X) EsseNnTIAL CALCULUS WITH APPLICATIONS, Richard A. Silverman. (0-486-66097-4) Interrotaion, LF, Steflensen, (0-486-45009-0) A Concise History of Maritemarics, Dirk J. Struik. (0-486-60255-9) PROBLEMS IN PROBABILITY THEORY, MATHEMATICAL. STATISTICS AND THEORY OF Ranpom Functions, A. A. Sveshnikov. (0-486-63717-4), ‘Tensor Catcuus, J. L. Synge and A. Schild. (0-486-63612-7) Mobern ALcesra: Two VotuMes Bounp As One, B.L. Van der Waerden. (0-486-44281-0) CALCULUS OF VARIATIONS WiTit APPLICATIONS TO PHYSICS AND ENGINEERING, Robert Weinstock. (0-486-63069-2) INTRODUCTION TO VECTOR AND TENSOR ANALYSIS, Robert C. Wrede. (0-486-61879-x) DistRIBUTION THEORY AND TRANSFORM AwNatysis, A. H. Zemanian. (0-486-65479-6) Paperbound unless otherwise indicated. Available at your book dealer, online at www.doverpublications.com, or by writing to Dept. 23, Dover Publications, Inc., 31 East 2nd Street,’ Mineola, NY 11501. For current price information or for free catalogs (please indicate field of Interest), write to Dover Publications or log on to www.doverpublications.com and see every Dover book In print. Each year Dover publishes over 500 books on fine art, music, crafts and needlework, antiques, languages, Ilt- erature, children’s books, chess, cookery, nature, anthropology, science, mathematics, and other areas. Manutfactured in the US.A.Mathematician, physicist, and astronomer, William H. McCrea conducted research in many areas and is best know: eo studied and tau, universities around the ok is based on a series of his lectures. Geoneny oF nee ‘NuMBers, Schwerdtfeger. 200pp. 5% x 8/A. 0-486-63830-8 For ‘current price information write to Dover Publications, or ow on onto www.dor see every Dover Raa ISBN O- NBb-45313-8 SLL-95 USA $17. a5 (GANALYTICAL GEOMETRY OF THREE DIMENSIONSANALYTICAL GEOMETRY OF THREE DIMENSIONS Second Revised Edition WILLIAM H. McCrREA Dover PUBLICATIONS, INC. MINEOLA, NEW YORKBibliographical Note This Dover edition, first published in 2006, is an unabridged republication of the Second Revised Edition first published in 1960 by Interscience, New York, from the 1947 edition published by Oliver and Boyd., Ltd., London and Edinburgh. International Standard Book Number: 0-486-45313-8 Manufactured in the United States of America Dover Publications, Inc., 31 East 2nd Street, Mineola, N.Y. 11501PREFACE Tuts book is based upon a short course of lectures to firat- year Honours students who have just completed a course on Algebra approximating to that covered by Dr Aitken’s Determinants and Matrices in these Texts, and who will have later in their curriculum a course of modern projective geometry. While the original lecture notes. have been drastically revised so that the book may, one hopes, meet the needs also of students whose courses are differently arranged, it ir still designed primarily for students at about the same stage as those to whom the lectures were addressed. This explanation, as well as the smallness of the book, will account for various omissions. For instance, homo- graphio correspondence is not introduced, largely on the ground that it plays its fundamental part in non-metrical geometry which the student will normally encounter at a later stage. On the other hand, it is hoped that the topics included will be found to be treated with a reasonable standard of rigour. In partioular, care has been taken to frame the theory so that it does strictly apply to real space. This explains the avoidance of certain familiar short-cuts, which actually depend on jumping difficulties about reality conditions, ‘The book purports to be » “University Text.” This presumably means that it will normally be used in con- junction with lectures or other personal instruction. It Vvvw PREFACE would therefore seem merely foolish to include the sort of additional explanation, illustrated by “trivial” examples, which is more appropriately given by word of mouth. At any rate, the author has felt justified in assuming that the majority of readers will not be working entirely without supervision, and has allowed this assumption to influence his manner of presenting the subject. I gratefully acknowledge the valuable suggestions I have received from the editors and from my colleague, Dr R. Cooper, and the careful cooperation I have received from the printers. W. H. McCREA Baxvasr, July 1942 PREFACE TO SECOND EDITION Tms edition differs from the first by the inclusion of a small number of new examples designed to amplify some of the discussion in the text. In these additions I have been glad to take advantage, in so far as my treatment of the subject had prepared the way for me to do so, of some suggestions kindly made by Professor H. 8. M. Coxeter and Professor A. G. Walker. W. H. McCREA Rovat Hottoway Conizon January 1947CONTENTS narra, PREFAOH . . . . . : . I. Coonpivatz System: Dmecrions. . . Tl. Puanzs anp Lirgs 6 wt eee Tl. SreerE . . . . . . IV. Homoaznrovs Coorpmates— Ports at V. Geert Equation or THE Szconp DEGREE . VI. Quaprio mw CarTesian CoorpinaTes; STAN- paRD Forms . . . . 7 VII. Inrersection oF Quaprics: SysTEMs oF QuanpzIcs. . : A . . fs Norz on Ansrgacr GEOMETRY . . . Supszcr Inpex - ww eee 41 89 125 139 143OHAPTERI COORDINATE SYSTEM: DIRECTIONS 4. Introductory We are going to study by algebraic methods the geometry of three-dimensional real euclidean space, usually regarded as “ordinary” space. We adopt the elementary view of analytical geometry, according to which it is merely a matter of convenience to introduce the algebraic method as a tool for the solution of problems having a well-defined meaning apart from the algebra. However, we shall observe that our tool guides us to those problems with which it is best fitted to deal. This elementary treatment is useful in applications to other parts of mathematics and to mathematical physica. But also, in geometry itself, the student can scarcely hope to appreciate modern abstract treatments without some such introduction. Moreover, the algebraic manipulation in abstract geometry is not essentially different. The form in which it is cast in this book is chosen partly to meet the requirements of the student wishing to pursue the subject further, and for his benefit a note on abstract geometry is given at the end. ‘The reader is assumed to be acquainted with elementary pure solid geometry, and with simple analytical geometry of two dimensions. The only other special mathematical equipment required is some knowledge of determinants and matrices; for this, reference will be made to Dr Aitken’s Determinants and Matrices in these Texts (quoted as “ Aitken”). 12 ANALYTICAL GEOMETRY OF THREE DIMENSIONS ‘The examples consist almost entirely of auxiliary results needed in the general development. In some cases proofs are indicated and should be completed by the reader; in others proofs should be supplied by the reader as he proceeds. In each section, examples are numbered consecutively, and subsequent reference made by giving the number of the section followed by that of the example. Formule are similarly treated, except that their numbers are put in brackets. The reader is urged to construct for himself numerical exercises; for “riders” he must consult larger textbooks and examination papers. Metrical geometry. The present geometry is metrical, which means that results are expressed, direotly or in- directly, in terms of distance and angle. Distance expresses arelationship between a pair of points; angle a relationship between a pair of directions. Both magnitudes have to be measured by comparison with selected standards. It may be remarked that, whereas there is in euclidean space a natural standard angle (the right angle being a convenient unit), the standard of length is quite arbitrary. However, we are here assuming the fundamental properties of these magnitudes, and our initial consideration in applying algebraic methods to the geometry is to find means first of labelling points and directions by algebraic symbols and then of expressing distances and angles in terms of these symbols. Such is the object of this chapter. Nomenclature. We call three-dimensional euclidean space &. Line will always mean straight line; any other sort of “line” will be called a curve. If A, B are any two distinct points, then we use the following notation: “The line AB,” or simply AB, means the whole line containing A, B, as distinct from ‘the segment AB”; AB denotes the same line when sense is relevant and is taken from A to B; | AB | denotes the length of the segment AB; (AB) denotes the distance from A to B, aense being relevant;COORDINATE SYSTEM: DIRECTIONS 3 ZB denotes the vector associated with the segment AB, in the senso from A to B. If A, B, O are any three non-collinear pointa, then: “The plane ABO,” or simply ABO, means the whole plane containing A, B, 0. We use the abbreviations: w.r.t. = “with respect to”; rhs, = “right-hand side”; Lhs. = “left-hand side”; and a few others introduced subsequently. 2. Cartesian Coordinates Consider any fixed point O and any three distinct planes through O. These planes meet in pairs in three non- coplanar lines through O; let X, Y, Z be fixed points, other than O, one on each line. Let P be any point. The lines through P parallel respectively to OX, OY, OZ meet the planes OYZ, OZX, OXY in points L, M, N, say (fig. 1). Write x= |ZP | if P, X are on the same side of OYZ, a=—| LP | if P, X are on opposite sides of OYZ, lengths being measured in terms of some selected unit; let y,zbe analogously related to | MP|,|NP|. Then, when P is given, the numbers 2, y, z are uniquely determined. Con- versely, it is seen that, given any three positive or negative numbers 2, y, z, there is a unique point P with which these pumbers can be associated in the manner described, So wo may speak of P as “the point (x, y, z).”4 ANALYTICAL GEOMETRY OF THREE DIMENSIONS When the points of & are labelled in this fashion, we say that they are referred to origin O and coordinate planes OYZ, OZX, OXY, or coordinate axes OX, OY, OZ. We call x, y, z the (cartesian) coordinates of P in this frame of reference, ‘Two features should be noted: (i) It must be realised that we can desoribe any point P only by its relationship to some particular set of points arbitrarily chosen es a system of reference, (ii) When we say that & is real we mean merely that every point of & can be labelled with three real numbers serving as coordinates. Consequently we must ensure that any algebraic theorems to which we attempt to give geo- metrical interpretations do in fact hold good in the field of real numbers. If OX, OY, OZ are mutually perpendicular, we call them rectangular or orthogonal axes. L, M, N are then the orthoggnal projections of P on the coordinate planes, and z, y,z the perpendicular distances of P from these planes, with appropriate signs attached. Unless otherwise stated, we shall use only such rectangular cartesian coordinates. Also, for definiteness, we shall use right-handed systems, i.e. viewing from O towards X, a rotation from Y towards Z would be that of a right-handed sorew, and so on in oyolio order. lL %, y, 2 are the rectangular components perallel to the axes of the vector OP, Length of a segment. If P,, P, are the points (as Yi» %1)s (as Yor 22), then [PiPs |? = (@—ee)*+(—e)*+ 1a)”. (1) This follows from an elementary application of Pytha- goras’s theorem to the rectangular parallelepiped having P,, P, a8 opposite vertices and edges parallel toOX, OY, OZ. 2. If the axes are oblique and the angles YOZ, ZOX, XOY are ©, ®, ¥, then [OP |* = at+y"-+2!-+ 2yz cos © + 22x cos O+-2ey cos Y.COORDINATE SYSTEM: DIRECTIONS s 8. A necessary and sufficient condition for 4 cartesian coordinate system to be rectangular is that | OP |* =2'-+y*-++e* for all positions of P(x, y, 2). Coordinates in general. We shall use the term “coordinates” with the following general meaning: Let E be e collection of geometrical objecta such that every member of is labelled by a unique ordered set of n numbers (&, 1, ...» 7) and such that every such set of numbers, in specified rango, is the label of a unique member of Z. Then &, 7, ..., 7 are called the coordinates of the corresponding member of &, in this system of labelling. If the objecta are points, we can when desirable distinguish their coordinates as “‘point-coordinates,” if planes, as “plane- coordinates,” and so on. However, we sometimes find it convenient to replace the n coordinates by the ratios of n+-1 other numbers, or of more than n+1 numbers connected by certain specified relations. We shall also call these new numbers “coordinates,” and shall find that we may do so without causing confusion. 3. Projections Projection will be restricted to mean orthogonal pro- jection, The projection of a point P on a plane II is the foot of the normal from P to II. The projection of any other figure on I is the aggregate of the projections of ite points, e.g. the projection of a line ¢ is the intersection of II with the plane through s perpendicular to II. The projection of @ point P on a line ¢ is the foot of the perpendicular from Pons, s.e.the meet of swith the plane through Pnormal tos. The angle @ between two planes II, A is the angle * between the normals from any point to II, A. Let a closed boundary in II enclose area a, and let its projection in A enclose area 8. Since lengths parallel to the intersection of TI, A are unchanged by the projection, while lengths perpendicular to it are multiplied by cos 6, we find B=a.cos 6. The angle ¢ between a line ¢ and a plane IT is the angle * * Tho acute angle, unless otherwise atated.6 ANALYTICAL GEOMETRY OF THREE DIMENSIONS between ¢ and its projection on II. Consequently the Projection on II of a segment PQ of ¢ has length | PQ | cos ¢. The angle ¢ between two skew lines ¢, ¢ is the angle * between lines through any point parallel to s, #. The Projection on ¢ of a segment PQ of a has length | PQ | cos x. But this should be carefully compared with the following paragraph. Sensed lines, There are two opposed senses of dis- placement along a line ¢; we arbitrarily call one positive and the other negative. The positive one is sometimes called merely the sense of s. When we wish to emphasise that ¢ has an assigned sense we denote itbys. P,, P, being a Fra. 2. points of s, we reckon the distance (P,P,) positive if the displacement from P, to P, is in the positive sense, and otherwise negative. Thus, if (P,P,) is positive, (PP) = | P\P3|=—(PsP,). If Q is another point of a, we say that @ divides the segment P,P, in the ratio (P,Q) : QP,). Let t be any other “sensed” line. We now define the * Tho acute angle, unlosa otherwise stated,COORDINATE SYSTEM: DIRECTIONS 7 angle x between s, t as the angle (0 < x <7) between lines 8’, t’ through any point parallel to, and in the same sense a8, 8, t, respectively (fig. 2, where arrows indicate the senses), 1. The ratio in which Q divides the segment P,, P, is positive if Q lies between P,, P,; negative and numerically less than unity if P, lies between Q, P,; negativo and numeri- cally greater than unity if P, lies botwoen Py, Q. 2. P, Q being points of s, P®, Q* their projections on t, (PQ) has a sign depending on the sense ascribed to 8, (P*Q*) one depending on the sense ascribed to t. We call (P*Q*) the projection of (PQ); in all cases (P*Q*) = (PQ) cos zx. [The problem is merely to see that, coupled with our definitions of (PQ), (P*Q®), x, this formula is equivalent to the standard definition of cos x.] 3. P,P, ... P,P, being any polygon, not necessarily plane, the sum of the projections of its sides P\P;, P:Ps, ...,P,P,0n any (sensed) line is zero. 4. The projection of a given vector on any (sensed) tine ia the sum of the projections of ite componente. 4, Direction-cosines and Direction-ratios Let v be any sensed line through O. We can con- veniently describe its orientation by its relation to S, the sphere with centre O and unit radius. For v meets $ at the ends of a diameter and one end, say V, is such that (OV) is positive (fig. 3). If v is given, V is a unique point of 3; conversely, if V is any given point of S, then v is uniquely determined as the line OV. Let V have coordinates 1, m, 2; it lies on S if and only if | OV | = 1, é.e. from 4 (1), Pimtnt=l. . . . (1) Therefore the preceding statement is equivalent to: If v is given, then l, m, n satisfying (1) are uniquely determined ; if 1, m, n satisfying (1) are given, then v is uniquely determined. The numbers 1, m, n specify completely the direction, including the sense, of v, and hence of any line 8 parallel tov8 ANALYTICAL GEOMETRY OF THREE DIMENSIONS and in the same sense. They are called the direction-cosines (d-c’s) of 8; (1) is the relation satisfied by every set of d-o’s, 1, 1, m, nare the cosines of the angles a, , y (say) between 8 end OX, OY, OZ. a, f, y are called the direction-angles of s. 2. 1, m, n are the components parallel to OX, OY, OZ of a unit vector along s. 3. If the senso of 8 is reversed, then the signs of its d-o’a are reversed, Now let A, 4, v be a set of numbers proportional to 1, m,n; then, using (1), Se eee AB VOR EEE ie, » (2) Amy Gm. TORE EA So, if A, 4, v are given, I, m, n are determined apart from sign, i.e. the direction of 8 is determined apart from sense. ‘These numbers are called direction-ratios (d-r'’s) of a. 4. 4, a, ¥ are the coordinates of some point of v, They are also the components of some veotor along #.COORDINATE SYSTEM: DIRECTIONS 9 &. D-r's of the join of Pity Yu %1)s Pale Ys 21) OPO Zy—Z yy Y¥s~Yu %—Z- The d-o’s of P,P, are theso quantities divided by | PiPal- Notation. We write ‘‘the direction (J, m, n)” for the sensed direction having d-o's 1, m, n; ‘the direction (4, #, »)” for the unsensed direction having d-t’s 4, 4, ». When occasion- ally other symbols are used, and we write, for instance, ‘the direction (a, 8, 0),’” then a, 6, o are to be interpreted as d-r’a unless the context shows that d-o’s are implied. Angle between two directions. The angle x between two directions (l,m, n), (I', m', n’) ia given by cos x =Hl'-+-mm'+nn’. (0< x <7). (3) Let V7, V’ be the points (2, m,n), (I, m’, n’); then OV, ov have the given directions. The projection of OV on OV" is cos x, since |OV |= 1. The z-component of OV is J, and the cosine of the angle between OX and ov'is U; hence the projection of the x-component of OV on OV’ is ll’, and so on. So (3) results from 34, It follows that (i, m,n), (', m’, 2’) are orthogonal if and only if U+mm' +nn' = 0. . . » 4 6. (Another proof.) Using 4 (1), 4(1), we have [00 |) C24 (mm) (n—n')t = 2—2(W! mm’ +n’). Again, from the triangle OV V’ in which ]OV | = |OV’| =1, LVvOV=% LUV’ |» =f OV |*-+] OV’ |*—2| OV | [OV | cos x = 2—2cos x. Comparing these values, we recover (3). 1. a, b, ¢, a’, b’, o’ being any numbers, (b0’ —b’c)*-+(ca’ ~0’a)* + (ab’—a’b)* = (a? +b*-+-08)(a"t +b/*-+0') —(aa’ +b! +0"). 8, From (3) and 7, sin x = + -/{mn! —m'n)+ (nl! —n'l)! + (Im —I'm)}}.10 ANALYTICAL GEOMETRY OF THREE DIMENSIONS a If $ is the angle between the directions (4, 1, »), (A’, 1’, ¥), cca Y= (AN bpp’ to) e/a tptper(at tp ty}; ain § = V[{(uv’—p'r)*-+ (v2! —¥'2)* Hap’ Hay (at-bat ota +p +e} [These being unsensed directions, the acute angle is taken; 80 the square roots are chosen so that cos ¥, sin ¢ are positive.) 10. The projection of the segment P,P, on the direction (A, 4, ¥) bas length 1 Alea —21) +4(¥s—9n) + ¥%s—A1) | /V (A+ ut+¥%). [Use 5, 9.] 11. The direction perpendicular to each of two distinct directions (4, 4, »), (4, p’, ») has der’s py’—p’v, vd'—v/A, Ap! — Xp. 12. x being the angle between (1, m, n), (I, m’, n’ the direction perpendicular to these has d-o’a a(mn’—m'n, nl’—n'l, Im’ —V'm)+sin x. The positive sign applies to the direction such that a right- handed screw travelling along it would turn from (I, m, n) towards (I’, m’, n’). 18. The directions (41, His ¥%1), (Ax Hs» Ya) (Aue Ha» 92) are parallel to a single plane if and only if Aw, yf =O Ay Bs % A bs &. Transformation of Coordinates Our object is to establish properties of geometrical figures which are independent of any particular labelling of the points. Consequently we choose at each stage the labelling which best facilitates the calculations, and so we want to be able to change from one system of labelling to another. Restricting ourselves to rectangular cartesian co- ordinates, the general transformation is a combination of (i), (i) as follows, provided the unit of length is unaltered. (i) Ghange of origin without rotation of axes. Let 8, S* be two coordinate systems, origins O, O*, corre-COORDINATE SYSTEM: DIRECTIONS " sponding axes being parallel and in the same sense. Let the coordinates of O* referred to G be &, », ¢. The coordinates of any point P being x, y,z; x*, y*, 2* referred to &, S* respectively, we have ea e+s, yyy, z=eetl. -. Cl) For these express merely the fact that the distance of P from OYZ is the sum of its distance from 0* Y®Z® and that of 0° Y®Z® from OFZ, and so on. 1, If p is the position vector of O* referred to O, r, r* tho position vectors of P referred to O, 0%, then r =r*+, yielding (1). (ii) Rotation of axes without change of origin. Let OX, OY,0Z (8), OX’, OY’, OZ (.") be two systems of axes, origin O. Let OX’, OY’, OZ' have d-c’s (, m, m), (ay tg» 19), (Fay Mm, 709) referred to S. We have Bmi +n = G+m}-+nj = B+-m}-+n} = 1, 2) and, the three directions being mutually perpendicular, aba tgs + Matty = Daly tng, +4, = Lyla tmyms +n, = 0. Also, since 1, 1, 1s are the cosines of the angles between OX and OX’, OY’, OZ’, they are the d-o’s of OX referred to ’. Similarly m,, mg, m3; m4, Mg, Ms are the d-o’s of OY, OZ referred to $'. So, corresponding to (2), (3), we have BHR+E = mitmitmi=nltntnj=1, (4) Magy Pmgttg bmegthy = Myby + Male t-Megl= Lym, +lyMmg-+ymmy = 0. 2. If Te|l, mn], then T*=]1, m ny hh &j=i Lmny I, my na] |] m, mm, im, ny Ae using (2), (3). Honce T = +1. 3. If B, coincide, then T = +1, 4, If S, S' are both right-handed (r.h), or both left-handed 8) (6)12. ANALYTICAL GEOMETRY OF THREE DIMENSIONS (Lh), then 7 = +1 for all G’. Suppose Gis rh. (2)-(6) hold whether or not 9’ is also rh. But if and only if $’ isr-h could it be rotated into coincidence with G. During the rotation, the value of the determinant 7, if it changes at all, must do so continuously. But by 2 it can take only the values +1; there being no possible intermediate values it cannot change continuously from one to the other. Hence it is always +1, or alwaye —1. In the final position of coincidence with S$, T = +1, from 3. Therefore T = +1 for every position of S’. 5. If G, @’ are one rh and ono 1.h system, then T = —1. 6. S, &’ being both r-h or both l.h, we have 1, = mn, —m,n,, and 80 on. ‘The coordinates of any point P being x, y, z; 2’, y’, 2’ referred to @, .$', x is the projection of OP on OX, 2’, y’, 2 the components of OP parallel to OX’, OY’, OZ’, so that Lz’, lay’, 2’ are the projections of these components on OX. Therefore, using 3 4, and treating y, z analogously, we have ashe’ thy +12’, y= mz’ +m +mz', } - « &® z=nz' +n,y' +n’, giving the required substitution. Symmetrically, or by solving (6) and using (2)-(5), we have 2 =hetmy+mz, y =he+my+nz, oo. (1) a = Le-+my tne, giving the inverse substitution. 7. Verify that (6), (7) both give 2*-ty! 42% = a/--y't42%, 8. Orthogonal matrices. From 23 it follows that the required transformation leaves z*+-y*+-z* invariant and eo is “orthogonal.” The above theory is then more concisely expressed in the language of orthogonal matrices, and the reader should compare Aitken, section 24, particularly examples 2, 6, on this topic. 9. Any change of axes transforms a polynomial of degree n in the original coordinates into one of degree n in the newCOORDINATE SYSTEM: DIRECTIONS 13 coordinates. [For the substitutions (1), (6) cannot raise the degree; neither can they lower the degree, for, if they did, their inverses, which are similar transformations, would raise the degree.} 10. A change of axes cannot alter the number of poly- nomial factors of a given polynomial in x, y, 2. 11. The general transformation of rectangular axes involves six independent parameters, or sevcn if the unit of length be changed.OHAPTER II PLANES AND LINES Tuts chapter supplies an elementary account of analytio geometry of planes and lines; certain more general con- siderations await Chapter IV. 6. Collinear Points The point Pa, Yy, 2) which divides the join of Py(%,; 915%)» Peas Yrs %) in the ratio dy : Ay ts given by atta y= tle, “FA” Mea? This is proved in the same way as the corresponding theorem in two dimensions (or seo 1 below). It follows that P lies on the join of P,, P, if and only if numbers A,, A, exist such that equations (1) are satisfied. Expressing this in a more symmetrical notation, points Pla, Yas %)s Palas Yor %)s Pa (tas Yss %) are collinear sf and only if numbers p11, po, jg, not all zero, exist such that Hirt, + Hatt Hats = 0, Hadi t+ Bats t Hays = 0, (2) Pati tHazet+Hs%3 = 0, Mt+HetH, == 0. Note that if P,, P,, P, are collinear and distinct, then none Of py» Ha» Hy 18 zero. Further, numbers j1, jg, 23, not all zero, satisfying (2) 4PLANES AND LINES 15 exist if and only if every 3-rowed determinant formed from the matrix mmm tl Pa pees eee aicsy % Ys % 1 is zero (Aitken, 28). This condition is commonly written % mw 4% I/=0, % Ye % 1 ieaeeceeeea | % Ys % 1 giving the most concise form of the collinearity condition. 1. Let P,P, have direction (I, m,n); let P? , P$, P® be the projections of P,, P,, Pon OX. Then z—z, = (PYP*) = (P,P), 2,—z = (P*P$) = (PP,)L, and the first of equations (1) follows from (P,P): (PP) =A, 2A Similarly for the others. What happens if 2 = 0? 2. The present section (but not 1) is still valid if the axes are oblique. 3. If in (3) the elements of no column are proportional to those of another, then the vanishing of any wo 8-rowed determinants formed from it ensures also the vanishing of the remaining two. In any case, if the three of these deter- minants each containing the last column are zero, then the determinant of the first three columns is also zero. 7. Coplanar Points We prove: Points P,(2,, Ys, 21), Pala» Ya» 2a)» Palas Yas 2s)» Pala Yu 2%) are coplanar if and only if numbers py pas Har Har not all zero, exist such that Mat tHetetHots tHe = 0, Hiab HaYe tT Has t Has =O, Q) bata heats thats tots = 9, bate test He =0.16 ANALYTICAL GEOMETRY OF THREE DIMENSIONS Suppose P,, Py, P,, P, are distinct; otherwise the theorem is trivial. The four points are coplanar if and only if the lines P,P,, P,P, have a point P(x, y, 2), say, in common, or are parallel. If P exists, then by 6 there exist numbers pt, (2's fy» Ma Hay pt Such that uxt pit, + pats = 0, wat peta tte = 0, py tii tHe =, | (py piytpaye tHe =O | (gy etme tpate =0, wetpets tpt =9, ett, = =05 w+pethy = =. Now p #0; otherwise it would follow from (2) that Py= Py. Similarly yx’ #0. Therefore, by absorbing 2 suitable con- stant into p’, ps, jy We can without loss of generality suppose p+p’=0. Then by adding corresponding equations in (2), (3) we obtain (1). Conversely, assuming (1), we can reverse the algebra and derive equations of the form (2), (3), thus establishing the existence of P. Hence the theorem is proved. Farther, by a well-known theorem of algebra, numbers Ha Ha» Ha» Hy Dot all zero, satisfying (1) exist if and only if Bn a 4 =0, my Ya 29 ? : : %% Ys % 1 ® mH % 1 giving the most concise form of the coplanarity condition. 1. P, is the centroid of “masses” (positive or negative) Hy Hy Hs at Py, Py, Po 2, If P,P,, P,P, are parallel, equations (1) hold good with Bytes = 0 = Bathe 8. General Equation of the First Degree ‘The general equation of the first degree (linear equation) in x, y, 2 is (a, y, 2) = aztbytoz-+d=0, » @ where a, 6, ¢, d are any given constants, a, 6, ¢ being not all zero. The values of =, y, 2 satisfying (1) depend only onPLANES AND LINES 7 the ratios a:b:c0:d, i.e. on three independent constante. There is an infinite nuniber of points satisfying (1); call their aggregate II. We say that IT is the locus of equation (1), and that (1) is the equation of IT. We here introduce the device of using a single symbol, in this case IT, for the locus of an equation and also for the L.b.s. of that equation, giving in this case the contracted form II = 0. This proves useful, and is found to introduce no confusion. 1. Not all points of II are collinear. The equation of every plane is linear. For let A be a given plane and let P,(2,, y,, 21), Pola Yar 22), Ps(tas Yas 24) be any three non-collinear points of A. Then P(z, y, z) belongs to A if and only if it is coplanar with P,, Ps, Ps, se. from 7 (4), if and only if x, y, z satisfy z yz 1j=0 mm m4 1 Ze. Ye 2% 1 . : » (2) % 3 2% 1 This is therefore the equation of A and is linear. The coefficients of z, y, z are not all zero; otherwise, from 6 (3), P,, Py, P, would be collinear. The locus of every linear equation is a plane. For, TI being the locus of (1), let Py(z, yy, 21), Palas Ya» 29), P(X, Ys, 2) be three non-collinear points of II (ef. 1), Then solving the equations Tey yrs 2%) = 0, TN (ayy Yay 29) = 0, 125 Ya, 29) = 0, for a, b, c, d we get id= nalli—laallslamdy:—lamal. Ye%1 2,1) [az y_) 292%) (3) Ys 231 %y2y1) |x ys 1 23 Ya 2% From 6, these determinants are not all zero since P,, Py, Ps are not collinear. Hence, substituting these values for18 ANALYTICAL GEOMETRY OF THREE DIMENSIONS a, 6, ¢, din (1), it assumes the form (2). 80 I is the plane P,P,P3. Rectified equation. Let II be a plane, P(x, y, 2) any point of II, N the foot of the normal from O to II. Also let (Z, m, n) be the direction of ON or NO, and let (ON) =p (this being positive or negative according as (, m, n) is the direction of ON or NO). Expressing the fact that (ON) is the projection of (OP) on ON, we have le-+my+nz=p. . . .- @ This is the equation of II in what is called rectified form. Conversely, if II is given by (1), then comparison with (4) gives Ya = mjb = nfe = —p/d, (ON) = —djy(at-+0t-+0%), . . (6) whence measured in the direction (a, 6, c) (a8 +5840). Therefore a, b, ¢ are direction-ratios of any normal to Il; if (1) is written so that d is negative, then (2, 6, c)-+/(a?-+b*-+e%) are the direction-cosines of ON. 2. M being the foot of the normal from any point Q(z’, y’, 2’) to the plane (4), we have (MQ) = lz'+my'-+nz’—p, measured in the direction (2, m,n). [For (Q) = projection of (OQ) on (J, m, n)—projection of (OM) on (I, m, n).] 3. Using (5) and 2, the perpendicular distance of Q from the plane (1) ie (ax! by’ +02’ +d)+ (at +b* +0"), this expression being positive for all Q in the region on one side of II, negative for all Q in the region on the other side.PLANES AND LINES 9 4. Form the equations of the planes bisecting the angles between two given planes. 6. The plane making intercepts a, £, y on OX, OY, OZ has equation alat+y/B-+e]y = 1. 6, The plane through (z, y, 2’) parallel to two given directions (Ay, fy ¥1)) (Ags Hg» %) is 2! y—y’ 2-2! A AO 4a 7. Area of triangle. Let a be the area of the triangle with vertices B(zs, yx, 21), Clzs, yx Zs) Dizu yu %); let the direction (2, m,n) be normal to BCD. Then (+)al is the area of the projection of ABOD on OFZ. But the projections of B, C, D on OYZ are the points (ys, 2%), (Ys 23), (Yas %) in O¥Z; 80, by o theorem of plane geometry, the area of the Projection is (£)t =0 ws 4% 1 ww % 1 Mu % 1 Using corresponding resulta for OZX, OXY, we have, since Bim+nt = 1, dat—ly, 2, 14/2. zy Lb) ae yy 1] % 2% 1) Jz 2, tm 1 (8) Mm % 1) lem 2 me % 1 8. Volume of tetrahedron. Let % be the volume of the tetrahedron with vertices Alty Yu 2) Blt Ys 24), Claas Ys» 24), Dlty yy 2%). The equation of BOD is l2 y 2 Lesalyaz, 1) tule. 0.1) +2/2. 9.1! —Japy, 4] = 0. Tr Yat) Yatl 24% 1 Tysd Ta Ya ty Fa ¥stsl Yel 2% tyl Re He Mead From 3, the perpendicular distance p, of A from BOD is got by writing 2, y1, 2, for 2, y, 2 in Lbs. of (7) and dividing by the ‘square root of the sum of the squares of the coefficienta of 7)20 ANALYTICAL GEOMETRY OF THREE DIMENSIONS 2, y, & By (6), the latter is 2c. Hence, since Y = tp.c, we obtain V=tln nu a m Ye te Ze Us te MM mM ‘This has been derived without regard to sign. An argument like that giving the sign of 7 in 5 4 shows: The axes being rh, if rotation in the sense B-> 0+ D carries a r.h screw towards A, then given by (8) is positive, in the contrary case negative. 9. Recover (8) by methods analogous to those used in plane geometry to get the corresponding result for the area of a triangle, 10. The volume of the tetrahedron whose feces have equations ae+b.y-+e2-+d,= 0, ae+...=0, az+...= 0, ag+...=0 is K*/6D,D,D,D, where K is the determinant |; d,.¢,d, | and Dy, ... the cofactors of d,, ... in K. [We use the familiar device of indicating a determinant by its principal diagonal. Express the coordinates of the vertices in terms of cofactors of K, substitute in (8), and apply Jacobi’s theorem (Aitken, 42).] 11. In the tetrahedron ABOD, let a, b, 0 bo the sides of AABO, d, ¢, f the edges joining A, B, C to D. Take D as origin, then (8) gives (8) wee 2889" = 82, y, 2% ).)a, oe oe Boy tal |y Ya Ye Ze Ys Ze] [21 Za =8) cltulted Bertyyr tats Mite t ye tees aetyantee th tyites Bers tyiye teres aetynten aetyytet | ateate 2a Peta a4 der—ot Ret efit a4f—b ot+ft—at ap thus expressing the volume of a tetrahedron in terms of its edges. (This example is given as an application of the theory to establish a result independent of any coordinate system.)PLANES AND LINES a Hence, or otherwise, expresa 2) in terms of three concurrent edges and the face-angles at the common vertex. 9. Incidence of Planes Let four given distinct planes be Tl, Saz+bytez+d,=0,. . (I) Tl, Sagz+byteztd,=0,. . (2) TI, Ss ayetbytezt+d=0,. . (8) TSogtoytoet+d=0.. . (4) Suppose IT,, II, are not parallel; then they have in common a line 4,9, say, and this must consist of all points satisfying (1), (2) simultaneously. Now consider the equation Tekl,+ml=—0, . . (6) where k,, k, are any constants. (5) is linear, so Il is a plane. Also TI = 0 for all (z, y, z) for which II, =0, 1, =0 simultaneously; therefore II contains 4,9. Further, Q(z’, y’, z’) being any point not on 4,;, II contains Q if Ky (aya!' +byy' +e2' +-0,) +hy(age’ +bay' +ey2' +d.) = 0, and this determines £,:%, uniquely. Since any plane through 8,3 is determined when one of its points not on 55 is given, it follows that every plane through the intersection of TI,, Tl, ts expressible in the form (6). 1. TI, T1,, II, contain a common line, or are parallel, if and only if non-zero numbers &,, &,, k, exist such that 4,11, +411, +411, = 0. 2. Ti, T1,, TI, are parallel to some-line if and only if la, be | =O. 3. The planes y—pz = p, de—ve = g, pz—dy = ¢ contain @ common line if and only if Ap+yq+sr = 0. This line lies also in the plane px+gy+rz = 0. Suppose II, is not parallel to 4,5. Then it meets a4, in a point P,, which is the only point common to II,, Ig, I; ita coordinates are got by solving (1), (2), (3).22 ANALYTICAL GEOMETRY OF THREE DIMENSIONS 4, Every plane II through P, can be written T=4,1,+411,+4II, = 0, where ,, ky, ks are constants. (Contrast 1.) Finally, II, also contains P,, i.e. the four planes have at least one common point, if and only if (Aitken, 30) [a o o d|=0. . > 6 5, If Ty, Ty My, I1,are not concurrent, and no three are parallel, then any plane II can be written Tsk, +e. +h thls = 0, where k,, ... ate constants. 6, Quadriplanar coordinates. II;, Iq, Ms, 11, being as in 5, constants a}, ay, , 6, exist such that esall,+al,tall,tall, . « vy=AM+61.+60+80y «5 (7-2) zevllt+yll.t+yIL+rM. 1=6,11,+6,11,+6,11,+6,I1,. In fact, the condition that (6) is not satisfied makes possible such a unique solution of a,2-+by-+co2+d,1 = Il, etc. (7.1)-(7.8) provide a transformation from 2, y, z to new “coordinates” II, Ty Ts TI, connected by (7.4). Any polynomial of degree n in z, y, z is expressible as a homogeneous polynomial of degree n in II, II, Ts, Ta ‘Now suppose (1)-(4) to be in rectified form, ao that IT, ... are the perpendicular distances of P(x, y, z) from these planes; let II, be positive when P, P, are on the same side of II, and soon. I],, ... are then called quadriplanar coordinates of P, analogous to trilinear coordinates in two dimensions. The four planes enclose a tetrahedron; 4,, 4,, 45, 4, being the areas of its faces and @) its volume, we have by elementary geometry (11,4,+11,4,+11,4,+11.4,)/3V = 1, showing the geometrical meaning of (7.4). 10. Equations of a Line We have seen that all the points which satisfy simul- taneously two linear equations form a line. Conversely,PLANES AND LINES 23 let @ be a given line and I,, II, any two distinct planes through 8. Then a point lies on s if and only if it satiafies simultaneously I,=0, M,=0. » « @) 8o we call these equations, taken together, the equations of 8. These equations are not unique, but may be replaced by those of any other pair of planes through e. Some standard forms are given in 2-4 below. ‘This is our first instance of a locus being specified by more than one equation. But a single algebraic equation cannot suffice to specify a line. For there are in general infinitely many values of z, y (say) satisfying such an equation for any given value of z, whereas on a line there is in general a unique point for which the coordinate z has a given value. Therefore at least two equations are required. So we could not specify line more simply than by two linear equations, and we have seen that these suffice. In the field of real numbers, a single equation AAR Hh =o is, in fact, equivalent to the set f, = 0, fy = 0, .. +2 fa = 0. But when we speak of a single equation wo shall exclude such cases.® 1, The equations of a line depend on four independent constants. 2. # being not parallel to z = 0, ita equations can be put in the form weeata, y= fze+f’. ‘These define ¢ by its projections in the planes y = 0, = 0. %. The line joining two given points Piz, yi, 2) Pity Yu 21) is Mmmm MMe mee emt ym oA * Except where they arise as certain special cases of the equation of the seoond dogree. eae .24 ANALYTICAL GEOMETRY OF THREE DIMENSIONS 4, Tho line through the point (2’, y’, 2’) in the direction (A, Hs, ») is, using 4 5, soe yoy oe 5. The line in 4 lies in the plane ax+by+ce+d = 0 if aatbuter=0, ax’ by’-per’+d = 0. [Algebraically: put each ratio in 4 equal to ¢ and make (@, y, 2) lie in the plane for allt. Geomstrically: the direction (A, #, ¥) must be parallel to the plane, and the point (2’, y’, 2’) must be in the plane,} 8, Btate conditions for the line in 2 to lie in a given plane, 7. Show how to write the intersection of aetbhytoz+d =0, a,2-+by+oz+d = 0 in the form given in 4, a, +, ¥) being perpendioular to (G1 bis €1)s (Gay bas 04) is given by (itr biey yG,—Cy, ab, —a yb). If then, say, » # 0, give z any value 2’ and solve for 2, y, thua obtaining a point (2’, y’, z’) on the line.) 8 ‘The perpendicular distance of Q(z", ¥, 2°) from the line through P(z’, y’, 2’) with direction (1, m, n) is VAZLy’—y"n—(2’—2")m}}}. [This is | PQ|ain y where ¥ is the angle between PQ and the line.) Shortest distance between two lines. Let two non-parallel lines 4,, 8, be given by @—a)/A = U-Wiih=@—a)oy (1) (e—2g)/%a = (Y—Yal/ta = (@—29)/vy- - (2) ‘The single direction § perpendicular to 4, 6 is (11,75 —psy, VyAg—VgAz, Aytg—Aggy). The plane Tl, containing a andPLANES AND LINES 3B parallel to 5, and the plane IT, containing 4, and parallel to 8 are, respectively, (8 6) =~ ™y yh az, |=0, A Mh “Yn aMs—HaMy MaMa Ved, Auta — Aaa @) t—2y y—Ys Z—2y =0. Ha Ve Bava PaY, MAg—VoA Aaa Apes Since I1,, IT, are both parallel to 5, their line of intersection is parallel to 5; since IT, contains 6,, ¢ meets 8,; similarly 8 meets 8. Therefore (3) are the equations of a unique line 2 perpendicular to 8, 8, and meeting them both (in Ny, Ng, say). P,, P, be any points of ¢,, 83 respectively; let p be the angle between P,P,, N\Ny. Then ys =0 only if P,=N,, P,=N,; otherwise P,P, would be a second line per- pendicular toe, ¢,and meeting both. Now P,N,, P,N, being perpendicular to N,N, the segment N,N, is the projection of the segment P,P, 0n N,N,. So| N,N,| = | P,P, | cos p, whence | V,N,|<| P,P, /| unless P,=N,,P,==N,. Therefore {4,2 | is the shortest distance between 6, 82, a8 is well known from elementary geometry. Taking in partioular P,, P to be (2, Ys. 21), (tas Yas %) and using 410, the projection of the segment P,P, on s is : (EYE —2 (Ha Har)} (La Ha), 4.2. (4) |[%s—%) eth %—% Mh yy Pe Va giving the length of the shortest distance. It follows that 84, 89 intersect if and only if ny WW A a ary vw Ha Vg AV (Hao Ha1)? (ry Ag —vgAy)® FOAaHa Aas), =0. or (4)6 ANALYTICAL GEOMETRY OF THREE DIMENSIONS 9. Recover (4) by forming the equation of the plane through a, parallel to e, and finding the perpendicular distance of (21) Ya» Z) from it. 10. Equating the ratios (1), (2) to ky, fy any points P,, P, Of By, By are (2+ Athy Yattaky thy) (at Ak Yat tale, Zq+¥ekq). Determining hy, %, to make P,P, perpendicular to &y 8, find the coordinates of Ny, Ny and the length and direction of N,N. The equations of 8,, 6, can now be put in a simple form which facilitates the solution of particular problems. Take origin O the midpoint of N,N,; OZ along N,N,; OX, OY parallel to the bisectors of the angles between 1, %, Then the equations of ¢,, 8, become y=mr, 2=ky y= —ma, 2=—k,. (8) where 2k=|N,N,|, and 2 tan“) m is the angle between By, 83. 11. Find the locus of a variable line # meeting fixed lines 4, 8 in P,, P, 80 that | NP, | =| NP, |, Ni, Ny being as defined above. [Taking axes as in (6), P,, P, have coordinates of the forms (z,, mx, k), (%» —mt%y, —k). The condition INP, 1 = [NPs | gives 2, = tz. With 2, =a, P,P, is (e—2,f0 = (y—m,)}me, = (eb), Eliminating 2, required locus is mzx—ky = 0, referred to this particular coordinate system. With z, = —2,, it is yz—kmz = 0. Later work will show that each locus is a hyperbolic paraboloid, and that this could be foreseen geometrically.) 41. Parametric Representation The reader will recall the introduction of parametric (or “‘freedom”’) equations of a locus in two dimensions and the advantages of their use. Familiar instances are the representation of a line by equations of the form z = a+pt, y = b-++qt, and of a parabola by z= af, y = 2at. Each of these is an example of a rational representation, having the fundamental property that it establishes a one-to-one correspondence between the points of the locus and the values of the parameter ¢. So we may specify completelyPLANES AND LINES 7 a point of the loous by the corresponding value of t, referring to it, if we wish, as “the point ¢.” One and only one parameter is needed because a point on the locus has just one degree of freedom. In three dimensions a curve (a line being the simplest case) is atill by definition a locus on which a variable point has o single degree of freedom. The coordinates of such & point must again be expressible in terms of one parameter. A surface (a plane being the simplest case) is by definition a locus on which a variable point has two degrees of freedom. The coordinates of such a point must be expressible in terms of two independent parameters, In each case the expression of the coordinates in terms of the parameter or parameters need not be rational, or even algebraic. But if it can be made rational, then the curve or surface is said to be rational. Obviously, not every parametric repre- sentation of a rational locus is rational. Various features of these conceptions will receive subsequent illustration. 1. An arbitrary point in @ has three degrees of freedom; a point satisfying a single equation has two degrees of freedom; a point satisfying two equations has in general one degree of freedom; point satisfying three equations has in general no freedom, but must be one of a finite number (if the equations are algebraic) of fixed pointe, [Interpreted geometrically these results are: the locus in & of a single equation is a surface; if two surfaces intersect they do so, in general, in a curve; if a curve and a surface intersect they do so, in general, in isolated points. There are, however, important cases of exception, ¢.g. two or more surfaces may have a portion of each surface in common, three or more surfaces may have a curve in common. But the analytical formulation of these cages is too difficult to be given here.] Line. Lets bea given lino; J,m,nited-o’s; P,(%, 41,2), Pula, Yor 2a) fixed points of ¢; Plz, y, z) any point of s. (P,P) =r, measured in the sense (i, m,n). Then thero is a one-to-one correspondence between the positions28 = ANALYTICAL GEOMETRY OF THREE DIMENSIONS of ‘P and the values of r(—o @ contain siz constants which do not depend quite simply on 2 set of four independent constants sufficient to determine a.PLANES AND LINES 31 In particular, (z', y’, z') may be any fixed point on s. However, we obtain from (1) vy —p2 = vy'—pa!, Neve = delve’, } @) pay = p2'—dy’, : so that, (x”, y’, 2”) being any other point of s, we have vy" —p2" = vy'—p2', ete. Hence, writing vy’ —p2' =p, de’ —v2e'=q, p2'—dy'=r,. (3) the values of p, g, r are the same for all positions of (2’, y’, 2’) ons. Also from (3) we have identically Aptpgtwr=0.. 2. . 4%) Finally, ¢ is unchanged if we multiply A, 1, v, and con- sequently p, 9,7, by any non-zero constant. Therefore, to ® given line there corresponds a set of six numbers A, y, v, P; ¥ 7, Whose ratios are uniquely determined and which satisfy the identity (4), thus involving four independent quantities. Conversely, given any set of values of A, Hv p,q, 7 satisfying (4), we can form the equations (2), 4.6. vy—pz =p, M—ve=q, pe—ly=r,. (8) and these determine a unique line (93). We call A, ba Y, P, 9, 7 the coordinates of the line.* 1, Express the line-coordinates of the join of two given points in terms of the coordinates of these points. 2, Express the line-coordinates of the meet of two given planes in terms of the coordinates of these planes. 3. The shortest distance between the lines a(4,, py vy Pur Qu tid 81(Aas Has Yar Par Yer Pa) 8 Apa tings trite tA tags teary a {(4:%2 Ha) + (da aA) atts — Ae 81, 8, intersect if and only if Aprt agers t Api tan tot, = 0. [Expand the determinant in 10 (4) using tho definitions (3).] * Appeal to a mechanical analogy clarifies tho significance of lino-coordinates; see Salmon, Analytic Geometry of Three Dimensions, § 63.OHAPTER III SPHERE Tus chapter is inserted in order to give elementary illustrations of the use of the results already found, and to provide some introduction to later general theory, of which it gives particular cases. 14, Sphere referred to its Centre Let S be ® sphere of radius a. Take axes through origin O at the centre of §. Then P(z, y, z) is on S if and only if | OP | =a, or e@tytetaa, , . . (1) which is therefore the equation of $. Let Q(2’, y’, 2’) be a fixed point, and consider the line s through Q with direction (1, m,n). Any point P(z, y, z) of is given by (42(1)) z= 2'4r, y=y'+mr, 2=2'-+nr, where r= (QP). P lies also on if (e! lr)? + (y’ inn)? +(e! -+nr)? = a3, #24-2(la my’ --n2'\r-+aty242"—a9 = 0, (2) This is a quadratic equation for r having roots r,, r, (say), both real or both complex. If r,, rz are real and distinct, 8 meets § in two points P,, P, (say). If r; = 1, (then necessarily real), s meets 9 in one point P, which may be regarded as given by the coincidence of P,, P,; we say that 8 touches S, or is tangent to S, with point of contact P,. If 11, Tg are not real, does not meet S. 32 fe.SPHERE 3 From (2) we have (QP,).(QPs) = rita = 2" +y"*+22—a*, . (8) Hence if any line through a point Q mecta a aphere S in points P,, P, the product (QP,).(QP;) depends only on Q, 8. We call this the power p(Q ; 8) of Q w.r.t. 8. Equation (3) may be interpreted as p(Q; 8) = (distance of Q from centre of S)*—(radius of S)*. Therefore p(Q; 8) is (i) <0, if Q is inside 8, and (3) shows that Q then lies between P,, Ps; (ii) = 0, if Q is on 8; (iii) > 0, if Q is outside S, and (3) shows that Q is then not between P,, Py. Ifr, =f, (3) gives p(Q;S) = (length of tangent from @ to 8)? > 0, and Q is then not inside S. ‘Again from (2) we have (QP,)+(QP,) = rte = —2(le'-+my'+nz’), (4) Qis the midpoint of (P,P,) if (QP,) = —(QPy), i.e. from (4), if le'+my'--nz' = 0, Therefore, if 1, m, m are fixed, Q must lie in the plane lz+-my+nz = 0. This contains O and is normal to (J, m, n). Hence the midpointa of chords of a ephere parallel to a fixed direction lie in the diametral plane normal to the latter. Now let @ bo on S, so that x’?+y"+2'? = a*, Then the roots of (2) become r; = 0, 1, = —2(lx'+-my'-+nz'). ‘The vanishing of r, means merely that every line through Q meets $ in Q, ie. P,=Q. Then P,=P,, making » tangent to S at Q, if and only if r, = 0, giving la! 4 my! +nz! =0. This expresses that (/, m, n) is perpendicular toOQ. Hence a line through o point Q on & is tangent to S (at Q) if and only if it is perpendicular to the radius through @. So34 ANALYTICAL GEOMETRY OF THREE DIMENSIONS every tangent line at Q lies in the plane through Q normal to OQ, and every line through Q in this plane is a tangent line at Q; this is called the tangent plane at Q. These properties show that ite equation is 2 e—2)tyy—-y) 42-2!) = 0, or, since 2’?+-y'2-+-2'? = a3, wa! +yy'+22' = a2, ‘ - (8) Suppose now that (5) passes through a fixed point R(x’, y’, 2’). Then 2’x"+y'y"+2'z” = a, showing that Q then lies in the fixed plane we" -+yy" +22" = a, . e (8) Hence the points of contact of tangent lines to 8 from a fixed point R lie in a fixed plane. This is called the plane of contact of R and the locus of these tangent lines the tangent cone from R, It is clear that the foregoing geometrical theorems, which are trivial in themselves and are reproduced merely to illustrate the algebraic method, are independent of the partioular choice of axes used in deducing them. The form of their algebraic expression, on the other hand, does depend on this choice. 1, Tangent lines of S pass through a given point R if and only if R ia not inside S. (For the plane (6) meets S if and only if its perpendicular distance from O is not greater than a, de. 2*+y"*+2"!—a > 0.] 2. Polar plane. Let P,(2,, yy, 2,), Palzss Ys 25) bo such that the sum of their powers ws.t. 8 ia equal to the equare of their mutual distance. Then Rta t ys bets = oh If P, is fixed, P, must therefore lie in the plane wetyyite, say. 6. (7) and conversely, if P, is any point of (7), P,, P, are related in this way. [(7) is called the polar plane of Py w.rt. 8. This is‘SPHERE 35 not @ standard definition, but moat such definitions fail to apply generally to every position of P, in the polar plane.) 8. Using (7), establish the properties of pole and polar in 33 for the particular case of a ephere. 45. General Equation of a Sphere The sphere with centro (a, f, 1) and radius a is seen to (z—a)*+(y—BP+e—y)* » @ This equation has the form ext toy? tez*+2uz+2oy+2uz+d=0, . (2) where c, d, u, v, w are constants (¢ # 0). Conversely, given any equation of the form (2), we may write it (w+ufe)?-+(y-+u/e)* +2 +u/e)* = k, where } - @) b= (ut-to8--wt—ed)/e2. Comparing (3) with (1), we see that ita locus is the sphere having centre C(—u/e, —v/e, —w/c) and radius /k, provided k>0. The case k =0 gives a sphere of zero radius (point-sphere) reducing to the single point C. There are no points which satiafy (3) if k< 0. Hence the equation of every aphere is of the form (2), and the locus, when it existe, of every such equation is a aphere. Wo usually assume (2) to be “normalised” so that c = 1. 1, Tho general equation of a sphere depends on four independent constants. One and only one sphere passes through four given non-coplanar points. By the methods of 14, or by merely changing the origin in the reaulta of 14, we prove: § being given by (2) with c= 1, and Q being (z', y’, 2’), the power of @ is given by PQ; 8) = 22 4y42'942ua! + 20y'+2u2'+d; (4)36 ANALYTICAL GEOMETRY OF THREE DIMENSIONS and the equation se! yy’ +22! ulate’) -+oly-+y’) +ole-+2')+d=0 gives the tangent plane at Q if Q is on 8, the plane of contact of Q if Q is outside S, the polar plane of Q for all Q. 16. Circle If a plane and a sphere intersect, they do so in a circle. So the locus of a'z+b'y+e'z+d' = 0, } @ w-+y?+zt+2ue+2oy+2uetd=0, f° taken together, when it exists, isacircle. Conversely, every circle can be represented by equations of these forms. As in the case of line, more than one equation is needed to specify a circle in &. We cannot specify a circle more simply than by one equation of the first degree and one of the second. In this specification the former equation gives the unique plane containing the circle, but the latter need not be that of a sphere; it may be a quadrio of any type possessing circular sections (44). 1, Find the condition for the circle (1) to exist, and find ite centre. [The distance of the plane from the centre of the sphere must not exceed its radius.) 47. Radical Plane Let 8, = x2+y*+2942u,2+20,y-+2w,2-+d, = 0, Sq = e8+-y? +22 +2ugr-+20,y -+20,2z-+d, = 0, be two given spheres, and Q(z’, y’, 2’) any point. Then P(Q; S;) = p(@ ; S,) if and only if, using 15 (4), a py9p2'8 4 2uye’ + 2ry! +202’ +d; Satya? ue! + ony’ +m’ +dySPHERE 37 i.e. if and only if Q lies in the plane Qtr, —teg} 2 +2(r —vg)y +2(0,—t)2-+d,—d, = 0. (1) This is normal to the direction (tt, —tg, 0g, %4—t,) and 80 to the line of centres of S,, 8,. Also, if R is any point common to 8, Sz, then p(R ; 9) = 0 = p(R; 8); 80, if 8, Sq intersect, their intersection lies in (1). Hence the locus of a point having equal powers w.r.t. two apherea is a plane normal to their line of centres and containing their intersection, if any. It is called their radical plane, 1. Let S, = 0 (¢ = 1, 2, 3) be the normalised equations of three given spheres. Tho three radical planes of Sy Sy Ss taken in pairs meet in a line [given by 8, = 8, = S, and sometimes called their radical axis), or sre parallel. 2, Let S; = 0 (i =1, 2, 3, 4) be the normalised equa- tions of four given spheres. The six radical plancs of 81, Sy S,, S, have in general 6 point in common [given by 8, = 8, = 8, =S,], and called the radical centre, Consider cases of exception. 418. Coaxial Spheres Asphere Sand a plane II being given by 46 (1), consider the equation S\(x, y, 2) = S+2ATT = xt y*+2242(ua’)x $2(v-+Ab')y +2 (ww-+Ae')e-+d-+-2Ad’ = 0, where A is any constant. It is the equation of a sphere 9, Let Q(z’, y’, 2) be any point of II so that II(z,’ y,’ 2’) = 0. Then (45 (4)) PQ5S,) = Se’, y', 2!) = Se’, y's 2) +2ATT(@’, y’, 2’) =S@',y'2')=% say, where g depends on Q but not on A. Hence as A varies the spheres S, form a family such that the power of Q w.r.t. every member is the same, é.e. I is the radical plane of every pair of members. It follows that the spheres have also a38 = ANALYTICAL GEOMETRY OF THREE DIMENSIONS common line of centres. They are said to constitute a 1. Give the coordinates of the centre of S,; verify that it lies on a fixed normal to I]. 2, There is one and only one sphere S, through any point not on II. Either no two spheres of the system intersect, or else they all contain a common cirele in II. 3. If any lino ¢ through Q, but not contained in IT, meeta Sin Pi PL then (QP,). (QP%) =q; hence, as A varies, P, Py generate an involution on 8. If Q is outside the spheres, two of them touch ¢; if @ is inside the spheres, none touches S. 4, There are two point-spheres (limiting pointe) of a non- intersecting coaxial system. 5. Orthogonal spheres. Spheres 8 =x wtty'+2"+ 2ux + Qvy + 202 +d = 0, SS z+ y8+294 2u'2 + 2v'y+2w'2s+d’= 0, intersect orthogonally at every common point if and only if 2uu’+2vv’+2ww’ =d+d’. [This is a simple extension of the corresponding result for circles in a plane.) 6. If a sphere S is orthogonal to each of two epheres S,, Sy, its centre C lies in the radical plane I of S,, 8, Conversely, if lies in II, and if S is orthogonal to S,, it is orthogonal to Sy end so to every member of the coaxial system by S,, II. 7. All the spheres orthogonal to three given spheres form in general coaxial system whose radical plane is the plane of centres of the given spheres and whose line of centres is their radical axis. 8. There is in general one and only one sphere orthogonal to four given spheres, 19, Parametric Representation Let T* be o fixed plane and let wu, » be cartesian coordinates of any point P® of II* referred to axes in II* itself. We saw (11 6) that the parametric representation of any other plane I yicldsa*‘map” of IT on II*. Conversely,SPHERE 39 if we can map any surface © on II* we shall obtain a parametric representation of Z. For, if P(x, y, z) is any point of Z and P®%(u, v) its “map” in II®, then x, y, 2 must ‘be expressible in terms of u, v. One way of mapping a sphere 9 (centre O, radius a) on Tl® is by stereographic projection, the simplest case being the following: Take as II® any plane through O. Let A be one extremity of the diameter normal to II*; then the join of A to any other point P of S meets II in one point P*, and conversely. This gives a one-to-one correspondence be- tween P, P®; so P® may be taken as the “map” of P. Now take axes with origin O and OZalong OA. Then Tl* is the plane z = 0; A is the point (0, 0, a); S is the sphere w+yi+2—a% . . sed) Let P* be the point (u, v, 0); so AP* has parametric equations z=u, y=u, z=a(l—t).. . (2) This meets 9 at points where, from (1), w2tv%+a%(1—1)? = a4, whence t = 0 or 2a?/(u2+v?+a%). The first root gives A; therefore the second must give P. Substituting for ¢ in (2) the coordinates of P are 2a%e _ __ 2a _ ete =p Yap pa 7 Ep pa® @) Also (2) gives u=azfa—z), v=ayfe—2). . (4) (8) is a parametric representation of 8. (3), (4) render explicit the (1-1) correspondence between P, P*; if u, v aro given, (3) determines unique values of , y, z satisfying (1); if x, y, 2 satisfying (1) are given, (4) determines unique values of u, ¥, With the single exception that no point of II* corresponds to A on 9.40 = ANALYTICAL GEOMETRY OF THREE DIMENSIONS 1. The ourves & = constant, v = constant, on § are two families of small circles through A, one member of each going through every point of § other than A. 2. A more familiar method of locating a point on 8 sphere is by ite latitude and longitude. The following slight modifica- tion of this is used in mathematica: In the frame of reference used above, call the angle ZOP the colatitude @ of P, and the angle between the planes OZX, OZP, in the r.h sense of rotation about OZ, the azimuth ¢ of P. Then, if P is given, 6, $ are uniquely determined ao that OC O< 9, OG ¢ < 2n; if any pair of values of 8, ¢ in this range is given, it locates & unique point P on §, Hence 6, ¢ serve as parameters of P. Projecting OP on the axes we obtain the parametric equations z=asinOeos$, y=asinOsing, 2=acos8, (5) This non-algebraic representation is not so useful as (3) in algebraic geometry. 3. Spherical polar coordinates. Let Q(z, y, 2) be any point of G. Let P be either of the two points in which 0Q meets S, and let P be specified by 6, $ asin 2. Taking OP as fixing the positive sense of OQ, let (OQ) =r. Then as in (5) we obtain z=rsinOcos¢, y=rsinOsing, 2 =rcosd. ‘Thus Q is determined uniquely when r, 0, ¢ are given; these are called spherical polar coordinates of Q in the frame of reference used. Note however that, if Q is given, r, 0, ¢ are not uniquely determined, owing to the ambiguity in the choice of P. 4, If in 3 we take rsin @ = p, Q is determined uniquely when p, ¢, 2 are given; these are called cylindrical polar coordinates of Q.CHAPTER IV HOMOGENEOUS COORDINATES—POINTS AT INFINITY Tue preceding work suffers from certain inelegancies of form and expression: of form, since we specify any point by unique set of three point-coordinates, and go on to specify any plane by a non-unique set of four plane- coordinates, and any line by a set of line-coordinates of non-uniform type; of expression, since we have to make continual mention of special cases depending on parallelism. The simple algebraic device now to be introduced removes these defects and simplifies the subsequent algebra. It does not mean that we pass to a different geometry. In particular, we do not discard the special cases due to parallelism; they merely cease to be exceptional as regards the general treatment, but can be isolated whenever necessary. This is well illustrated by the successive stages in the classification of quadrics (Ch. V). 20. Homogeneous Coordinates Suppose we have a collection @ of geometrical objects and we label them with sets of numbers (9,, 92, ...,9,). If {i) to every sot of values of g;, ..., ga in a stated range there corresponds a unique member of &, which we call “the object (91, «. .» dn),” (ii) the objects (91, .-+ 95)» (Ji ---+ 95) are the same if and only if g;/91 = ... =g)/9%, (iii) this system of labelling includes all the members of @, then we call g;, ..., gq homogeneous coordinates of the corro- sponding object in 2. al42° ANALYTICAL GEOMETRY OF THREE DIMENSIONS Since the set of values 0, 0, ..., 0 could be reckoned proportional to any other set, it must be excluded from the range of possible values of g,, ...5 Jn Now let some geometrical relation concerning the object (91, »++s 94) be expressed by an algebraic equation in- volving these coordinates. Then that equation must be satisfied when these coordinates are replaced by the “equivalent” set (gj, ..., 94), where 93/91 =... = gi /g%, since this set specifies the same geometrical object. Hence the equation must be homogeneous in gi, ..., Gas 24, Linear Dependence We denote a set (93, ...,9n) by the symbol g, and a set (gy, .-.+5 gn) by kg. We can without confusion denote also by g the geometrical object corresponding to the set g; then g and kg denote the samo object (k 0). Sets g’, g", ..., g!7 are said to be linearly dependent if there exist numbers £’, k*, &, not all zero, such that Kg! +h'g’+...+hgit=0,. . (1) where this implies the » equations GAM OG +. HOG) = 0. (= 1,...,0) (2) If in particular k), say, (1

) is linearly dependent on the remaining g’s in (1). Sets which are not linearly dependent are called linearly independent. We require the following lemmas: (i) There exists a g linearly dependent on u, v, ..., and if and only if u, v,...,y, % ... are linearly ongst themeclves. Thia follows from the (i) Any m of the g’s are linearly dependent if m>n (Aitken, 28 3). (iii) At least t10 of the X's in (1) are not zero, For if all the k's except k) are zero, (2) reduces to Kg? = 0 (i = 1,....0)-HOMOGENEOUS COORDINATES—POINTS AT INFINITY 43 But we have excluded the possibility g? =... = gM = 0. Hence k(5) = 0, and wo have a contradiction. (iv) Two geometrical objects g’, g” are identical if and only if g’, g” are linearly dependent, For, if k’g’+k’g” = 0, then, by (iii), &', k” # 0, 80 gil? =... = g'/9%, and conversely. Bearing in mind this lemma, we may speak of linear dependence of sete of coordinates in terms of the corresponding geometrical objects. 22. Point-coordinates : Space é Let P(z, y, 2) be any point of @. We shall now label P by four numbers 2, 23, #3, 2, such that ce afm, yeas, t= fy. ~ (1) If, in the definition in 20, @ is the aggregate of points of & and 2%, .., % can take all values, then we see that %y» Zqy %y, Zq are homogeneous coordinates of P in @, provided we exclude the possibility x, = 0. Suppose now wo define a collection & of objects as those having homogeneous coordinates 2, 2, %, 2%, where these range unrestrictedly over all real values (except only the set 0, 0, 0,0). We shall, as is possible from what has just been said, identify those objects of & for which x, # 0 with the points of &. We shall then choose to call all the objects of & the “points” of €. When it is necessary to make the distinction, we shall call points having x, # 0 “ordinary” points of &, and those having x, = 0 “special” points. The latter are not points of @; so we have hitherto no geometrical definitions concerning them and we are consequently now free to impose any we please (provided they are self-consistent). Their usefulness in respect to & will become evident in 27. Reversion to cartesian coordinates. If =x is any ordinary point P, then, kx being the same point for all & and x,, being not zero, we can take k = 1/z, and so obtain for P, from (1), the particular set of coordinates (x, y, z, 1). We can thus make the 2,, 2, x-coordinates equal to the