r O V^ I o> Digitized by the Internet Archive in 2008 with funding from Microsoft Corporation http://www.archive.org/details/elementsofprojecOOcremrich ELEMENTS OP PROJECTIVE GEOMETRY CREMONA HENRY FROWDE Oxford University Press Warehouse Amen Corner, E.C. ELEMENTS OP PROJECTIVE GEOMETRY BY LUIGI CREMONA LL.D. EDIN., FOE. MEMB. E. S. LOND., HON. F.B.S. EDIN. HON. MEMB. CAMB. PHIL. 80C. + PEOFESSOB OF MATHEMATICS IN THE HNIVEESITT OF BOMB TRANSLATED BY CHARLES LEUDESDORF, M.A. FELLOW OF PEMBEOKE COLLEGE, OXFOED AT THE CLARENDON PRESS 1885 r I All rights reserved ] 1T £7 3 who enunciated and proved the theorem concerning triangles and quadrilaterals in perspective or homology. This theorem, for the particular case of two triangles (Art. 1 7), is however really of much older date, as it is substantially identical with a celebrated porism of Euclid * ' On ne peut se flatter d' avoir le dernier mot d'une theorie, tant qu'on ne peut pas l'expliquer en peu de paroles a u'n passant dans la rue' (cf. Chasles, Apercu Mstorique, p. 115). + (Euvres de Desargues, reunies et analysies par M. Poudra (Paris, 1864), tome i. Brouillon-projet d'une atteinte aux ivinements des rencontres d'un cdne avec unplan (1639), pp. 104, 105, 205. % Loc. cit., pp. 105, 106. § Traite des proprietes projectives des figures (Paris, 1822), Arts. 96, 580. || Loc. dt., p. 210. 1T Philosophiae naturalis principia matliematica (1686), lib. i. prop. 27, scholium. ** Freie Perspective, 2nd edition (Zurich, 1774). tf Loc. cit., pp. 413-416. X AUTHOK S PREFACE TO THE FIRST EDITION. (Art. 114), which has been handed down to us by Pappus*. Homological figures in space were first studied by Pon- CELET f. The law of duality, as an independent principle, was enun- ciated by Gergonne % ; as a consequence of the theory of reciprocal polars (under the name jjrincipe de reciprocite polaire) it is due to Poncelet§. The geometric forms (range of points, flat pencil) are found, the names excepted, in Desargues and the later geometers. Steiner II has defined them in a more explicit manner than any previous writer. The complete quadrilateral was considered by CarnotIT; the idea was extended by Steiner ** to polygons of any number of sides and to figures in space. Harmonic section was known to geometers of the most remote antiquity ; the fundamental properties of it are to be found for example in Apollonius ft- De la Hire \% gave the construction of the fourth element of a harmonic system by means of the harmonic property of the quadrilateral, i. e. by help of the ruler only. From 1832 the construction of projective forms was taught by Steiner §§. The complete theory of the anharmonic ratios is due to Mobius mi, but before him Euclid, Pappus! If, Desargues****, and Brianchon ttt had demonstrated the fundamental pro- position of Art. 63. Desargues %%% was the author of the theory * Chasles, Les trots livres deporismes d'Euclide, &c. (Paris, i860), p. 102. t Loc. cit., pp. 369 sqq. X Annates de Mathdmatiques, vol. xvi. (Montpellier, 1826), p. 209. § Ibid., vol. viii. (Montpellier, 1818), p. 201. || Systematische Entwichelung, pp. xiii, xiv. Collected Works, vol. i. p. 237. f De la correlation des figures de Geome'trie (Paris, 1801), p. 122. ** Loc. cit., pp. 72, 235; §§ 19, 55. +t Conicorum lib. i. 34, 36, 37, 38. XX Sectiones conicae (Parisiis, 1685), i. 20. §§ Loc. cit, p. 91. IHI Der barycentrische Calcul (Leipzig, 1827), chap. v. ITIf Mathematicae Collectiones, vii. 129. *** Loc. cit., p. 425. ttt Mimoire sur les lignes du second ordre (Paris, 181 7), p. 7. XX+ Loc. cit., pp. 119, 147, 171, 176. AUTHORS PREFACE TO THE FIRST EDITION. XI of involution, of which a few particular cases were already- known to the Greek geometers *, The generation of conies by means of two projective forms was set forth, forty years ago, by Steiner and by Chasles ; it is based on two fundamental theorems (Arts. 149, 150) from which the whole theory of these important curves can be deduced. The same method of generation includes the organic description of Newton f and various theorems of Maclaurin. But the projectivity of the pencils formed by joining two fixed points on a conic to a variable point on the same had already been proved, in other words, by Apollonius %. When only sixteen years old (in 1640) Pascal discovered his celebrated theorem of the mystic hexagram §, and in 1 806 Brianchon deduced the correlative theorem (Art. 153) by means of the theory of pole and polar. The properties of the quadrilateral formed by four tangents to a conic and of the quadrangle formed by their points of contact are to be found in the Latin appendix (De linea- rum geometricarum projorietatibus generalibus tractatus) to the Algebra of Maclaurin, a posthumous work (London, 1748). He deduced from these properties methods for the con- struction of a conic by points or by tangents in several cases where five elements (points or tangents) are given. This problem, in its full generality, was solved at a later date by Brianchon. The idea of considering two projective ranges of points on the same conic was explicitely set forth by Bellavitis ||. To Carnot^[ we owe a celebrated theorem (Art. 385) con- cerning the segments which a conic determines on the sides of * Pappus, Mathematical Collectiones, lib. vii. props. 37-56, 127, 128, 130-133. + Loc. cit., lib. i. lemma xxi. t Conicorum lib. iii. 54, 55, 56. I owe this remark to Prof. Zeuthen (1885). § Letter of Leibnitz to M. Perier in the (Euvres de B. Pascal (Bossut's edition, vol. v. p. 459). || Saggio di geometria derivata (Nuovi Saggi dell' Accademia di Padova, vol. iv, 1838), p. 270, note. U Geometrie de position (Paris, 1803), Art. 379. xii author's preface to the first edition. a triangle. Of this theorem also certain particular cases were known long before *. In the Freie Perspective of Lambert we meet with elegant constructions for the solution of several problems of the first and second degrees by means of the ruler, assuming however that certain elements are given ; but the possibility of solviDg all problems of the second degree by means of the ruler and a fixed circle was made clear by Poncelet ; afterwards Steiner, in a most valuable little book, showed the manner of practically carrying this out (Arts. 238 sqq.). The theory of pole and polar was already contained, under various names, in the works already quoted of Desargues f and De la Hire J ; it was perfected by Monge §, Brian- chon ||, and Poncelet. The last-mentioned geometer derived from it the theory of polar reciprocation, which is essentially the same thing as the law of duality, called by him the ' prin- cipe de reciprocite polaire.' The principal properties of conjugate diameters were ex- pounded by Apollonius in books ii and vii of his work on the Conies. And lastly, the fundamental theorems concerning foci are to be found in book iii of Apollonius, in book vii of Pappus, and in book viii of De la Hire. Those who desire to acquire a more extended and detailed knowledge of the progress of Geometry from its beginnings until the year 1 830 (which is sufficient for what is contained in this book) have only to read that classical work,Hhe Apergu historique of Chasles. * Apollonius, Conicorum lib. iii. 16-23. Desargues, loc. cit., p. 202. De la Hire, loc. cit., book v. props. 10, 12. Newton, Enumeratio linearum tertii ordinis (OpticJcs, London, 1704), p. 142. f Loc. cit., pp. 164, 186, 190, sqq. X Loc. cit., i. 21-28; ii. 23-30* § Oeome'trie descriptive (Paris, 1 795), Art. 40. II Journal de VEcole Poly technique, cahier xiii. (Paris, 1806). AUTHOR'S PREFACE TO THE ENGLISH EDITION. In April last year, when I was in Edinburgh on the occasion of the celebration of the tercentenary festival of the University there, Professor Sylvester did me the honour of saying that in his opinion a translation of my book on the Elements of Projec- tive Geometry might be useful to students at the English Uni- versities as an introduction to the modern geometrical methods. The same favourable judgement was shown to me by other mathematicians, especially in Oxford, which place I visited in the following month of May at the invitation of Professor Syl- vester. There Professor Price proposed to me that I should assist in an English translation of my book, to be carried out by Mr. C. Leudesdorf, Fellow of Pembroke College, and to be published by the Clarendon Press. I accepted the proposal with pleasure, and for this reason. In my opinion the English excel in the art of writing text-books for mathematical teach- ing ; as regards the clear exposition of theories and the abundance of excellent examples, carefully selected, very few books exist in other countries which can compete with those of Salmon and many other distinguished English authors that could be named. I felt it therefore to be a great honour that my book should be considered by such competent judges worthy to be introduced into their colleges. Unless I am mistaken, the preference given to my Elements over the many treatises on modern geometry published on the Continent is to be attributed to the circumstance that in it I have striven, to the best of my ability, to imitate the English XIV AUTHOR S PREFACE TO THE ENGLISH EDITION. models. My intention was not to produce a book of high theories which should be of interest to the advanced mathe- matician, but to construct an elementary text-book of modest dimensions, intelligible to a student whose knowledge need not extend further than the first books of Euclid. I aimed there- fore at simplicity and clearness of exposition; and I was careful to supply an abundance of examples of a kind suitable to encourage the beginner, to make him seize the spirit of the methods, and to render him capable of employing them. My book has, I think, done some service in Italy by helping to spread a knowledge of projective geometry; and I am encouraged to believe that it has not been unproductive of results even elsewhere, since I have had the honour of seeing it translated into French and into German. If the present edition be compared with the preceding ones, it will be seen that the book has been considerably enlarged and amended. All the improvements which are to be found in the French and the German editions have been incor- porated ; a new Chapter, on Foci, has been added ; and every* Chapter has received modifications, additions, and elucidations, due in part to myself, and in part to the translator. In conclusion, I beg leave to express my thanks to the eminent mathematician, the Savilian Professor of Geometry, who advised this translation ; to the Delegates of the Clarendon Press, who undertook its publication; and to Mr. Leudesdorf, who has executed it with scrupulous fidelity. L. CREMONA. Rome, May 1885. TABLE OF CONTENTS. Peeface PAGE V CHAPTER I. Definitions (Arts. 1-7) . i CHAPTER II. CENTRAL PROJECTION; FIGURES IN PERSPECTIVE. Figures in perspective (Arts. 8-11) ...... 3 Point at infinity on a straight line (12) - . . . . . 4 Line at infinity in a plane (13) -5 Triangles in perspective; theorems of Desargues (14-17) . 6 CHAPTER III. HOMOLOGY. Figures in homology (Arts. 18-21) 9 Locus of the centre of perspective of two figures, when one is turned round the axis of perspective (22) . . . .12 Construction of homological figures (23) 13 Homothetic figures (23) . . . . . . . . 18 CHAPTER IV. HOMOLOGICAL FIGURES IN SPACE. Relief-perspective (Art. 24) 20" Plane at infinity (26) . . . . . . . .21 XVI TABLE OP CONTENTS. CHAPTER V. GEOMETRIC FORMS. PAOK The geometric prime-forms (Arts. 27-31) . . . .22 Their dimensions (32) 24 CHAPTEE VI. THE PRINCIPLE OF DUALITY. Correlative figures and propositions (Arts. 33-38) . . .26 CHAPTER VII. PROJECTIVE GEOMETRIC FORMS. First notions (Arts. 39-41) 33 Forms in perspective (42, 43) . . . . . . -35 Fundamental theorems (44,45) 36 CHAPTER VIII. HARMONIC FORMS. Fundamental theorem (Arts. 46, 49) 39 Harmonic forms are projective (47, 48, 50, 51) . . . -41 Elementary properties (52-57) ...... 44 * Constructions (58-60) 47 CHAPTER IX. ANHARMONIC RATIOS. Distinction between metrical and descriptive (Art. 61) . . 50 Rule of signs : elementary segment-relations (62) . . .50 Theorem of Pappus, and converse (63-66) . . . .52 Properties of harmonic forms (68-71) . . . . • 57 T J3 twenty-four anharmonic ratios of a group of four elements (72) 59 In two projective forms, corresponding groups of elements are ^ [ equianharmonic (73) . . . . . . .62 Metrical property of two projective ranges (74) . . .62 Properties of two homological figures (75-77) . . . -63 TABLE OF CONTENTS. xvii CHAPTER X. CONSTRUCTION OF PROJECTIVE FORMS. PAOE Two forms are projective if corresponding groups of elements are equianharmonic (Art. 79) . . . . . . .66 Forms in perspective ; self-corresponding elements (80) . .67 Superposed projective forms (81, 82) . . . . .68 A geometric form of four elements, when harmonic (83) . . 69 Constructions (84-86, 88-90) . . . . . . . 70 Hexagon whose vertices lie on two straight lines ; theorem of Pappus (87) 75 Properties of two projective figures (91-94) . . . -79 Construction of projective plane figures (95, 96) . . . 8 r Any two such figures can be placed so as to be in homology (97) 84 CHAPTER XI. PARTICULAR CASES AND EXERCISE?. Similar ranges and pencils (Arts. 99-103) . . . .86 Equal pencils (104-108) . . . ■ 89 Metrical properties of two collinear projective ranges (109) ' . 91 Examples (110-112) 93 Porisms of Euclid and of Pappus (113, 114) . . . -95 Problems solved with the ruler only (115-118) . . .96 Figures in perspective ; theorems of Chasles (119, 120) . . 98 CHAPTER XII. involution. Definition; elementary properties (Arts. 121-124) . . . too Metrical property ; double elements (125, 126) . . .102 Two pairs of conjugate elements determine an involution (127) 104 The two kinds of involutions (128) ..... 105 b XV111 TABLE OF CONTENTS. PAGE Another metrical property (130) . . . . . .106 Property of a quadrangle cut by a transversal (131, 132) . 107 The middle points of the diagonals of a complete quadrilateral are collinear (133) ........ 108 Constructions (134) ........ 109 Theorems of Ceva and Menelaus (136-140) . . . .110 Particular cases (142) .113 CHAPTER XIII. PKOJECTIVE POEMS IN KELATION TO THE CIECLE. Circle generated by two directly equal pencils (Art. 143) . 114 Fundamental property of points on a circle (144) . . .114 \s Fundamental property of tangents to a circle (146) . . .115 Harmonic points and tangents (145, 147, 148) . . .115 CHAPTER XIV. PROJECTIVE FORMS IN RELATION TO THE CONIC SECTIONS. Fundamental theorems (Art. 149) Generation of conies by means of two projective forms (150) ^ Anharmonic ratio of four points or tangents of a conic (151) Five points or five tangents determine a conic (152) / Theorems of Pascal and Brianchon (153, 154) . Theorems of Mobius (155) and Maclaurin (156) . Properties of the parabola (157, 158) .... CHAPTER XV. constructions and exercises. 118 119 122 123 124 126 127 Properties of the hyperbola ; theorems of Apollonius (158-160) 129 Pascal's and Brianchon's theorems applied to the construction of a conic by points or by tangents (Art. 161) . . . 131 Cases in which one or more of the elements lie at infinity (162, 163) 132 TABLE OF CONTENTS. XIX CHAPTER XVI. DEDUCTIONS EBOM THE THEOBEMS OF PASCAL AND BEIANCHON. PAGE Theorem on the inscribed pentagon (Art. 164) . . -136 Application to the construction of conies (165) . . . 137 Theorem on the inscribed quadrangle (166, 168) . . .138 Application to the construction of conies (167, 168) . . 139 The circumscribed quadrilateral and the quadrangle formed by the points of contact of the sides (169-173) . . . 140 Theorem on the inscribed triangle (174) . . . . 143 Application to the construction of conies (175) . . . 143 The circumscribed triangle and the triangle formed by the points of contact of the sides (176-178) . . . 144 Theorem on the circumscribed pentagon (179) . . 145 Construction of conies subject to certain conditions (180-182) . 146 CHAPTER XVII. DESAEGUES' THEOEEM. Desaegues' theorem and its correlative (Art. 183) . . .148 Conies circumscribing the same quadrangle, or inscribed in the same quadrilateral (184) 149 Theorems of Poncelet (186-188) 151 Deductions from Desaegues' theorem (189-194) . . . 152 Group of four harmonic points or tangents (195, 196) . 157 Property of the hyperbola (197) 158 Theorem 'ad quatuor tineas' quoted by Pappus (198) . .158 Correlative theorem (199) . 159 CHAPTER XVIII. SELF-COEEESPONDING ELEMENTS AND DOUBLE ELEMENTS. Projective ranges of points on a conic (Art. 200) . . 161 Projective series of tangents to a conic (201) . . . .163 Involution of points or tangents of a conic (202, 203) . . 165 TABLE OF CONTENTS. PAGl Harmonic points and tangents (204, 205) . . . .168 Construction of the self-corresponding elements of two super- posed projective forms, and of the double elements of an involution (206) . . . . . . . .169 Orthogonal pair of rays of a pencil in involution (207) . .172 Construction for the common pair of two superposed involutions (208,209) 173 Other constructions (210, 211) . . . . . .174 CHAPTER XIX. PROBLEMS OF THE SECOND DEGEEE. Construction of a conic determined by five points or tangents (Arts. 212,213, 216, 217) 176 Particular cases (214, 215) 178 Construction of a conic determined by four points and a tangent, or by four tangents and a point (218) . . . .180 Case of the parabola (219, 220) 181 Construction of a conic determined by three points and two tan- gents, or by three tangents and two points (221) . .182 Construction of a polygon satisfying certain conditions (222- 225) 184 Construction of the points of intersection, and of the common tangents, of two conies (226-230) ..... 188 Various problems (231-236) . . . . . . .190 Geometric method of false position (237) . . . 193 Solution of problems of the second degree by means of the ruler and a fixed circle (238) . . . . . . . 194 Examples of problems solved by this method (239-249) . .194 CHAPTER XX. POLE AND POLAR. Definitions and elementary properties (Arts. 250-254) . .201 Conjugate points and lines with respect to a conic (255, 256) . 204 Constructions (257) 205 TABLE OF CONTENTS. XXI Self-conjugate triangle (258-262) . Involution of conjugate points and lines (263, 264) Complete quadrangles and quadrilaterals having the same diagonal triangle (265-267) Conies having a common self-conjugate triangle (268, 269) Properties of conies inscribed in the same quadrilateral, or cir cumscribing the same quadrangle (270-273) . Properties of inscribed and circumscribed triangles (274, 275) CHAPTER XXI. THE CENTKE AND DIAMETERS OF A CONIC. The diameter of a system of parallel chords (Arts. 276-278) Case of the parabola (279-280) Centre of a conic (281-283, 285) . Conjugate diameters (284, 286-288, 290) Case of the circle (289) .... Theorem of Mobius (291) Involution property of a quadrangle inscribed in a conic (292) Ideal diameters and chords (290, 294) . Involution of conjugate diameters : the axes (296-298) . Various properties of conjugate diameters ; theorems of Apol- lonius (299-315) .... Conies inscribed in the same quadrilateral ; theorem of Newton (317, 318) Constructions (285, 290, 293, 301, 307, 311, 316, 319) 206 209 210 212 213 215 217 218 218 219 222 224 224 226 227 228 236 238 CHAPTER XXII. POLAR RECIPROCAL FIGURES. Polar reciprocal curves (Arts. 320, 321) 239 The polar reciprocal of a conic with regard to a conic (322, 323) 240 Polar reciprocal figures are correlative figures (324, 325) . . 241 Two triangles which are self-conjugate with regard to the same conic (326) 242 XX11 TABLE OF CONTENTS. PAGK Triangles self-conjugate with regard to one conic, and inscribed in, or circumscribed to, another (327-329) . . -243 Two triangles circumscribed to the same conic (330, 331) . 244 Triangles inscribed in one conic and circumscribed to another (332) 244 Hesse's theorem (334) 245 Reciprocal triangles with regard to a conic are homological (336) 246 Conic with regard to which two given triangles are reciprocal (338) 247 Polar system (339) . . ,. • 248 CHAPTER XXIII. FOCI. Foci of a conic defined (Arts. 340-341) 249 The involution determined by pairs of orthogonal conjugate rays on each of the axes (342) ...... 250 The foci are the double points of this involution (343) , . . 251 Focal properties of tangents and normals (344-346, 349, 360- 362) 252 The circle circumscribing the triangle formed by three tangents to a parabola passes through the focus (347) . . -253 The directrices (348, 350, 351) 254 The latus rectum (352) 257 The focal radii (353) • 258 The eccentricity (354, 355) 259 Locus of the feet of perpendiculars from a focus on the tangents to a conic (356-359) . 260 Constructions (363, 364, 366, 367) 264 Confocal conies (365) . . . . . . . .266 Locus of intersection of orthogonal tangents to a conic (368, 369) 268 Property of the director circle in relation to the self-conjugate triangles of a conic (370-375) . ..... 270 The polar reciprocal of a circle with respect to a circle (376-379) 273 TABLE OF CONTENTS. XX111 CHAPTER XXIV. COKOLLARIES AND CONSTRUCTIONS. Various properties and constructions connected with the hyper- bola and the parabola (Arts. 380-384) Caenot's theorem, and deductions from it (385-389, 391) Constructions of conies (390, 392-394) .... The rectangular hyperbola (395) Method of determining to which kind of conic a given arc be longs (396) Constructions of conies (397-404) ..... Trisection of an arc of a circle (406) .... Construction of a conic, given three tangents and two points, or three points and two tangents (408) Newton's organic description of a conic (409, 410) . Various problems and theorems (411-418, 421) Problems of the second degree solved by means of the theory of pole and polar (419) Problems solved by the method of polar reciprocation (420) Exercises (422) Index 303 277 279 284 285 289 289 294 295 296 297 300 301 302 UNIVEB ELEMENTS OF PROJECTIVE GEOMETRY. CHAPTEK I. DEFINITIONS. 1. By a figure is meant any assemblage of points, straight lines, and planes ; the straight lines and planes are all to be considered as extending to infinity, without regard to the limited portions of space which are enclosed by them. By the word triangle, for example, is to be understood a system consisting of three points and three straight lines connecting these points two and two ; a tetrahedron is a system consisting of four planes and the four points in which these planes inter- sect three and three, &c. In order to secure uniformity of notation, we shall always denote points by the capital letters A,B ,C , ... , straight lines by the small letters a , b , c , ... , planes by the Greek letters a , /3 , y , ... . Moreover, AB will denote that part of the straight line joining A and B which is comprised between the points A and B ; Aa will denote the plane which passes through the point A and the straight line a ; aa the point common to the straight line a and the plane a; a/3 the, straight line formed by the intersection of the planes a, £ ; ABC the : plane of the three points A , B , C ; afiy the point common to the three pJrne.s a , /3 , y ; a.BC the point common to the plane a and the straight line BC ; A.fiy the plane passing through the point A- and the straight line /3y ; a. Be the straight line common to the plane a and the plane Be ; A. fie the straight line joining the point A to the point fie, &c. The notation a.BC = A / we shall use to express that the point common to the plane a and the straight line BC coincides with the point A' ; u — ABC will express that the straight line u contains -the points A,B,C,&c. 2. To project from a fixed point 8 (the centre of projection) a figure ABCD..., abed...) composed of points and straight lines, is to construct the straight lines or projecting rays 2 DEFINITIONS. [3-7, 8 A , 8B , SC , #Z) , ... and the planes (projecting planes) 8a ,Sb,Sc ,Sd , ... . We thus obtain a new figure composed of straight lines and planes which all pass through the centre 8. 3. To cut by a fixed plane a (transversal plane) a figure (a/3yS,... abed...) made up of planes and straight lines, is to construct the straight lines or traces B'C', . . of another plane figure or' * 9 in such a way that the corresponding lines AB and A'B', AC and A'C,..., BC and B'C,... meet in points lying on the line of intersection (&&'), of the planes .a-, and lane, is parallel to the plane tt, since all the vanishing lines of the planes o- are parallel to the same plane n. The vanishing plane ' is thus the locus of the straight lines which correspond to the straight lines at infinity in all the planes of space, and is consequently also the locus of the points which correspond to the points at infinity in all the straight lines of space : for the line at infinity in any plane a is the same thing as the line at infinity in the plane through parallel to a ; so also the point at infinity on any straight line a coincides with the point at infinity on the straight line drawn through parallel to a. 26. The infinitely distant points of all space are then such that their images are the points of one and the same plane ' (the vanishing plane). It is therefore natural to consider all the infinitely distant points in space as lying in one and the same plane $ (the plane at infinity) of which the plane <£' is the image f. The idea of the plane at infinity being granted, the point at infinity on any straight line a is simply the point a, and the straight line at infinity in any plane a is the straight line a(f>. Two straight lines are parallel if they intersect in a point of the plane ; two planes are parallel if their line of intersection lies in the plane $, &c. * Since c' cuts both a! and b' without passing through the point a'b', therefore c' has two points in common with the plane a'b', and consequently lies entirely in the plane a'b'. And similarly for the other straight lines. t Poncelet, Prop. proj. 580. CHAPTER V. GEOMETRIC FORMS. 27. A range or row of points is a figure A,B,C,... composed of points lying on a straight line (which is called the base of the range) ; such is, for example, the figure resulting from the operations of Art. 5 or Art. 7. An axial pencil is a figure a, /3,y, ... composed of planes all passing through the same straight line (the axis of the pencil) ; such is the figure resulting from the operations of Art. 4 or Art. 6. Aflat pencil is a figure a, b, '; these latter straight lines lie therefore in the same plane (that of the triangle). So also the proof for the case where the two trihedral angles have the same vertex S will be correlative to that for the analogous case of two triangles A'B'C and A"B"C" which lie in the same plane (Art. 17). The theorem may also be established by projecting from a point S the figure corresponding to the theorem of Art. 16. The proof of the theorem correlative to that of Arts. 14 and 16 is left as an exercise for the student. It may be enunciated as follows : If two trihedral angles a / /3 / y / , a fr $"y" are such that tlie straight lines a' a", fi'fi", y'y" lie in tlie same plane, then the pairs of edges ft'y' and &"y", yV and y"a", a'$ f and d'$" determine three planes zvhich pass all through the same straight line. 36. In the Geometry of the plane, two correlative proposi- tions are deduced one from the other by interchanging the words point and line, as in the following examples : 36] THE PRINCIPLE OF DUALITY. 29 1. Two points A,B determine a straight line, viz. the line AB. 2. Four points A,B,G,D (Fig. 13), no three of which are col- linear, form a figure called a com- plete quadrangle*. The four y> G Fig. 13. points are called the vertices, and the six straight lines joining them in pairs are called the sides of the quadrangle. Two sides which do not meet in a vertex are termed opposite-^ there are accordingly three pairs of opposite sides, BO and AD, CA and BD, AB and CD. The 1 . Two straight lines a , b de- termine a point, viz. the point ab. 2. Four straight lines a,b,c,d (Fig. 14), no three of which are concurrent, form a figure called a complete quadrilateral*. The four Fig. 14. straight lines are called the sides of the quadrilateral, and the six points in which the sides cut one another two and two are called the vertices. Two vertices which do not lie on the same side are termed op- 2)0site; there are accordingly three pairs of opposite vertices, be and ad, ca and bd, ab and cd. Fig. 16. points E,F,Gin which the oppo- The straight lines e,f,g which site sides intersect in pairs are join pairs of opposite vertices are * The complete quadrangle has also been called a tetrastigm, and the complete quadrilateral a tetragram. Townsend, Modern Geometry, ch. vii. 30 THE PRINCIPLE OP DUALITY. [36 termed the diagonal 2>oints ; and the triangle EFG is termed the diagonal triangle of the complete quadrangle. The complete quad- rangle includes three simple quadrangles, viz. ACBD, A BCD, . and ABDC (Fig. 15). 3. And so, in general : A complete polygon (complete n-gon, or n-point *) is a system of n points or vertices, with the — straight lines or sides called the diagonals ; and the triangle efg is termed the diagonal triangle of the complete quadri- lateral. The complete quadri- lateral includes three simple/ quadrilaterals, viz. acbd, adcb, and acbd (Fig. 16). A complete multilateral (or n-side t) is a system of n straight lines or sides, with the — ' points or vertices in which they intersect one another two and two. 4. The theorems of Arts. 16 and 17 are correlative each to the other. • ' Theorem. If two complete quadrilaterals abed, a'b'c'd' are such that five pairs of vertices ab and a'b' ', be and b'c', ca and c / a / , ad and a'd f , bd and b'd' lie upon five straight lines which meet in a point S, then the re- maining pair cd and c'd' will also lie on a straight line through S (Fig. 18). which join them two and two. Vz 5. Theorem. If two complete ^quadrangles ABC I), A'B'C'D' are such that five pairs of sides AB and A'B', BC and B'C, CA and C'A ',AD and A'D', BD and B'D' cut one another in five points lying on a straight line s, then the remaining pair CD and CD' will also intersect one ano- ther on s (Fig. 17). - Fig. 17. Since the triangles ABC, A' B'C' are by hypothesis in * Or polystigrn ; Townsend, loc. < it. Fig. 18. Since the triangles (tri- lateral) abc, a'b'c' are by f Or polygram. 37] THE PRINCIPLE OF DUALITY. 31 perspective (Arts. 17, 18), the straight lines AA' , BB', CC will meet in one point S. So toxuthe triangles ABD, J.'B'D' are in perspective ; there- fore DD' also will pass through S, the point common to AA' and BB' . It follows that the triangles BCD, B'C'D'are also in perspective : therefore CD and CD' meet in a point on the straight line s, which is deter- mined by the point of intersec- tion of BC and B'C and by that of BD and B'D' *. hypothesis in perspective (Art. 18), the points aa', bb', cc' will lie on one straight line s. So too the triangles abd, a'b'd' are in perspective; therefore the point dd' lies on the straight line 8 which passes through the points aa', bb'. It follows that the triangles (trilaterals) bed, b'e'd' are also in perspective ; therefore cd and c'd' lie on a straight line through the point S, which is determined by the straight lines (6c) (6V) and (bd) (b'd') *. 37. In the Geometry of space the following are correlative : A complete n-gon (in a plane). A complete n-flat (in a sheaf) ; i.e. a figure made up of rt planes (or faces) which all pass through the same point (or vertex), toge- ther with the w(n— i) edges in A complete multilateral of n sides, or n-side (in a plane). which these planes intersect two and two. A complete n-edge (in a sheaf) ; i.e. a figure made up of n straight lines radiating from a common point (or vertex), together with the n in— i) planes (or faces) which pass through these straight lines taken in pairs. Thus the following theorems are correlative, in the Geometry^of space, to the two theorems above (Art. 36, No. 5), which latter are themselves correlative to each other in the Geometry of the plane. If two complete four-flats in a sheaf (be their vertices coincident or not) apyb, a'fi'y'd' are such that five pairs of corresponding If two complete four-edges in a sheaf (be their vertices coincident or not) abed, a'b'c'd' are such that five pairs of corresponding faces cut * These two theorems hold good equally when the two quadrangles or quadri- laterals lie in different planes ; in fact, the proofs are the same as the above, word for word. 32 THE PRINCIPLE OF DUALITY. [38 edges lie in five planes which one another in five straight lines pass all through the same straight which lie all in one plane [ EFGB be aprojection of thesSfteints from a centre M on a straighijuiine BF passing through B. If AF , CM meet in N, then MNGC will be a projection of EFGB from centre A, and BABC a projection of MNGC from centre F; therefore (Arts. 40, 41) the form BABC is pro- jective with ABCB. In a similar manner it can be shown that CBAB and BCBA are projective with ABCB*. From this it follows for example that if a flat pencil abed is projective with a range ABCB, then it is projective also with BABC, with CBAB, and with BCBA; i.e. if two geometric forms, each consisting of four elements, are projectively related, then the elements of the one can be made to correspond respectively to the elements of the other in four different ways. * Staudt, Geometric der Lage, Art. 59. W Fig. 25. CHAPTER VIII. HARMONIC FORMS. J 46. Theorem*. Given three points A, B, C on a straight line s) if a complete quadrangle (KLMX) be con- structed (in any plane through s) in such a manner that two oppo- site sides (KL , MX) meet in A y two otheropposite sides (KX, ML) meet in B, and the fifth side (LX) passes through C, then the sixth side (it J/) will cut the straight line 8 in appoint D which is de- termined by the three given points ; i.e. it does not change its position, in whatever manner the arbitrary elements of the quad- rangle are made to vary (Fig. 26). Given in a plane three straight lines a, b. c which meet in a point S ; if a complete quadrilateral (Jclmn) be constructed in such a manner that two opposite vertices (kl, mn) lie on a, two other oppo- site vertices (hi , ml) lie on b, and the fifth vertex (nl) lies on c, then the sixth vertex (&;?i) will lie on a straight line d which passes through S, and which is determinate ; t. e. it does not change its position, in whatever manner the arbitrary elements of the quadrilateral are made to vary (Fig. 27). Fig. 2;. For if a second complete For if a second complete quadrangle (K'L'M'X') be cen- quadrilateral {k'l'm'n ') be con- * Staudt, loc. cit., Art.. 93. 40 HARMONIC FORMS. [46 < structed (either in the same plane, or in any other plane through s), which satisfies the prescribed con- ditions, then the two quadrangles will have five pairs of correspond- ing sides which meet on the given straight line ; therefore the sixth pair will also meet on the same line (Art. 36, No. 5, left). From this it follows that if the first quadrangle be kept fixed while the second is made to vary- in every possible way, the* point D will remain fixed ; which proves the theorem. ^ The four points A BCD are called harmonic, or we may say that the group or the geometric form constituted by these four points is a harmonic one, or that A BCD form a harmonic range. Or again: Four points A BCD of a straight line, taken in this order, are called harmonic, if it is pos- sible to construct a complete quad- rangle such that two opposite sides pass through A, two other opposite sides through B, tlie fifth side through C, and the sixth through D. It follows from the preceding theo- rem that when such a quadrangle exists, t. e. when the form ABCD is harmonic, it is possible to con- struct an infinite number of other quadrangles satisfying the same conditions. It further follows that, given three points ABC of a range (and also the order in which they are to be taken), the fourth point D, which makes with them a harmonic form, is determinate and unique, and is found by the construction of one of the quad- rangles (see below, Art. 58). structed which satisfies the pre- scribed conditions, then the two quadrilaterals will have five pairs of corresponding vertices collinear respectively with the given point ; therefore the sixth pair will also lie in a straight line passing through the same point (Art. 36, No. 5, right). From this it follows that if the first quadrilateral be kept fixed while the second is made to vary in every possible way, th straight line t&^ill remain fixed ; which proves the theorem. The four straight lines or rays abed are called harmonic, or we may say that the group or the geometric form constituted by these four lines is a harmonic one, or that abed form a harmonic pencil. Or again : Four rays abed of a pencil, taken in this order, are called harmonic, if it is possible to construct a complete quadrilateral such that two oppo- site vertices lie on a, two other opposite vertices on b, the fifth vertex on c, and the sixth on d. It follows from the preceding theo- rem that when such a quadri- lateral exists, i.e. when the form abed is harmonic, it is possible to construct an infinite number of other quadrilaterals satisfying the same conditions. It further follows that given three rays abc of a pencil (and also the .order in which they are to be taken), the fourth ray d, which makes with them a harmonic form, is deter- minate and unique, and is found by the construction of one of the quadrilaterals (see below, Art. 58). 48] HARMONIC FORMS. 41 rf47. If from any 'point S the harmonic range ABCB be projected- / upon any other straight line, its projection A'B'C'B' will also be a- 1 l * harmonic range (Fig. 28). Imagine two planes drawn one through each of the straight lines AB , A'B', and suppose that in the first of these planes is constructed a complete quadrangle of which two opposite sides meet in A, two other opposite sides meet in B, and a fifth side passes through C\ then the sixth side will pass through D (Art. 46), since by hypothesis ABCB Fi gt 2 8. is a harmonic range. Now project this quadrangle fr om A r point S on to the second plane ; then a new quadrangle is obtained of which two opposite sides meet in A f , two other opposite sides meet in B', and whose fifth and sixth sides pass respectively through C f and B' ; therefore A'B'C'B' is a harmonic range^^ 48. An examination of Fig. 27 will show that the harmonic pencil abed is cut by any transversal whatever in a har- monic range. For let 8 be the centre of the pencil and m be any transversal ; in a take any point R ; join R to B by the straight line k and to B by the straight line I; and join A to hb or P by the straight line n. As abed is a harmonic pencil and five vertices of the complete quadrilateral Hmn lie on a, b, and d, the sixth vertex In or Q must lie on the fourth ray c. Then from the complete quadrangle PQRS it is clear that ABCB is a harmonic range. Conversely, if the harmonic range ABCB (Fig. 27) be given, and any centre whatever of projection S be taken, then the four projecting rays S(A,B,C,B) will form a harmonic pencil. For draw through A any straight line to cut SB in P and SC in Q, and join BQ, cutting AS in R. The quadrangle PQRS is such that two opposite sides meet in A, two other opposite sides in B, and the fifth side passes through C ; consequently the sixth side must pass through B (Art. 46, left), since by hypothesis the range ABCB is harmonic. But then we have a complete quadrilateral Tclmn which has two opposite vertices A and R lying on SA, two other opposite vertices B and P on SB, a fifth vertex Q on SC, and the sixth B on SB; therefore 42 HARMONIC FORMS. [40 (Art. 46, right) the four straight lines which project the range ABCD from S are harmonic. We may therefore enunciate the following proposition : A harmonic pencil is cut by any transversal whatever in a harmonic range ; and, conversely, the rays ivhich project a harmonic range from any centre whatever form a harmonic pencil. Corollary. In two homological figures, to a range of four harmonic points corresponds a range of four harmonic points ; and to a pencil of four harmonic rays corresponds a pencil of four harmonic rays. 49. The theorem on the right in Art. 46 is correlative to that on the left in the same Article. In this latter theorem all the quadrangles are supposed to lie in the same plane ; but from the preceding considerations it is clear that the theorem is still true and may be proved in the same manner, if the quadrangles are drawn in different planes. Considering accordingly this latter theorem (Art. 46, left) as a proposition in the Geometry of space, the theorem corre- lative to it will be the following : If three planes a , /3 , y all pass through one straight line s, and if a complete four -flat (see Art. 37) k\\w he cofistructed, of which two opposite edges k\ , \lv lie in the plane a, two other opposite edges kv , \fx lie in the plane ft, and the edgekv lies in the plane y ; then the sixth edge kjjl will always lie in a fixed plane b (passing through s), which does not change, in whatever manner the arbitrary elements of the four-flat he made to vary. For if we construct (taking either the same vertex or any other lying on s) another complete four-flat which satisfies the prescribed conditions, the two four-flats will have five pairs of corresponding edges lying in planes which all pass through the same straight line s; therefore (Art. 37, left) the sixth pair also will lie in a plane which passes through s. The four planes, a , /3 , y , 5 are termed harmonic planes ; or we may say that the group or the geometric form constituted by them is harmonic ; or again that they form a harmonic (axial) pencil. 50. If a complete four-flat k\\xv be cut by any plane not passing through the vertex of the pencil, a complete quadri- lateral is obtained ; and the same transversal plane cuts the planes a , j3 , y , 5 in four rays of a flat pencil of which the first 51] HARMONIC FORMS. 43 two rays contain each a pair of vertices of the quadrilateral while the other two pass each through one of the remaining vertices. Consequently (Art. 46, right) an axial pencil of four harmonic planes is cut by any transversal plane in a flat pencil of four harmonic rays. Similarly, if the harmonic axial pencil of four planes a, (3, y,b is cut by any transversal line in four points A, B, C, B, these form a harmonic range. For if through the transversal line a plane be drawn, it will cut the planes a , /3 , y , 5 in four straight lines a , b , c , d. This group of straight lines is har- monic, by what has just been proved; but ABCB is a section of the flat pencil a, b,c, d; consequently (Art. 48) the four points A , B , C , D are harmonic. Conversely, if four points forming a harmonic range be projected from an axis, or if four rays forming a harmonic pencil be projected from a point, the resulting axial pencil is harmonic. 51. If then we include under the title of harmonic form the group of four harmonic points (the harmonic range), the group of four harmonic rays (the harmonic flat pencil), and the group of four harmonic planes (the harmonic axial pencil), we may enunciate the theorem : Every projection or section of a harmonic form is itself a harmonic form: or, Every form tvhich is projective with a harmonic form is itself harmonic. Conversely, two harmonic forms are ahc ays projective with one another. To prove this proposition, it is enough to consider two groups each of four harmonic points ; for if one of the forms were a pencil we should obtain four harmonic points on cut- ting it by a transversal. Let then ABCB, A'B'C'B' be two harmonic ranges, and project ABC into A'B'C in the manner explained in Art. 44; the same operations (projections and sections) which serve to derive A'B'C from ABC will give for B a point B x \ from which it follows that the range A'B'C'B t will be harmonic, since the range ABCB is harmonic. But A'B'C'B' are also four harmonic points, by hypothesis ; there- fore B x must coincide with B', since the three points A'B'C determine uniquely the fourth point which forms with 'them a harmonic range (Art. 46, left). 44 HARMONIC FORMS. [52 We may add here a consequence of the definitions given in Arts. 49 and 50 : The form which is correlative to a harmonic form is itself 52. If a , b, c , d are rays of a pencil (Fig. 28), then a and b are said to be separated by c and d, when a straight line pass- ing through the centre of the pencil, and rotating so as to come into coincidence with each of the rays in turn, cannot pass from a to b without coinciding with one and only one of the two other rays c and d *. The same definition applies to the case of four planes of a pencil, and to that of four points of a range (Fig. 26) ; only it must be granted that we may pass from a point A to a point B in two different ways, either by describing the finite segment AB or the infinite segment which begins at A, passes through the point at infinity, and ends at B. This definition premised, the follow- ° ° ing property may be enunciated as at < £ ° < once evident: Four elements of a one- Fig. 29. dimensional geometric form [i. e. four points of a range, four rays of a pencil, &c.) can always be so divided into two pairs that one pair is separated by the other, and this can be done in one way only. In Fig. 26, for example, the two pairs which separate one another are AB , CD ; and if A'B'G'D' is a form projective with ABC I), the pair A'B' will be separated by the pair CD' \ for the operations of projection and section do not change the relative position of the elements. 53. Let now ABCJD (Fig. 30) be four harmonic points, i. e. four points obtained by the construction of Art. 46, left. This allows us to draw in an infinite number of ways a complete quadrangle of which A and B are two diagonal points (Art. 36, No. 2, left), while the other two opposite sides pass through C and B. It is only necessary to state this con- struction in order to see that the two points A and B are precisely similar in their relation to the system, and that the same is true with regard to C and B. It follows from this that if ABCB is a harmonic range, then BACB, ABBC, BABC, which are obtained by permuting the letters A and B or C and B, or both at the same time, are harmonic ranges also. * a and b, c and d, may also be termed alternate pairs of rays. 55] HARMONIC FOEMS. 45 Consequently (Art. 51) the harmonic range AB CD for example is projective with BACB, i. e. we can pass from one range to the other by a finite number of projections and sections. In fact if the range ABCB be projected from K on CQ, we obtain the range LNCQ, which when projected from M on AB gives BACB. 54. If A , B , C , B are four harmonic points, then A and B are necessarily separated by C and B. For if (Fig. 30) the group ABCDbe projected on the straight line KM, first from the centre L and then from the centre N, the projections are KMQB A c b d and MKQB respectively. * ~" "^^T" ~T\ ~/f[ ~^^ These two groups of \/ ^^tyds^'s' points, consisting as they \ y^^^y^ do of the same elements, mx" J/Jr y must show the same kind vC \ I of arrangement; therefore . Nl the points K and M are Fig 3 , separated by Q and B, and therefore A and B are separated by C and B. 55. Let the straight lines AQ , BQ be drawn (Fig. 31), the. former meeting MB in U and NB in S, while the latter meets KL in T and MN in V. The complete quadrangle LTQU has two opposite sides meeting in A, two other opposite sides meeting in B, and a fifth side (LQ or LN) passes through C ; therefore the sixth side UT will pass through B (Art. 46). In like manner the sixth side V8 of the complete quadrangle NYQ8 must pass through B, and the sixth sides of the com- plete quadrangles KSQT , M UQ V through C. We have thus a quadrangle STUV two of whose opposite sides meet in C, two 46 HARMONIC FORMS. [56 other opposite sides in B, while the fifth and sixth sides pass respectively through A and B. This shows that the relation to which the points C and B are subject (Art. 53) is the same as the relation to which the points A and B are subject ; or, in other words, that the pair A , B may be interchanged with the pair C , B. Accordingly, if ABCB is a harmonic range, then not only the ranges BACB , ABBC , BABC } but also CBAB,BCAB, CBBA,BCBA are harmonic*. s The points A and B are termed conjugate points, as also are C and B. Or either pair are said to be harmonic conjugates with respect to the other. The points A and B are said to be harmonically separated by the points C and B, or the points G and B to be harmonically separated by A and B. We may also say that the segment AB is divided harmonically by the segment CB, or that the segment CB is divided harmonically by AB. If two points A and B (Fig. 30) are separated har- monically by the points C and B in which the straight line AB is cut by two straight lines QC and QB, we may also say that the segment AB is divided harmonically by the straight lines QC , QB, or by the point Cand the straight line QB,&c; and that the straight lines QC , QB are separated harmonically by the points A, B\ &c. Analogous properties and expressions exist in the case of four harmonic rays or four harmonic planes. [Note. — In future, whenever mention is. made of the harmonic system ABCB, it is always to be understood that A and B, C and D, are conjugate pairs ; it being at the same time remembered that (Art. 54) A and B, C and B, are necessarily alternate pairs of points.] ^rJ 56. The following theorem is another consequence of the proposition of Art. 46, left: hi a complete quadrilateral, each diagonal is divided harmonically by the other two f. Let A and A', B and B', C and C be the pairs of opposite vertices of a complete quadri- lg ' 32 " lateral (Fig. 32), and -let the diagonal AA'he cut by the other diagonals BB' and CC in F * Reye, Geometric der Lage (Hanover, 1866), vol. i. p. 34. t Carnot, Gtomttrie de position (Paris, 1803), Art. 225. 58] HARMONIC FORMS. 47 and E respectively. Consider now the complete quadrangle BB'CC '; one pair of its opposite sides meet in A, another such pair in A', a fifth side passes through E, the sixth through F. The points A , A' are therefore harmonically separated by F and E. Similarly a consideration of the two complete quadrangles CC'AA' and AA'BB' will show that B , B f are harmonically separated by i^and B\ and C , C by B and E. 57. In the complete quadrangle BB'CC' the diagonal points are A , A', and J^;..also since the range BB'FB is harmonic, so too is the pencil of four rays which project it from A (Art. 48) ; therefore : In a complete quadrangle, any two sides which meet in a diagonal point^are divided harmonically hy the two other diagonal points. This theorem is merely however the correlative (in accord- ance with the principle of Duality in plane Geometry) of that proved in the preceding Article. 58. The theorems of Art. 46 can be at once applied to the solution, by means of the ruler only, of the following pro- blems : Given three points of a har- monic range, to find the fourth. Solution. Let A , B , C (Fig. 33) be the given points (lying on a given straight line) and let Given three rays of a har- monic 'pencil, to construct the fourth. Solution. Let a , b , c (Fig. 34) be the given rays (lying in one plane and passing through a Fig. 33- A and B be conjugate to each other. Draw any two straight lines through A, and a third through C to cut these in L and .Fig. 34- given centre S), and let a and b be conjugate to each other. Through any point Q lying on c draw any two straight lines to 48 HARMONIC FORMS. [59 N respectively. Join BL cutting AN in M, and BN cutting AL in K ; then if KM be joined it will cut the given straight line in the required point D, conjugate to c*. cut a in A and R, and b in P and B, respectively. Join AB and RP ; these will cut in a point D, the line joining which to S is the required ray d, conjugate to c. Kg. 35- 59. In the problem of Art. 58, left, let C lie midway between A and B. We can, in the solution, so arrange the arbitrary elements that the points K and M shall- move off to infinity ; to effect this we must con- struct (Fig. 35) a parallelogram ALBN on AB as diagonal ; then since the other diagonal LN passes through C, the point D will lie at infinity. If, conversely, the points A , B , D are given, of which the third point D lies at infinity, we may again construct a parallelogram ALBN on AB as diagonal ; then the fourth point C, the conjugate of D, must be the point where LN meets the given straight line : that is, it must be the middle point of AB. Therefore,; If in a harmonic range ABCD the point C lies midway between the two conjugates A and B, then the fourth point D lies at an infinite distance; and conversely, if one of the points D lies at infinity, its conjugate C is the point midway between the two others, A and B. 60. In the problem of Art. 58, right,, let c be the bisector of the angle between a and b. (Fig. 36). If Q be taken at infinity on c, the segments AB , PR become equal to one another and lie between the parallels A P , BR ; consequently the ray d will be perpendicular to c, i.e. given a harmonic pencil of four rays, abed; if one of them c bisect the angle between tJie two conjugates a and b, the fourth ray d will be at right angles to c. Conversely : if in a harmonic pencil abed (Fig- 37) two conjugate rays c , d are at right angles, then they are the bisectors, internal and external, of the angle between the oilier two rays a,b. * De la Hire, Sectiones Conicae (Parisiis, 1685), lib. i, prQp. ao. Fig. 36. d N^X d h/ C \B £ V Fig. 37. 60] HARMONIC FORMS. 49 For if the pencil be cut by a transversal AB drawn parallel to d, the section A BCD will be a harmonic range (Art. 48) ; and as D lies at infinity, C must lie midway between A and B (Art. 59) ; conse- quently, if S be the centre of the pencil, A SB is an isosceles triangle and SC the bisector of its vertical angle, CHAPTEK IX. ANHARMONIC RATIOS. J 61. Geometrical propositions divide themselves into two classes. Those of the one class are either immediately con- cerned with the magnitude of figures, as Euc. I. 47, or they involve more or less directly the idea of quantity or measure- ment, as e.g. Euc. I. 12. Such proposition : called nietrieal. The other class of propositions relate merely to the position of the figures with which they deal, and the idea of quantity does not enter into them at all. Such propositions are called descriptive. Most of the propositions in Euclid's Elements are metrical, and it is not easy to find among them an example of a purely descriptive theorem. Prop. 2, Book XI, may serve as an instance of one. Projective Geometry on the other hand, dealing with projective properties (i. e. such as are not altered by projection), is chiefly concerned with descriptive properties of figures^ In fact, since the magnitude of a geo- metric figure is altered by projection, metrical properties are as a rule not projective. But there is one important class of metrical properties (anharmonic properties) which are pro- jective, and the discussion of which therefore finds a place in the Projective Geometry. To these we proceed; but it is necessary first to establish certain fundamental notions. 62. Consider a straight line ; a point may move along it in two different directions, one of which is opposite to the other. Let it be agreed to call one of these the positive direction, and the other the negative direction. Let A and B be two points on the straight line ; and let it be further agreed to represent by the expression AB the length of the segment comprised between A and B, taken as a positive or as a negative number of units according as the direction is positive or negative in which a point must move in order to describe the segment ; this point starting from A (the first letter of the expression AB) and ending at B. 62] ANHARMONIC RATIOS. 51 In consequence of this convention, which is termed the rule of signs, the two expressions AB , BA are quantities which are equal in magnitude but opposite in sign, so that BA = — AB, or AB+BA = (1) Now let A, B, C be three points lying on a straight line. If C lies between A and B (Fig. 38 a), (a) c B B c A (i) K B C c B A (c)° A B B AC Fig. 38. we have AB = AC+ CB ;- whence - CB-AC+ AB = 0, or BC+CA + AB=:0. Again, if iTlies between A and C (Fig. 38 b), AC=AB + BC; whence BC-AC+AB = 0, or BC+CA + AB=0. Lastly, if A lies between B and C (Fig. 38 c), CB=CA + AB; whence -CB+CA + AB = 0, or BC+CA + AB= 0. Accordingly : If A,B, C are three collinear points, then whatever their relative positions may be, the identity BC + CA + AB=zO (2) always holds good. From this identity may be deduced an expression for the distance between two points A and B in terms of the distances E % 52 ANHARMONIC RATIOS. [63 of these points from an origin chosen arbitrarily on the straight line which joins them. For since OA + AB + B0 = 0, .-. AB = OB-OA;. (3) or again, AB = AO+0B*. * The results (l) and (2) may be extended ; they are in fact particular cases of the following general proposition : If A x , A 2 ,... A n be n collinear points, then A X A 2 + A 2 A Z + ... + A n _ x A n + A n A x = 0, the truth of which follows at once from (3), since the expres- sion on the left hand is equal to . . (OA 2 - OA,) + (OA 3 - OA 2 ) + . . . +(OA 1 -OA n ), ' which vanishes. Another useful result is that if A, B, C, B be four collinear points, BC.AB+CA.BB + AB.QB= 0. This again follows from (3), since the left-hand side = (BCr-BB)AB+. .. + ... = 0. Many other relations of a similar kind between segments might be proved, but they are not necessary for our purpose. We will give only one more, viz. If A, B , C , be any four collinear points, then OA 2 . BC+ OB 2 . CA+OC 2 . AB = -BC. CA . AB. For by (3) the left-hand side is equal to ( oa 2 - oc 2 ) bc +(ob 2 - oc 2 ) ca = ca(oa+oc)bc+cb(ob+oc)ca = bc.ca(oa-ob) = -BC.CA.AB. It may be noticed that this last theorem is true even if do not lie on the straight line ABC, but be any point whatever. For if a perpendicular 00' be let fall on ABC, OA 2 .BC+OB 2 . CA+OC 2 . AB = (00 /2 + O'A 2 ) BC "+ ... + ... = O'A 2 . BC + O'B 2 .CA+O'C 2 . AB + 00' 2 (BC + CA + AB) = -BC.CA.AB, by what has just been proved. 63. Consider now Fig. 39, which represents the projection * Mobius, Barycentrische Calcul, § 1. 63] ANHA.KMONIC BATIOS. 53 J from a centre S of the points of a straight line a on to another straight line a' '; let us examine the relation which exists between the lengths of two corresponding segments AB } A'B'. fa I - r v Eig. 39. Fig. 40. From the similar triangles SAJ , A f SV JA:J£::rS:FA';* so from the similar triangles SB J, B'SI\ JB: JS ::I'S \TB'\ .-. J A J'A' = JB. TB' = JS . I'&i i.e. the rectangle JA.I'A' has a constant value for all pairs of corresponding points A and A '. If the constant JS.I'S be denoted by k, we have therefore by subtraction, I'V-FA' = '<%-??. J A .JB But I'B'-I'A' = ^'5', and JA-JB = BA = -AB; ••• ^'=jA-^- If we consider four points A , B , C , D (Fig. 40) of the straight line a and their four projections, A\B\C\1)\ we obtain, in a similar manner, * We suppose that in all equations involving segments the rule of signs is observed. See Mobius, Baryc. Calcul, § 1 ; Townsend, Modem Geometry, chapter v. • 54 ANHARMONIC RATIOS. [63 —h B ' c ' = 7STTc' BC ' B '"-jF!n>' BI) > whence by division A'C\A'B' _ AC % AB ~Wc r 'B r B , ~~ BC' BB % This last equation, which has been proved for the case of projection from a centre S, holds also for the case where ABC I) and A'B'C'B' are the intersections of two transversal lines s and *' (not lying in the same plane) with four planes a, (3, y , b which all pass through one straight line u ; in other words, when A'B'C'B' is a projection of ABCB made from an axis u (Art. 4). For let the four planes a , j3 , y , b be cut in A", B", C", B" respectively by a straight line %" which meets s and /. The straight lines AA'\ BB" y C€", BB" are [the intersections of the planes a , (3 , y , b respectively by the plane mt' % and therefore meet in a point S ; that namely in which the plane ss" is cut by the axis u. So also A' A", B'B", C'C'\ B'B" are four straight lines lying in the plane *V' and meeting in a point 8' of the axis u (that namely in which the plane //' is cut by the axis u.) Therefore A"B"C"B" is a projection of ABCB from centre S and a projection A'B'C'B' from centre S' ; so that A"C'\ A"B" _ AC , AB _ A'C\A'B f B"C" '' B"B" ~ BC : BB~~ B'C ' B'B' ' _ , AC AB Ine number -jr^ : -^r- is called the anharmonic ratio of the four collinear points A , B , C , B. The result obtained above may therefore be expressed as follows : The anharmonic ratio of four collinear points is unaltered ly any projection whatever *. * Pappus, Mathemalicae Collecliones, book vii. prop. 129 (ed. Hultsch, Berlin, 1877, vol.ii. p. 871). 65] ANHARMONIC RATIOS. 55 Or again : If two ranges, each of four points, are projective, they have the same anharmonic ratio, or, as we may say, are equianharmonic *. 64. Dividing one by the other the expressions for A'C and B'C, we have A'C _ AC.AJ B'C~ BC'BJ' In this equation the right-hand member is the anharmonic ratio of the four points A , B , C ,J ; consequently the left-hand member must be the anharmonic ratio of A', B', C, J'-, thus the anharmonic ratio of four points A',B',C, J', of which the last lies at infinity, is merely the simple ratio A'C : B'C. This may also be seen by observing that if A' and B' remain fixed while JD' moves off to infinity on the line A'B', then A'B' limiting value of wrfy = 1 5 ,. ... , . A'C A'B' A'C 1 .-. limiting value of -^, : j^, = ^/ • Similarly, on the same supposition, - ., ... , £ A'B' A'C B'C limiting value of j^, : ^j^- f = j^ ; i.e. the anharmonic ratio of the four points A',B' ,B', C, of which the third lies at infinity, is equal to the simple ratio B'C : A'C. 65. From this results the solution of the following Problem. — Given three collinear points A , B, C; to find a fourth B so that the anharmonic ratio of the range ABCB may be a numher A given in sign and magnitude (Fig. 41). Solution. — Draw any transversal through C, and take on it two points A', B' such that the ratio CA' : CB' is equal to A :. 1^ the given value of the anharmonic ratio ; the two points A' and B' lying on the same or on opposite sides of C according as A is positive or negative. Join AA ', BB ', meeting in S; the straight line through S parallel to A'B' will cut AB in the point B required f. For if B' be the point at infinity on * Townsend, Modern Geometry, Art. 278. •j* Chasles, Geometrie superieure (Paris, 1852), p. 10. 56 ANHARMONIC RATIOS. [66 A'B', and we consider ABCB as a projection of A'B'C'B' (C coincides with C) from the centre S, then the anharmonic ratio of ABCB is equal to that of A'B'C'B', that is, to the simple ratio A'C : B'C or A. The above is simply the graphical solution of the equation AC AB _ BC : BB~ ' AB AC % or -^ji = ~Wn : A = f*, BB BC or in other words of the problem : Given two points A and B } to find a point B collinear with them such that the ratio of the segments AB, BB to one another may he equal to a number given in sign and magnitude. As only one such point B can be found, the proposed problem admits of only one solution ; this is also clear from the construction given, since only one line can be drawn through 8 parallel to A'B'. Consequently there cannot be two different points B and B L such that ABCB and ABCB^ have the same anharmonic ratio. Or : If the groups ABCB, ABCB 1 are eqziianharmonic, the point B x must coincide with B. 66. Theorem. (Converse to that of Art. 63.) If two ranges ABCB, A'B'C'B', each of four points, are equianharmonic, they afe projective with one another. For (by Art. 44) we can always pass from the triad ABC to the triad A' B'C by a finite number of projections or sections ; let B" be the point which these operations give as corresponding to B. Then the anharmonic ratio of A'B'C'B" will be equal to that of ABCB, and consequently to that of A'B'C'B'-, whence it follows that B " coincides with B ', and that the ranges ABCB , A'B'C'B' are projective with one another. 67. It follows then from Arts. 63 and 66 that the necessary and sufficient condition that two ranges ABCB , A'B'C'B\ consisting each of four points, should be projective, is the equality (in sign and magnitude) of their anharmonic ratios. The anharmonic ratio of four points ABCB is denoted by the symbol (ABCB)*; accordingly the projectivity of two forms ABCB and A'B'C'B' is expressed by the equation (ABCB) = (A'B'C'B'). * Mobius, Barycentrische Calcul, § 183. 69] ANHARMONIC RATIOS. 57 From what has been proved it is seen that if two pencils each consisting of four rays or four planes are cut by any two transversals in ABCD and A'B'C'D' respectively, the equation (ABCD) = (A'B'C'D') is the necessary and sufficient condition that the two pencils should be projective with one another. v/ The anharmonic ratio of a pencil of four rays a ,b , c ,d or four planes a , /3 , y , 8 may now be defined as the constant anharmonic ratio of the four points in which the four elements of the pencil are cut by any transversal, and may be denoted by (abed) or (a(3yb). This done, we can enunciate the general theorem : If two one-dimensional geometric forms ; consisting each of four elements^ are projective, they are equianharmonic ; and if they are equianharmonic, they are projective. y 68. Since two harmonic forms are always projectively related (Art. 51), the preceding theorem leads to the con- clusion that the anharmonic ratio of four harmonic elements is a constant number. For if ABCD is a harmonic system, BACD is also a harmonic system (Art. 53), and the two systems ACIID and BC4-D are projectively related*; thus (AC^D) = (BCAB), AB \AB _BABB % l ' 6 ' CB' CB~ PA : CB ; , AC AB Wh6nce BC : BB=- 1 > t i.e. (ABCB) = -1; therefore the anharmonic ratio of four harmonic elements is equal to -If. 69. The equation (ABCD) = - 1, or AC AD _ M BC + BD = °> W which expresses that the range ABCD is harmonic, may be put into two other remarkable forms. Since AD = CD-CA (Art. 62) and BD = CD-CB, the equation (1) gives CA (CD-CB) + CB (CD-CA) = 0, UD=Hk+m> (2 > * In Fig. 30 ACBD may be projected (from K on NC) into LCNQ ; and then LCNQ may be projected (from M on AD) into BCAD. t Mobius, loc. cit. p. 269. 58 AXHARXOXIC RATIOS. .70 t. e. CD is the harmonic mean between CA and CB ; a formula which determines the point D when A , B , C are given. Again, if is the middle point of the segment CD, so that we have 0D = CO = -0C, then = OC-OA ; AD = OD-OA = -(0C+ Oi) ; BC=OC-OB; BD = -(0C+OB). Substituting these values in (1) or in we have OC-OA Fig. 4*. OC+OA qc *' o^ or 0C*=0A.0B, (3) t. e. half the segment CD is a mean proportional between the distances of A and B from the middle point of CD. The equation (3) shows that the segments OA and OB must have the same sign, and that therefore can never lie between A and B. If now a circle be drawn to pass through A and B (Fig. 42), O will lie outside the circle, and OC will be the length of the tangent from to it* (Euc. HL 37). The circle on v CZ) as diameter will therefore cut the first circle (and aU circles through *i and B) orthogonally. Conversely, if two circles cut each other orthogonally, they will cut any diameter of one of them in two pairs of harmonic points t. 70. The same formula (3) gives the solution of the following pro- blem: Given two collinear segments AB and A'B'; to determine another segment CD which shall divide each of them harmonically (Figs. 43, 44). Take any point G not lying on the common base AB', and draw the circles GAB . GA'B' meeting * If through a point any chord be drawn to cut a circle in P and Q, the rectangle OF . OQ is, called the potctr of the point with regard to the circle. Stkikkr, CrtUJs Jowrwdy vol. i. (Berlin, i8a6) ; Collected Works, toL i. p. aa. We may then say that OC* is the povar of the point O with regard to the circle in Fig. 4^. f Poxcelet, Propr.proj. Art. 79. Rg; 43. no AXHARMOXIC RATIOS. 59 Fig. 44- again in H. Join GH *, and produce it to cut the axis in 0. Then from the first circle OA.OB=OG. OH (Euc. m. 36), and from the second oa'.qb'=og.oh; * oa.ob = oa'.ob: is therefore the middle point of the segment required; the points C and D will be the intersections with the axis of a circle described from the centre with radius equal to the length of the tangent from to either of the circles GAB, G' A'B'. The problem admits of a real solution when the point falls outside both the segments AB, A'B', and consequently outside both the circles GAB. GA'B' (Figs. 43, 44). There is no real solution when the segments AB, A'B' overlap (Fig. 45) ; in this case lies within both segments. 71. Let ABCD be a harmonic range, and let A and B (a pair of conjugates) approach indefinitely near to one another and ultimately coin- cide. If C lie at an infinite distance, then D must coincide with A and B, since it must lie midway between these two points (Art. 59). If C lie at a finite distance, and assume any position not coinciding with that of A or B. then equation (2) of Art. 69 gives CD = CA= CB, i.e.!) coincides with A and B. Again, let A and C (two non-conjugate points) coincide, and B (the conjugate of A) lie at an infinite distance. In this case A must lie midway between C and D, so that D will coincide with A and C. If B lie at a finite distance , and assume any position not coinciding with that of A or C, then equation (1) of Art. 69 gives AD = 0, i.e. the point D coincides with A and C. So that : If of four points forming a harmonic range, any two coincide, one of the other two points will also coincide with them, and the fourth is indeterminate. 72. The theorem of Art. 45 leads to the following result : given four elements .4 , B , C , D of a one-dimensional geometric form, the * GH is the radical axis of the two circles, and all points on it are of equal power with regard to the circles. Fig- 45- 60 ANHARMONIC RATIOS. [72 anharmonic ratios (ABCD) , (BADC) , (CDAB) , (DCBA) are all equal \j to one another. I. Four elements of such a form can be permuted in twenty-four different ways, so as to form the twenty-four different groups ABCD , BADC , CDAB , DCBA , ABDC , BACD , DCAB , CDBA , ACBD , CADB , BDAC , DBCA , ACDB , CABD , DBAC , BDCA , ADBC , DACB , BCAD , CBDA , ADCB , DABC , CBAD , BCD A , here arranged in six lines of four each. The four groups in each line are projective with one another (Art. 45), and have therefore the same anharmonic ratio . In order to determine the anharmonic ratios of all the twenty-four groups, it is only necessary to consider one group in each line ; for example, the six groups in the first column. These six groups are so related to each other that when any one of them is known the other five can be at once determined. II. Consider the two groups ABCD and ABDC, which are derived one from the other by interchanging the last two elements. Their anharmonic ratios and (ABDC) or ^ : § are one the reciprocal of the other ; thus (ABCD) (ABDC) =1 (1) Similarly, (ACBD) (ACDB) = 1, (2) and (ADBC) (ADCB) =1 (3) III. Now if A, B, C, D are four collinear points, it has been seen (Art. 62) that the identical relation BC. AD + CA . BD + AB . CD = always holds. Dividing by BC .AD, we have ^ -r jj ' Jy ACBD AB.CD . * • ~* im7AD + CB7AD= l > AC AD AB AD_ ° r BC '' BD + CB '' CD ~ ' that is (Arts. 63, 67), (ABC 'D) + (ACBD) = 1 (4) Similarly, (ABDC) + (ADBC) = 1, . (5) and (ACDB) + (ADCB) = 1 (6) 72] ANHARMONIC RATIOS. 61 IV. If X denote the anharmonic ratio of the group ABCD, i.e. if (ABCD) = A, the formula (1) gives (ABDC) = -» and (4) gives. (ACBD) = 1 - A ; then by (2) (ACDB) = -?— , 1 — A and by (6) (AJDCB) = 1 _ ^ = JL ; and finally, by (3) or (5) (ADBC) = ^1 . * A V. The six anharmonic ratios may also be expressed in terms of the angle of intersection 6 of the circles described on the segments AB, CD as diameters ; it being supposed that A and B are separated by C and D. It will be found that (ABCD) = ■ -tan 2 -, ^Z)(7) = -cot 2 |, (ACBD) = a sec 2 -, (ACDB) = cos 2 |, (ADCB) = SIB'-, (ADBC) = cosec 2 |-t VI. If in the group ABCD two points A and B coincide, then AC = BC i AD = BD, and therefore (ABCD) = (AACD) = 1. But if X = 1 , the other anharmonic ratios become (ACAD) =1-1=0, and (ACDA) = oo ; \J thus when of four elements two coincide, the anharmonic ratios have the values 1, 0, oo. If (ABCD) = — 1, ». e. if the range ABCD is harmonic, the formulae of (IV) give (iM) = 2 and ( ACDB ) = i . so that when the anharmonic ratio of four points has the value 2 or ^, these points, taken in another order, form a harmonic range. VII. Conversely, the anharmonic ratio of a range ABCD, none of whose points lies at infinity, cannot have any of the values 0, 1, oo, without some two of its points coinciding. AC AD For if in (IV) A = 0, -^ : -^ = , and either AC or BD must vanish ; i. e. either A coincides with C, or B with D. * Mobius, loc. cit. p. 249. f Casey, On Cyclides and S2>hero~quariics (Phil. Trans. 1871), p. 704. 62 ANHARMONIC RATIOS. [73 If X = 1, (ACBD) = 1— X = 0, so that either A coincides with B, op C with D. And if X = oo, (ABDC) = - = 0, so that either A coincides with D, or B with C. VIII. By considering the expressions given for the six anharmonic ratios in (IV) it is clear that whatever be the relative positions of the points A, B , C , D , two of the ratios (and their two reciprocals) are always positive and a third (and its reciprocal) negative ; and thus we see that the anharmonic ratios of four points no two of which coincide may have all values positive or negative except + 1, 0, or oo. 73. From the theorems of Arts. 63 and 66, which express the necessary and sufficient condition that two ranges, each consisting of four elements, should be protectively related, we conclude that If two geometric forms of one dimension are projective^ then any two corresponding groups of four elements are equianharmonic *. As a particular case, to any four harmonic elements of the one form correspond four harmonic elements of the other (Art. 51). 74. Let A, A' and B,B' be any two pairs of corresponding points of two projective ranges (Fig. 46) ; let / be the point at infinity belonging to the first range, and V the point corresponding to it in the second range ; so let J' be the point at infinity belonging to the second range, and J its correspondent in the first range. By Art. 73 (AMJ) = (A'B'rJ')', .% (BAJI) = {A'BTJ') (Art. 72); from which, since / and J' lie at infinity, BJ:AJ= AT : B f V (Art. 64), and JA.I'A' = JB.I'B'; i.e. the product JA.I'A' has a constant value for all pairs of corresponding points f. [This proposition has already been proved in Art. 63 for the particular case of two ranges in perspective.] * Steiner, Systematische Entwickelung . . (Berlin, 1832), p. 33, § 10; Collected Works, ed. Weierstrass (Berlin, 1881), vol. i. p. 262. f Steineu, loc. cit. p. 40, § 12. Collected Works, vol. i. p. 267. 76] ANHARMONIC RATIOS. 63 ' (UNI 75. In two homological figures, four collinear points or four concurrent straight lines of the one figure form a group which is equianharmonic with that consisting of the points or lines corresponding to them in the other figure (Art. 73). Let be the centre of homology, 31 and M ' any pair of corre- sponding points in the two figures, N and N' another pair of corresponding points lying on the ray 03131', and X the point in which this ray meets the axis of homology. Since the points 031 NX , 031' N'X correspond severally to one another, (0X31N) = (OXM'N'), OM ON OM' ON' or MX 'NX ~ M'X' N'X 9 031 OM' ON ON' " MX 'M'X NX 'N'X and consequently the anharmonic ratio (0XM31') is con- stant for all pairs of corresponding points M and 31' taken on a ray OX passing through the centre of homology. Next let L and L' be another pair of corresponding points, and Y the point in which the ray OLL' cuts the axis of homology. Since the straight lines LM , L'31' must meet in some point Z of the axis XY, it follows that OYLL' is a pro- jection of 0X31M' from ^as centre, and therefore (0YLL') = (0XM31'); consequently the anharmonic ratio (0XM31') is constant for all pairs of corresponding points in the plane. Consider now a pair of corresponding straight lines a and a', the axis of homology s, and the ray o joining the centre of homology to the point aa'. The pencil osaa' is cut by every straight line through in a range of four points analogous to 0XM31'; consequently the anharmonic ratio (osaa') is constant for all pairs of corresponding straight lines a and a', and is equal to the anharmonic ratio (0X31M'). This anharmonic ratio is called the coefficient or parameter of the homology. It is clear that two figures in homology can be constructed when, in addition to the centre and axis, we are given the parameter of the homology. 76. When the parameter of the homology is equal to — 1, all ranges and pencils similar to 0X31M', osaa', are harmonic. 64 ANHARMONIC RATIOS. [77 In this case the homology is called harmonic* or involutorial, and two corresponding points (or lines) correspond to one another doubly ; that is to say, every point (or line) has the same correspondent whether it be regarded as belonging to the first or the second figure. (See below, Arts. 122, 123.) Harmonic homology presents two cases which deserve special notice : ( i ) when the centre of homology is at an infinite distance, in the direction perpendicular to the axis of homology; (2) when the axis of homology is at an infinite distance. In the first case we have what is called symmetry with respect to an axis ; the axis of homology (in this case called also the axis of symmetry) bisects orthogonally the straight line joining any pair of corresponding points, and bisects also the angle included by any pair of corresponding straight lines. The second case is called symmetry with resjyect to a centre. The centre of homology (in this case called also the centre of symmetry) bisects the distance between any pair of corresponding points, and two corresponding straight lines are always parallel. In each of these two cases the two figures are equal and similar (congruent) t ; oppositely equal in the first case, and directly equal in the second. 77. Considering again the general case of two homological figures, let a , b , m, , n be four rays of a pencil in the first figure, and a\ b\ m\ n' the straight lines corresponding to them in the second. Then (mnab) = (m'n'a'b'). Now let an arbitrary transversal be drawn to cut mnab in MNAB, and draw the corresponding (or another) transversal to cut m'n'a'b' in M'N'A'B' ; then {MNAB) = (M'N'A'B'), MA M'A' _ NA N'A' ° r MB '' M'B' ~ NB '' N'B' ' MA M'A' Consequently, the ratio -^^ : ,, /7?/ depends only on the straight lines ah (and a'b'), and not at all on the straight line m (or m'). The ratio MA : NA is equal to that of the distances of the points M, iV'from the straight line a, which distances we may denote by (M, a), (N,.a); thus * Bellavitis, Saggio di Geometria derivata (Nuovi Saggi of the Academy of Padua, vol. iv. 1838), § 50. f Two figures are said to be congruent when the one may be superposed upon the other so as exactly to coincide with it. 77] ANHARMONIC RATIOS. 65 (M,a) (M', a') , that is to say * : In two homological figures (or, more generally, in two protectively related figures) the ratio of the distances of a variable point Mfrom two fixed straight lines a ,b in the first figure hears a constant ratio to the analogous ratio of the distances of the corresponding point M' from the corresponding straight lines a' , b f in the other f,gure. Suppose b to pass through the centre of homology ; then M and M' are collinear with and V coincides with b, so that (M, b):(M',b') = OM: OM' \ and therefore OM {M,a) , Gwwy) = constaat - If N and N' are another pair of corresponding points, we have then OM m (M, a) ON m (N,a) ' 031' : (if', a') ~ ON' '' (N', a')' Now suppose the straight line a f to move away indefinitely ; then a becomes the vanishing line in the first figure ; the ratio \jqr — 7v will in the limit become equal to unity, and thus OM /ll/r . ON '. x — ,:(if,.) = _: W .) = constant ; in other words f : In two homological figures, the ratio of the distances of any point in the first figure from the centre of homology and from the vanishing line respectively, bears a constant ratio to the distance of the corre* sponding point in the second figure from the centre of homology. * Chasles, Gdometrie supirieure, Art. 512. f Chasles, Sections conigues, Art. 267. CHAPTER X. CONSTRUCTION OF PROJECTIVE FORMS. 78. Let ABC and A'B'C be two triads of corresponding elements of two projective forms of one dimension (Fig. 47), and imagine any series of operations (of projection and section) by which we may have passed from ABC to A'B'C. Then whatever this series be*, it will also lead from any other element D of the first form to the element D' which corresponds to it in the second. For if D could give, as the result of these operations, an element D" different from D', then^lhe anharmonic ratios {ABCD) and {A'B'C'D") would be equal; but by hypothesis {ABCD) = {A'B'C'D') ; therefore {A'B'C'D') = {A'B'C'D"), which is impossible unless D" coincide with D' (Art. 65). 79. Theorem (converse to that of Art. 73) : Given two forms of one dimension ; if to the elements A ,B,C,D,... of the one correspond respectively the elements A', B', C, D', ... of the other in such a manner that any four elements of the first form are equianharmonic with the four corresponding elements of the second, then the two forms are projective. For every series of operations (of projection or section), which leads from the triad ABC to the triad A'B'C, leads at the same time from the element D to another element D" such that {ABCD) = {A'B'C'D"). But {ABCD) = {A'B'C'D') by hypothesis ; therefore {A'B'C'D') = {A'B'C'D"), and D" must * In Fig. 47 the series of operations is : a projection from S, a section by u", a projection from S', and a section by vf. 80] CONSTRUCTION OF PROJECTIVE FORMS. 67 coincide with B' (Art. 65). And since the same conclusion is true for any other pair whatever of corresponding elements, it follows that the two forms are projective (Art. 40). 80. From Art. 78 the following may be deduced as a par- ticular case : If among the elements of two projective forms of one dimension titer e are two corresponding triads ABC and A'B'C which are in perspective, then the two forms themselves are in perspective. (1). If, for example, the forms are two ranges ABCB ... and A'B'C'B' ...; then if the three straight lines AA', BB' , CC meet in a point S, the other analogous lines BB',... will all pass through S (Figs. 19, 40). Suppose, as a particular case, that the points A , A' coincide (Fig. 22), so that the two ranges have a pair of corresponding points A and A' united in the point of intersection of their bases *. The triads ABC, A'B'C are in perspective, their centre of perspective being the point where BB' and CC meet; accordingly : \f If two projective ranges have a self-corresponding point 3 they are in perspective. Conversely it is evident that two ranges which are in per- spective have always a self-corresponding point. (2). Again, if the two forms are two flat pencils abed ... and a'b'c'd'... lying in the same plane; then if the three points aa', W, cc' lie on one straight line s, the analogous points dd'. . . will all lie on the same straight line (Fig. 20). If the line s he altogether at infinity, we have the following property : If in two projective flat pencils, three pairs of corresponding rays are parallel to one another, then every pair of corresponding rays are ■ parallel to one another. The hypothesis is satisfied in the particular case where the Fi g rays a and a' coincide (Fig. 48), so that the two pencils have a self-corresponding ray in the * In the case of two projective forms we shall in future employ the term self-corresponding to denote an element which is such that it coincides with its correspondent ; thus A or A' above maybe called a self -corresponding point of the two ranges. F 2 68 CONSTRUCTION OF PROJECTIVE FORMS. [81 straight line which joins their centres ; then s is the straight line joining lib' and cc'. Accordingly: When two projective fiat pencils {lying in the same plane) have a self-corresponding ray t they are in perspective. Conversely, two coplanar flat pencils which are in perspec- tive have always a self-corresponding ray. (3). If one" of the systems is a range ABC I) ... and the other a flat pencil abed ... (Fig. 28), the hypothesis amounts to assuming that the rays a , b, c pass respectively through the points A,B ,C ; then we conclude that also d, ... will pass through D, . . . &c. 81. Two ranges may be superposed one upon the other, so as to lie upon the same straight line or base, in which case they may be said to be collinear. For example, if two pencils (in the same plane) S = abc ... and = a'b'c' ... (Fig. 49) are cut by the same transversal, they will determine upon it two ranges ABC ... , A'B'C ... which will be projectively related if the two pencils are so. The question arises whether there exist in this case any self-corresponding points, i.e, whether two corresponding points of the two ranges coincide in any point of the transversal. If, for instance, the transversal s be drawn so as to pass through the points aa' and bb\ then A will coincide with A\ and B with B' ; in this case consequently there are two self-corresponding points. Again, if a range u be projected (Fig. 50) from two centres S and (lying in the same plane with u), two flat pencils abc ... and a'b'c'. . . will be formed, which have a pair of corresponding rays a , a' united in the line SO. And if a transversal s be drawn through the point in which this line cuts u, we shall obtain two projective ranges ABC ..., A'B'C ... lying on a common base s, and such that they have one self-corresponding point AA'. c c Fig. 49. Fig. 50- 83] CONSTRUCTION OF PROJECTIVE FORMS. 69 And lastly, we shall see hereafter (Art. 109) that it is possible for two collinear projective ranges to be such as to have no self-corresponding point. So also two flat pencils (in the same plane) may have a common centre, in which case they may be termed concentric ; such pencils are formed when two different ranges are pro- jected from the same centre (Fig. 51). And two axial pencils may have a common axis ; such pencils are formed when we project two dif- ferent ranges from the same axis, or the same flat pencil from different centres. Again, if two sheaves are cut by the same plane, two plane figures are obtained; if, on the other hand, Fig. 51. two plane figures are projected from the same centre, two concentric sheaves are formed. In all these cases the forms in question may be said to be superposed one upon the other; and the investigation of their self- corresponding elements, when the two forms are projectively related, is of great importance. 82. Theorem. Two superposed projective {one-dimensional} forms either have at most two self -corresponding elements, or else every element coincides with its correspondent. For if there could be three self-corresponding elements A, B , C suppose ; then if B and B f are any other pair of cor- responding points, we should have (Art. 73) (ABCB) = (ABCB'), and consequently (Art. 65) B would coincide with B'. Unless then the two forms are identical, they cannot have more than two self-corresponding elements. \ 83. Theorem (converse to that of Art. 53). If a one-dimen~ sionalform consisting of four elements A,B ,C ,B is projective with a second form deduced from it by interchanging two of the elements (e.g. BACB), then the form will be a harmonic one, and the two interchanged elements will be conjugate to each other. First Proof. If (ABCB) = (BACB), then (Art. 72. IV) A = £ ; A .-. A 2 = i, and since we cannot take A = + 1 (Art. 72. VIII) we must have A = — 1, i.e. the form is a harmonic one. Second Proof Suppose, for example, that A , B , C , B are four collinear points (Fig. 52). Let K , M , Q , B be a projection of J 70 CONSTRUCTION OF PROJECTIVE FORMS. [84 these points on any straight line through B, made from an arbitrary centre L. Since ABCB is projective with KMQB and also (by hyp.) with BACB, the forms KMQD and BACB are projective with one another. And they have a self-corre- sponding point D ; consequently they are in perspective (Art. 80), and KB, MA, QC will meet in one point A 7 . But this being the case, we have a complete quadrangle KLMN, of which one pair of opposite sides meet in A, another such pair in B, while the fifth and sixth sides pass respectively through C and B. Accordingly (Art. 46) ABCB is a harmonic range. 84. Let there be given two projectively related geometric forms of one dimension. Any series of operations which suf- fices to derive three elements of the one from the three corre- sponding elements of the other will enable us to pass from the one form to the other (Art. 78); and any two given triads of elements are always projective, i.e. can be derived one from the other by means of a certain number of projections and sections. Hence we conclude that : Given three pairs of corresponding elements of two projective forms of one dimension, any number of other pairs of corresponding elements can he constructed. We proceed to illustrate this by two examples, taking (i) two ranges and (s) two flat pencils; the forms being in each case supposed to lie in one plane. Given (Fig. 53) three pairs of Given (Fig. 54) three pairs corresponding points A and A' , of corresponding rays a and a', B and B',C and C / ', of the pro- b and b',c and c', of the projec- jective ranges u and u' \ to con- tive pencils U and U' '; to con- struct these ranges. struct these pencils. We proceed as in Art. 44. On Through the point of infer- tile straight line which joins any section of any two of the corre- two of the corresponding points, sponding rays, say a and a f , say A and A f , take two arbitrary draw two arbitrary transversals points S and S'. Join SB , S'B' s and s'. Join the points sb and cutting one another in B", and sfb' by the straight line b" , and SC , S'C cutting one another in the points sc and sV by the C"; join B /f C /f , and let it cut straight linec"; and let a" be the 84] CONSTRUCTION OF PROJECTIVE FORMS. 71 A A' in A". The operations which straight line joining the points enable us to pass from ABC to b"o" and aaf. The operations Fig- 53- *V 54- A'B'C are : i. a projection from S; 2. a section by u" (the line on which lie the points.4", B", C") ; 3. a projection from S' ; 4. a section by u. The same opera- tions lead from any other given point D on u to the correspond- ing point D' on u', so that the rays SD and S'D' must intersect in a point D" of the fixed straight line u". In this manner a range u"=A"B"C"D"... is obtained which is in perspec- tive both with u and with vf. which enable us to pass from abc to a'b V are : 1 . a section by s ; 2. a projection from the point U" where a", b", c" meet; 3. a section by s' ; 4. a projection from U'. The same operations lead from any other given ray d of the pencil U to the correspond- ing ray df of the pencil TJ'\ so that the points sd and ttdf must lie on a straight line d" which passes through the fixed point U". In this manner a pencil is obtained which is in perspec- tive both with U and with U'. In the preceding construction (left), D is any arbitrary point on w. If D be taken to be the point at infinity on u, then (Fig. 53) SD will be parallel to u; in order therefore to find the point on vf which corresponds to the point at infinity on u, draw SI" parallel to 72 CONSTRUCTION OF PROJECTIVE FORMS. [85 u to cut u" in I" ; then join S'l", which will cut u' in the required point I'. Similarly, if the ray through S' parallel to u cuts u" in J", and SJ" be joined, this will cut u in J, the point on u which corresponds to the point at infinity on u'. If D be taken at P, the point where u and u" meet, then B" also coincides with P, and the point P' on w' corresponding to the point P on ^ is found as the intersection of S'P with u'. Similarly, if Q' be the point of intersection of u' and u", the point on u corresponding to Q' on v! is Q, where SQ' cuts w. 85. The only condition to which the centres S and S' are subject is that they are to lie upon the straight line which joins a pair of corresponding points ; in other respects their position is arbitrary. We may then for in- stance take S at A' and S' at A ( Fi g- 55). Then the ray S'P co- incides with u, and P' is accord- ingly the point of intersection of u and vf. So too the ray SQ' coincides with u, and Q also lies at the point uu'. If then we take the points A' and A as the centres S and S' respectively, the straight line u" will cut the bases u and u re- spectively in P and Q\ the points which correspond to the point uu' regarded in the first instance as the point P / of the line u' and in the second instance as the point Q of the line u. Now in the construction of the preceding Art., the straight line u" was found as the locus of the points of intersection of pairs In the preceding construction (right), d is any arbitrary ray passing through U. If it be taken to be p, the line joining U to U" 9 then the corresponding ray p' of the pencil U f is the line joining the point TJ f to the point s'p. Similarly, if (f be the ray U'U" of the pencil TJ\ the ray q corresponding to it in the pencil U is that which joins the points U and sq'. The only condition to which the transversals s and $' are sub- ject is that they are to pass through the point of intersection of a pair of corresponding rays ; in other respects their position is arbitrary. We may then for in- stance take a' for s and a for s' (Fig. 56). Then the point /p coincides with U, and yf is ac- cordingly the straight line UU'. So too the point sq' coincides with U\ and q also must be the straight line UU'. If then we take the rays a' and a as the transversals s and s' respectively, the point U" will be the intersection of the rays p and ),(B,By(C,C') ..., then the points of intersection of the pairs of straight lines AB ' and A'B , AC and A'C , BC and B'C, ... lie on one and the same straight line u", which passes through the point uu' ; and the straight line joining uu' to the centre of the pencil is the har- monic conjugate of u" with re- spect to u and u'. From this follows the solution of the problem : To draw the straight line con- necting a given point M-with the each pencil to the straight line joining the centres of the pencils when regarded as a ray of the other pencil. If the two pencils U and U' are in perspective (Fig. 59) the rays p and q' will coincide with the straight line UU'; and since through the point of intersection of the rays (ab' , a'b), (acf , a'c), (ad' , a'd), ... and through the point of intersection of the rays (ba' , b'a\ (be' , b'c), (bd' , I'd), ... pass two different straight lines, viz. UU' and (ab', a'b), these points must coincide. This being so, aa'bb' is a complete quadri- lateral, whose diagonals are UU', s (the straight line on which aa' , bb', . . . intersect), and m (the straight line which joins ab r and a 7 6); consequently (Art. 56) the points U and U' are harmonic conjugates with regard to U" and the point in which s meets UU'. If therefore a range be projected from two points U and U' by the rays (a , a'), (b , b'), (c , c') . . ., then the straight lines which join the pairs of points (ab' , a'b), (ac , a'c), (be , b'c), ... meet in one and the same point U", which lies on the line UU' ; and the point where the straight line UU' cuts the base of the range is the harmonic con- jugate of U" with respect to U and U'. From this follows the solution of the problem : To construct the point where a given straight line m would be in- 87] CONSTRUCTION OP PROJECTIVE FORMS. 75 inaccessible point of intersection of two given straight lines u and u'. tersected by a straight line ( U U') which cannot be drawn, but which is determined by its passing through two given points U and U'. Through M (Figs. 57 and 58) draw two straight lines to cut u in A and B, and u r in B' and A' On m (Fig. 59) take two points, and join them to U by the straight lines a and b, and to U' u" M N W A // / XV W B' C * A' W Fig. 58. respectively ; join AA',BB' meet- ing in S. Through S draw any straight line to cut u in C and u' in C", and join B£' , B'C, intersecting in N. The" straight line joining M and N will be the line u" required. Fig. 59- by the straight lines V and a' ; let s be the straight line joining the points of intersection of a , a' and b , b f . On s take any other point and join it to U, U' by the straight lines c , S" at the point where A A' meets CC (Fig. 62). Then since SB , S'B' meet in B', and SC, S'C in C, therefore B'C is the straight line u". Consequently any other pair of corresponding points D and D' are constructed by observing that the straight lines SD, S'D' must m/et on B'C. From a consideration of the figure SS'CDD'B, which is a >n, we derive the theorem : line joining the points aa', cc' be taken as the transversal s, and that joining the points aa', W as the transversal / (Fig. 63). Then since the line joining the points sb , s'b' is b, and the line joining the points sc , sV is c', therefore be' is the point U". Consequently any other pair of corresponding rays d and d' are constructed by observing that the points sd , s'd' must be collinear with be'. From a consideration of the figure ss'cdd'b, which is a hexa- gon (six - side) we derive the theorem : In a hexagon, of which two ver- tices are the centres of two pro- jective pencils, and the four others are the points of intersection of four pairs of corresponding rays, the three points in which the pairs of op>posite sides meet one another are collinear. If the three points aa', bb', ccf in Art. 84 (right) lay on the same straight line s (if, for ex- ample, a and a' coincided), then the two pencils would be in per- spective ; we should therefore only have to connect the two centres of the pencils with every point of s in order to obtain any number of pairs of corresponding rays (Fig. 20). 90. If the two ranges u and u' (Art. 84, left) are superposed one upon the other, i.e. if the six given points AA'BB'CC lie on the same straight line (Fig. 64), we first project u from an arbitrary centre S' on an arbitrary straight line u x , and then proceed to make the construction for the case of the ranges u~{ABC...) and «i== (A 1 B X C 1 ...), i.e. to construct with regard to the pairs of points (AA X ), (BBJ, (CCd in the way shown in Art. 84. A pair of corre- sponding points D and D x of the ranges u and u x having been found, In a hexagon, of which two sides are segments of the bases of two projective ranges, and the four others are the straight lines con- necting four pairs of correspond- ing points, the straight lines which join the three pairs of opposite vertices are concurrent. 89. If in the problem of Art. 84 (left) the three straight lines AA', BB', CC passed through the same point S (if, for example, A and A' coincided), then the two ranges would be in perspective ; we should therefore only have to draw rays through JS in order to obtain any number of pairs of cor- responding points (Fig. 1 9). 78 CONSTRUCTION OF PROJECTIVE FORMS. [90 the ray S'D X determines upon vf the point D' which corresponds toi>. The construction is simpler in the case where two corresponding points A and A' coincide (Fig. 65). When this is so, if u x be drawn through .4, the range u x will be in per- spective with u ; thus, after having projected u' upon u x from an arbi- trary centre JS' 9 if S be the point where BB' and CG X meet, it is only necessary further to project u from S upon u x , and then u x from S' upon u'. The two collinear ranges u and u' have in general two self-corre- sponding points; one at A A', and a second at the point of inter- Fig- 6 4- section of their common base line with the straight line SS'. If then SS' passes through the point uu x , the two ranges u and u' have only one self-corresponding point. If it were desired to con- struct upon a given straight line two collinear ranges having A and A' for a pair of corresponding points, and a single self-corre- sponding point at M (Fig. 66), we should proceed as follows. Take any point S / , and draw any straight line u x through M ; project A / from S' on u x ', join the point A x so found to A, and let A A x meet S'M in S. Then to find the point on uf which corresponds to any point B on u, project B from S into B x , and then B x from S' into B' ; this last is the point required. If the two pencils U , U' (Art. 84, right) are concentric, i.e. if the six rays aa'bb'cc' pass all through one point, we first cut a'b'c' by a transversal and then project the points of intersection from an arbitrary centre U x . If a x b x c x are the projecting rays, we have then 03] CONSTRUCTION OF PROJECTIVE FORMS. 79 only to consider the non-concentric pencils U and U l =(a 1 b 1 c 1 ). Or we may cut abc by a transversal in the points ABC, and a'b'c' by another transversal in A 'B'C, and then proceed with the two ranges ABC ..., A'B'C ... in the manner explained above. The figures corresponding to these constructions are not given; the student is left to draw them for himself. He will see that in these cases also the constructions admit of considerable simplification if, among the given rays, there be one which is self-corresponding ; if, for example, a and a' coalesce and form a single ray, &c. 91. Consider two projective (nomographic) plane figures tt and 7/ ; as has already been seen (Art. 40), any two corresponding straight lines are the bases of two projective ranges, and any two correspond- ing points are the centres of two projective pencils. If the two figures have three self-corresponding points lying in a straight line, this straight line s will correspond to itself; for it will contain two projective ranges which have three self-corresponding points, and every point of the straight line s will therefore (Art. 82) be a self-corresponding point. Consequently every pair of corresponding straight lines of n and i/ will meet in some point on s, and therefore the two figures are in perspective (or in homology in the case where they are coplanar). 92. If two projective plane figures which are coplanar have three self-corresponding rays all meeting in a point 0, this point will be the centre of two corresponding (and therefore projective) pencils which have three self-corresponding rays ; therefore (Art 82) every ray through will be a self-corresponding one. Hence it follows that every pair of corresponding points will be collinear with ; therefore the two figures are in homology. 93. If two projective plane figures which are coplanar have four self-corresponding points A , B , C , D, no three of which are collinear, then will every point coincide with its correspondent. For the straight lines AB , AC , AD , BC , BD , CD are all self- corresponding ; therefore the points of intersection of AB and CD, AC and BD, BC and AD, i. e. the diagonal points of the quadrangle A BCD, are all self-corresponding. Since the three points A , B, and (AB) (CD) are self-corresponding, every point on the straight line AB coincides with its correspondent ; and the same may be proved true for the other five sides of the quadrangle. If now a straight line be drawn arbitrarily in the plane, there will be six points on it which are self-corresponding, those namely in which it is cut by the six sides of the quadrangle ; and therefore every point on the straight line is a self-corresponding one ; which proves the proposition. In a similar manner it may be shown that if two coplanar pro- jective figures have four self-corresponding straight lines a, b, c, d, 80 CONSTRUCTION OF PROJECTIVE FORMS. [94 forming a complete quadrilateral (i.e. such that no three of them are con- current), then every straight line will coincide with its correspondent. 94. Theorem. Two plane quadrangles ABCD , A'B'C'D' are always projective. (1). Suppose the two quadrangles to lie in different planes it , i/. Join A A ', and on it take an arbitrary point S (different from A'), and through A draw an arbitrary plane it" (distinct from it) ; then from S as centre project A', B', C, D' upon it" and let A", B", C", D" be their respective projections (A" therefore coinciding with A). In the plane it join AB , CD, and let them meet in E ; so too in the plane w" join A"B" , C"D", and let these meet in E". The straight lines ABE , A"B"E" lie in one plane since they meet each other in the point A == A" ; therefore BB" and EE" will meet one another in some point S v Now let a new plane it'" (distinct from 77) be drawn through the straight line ABE, and let the points A", B" , C", D", E" be pro- jected from B x as centre upon it'". Let A'", B'", C", D"' , E'" be their respective projections, where A"', B'" , E'" are collinear and coincide with A , B ,E respectively, and C", D'", E'" are collinear also, since their correspondents C", D", E" are collinear. The straight lines CDE , C"'D'"E'" lie in one plane since they meet each other in the point E~E"'; therefore CC" and DD'" will meet one another in some point S 2 . If now the points A f ", B'" , C" , D f " be projected from S 2 as centre upon the plane it, their projections will evidently be A , B , C , D. The quadrangle ABCD may therefore be derived from the quad- rangle A'B'C'D' by first projecting the latter from JS as centre upon the plane it", then projecting the new quadrangle so formed in the plane it" from S x upon it'", and lastly projecting the quadrangle so formed in the plane it"' from >^ 2 upon it ; that is to say, by means of three projections and three sections *. (2). The case of two quadrangles lying in the same plane reduces to the preceding one, if we begin by projecting one of the quadrangles upon another plane. (3). If the two quadrangles (lying in different planes) have a pair of their vertices coincident, say D and D f , then two projections will suffice to enable us to pass from the one to the other; or, what amounts to the same thing, a third quadrangle can be constructed which is in perspective with each of the given ones ABCD, A'B'C'D'. For let there be drawn through D two straight lines s and s', one in each of the planes ; let s cut the sides of the triangle ABC in * Gbassmann, Die stereometrischen Gleichungen dritten Grades und die dadurch erzeugten Oberfiachen ; Crelle's Journal, vol. 49. § 4 (Berlin, 1855). 96] CONSTRUCTION OF PROJECTIVE FORMS. 81 L,M , iV respectively, and let s' cut the sides of the triangle A'B'C'm L',M',N' respectively. Then in the plane ss' the straight lines LL', MM', NN' will form a triangle which is in perspective at once with ABC and with A'B'C. (4). If the quadrangles (still supposed to lie in different planes) have two pairs of their vertices C = C, D == D' coincident, then if the straight lines AA' , BB' meet one another the quadrangles will be directly in perspective, the point of intersection of AA / and BB' being the centre of projection ; so that we can pass at once from the one quadrangle to the other by one projection from 0. If A A ' , BB' are not in the same plane, so that they do not meet one another, then through CD let an arbitrary plane n" be drawn, and in it let the straight line be drawn which meets AB and A'B'. If in this straight line two arbitrary points A", B" be taken, then A"B"C"D" will be a quadrangle which is in perspective at once with ABCD and with A'B'C'D'. 95. From the theorem just proved it follows that two projective plane figures n and 7/ can be constructed when we are given two corresponding quadrangles ABCD, A'B'C'D' ; for the operations (projections and sections) which serve to derive A'B'C'D' from ABCD will lead from any point or straight line whatever of v to the corresponding point or straight line of 7/ ; and vice versa. Or, again, it may be supposed that two corresponding quadrilaterals are given. For if in these two corresponding pairs of opposite ver- tices be taken, we have thus two corresponding quadrangles ; and the operations (projections and sections) which enable us to derive one of these quadrangles from the other will also derive the one quadrilateral from the other. 96. Two plane figures may also be made projective in another manner ; leaving out of consideration the relative position of the planes in which .they lie, we may operate on each of the figures separately *. Suppose that we are given, as corresponding to one another, two complete quadrilaterals abed, a'b'c'd'. We begin by constructing, on each pair of corresponding sides, such as a and a', the projective ranges which are determined by the three pairs of corresponding points ab and a'b', ac and a'c', ad and a'd'. Thi§ done, to every point of any of the four straight lines a,b,c,d will correspond a determinate point of the corresponding line in the other figure. (1). Now let in the first figure a transversal m be drawn to cut a, b, c, d in A , B , C , D respectively ; then the points^', B', C, D' which correspond to these in the second figure will in like manner lie on a straight line m'. * Staudt, Geom. der Lage, Art. 130. G 82 CONSTRUCTION OP PROJECTIVE PORMS. [90 For, considering the triangle abc, cut by the transversals d and m, the product of the three anharmonic ratios a {bcdm) , b (cadm) , c (abdm) is equal to + i (Art. 140); but these anharmonic ratios are equal respectively to the following : a'(b'c'd').A', b'{c'a'd').B' , c'{a'b' 'd').C ', so that the product of these last three is also equal to + i. And therefore, since the points a'd', b / d / , c'd' are collinear, the points A', B', C are also collinear (Art. 140). By considering in the same manner the triangle abd, cut by the transversals c and m, it can be shown that A', B', D' are collinear ; it follows then that the four points A' ', B', C, D' all lie on the same straight line m', the correspondent of m. This proof holds good also when m passes through one of the vertices of the quadrilateral abed ; if for example m pass through cd, the anharmonic ratios c (abdm), d(abcm) will each be equal to + i ; the reasoning, however, remains unaltered. Thus every pair of corresponding vertices of the quadrilaterals abed , a'b'c'd' (for example cd and c'd') become the centres of two projective pencils, in which to c , d, (cd)(ab) correspond c', d', (c'd')ia'b') respectively, and to any ray cutting a , b in two points P , Q cor- responds a ray cutting a', b' in the two corresponding points P',Q'. (2). The two ranges ABCD , A'B'C'D' in which the sides of the quadrilaterals abed, a'b'c'd' are respectively cut by two corresponding straight lines m, m' are projective. For, considering the triangle bem, cut by the transversals a and d, the product of the anharmonic ratios of the three ranges be , B , ba , bd G , cb , ca , cd B , C , A , D is equal to + i. And considering in like manner in the other plane the triangle b'c'm! \ cut by the transversals a' and d', the product of the anharmonic ratios of the three ranges b'c\ B\ b'a', b'd' B' , C , A* , D' is also equal to + i. But the range in which b is cut by the pencil cmad is equianharmonic with the range in which b' is cut by the pencil c'm'a'd' ; i.e. the ranges be j B , ba , bd 6V, B\ 6V, b'd' are equianharmonic ; and for a similar reason the ranges 90] CONSTRUCTION OF PROJECTIVE FORMS. 83 C , cb , ca , ed C, e'b\ c'a' s c'd' are equianharmonic. Therefore the ranges B , C , A , D B', Q\ A', D' will be equianharmonic and therefore projective ; whence it follows that the projective ranges m and m' are determined by means of the pairs of corresponding points lying on a and a, b and b', c and c'. (3). If the straight line m turn round a fixed point M, then m' also will revolve round a fixed point. For by hypothesis the points A and B, in which m cuts a and b, describe two ranges in perspective whose self-corresponding point is ab. Similarly the points A', B' describe two ranges, which, being respectively projective with the ranges on a , 6, are projective with one another ; and which are further seen to be in perspective, since they have a self-corresponding point a'b'. Consequently the straight line m' will always pass through a fixed point M', the correspondent of M ; and will therefore trace out a pencil. The pencils generated by m and m are projective, since the ranges are projective in which they are cut by a pair of corresponding sides of the quadrilaterals, e.g. by a and a'. To the rays of the pencil M which pass respectively through the vertices ah , ac , ad, be , bd , cd of the quadrilateral abed correspond the rays of the pencil M f which pass respectively through the vertices a'V, a'cf, a'd' , 6V, b'd', c'd' of the quadrilateral a / b / c / d / . This reasoning holds good also when the point M, round which m turns, lies upon one of the sides of the quadrilateral, on c for example ; because we still obtain two ranges in perspective upon two of the other sides. - Since c is now a ray of the pencil M , c' will be the corresponding ray of the pencil M'\ that is to say, M f will lie on c'. If M be taken at one of the vertices, as ed, then M f will coincide with c'd', &c. (4). Now suppose the pencil Mto be cut by a transversal n, and the pencil M ' to be cut by the corresponding straight line n'. "While the point mn describes the range n, the corresponding point m'n' will describe the range n' ; and these two ranges will be projective since they are sections of two projective pencils. When the point mn falls on one of the sides of the quadrilateral abed, the point mn' will fall on the corresponding side of the quadrilateral a'b'c'd' ; therefore the two projective ranges are the same as those which it has already been shown may be obtained by starting from the pairs of correspond- ing points on a and a', b and b', c and e' . G 2 84 CONSTRUCTION OF PROJECTIVE FORMS. [97 In this manner the two planes become related to one another in such a way that there corresponds uniquely to every point in the one a point in the other, to every straight line a straight line, to every range a projective range, to every pencil a projective pencil. The two figures thus obtained are the same as those which can be obtained, as explained above (Art. 95) by means of successive projections and sections, so arranged as to lead from the quadrilateral abed to the quadrilateral a'b'c'd'. For the two figures 7/ derived from it by means of these two processes have four self-corresponding straight lines a', b', c', d f forming a quadrilateral, and therefore (Art. 93) every element (point or straight line) of the one must coincide with the corresponding element in the other ; i.e. the two figures must be identical. 97. Theorem. Any two projective plane figures {the straight lines at infinity in which are not corresponding lines) can be superposed one upon the other so as to become homological. Let i , j f be the vanishing lines of the two figures — i.e. the straight lines in each which correspond respectively to the straight line at infinity in the other. In the first place let one of the figures be superposed upon the other in such a manner that i and j' may be parallel to one another. Since to any point M on i corresponds a point at infinity in the second figure, to the pencil of straight lines in the first figure which meet in M corresponds in the second figure a pencil of parallel rays. Through M draw the straight line m parallel to these rays ; then m will be parallel to its correspondent m'. Similarly let a second point N be taken on i and through N let the straight line n be drawn which is parallel to its correspondent n' ; let m and n meet in S, and rnf and n' in JS'. If through S a straight line I be drawn parallel to *, its correspondent V will pass through S' and will also be parallel to i, since the point at infinity on i corre- sponds to itself. The corresponding pencils S and S' are therefore such that three rays I , m , n of the one are severally parallel to the three corresponding rays V, m', n' of the other ; and consequently (see below, Art. 104) the two pencils are equal. Now let one of the planes be made to slide upon the other, without rotation, until S' comes into coincidence with JS; then the two pencils will become concentric ; and since they are equal, every ray of the one will coincide with the ray corresponding to it in the other. This being the case, every pair of corresponding points will be collinear with S, and the two figures will be homological, S being the centre of homology. 98. Suppose that in a plane n is given a quadrangle A BCD, and in a second plane it' a quadrilateral a'b'c'd'. By means of construc- tions analogous to those explained in Arts. 94-96, the points and straight lines of the one plane can be put into unique correspondence 98] CONSTRUCTION OF PROJECTIVE FORMS. 85 with those of the other, so that to any range in the first plane cor- responds in the second plane a pencil projective with the said range, and to any pencil in the first plane corresponds in the second plane a range projective with the said pencil. Two plane figures related to one another in this manner are called correlative or reciprocal. CHAPTEK XI. PARTICULAR CASES AND EXERCISES. 99. Two ranges are said to be similar, when to the points A,B,C, D,... of the one correspond the points^', B', C',D', ... of the other, in such a way that the ratio of any two corre- sponding segments AB and A'B', AC and A'C , ... is a con- stant. If this constant is unity, the ranges are said to be equal. Two similar ranges are projective, every anharmonic ratio such as (ABC I)) being equal to the corresponding ratio (A'B'C'B'). For suppose the bases of the two ranges to lie in the same plane (Fig. 6j) and let their point of inter- section be denoted by P f when considered as a point be- longing to u' and by Q when Fig 6 7> considered as a point belong- ing to u. Let A, A' be any pair of corresponding points ; P that point of u which corre- sponds to P\ and Q' that point of u' which corresponds to Q. Draw A A" parallel to u', and A' A" parallel to u. - The triangles PQQ', PA A" have the angles at Q and A equal and the sides about these equal angles proportionals, since by hypothesis PQ PA PA_ P'Q' ~~ P'A' ~ AA" ' Therefore the triangles are similar, and the angles QPQ' and APA" are equal; and consequently the points P,Q[, A" are collinear. If then the range ABC ... be projected upon PQ', by straight lines drawn parallel to n f , we shall obtain the range A" B" C" ... ; and from this last, by projecting it upon 102] PARTICULAR CASES AND EXERCISES. 87 u' by straight lines drawn parallel to u, the range A' B' C ... may be derived. If PQ = P' Q', i.e. if the straight line PQ' makes equal angles with the bases of the given ranges, the ranges are equal. To the point at infinity of u corresponds the point at infinity of u\ 100. Conversely, if the points at infinity I and V of two projective ranges u and u' correspond to each other, the ranges will he similar. For if (Fig. 67) u be projected from /', and u f from /(as in Art. 85, left), two pencils of parallel rays will be formed, corresponding pairs of which intersect upon a fixed straight line u". The segments A" B" of »" will be propor- tional to the segments AB of u and also to the segments A' B' of u' ; consequently the segments AB of u will be proportional to the segments A'B' of nf. Otherwise: if AA\ BB\ CC are three pairs of corre- sponding points, and / , V the points at infinity, we have (by Art. 73) (ABCI) = A'B' CI') ; or (by Art. 64), since /and /' are infinitely distant, AC A'C BC ~ B'C' an equation which shows that corresponding segments are proportional to one another. Examples. If a flat pencil whose centre lies at a finite distance be cut by two parallel straight lines, two similar ranges of points will be obtained. Any two sections of a flat pencil composed of parallel rays are similar ranges. In these two examples the ranges are not only projective, but also in perspective : in the first case the self-corresponding point lies at infinity ; in the second case it lies (in general) at a finite distance. 101. Two flat pencils, whose centres lie at infinity, are pro- jective and are called similar, when a section of the one is similar to a section of the other. When this is the case any other two sections of the pencils will also be similar to one another. 102. From the equality of the anharmonic ratios we con- clude that two equal ranges are projective (Art. 79), and that 88 PARTICULAR CASES AND EXERCISES. [103 conversely two projective ranges are equal (Art. 73), when the corresponding segments which are bounded by the points of two corresponding triads ABC and A'B'C are equal ; i.e. when A'B' = AB, A'C'=AC, (and consequently B'C'=BC). Examples. If a flat pencil consisting of parallel rays be cut by two transversals which are equally inclined to the direction of the rays, two directly equal ranges of points will be obtained *. If a flat pencil of non-parallel rays be cut by two transversals which are parallel to one another,. and equidistant from the centre of the pencil, two oppositely equal ranges will be obtained *. 103. Two similar ranges lying on the same base, and which have one self- corresponding point N at infinity, have also a second such point M, which is in general at a finite distance. If A A ', BB' are two pairs of corresponding points, MA : MA' = AB : A'B' = a constant. To find M therefore it is only necessary to divide the segment A A' into two parts MA , MA' which bear to one another a given ratio. This ratio MA : MA' is equal (Art. 64) to the anharmonic ratio (AA'MN). If its value is — i, the points AA'MN are harmonic (Art. 68), i.e. M is the middle point of AA' , and similarly also that of every other corresponding segment BB',... ; in other words, the two ranges, which in this case are oppositely equal, are composed of pairs of points which lie on opposite sides of a fixed point M, and at equal distances from it. But if the constant ratio is equal to + i , i. e. if MA and MA ' are equal in sign and magnitude, the point M will lie at infinity. For since (AA'MN)=i, .-. (NMA'A)=i (Art. 45); consequently the points M and N coincide. It follows also from the construction of Art. 90 (Fig. 66) that two ranges on the same base, which Jiave a single self-corres2)onding point lying at infinity, are directly equal. For if in Fig. 66 the point M move off to infinity, the straight lines SS' and A X B X become parallel to the a"; Bi given straight line u or u' on which Yig, 68. t'he ranges lie (Fig. 68), and as the triangles SA 1 B 1 and S'A 1 B l lie upon the same base and between the same parallels, the segments * Imagine a moving point P to trace out a range ABC... and its correspondent P' to trace out simultaneously the equal range A'B'C'.... Then if P and P' move in the same direction, the two ranges are said to be directly equal ; if P and P' move in opposite directions, the ranges are said to be oppositely equal. 106] PARTICULAR CASES AND EXERCISES. 89 which they intercept upon any parallel to the base are equal ; thus JB=A / B / , or two corresponding segments are equal', consequently AA'=BB', i.e. the segment bounded by a pair of corresponding points is of constant length. We may therefore suppose the two ranges to have been generated by a segment given in sign and magnitude, which moves along a given straight line ; the one extremity A of the segment describes the one range, and the other extremity A / describes the other range. Conversely it is evident that if a segment A A', given in sign and magnitude, slide along a given straight line, its extremities A and A' will describe two directly equal (and consequently projective) ranges, which have a single self-corresponding point, lying at an infinite distance. 104. Two flat pencils are said to be equal when to the elements of the one correspond the elements of the other in such a way that the angle included between any two rays of the first pencil is equal in sign and magnitude to the angle included between the two corresponding rays of the second. It is evident that two such pencils can always be cut by two transversals in such a way that the resulting ranges are equal ; but two equal ranges are always projective ; therefore also two equal flat pencils are always projective. Conversely, two projective flat pencils abed... and a'b'c'd' ... will be equal if three rays abc of the one make with each other angles which are equal respectively to those which the three corresponding rays make with each other. This theorem may be proved by cutting the two pencils by two transversals in such a way that the sections ABC and A'B'C of the groups of rays abc smd^a'b'c' may be equal. The projective ranges so formed will be equal (Art. 102); con- sequently also the other corresponding angles ad and a'd\ ... of the given pencils must be equal to one another. 105. Since two equal forms (ranges or flat pencils) are always projective with one another, it follows that if a range or a flat pencil be placed in a different position in space, without altering the relative position of its elements, the form in its new position will be projective with regard to the same form in its original position. 106. Consider two equal pencils abed... and a'b'c'd'... in the same plane or in parallel planes ; and suppose a ray of the one pencil to revolve about the centre and to describe the 90 PARTICULAR CASES AND EXERCISES. [107 pencil ; then the corresponding ray of the other pencil will describe that other pencil, by revolving about its centre. This revolution may take place in the same direction as that of the first ray, or it may be in the opposite direction ; in the first case the pencils are said to be directly equals and in the second case to be oppositely equal to one another. In the first case the angles aa', W ', cc',... are evidently all equal, in sign as well as in magnitude ; consequently a pair of corresponding rays are either always parallel or never parallel. In the second case two corresponding angles are equal in magnitude, but of opposite signs. If then one of the pencils be shifted parallel to itself until its centre coincides with that of the other pencil, the two pencils, now concentric, will still be projective (Art. 105) and will evidently have a pair of corresponding rays united in each of the bisectors (internal and external) of the angle included between two correspond- ing rays a and a'. It follows that these rays are also the bisectors of the angle included between any other pair of corresponding rays. If the first pencil be now replaced in its original position, so that the two pencils are no longer con- centric, we see that there are in each pencil two rays, each of which is parallel to its correspondent in the other pencil ; and these two rays are at right angles to each other, since they are parallel to the bisectors of the angle between any pair of correspond- ing rays. 107. If two flat pencils abed... and a'b'c'd' ... are projective, and if the angles aa' , W ', cc' included by three pairs of corresponding rays are equal in magnitude and of the same sign, then the angle dd' included by any other pair of corresponding rays will have the same For if we shift the first pencil parallel to itself until it becomes concentric with the second, and then turn it about the common centre through the angle aa', the rays a,b,c will coin- cide with the rays a', b', c' respectively. The two pencils, which are still projective (Art. 105), have then three self-correspond- ing rays; consequently (Art. 82) every other ray will coincide with its correspondent. If now the first pencil be moved back into its original position, the angle dd' will be equal to aa'. 108. As the angles aa', bb' , cc', ... of two directly equal 109] PARTICULAR CASES AND EXERCISES. 91 pencils are equal to one another, such pencils, when concentric and lying in the same plane, may be generated by the rotation of a constant angle aa! round its vertex 0, supposed fixed ; the one arm a traces out the one pencil, while the other arm a' traces out the other pencil. Conversely, if an angle of constant magnitude turn round its vertex, its arms will trace out two (directly) equal and therefore projective pencils. Evidently these pencils have no self-corresponding rays. A transversal cutting these two pencils determines on itself two collinear ranges having no self-corresponding points. What has been said in Arts. 104-108 with respect to two pencils in a plane might be repeated without any alteration for the case of two axial pencils in space. 109. (1). Let ABC ..., A'B'C ' ... be two projective ranges lying upon the same base, and let them, by means of the pencils abc ..., a'b'c' '..., be projected from different points U ,U'. Let i,j' be those rays passing through U, U' respectively, which are parallel to the given base, and let i' ', j be the rays corresponding to them. The points I', J in which these last two rays cut the given base will then be those points which correspond to the point at infinity (I or J') of the base, according as that point is regarded as belonging to the range ABC ... or to the range A'B'C' ... The fact that the two corresponding groups of points are pro- tectively related gives an equation between the anharmonic ratios, from which we deduce (as in Art. 74) JA.rA'=JB.I'B'= a constant; . . . . (1) i.e. the product J A . FA' is constant for every pair of points A, A'. Let be the middle point of the segment JI', and 0' the point corresponding to regarded as a point belonging to the first range. Since the equation (1) holds for every pair of corresponding points, and therefore also for and 0', we have jA.rA'=jo.ro\ (2) or (OA-OJ) (pA'-0I') + 0J(00'-0r) = o ; or since 01' = — OJ, 0A.0A'-0r(0A-0A'+00') = o (3) Let us now enquire whether there are in this case any self- corresponding points. If such a point exist, let it be denoted by E ; then replacing both A and A' in (3) by E, we have 0E 2 =0r.00' (4) We conclude that when 01', 00' is positive, i.e. when does not 92 PARTICULAR CASES AND EXERCISES. [109 lie between I' and 0' , there are two self-corresponding points E and F, lying at equal distances on opposite sides of 0, and dividing the segment I'O' harmonically (Art. 69). When lies between I' and 0' , there are no such points. When 0' coincides with 0, there is only one such point, viz. the point itself. (2). Imagine each of the given ranges to be generated by a point moving always in one direction*. If the one. range is described in the order ABC, the other range will be described in the order A'B'C; this order may be the same as the first, or may be opposite to it. If the order of ABC is opposite to that of A'B'C, the same will be the case with regard to the order of IJA and that of I' J' A ', and again with regard to the finite segment J A and the infinite segment J 'A'; i.e. the finite segments J A and I 'A' have the same sign. In con- sequence therefore of equation (2), JO and I'O' have the same sign ; so that does not fall between I' and / (Fig. 6ga) ; there are there- fore two self-corresponding points. And these will lie outside the finite segment JI', since OE is a mean proportional between 01' and 00'. If the order of ABC is the same as that of A 'B'C, we arrive in a similar manner at the con- clusion that J A and I'A\ and again JO and I'O', have opposite signs. In this case Fig. 69. then,self-correspondingpoints exist if does not lie be- tween I' and 0'; that is, if 0' lies between and I' (Fig. 696). And these will lie within the segment JI', since OE is a mean proportional between 01' and 00'. (3). Suppose that there are two self-corresponding points E and F (Fig. 70) ; draw through E any straight line, on which take two points S, S'; and project one of the ranges from S and the other from S'. The two pencils which result are in perspective, since they have a self-corre- sponding ray SES'; accordingly the corresponding rays SA and S'A', SB and S'B', ...SF and Fig. 70. ( S'F' will intersect in points lying on a straight line u' which passes through F. Let E" be the point where this straight line u" meets SS'. Then * Steiner, loc. cit. p. 61. § 16, II. Collected Works, vol. i. p. 280. « J u u' u J 0' u tt' 110] PARTICULAR CASES AND EXERCISES. 93 EFAA' and EFBB' are the projections of EE"SS' from the centres A" and B" respectively ; therefore EFAA f and EFBB' are projective with one another ; thus the anharmonic ratio of the system consisting of any two corresponding points together with the two self-corre- sponding points is constant. In other words : two projective forms which are superposed one upon the other, and which have two self-corresponding elements, are composed of pairs of elements which give with two fixed ones a constant anhar- monic ratio *. (4). Next suppose that there are no self-corresponding points ; so that lies between 0' and I' (Fig. 71). Draw from a straight line U at right angles to the given base and make U the geometric mean between I'O and 00'; thus I'UO' will be a right angle. Again, draw through Uthe straight line IUJ' parallel to the given base; then the angle IUI' will be equal to JUJ', and the angle 0U0' will be equal to 01' U and therefore to IUI'. Thus in the il____^_ L two projective pencils which pro- •yn^> ject the two given ranges from U, ' Sy | \ \X the angles ITJV, JUJ', OUO' % f 2 ^ J-^ ^ — £ included by three pairs of cor- responding rays are all equal ; lg ' * ' consequently (Art. 107) the angles A UA', BUB', ... are also all' equal to them and to one another, and are all measured in the same direction t. Thus : two collinear ranges which have no self- cor responding points can always be regarded as generated by the intersection of their base line with the arms of an angle of constant magnitude which revolves, ahvays in the same direction, about its vertex. 110. We have seen (Art. 84) the general solution of the problem : Given three pairs of corresponding elements of two projective one- dimensional forms, to construct any desired number of pairs ; or, in other words, to construct the element of the one form which corre- sponds to a given element of the other. The solution of the following particular cases is left as an exercise to the student : 1. Suppose the two forms to be two ranges u and u' which lie on different bases ; and let the given pairs of elements be (a) P and P', Q and Q% A and A' \ * The above construction gives the solution of the problem: Given two pairs A , A and B , B f of corresponding points, and one of the self-corresponding points E, to find the other self-corresponding point. t Chasles, loc. cit. p. 119. t P , P / , Q, Q', IfJft J, J' have the same meaning as in Art. 84 ; A ,B , ... are any given points. 94 PARTICULAR CASES AND EXERCISES. [Ill G>) P and P', .4 and .4', P and B' (c) / and /', »/ and J", P and P' (d) / and /', »7 and A', A" in moving describe three ranges which are two and two in perspective. Or the theorem of Art. 16 may be applied to two positions of the variable triangle. This proposition proved, the following corollary may be at once deduced : If the four vertices A,A f ,A'^A ffr of a variable quadrangle slide re- spectively upon four fixed straight lines which all pass through the same point 0, while three of its sides AA',A'A", A"A "' turn respectively round three fixed points C', B"' t B f , then will the fourth side A"' A and the diagonals A A", A' A"' pass respectively through three other fixed points C", C", B", which are deter- mined by the three former ones. The six fixed points ^are the vertices of a complete quadrilateral, i.e. they lie three by three on four straight lines (Fig. 7 2). In a similar manner may be deduced the analogous corollary relating to a polygon of n vertices. 112. Theokem. If a triangle 1 2 3 circumscribes another triangle U 1 U 2 U 3 , there exist an infinite number of triangles each of ivhich is circumscribed about the former and inscribed in the latter (Fig. 73). The two pencils 0,(^,^,0;...) and O z (U lt U„U z ...) Fig. 72. 114] PARTICULAR CASES AND EXERCISES. 95 obtained by projecting the range U 2 U 3 ... from 2 and from 3 , are evidently in perspective. Similarly the pencils O x (U v U 2 , U 9 ... ) and 3 (tr u TJ % , U 3 ... ) obtained by projecting the range U x U 3 ... from O x and from 3 , are in perspective. Therefore the pencils O x (U x ,U 2 ,U 3 ...) and 2 (U X , U 2 ,U 3 ...) are projective (Art. 41); but the rays O x U 3 and 2 U 3 coincide; therefore (Art. 62) the pencils are in perspective, and their corre- sponding rays intersect in pairs on U X U 2 . There are then three pencils O x , 2 , 3 , which are two and two in perspective ; corresponding rays of the first and second, second and third, third and first, intersecting in pairs on the straight lines U X U 2 , U 2 U 3 , U 3 U X respectively. This shows that every triad of corresponding rays will form a triangle which is cir- cumscribed about the triangle O x 2 3 , and inscribed in the triangle U X U 2 U*. Fig. 73. 113. Theorem. A variable straight line turning about a fixed point U cuts two fixed straight lines u and u' in A and A ' respectively ; if S , S' are two fixed points collinear with uu f , and SA. , S'A' be joined, the locus of their point of intersection M will be a straight line t. 1 To prove this, we observe that the points A and A' trace out two ranges in perspective with one another, and that consequently the pencils generated by the moving rays SA , S'A ' are in perspective (Arts. 41, 80). The demonstration of the correlative theorem is proposed as an exercise to the student. 114. Theorem. U,S,S / are three collinear points ; a transversal turning about U cuts two fixed straight lines u and u' in A and A' respectively ; if SA , S / A / be' joined, their point of intersection M will describe a straight line passing through the point uu'\ K The proof is analogous 'to that of the preceding theorem. The proposition just stated may also be enunciated as follows : If the three sides of a variable triangle A A 'M turn respectively about three fixed collinear points U, S , #', while two of its vertices A , A / * Steiner, loc. cit. p. 85. § 23, II. Collected Works, vol. i. p. 297. t Pappus, loc. cit., book VII. props. 123, 139, 141, 143. Chasles, loc. cit. pp. 241, 242. X Chasles, loc. cit. p. 242. 96 PARTICULAR CASES AND EXERCISES. [115 Fig. 74. slide respectively upon two fixed straight lines u , u' , then will the third vertex M also describe a straight line *. In a like manner may be demonstrated the more general theorem : If a polygon of n sides displaces itself in such a manner that each of its sides passes through one of n fixed collinear j>oints, while n—i of its vertices slide each on one of n — i fixed straight lines, then will also the remaining vertex, and the point of intersection of any two non-consecutive sides, describe straight lines t. The correlative proposition is indicated in Art. 85. 115. Problem. Given a parallelogram ABC D and a point P in its plane, to draw through P a parallel to a given straight line EF also lying in the plane, making use of the ruler only. First Solution. — Let E and F (Fig. 74) be the points where the given straight line is cut by AB and AD respectively. On AC take any point K\ join EX, meeting CD in G, and FK, meeting BC in H. The triangles AEF , CGH are homological (Art. 18), since AC , EG, FH meet in the same point K ; and the axis of homology is the straight line at infinity, since the sides AE , AF of the first triangle are parallel respectively to the cor- responding sides CG, CH of the second. Therefore also the remaining sides EF and GH are parallel to one another £. The problem is thus reduced to one already solved (Art. 86), viz. given two parallel straight lines EF and GH, to draw through a given point P a parallel to them. Second Solution §. — Produce (Fig. 75) the sides AB, BC, CD, DA q P and a diagonal AC of the given parallelogram to meet the given straight line EF in E, F, G, H, I respectively, and join EP, GP. Through I draw any straight line cutting EP in A' and GP in C, and join HA', FC / ; if these meet in Q, then will PQ be the required straight line. For if B' denote the point where EP cuts FQ, and D' the point * This is one of Euclid's porisms. See Pappus, loc. cit., preface to book VII. f This is one of the porisms of Pappus; loc. cit., preface to book VII. % Poncelet, Propriitis projectives, Art. 198. § Lambert, Freie Perspective (Zurich, 1774), vol. ii. p. 169. Pig- 75. 117] PARTICULAR CASES AND EXERCISES. 97 where GP cuts IIQ, the parallelograms ABCD and A'B'C'D' are homological, EF being the axis of homology. The point P corre- sponds to the point of intersection of AB and CD, and the point Q to that of BC and AD ; therefore PQ corresponds to the line at in- finity in the first figure ; accordingly it is the vanishing line of the second figure, and consequently PQ is parallel to EF (Art. 18). 116. Pkoblem. Given a circle and its centre; to draw a perpen- dicular to' a given straight line, making use of the ruler only. Draw two diameters AC , BD of the circle (Fig. 76) ; the figure ABCD is then a rectangle. Accordingly, if any point K be taken on the circumference, then by means of the last proposition (Art. 1 1 5) a parallel KL can be drawn to the given straight line EF. If the point L where this parallel again meets the circumference be joined to the other extremity M of the diameter through K, then evidently LM will be perpendicular to KL, and therefore also to the given straight line. N pj g# ^5. 117. Problem. Given a segment AC and its point of bisection B, to divide BC into n equal parts, making use of the ruler only. Construct a quadrilateral JJLDN (Fig. 77) of which one pair of opposite sides DL , NU meet in A, the other pair LTJ , DN in C, and of which one diagonal DU passes through B ; the other diagonal LN will be parallel to AC (Art. 59), and will be bisected in M by DU. Fig. 77. Now construct a second quadrilateral VMEO which satisfies the same conditions as the first, and which moreover has M for an extremity and N for middle point of that diagonal which is parallel to AC. To do this it is only necessary to join AM and BN, meeting in E, and to join CE ; this last will cut LN produced in a point H 98 PARTICULAR CASES AND EXERCISES. [118 sucli that NO=.MN—LM. Now construct a third quadrilateral analogous to the first two, and which has N for an extremity and for middle point of that diagonal which is parallel to AC. If P is the other extremity of this diagonal, then OP = NO=MN=LM. Proceed in a similar manner, until the number of the equal segments LM ,MN ,N0, OP, ... is equal to n. If PQ is the segment last obtained, join LB, meeting QC in Z) the straight lines which join Z to the points M , N ,0 ,P, ... will divide BC into n equal parts *. 118. The following problems, to be solved by aid of the ruler only, are left as exercises to the student : Given two parallel straight lines AB and u ; to bisect the seg- ment AB (Art. 59). Given a segment AB and its point of bisection C ; to draw through a given point a parallel to AB (Art. 59). Given a circle and its centre ; to bisect a given angle (Art. 60). Given two adjacent equal angles AOC, COB; to draw a straight line through at right angles to OC (Art. 60). 119. Theorem. If two triangles ABC , A' B'C, lying in different planes a , te where they intersect the other given ray. The straight lire u'\ the locus of the points of intersection of the "-pairs- of 3mea such as MN', M'N, formed by joining crosswise any two pairs of corresponding points of the ranges u , u\ (Art. 85), will pass through 0, since the lines AB' , A'B meet in that point. If now there be drawn through any other ray, cutting the transversals say in C and B', then will C'B also pass through 0, i.e. the rays OCB' and OBC also correspond doubly to each other. We conclude that When two superposed projective forms of one dimension are such that any one element has the same correspondent, to whichever form it be regarded as belonging, then every element possesses this 123. This particular case of two superposed projective forms of one dimension is called Involution*. We speak of an involution of points, of rays, or of planes, according as the v elements are points of a range, rays of a flat pencil, or planes of an axial pencil. In an involution, then, the elements are conjugate to one another in pairs; i.e. each element has its conjugate. To I whichever of the two forms a given element be considered to « * Desargues, Brouillon projet d'une atteinte aux evenements des rencontres oVun cdne avec unplan (Paris, 1639) : edition Poudba (Paris, 1864), vol. i. p. 119. 102 INVOLUTION. [124 belong, the element which corresponds to it is the same, viz. its conjugate. It follows from this that it is not necessary to regard the two forms as distinct, but that an involution may he considered as a set of elements which are conjugate to one another in pairs. When AA',BB', CC',... are said to form an involution, it is to be understood that A and ^',i?andi?', Cand C\... are pairs of conjugate elements ; moreover, any element and its con- jugate may be interchanged, so that AA' BB' CC ... and A' A B'B CC. . . are projective forms. 124. Since an involution is only a particular case of two superposed projective forms, every section and every projection of an involution gives another involution *. Two conjugate elements of the given involution give rise to two conjugate elements of the new involution. It fellows (Art. 18) that^the figure homological with an involution is also an\ involution. . 125, When two collinear projective ranges form an involu- tion, tMeiie; corresponds to each point (and consequently also to the point at infinity i" or /') a single point (1' or J) ; i.e. the two vanishing points coincide in a si ngle point. Let this point, the conjugate of the point at infinity, be denoted by 0. The equation (1) of Art. 109 then becomes OA . OA' = constant. In other words, an involution of points consists of pairs of points A, A' which possess the property that the rectangle contained by their distances from a fixed point 0, lying on the base, is constant f- This point is called the centre of the involution. The self-corresponding elements of two forms in involution ! are called the double elements of the involution. In the case of 1 the involution of points AA', BB',... we have OA.OA'=OB.OB'=,..= constant. If this constant is positive, i. e. if does not lie between two conjugate points, there are two double points E&ndF, such that OE 2 = OF 2 = OA .OA' = OB .OB' = ... ; * Desargues, he. cit. p. 147. t Ibid. pp. 112, 119. 126] INVOLUTION. 103 therefore lies midway between E and F, and the segment FF divides harmonically each of the segments AA',JBF', ... (Art. 69. [3]). Accordingly: If an involution has two double elements, these separate har monically any pair of conjugate elements; or: An involution is mack up of pairs of elements which are harmonically conjugate tvith regard to two fixed elements. If, on the other hand, the constant is negative, i.e.\i falls between two conjugate points, there are no double points. In this case there are two conjugate points situated at equal distances from and on opposite sides of it, such that OE= - OF', and OF 2 = OF' 2 = -OE.OF'=-OA. OA'. If the constant is zero, there is only one double point ; but in this case there is no involution properly so called. For since the rectangle OA. OA' vanishes, one out of every pair of conjugate points must coincide with 0. 126. The proposition that if an involution has two double elements, these separate harmonically any pair of conjugate elements, may also be proved thus : Let F a,nd F be the double elements, A sand A' any pair of conjugate elements ; since the systems FFAA', EFA'A are pro- jective, therefore (Art. 83) each of them is harmonic. The following is a third proof. Consider EAA f ... and FA' A... as two projective ranges, and project them respectively from two points SandS' collinear with F (Fig. 8i). The projecting pencils S(FAA'...) and. S'(EA'A...) are in perspective (since they have a self-corresponding ray in SS'F) ; therefore the straight line which joins the point of intersection of SA.tmd S'A' to that of SA' and S'A will contain the points of inter- section of all pairs of corresponding Fig. 8i. rays, and will consequently meet the common base of the two ranges at the second double point F. But from the figure we see that we have now a complete quadrilateral, one diagonal of which, AA', is cut by the other two in F and F; consequently (Art. 56) FFAA' is a harmonic range. 104 INVOLUTION. [127 The proposition itself is a particular case of that proved in Art. 109 (3). From this we conclude that the pairs of elements (points of a range, rays or planes of a pencil) which, with two fixed elements, give a constant anharmonic ratio, form two superposed projective forms, which become an involution in the case where the anharmonic ratio has the value — i (Art. 68). 127. An involution is determined ly two 'pairs of conjugate For let A , A 'and B,B'be the given pairs. If any element C be taken, its conjugate is determinate, and can be found as in Art. 84, by constructing so that the form A'AB'C' shall be projective with AA'BC. We then say that the six elements AA\BB\CC T are in involution; i. e. they are three pairs of an involution. Suppose that the involution with which we have to deal is an involution of points. Take any point G (Fig. 82) outside the base, and describe circles round GAA f 'and GBB'; if His the second point in which these circles meet, join GH, and let it cut the base in 0. Since GHAA' lie on a circle, G. OH=OA.OA'; and since GHBB' lie on a circle, OG.OH=OB.OB'; /. OA.OA'=OB.OB'. is therefore the centre of the involution determined by the pairs of points A , A 'and B ,B'. If any other circle be drawn through 6r and ZT, and cut the base in CandC", we have OG.OE= OC.OC'\ :. OC . OC' = OA . OA' = OB .OB', and C, C are therefore a pair of conjugate points of the invo- In other words, the circle which passes through two lution. 128] INVOLUTION. 105 conjugate points C, C or D ,1)' and through one of the points G , H always passes through the other. Accordingly : The pairs of conjugate points of the involution are the points of intersection of the base with a series of circles passing through the points G and H. 128. From what jxrecedes itis^evident that if the involution has double points , t he s e wil l be the points of contact ofjhe base wit h the two^ircleswhich ca n be draw n to pass through G and i/ _and to touch the ha,sft. It has already been seen (Art. 125) that these points are harmonically conjugate with regard to A and A\ and also with regard to B and B'. Con- sequently (Art. 70) the involution has double points when one of the pairs AA f , BB / lies entirely within or entirely without the other \ i.e. when the segments AA r and BB' do not overlap (Fig. 82)-; and the involution has 'no double points when one pair is alternate to the other, [^ and BB' overlap (Fjg. 83) *, In the first case, the involution (as already seen) consists of an infinite number of pairs of points which are harmonically conjugate with regard to a pair of fixed points. In the second case, on the other hand, the involution is traced out on the base by the arms of a right angle which revolves about its vertex. For since (Fig. 84) the segments A A' and BB' overlap, the circles described onAA' and BB f respectively as diameters Fi g> 84. will intersect in two points G and H which lie symmetrically with regard to the base ; GH being perpendicular to the base, which bisects it at } the centre of the involution. It follows that Fig. 83. e. when the segments AA r * An involution of the kind hich has double points is often called a hyperbolic involution ; one of the kind which has no double points being called an elliptic involution. 106 INVOLUTION. [129 OG 2 = 0H 2 = AO .OA' = BO .OB', and that all other circles passing through G and H and cutting the base in the other pairs CC\ BB\... of the involution will have their centres also on the base, and will have CC\ DD\... as diameters. If then we project any of the segments AA\ BB\ CC\... from G (or H ) as centre, we shall obtain in each case a right angle AGA', BGB', CGC\ ... (or AEA\ BEB\ CHC ,...). We conclude that when an involution of points AA ', BB',... has no double points, i. e. when the rectangle OA.OA f is equal to a negative constant — & 2 , each of the segments AA\BB\... subtends a right angle at every point on the circumference of a circle of radius k, whose centre is at and whose plane is perpendicular to the base of the involution. This last proposition is a particular case of that of Art. 109 (4). If then an angle of constant magnitude revolve in its plane about its vertex, its arms will determine on a fixed transversal two projective ranges, which are in involution in the case where the angle is a right angle. 129. Consider an involution of parallel rays ; these meet in a point at infinity, and the straight line at infinity is a ray of the involution. The ray conjugate to it contains the centre of the involution of points which would be obtained by cutting the pencil by-any transversal ; it may therefore be called the central ray of the given involution. If, reciprocally/ we project an involution of points by means of parallel rays, these rays will form a new involution, whose central ray passes through the centre of the given involution. When one involution is derived from another involution by fteans of projections or sections (Art. 124), the double elements of the first always give rise to the double elements ofj;he setqpd. 130. Since in an involution any groupof elements is projective with the group of conjugate elements, it follows that if any four points of the involution be taken, theirmnharmonic ratio will be equal to that of their four conjugates. In the involution A A ', BB' , CC',. . . the groups of points ABA'C and A 'B'AC, for example, will be projective ; therefore 4 AA' AC^_A^A A'C BA ,: BC '~ B / A : B 7 C > whence ^ AB\ BC. CA' + A'B.B'C. C'A = o. Conversely, if this relation hold among the segments determined by sixcollinear points AA'BB'CC^ these will be three conjugate pairs of 131] INVOLUTION. 107 an involution. For the given relation shows that the anharmonic ratios (ABA'C) and (A'B'AC) are equal to one another; the groups ABA'C and A'B'AC are therefore projective. But A and A' corre- spond doubly to each other ; therefore (Art. 122) AA / ', BB / \ GC are three conjugate pairs of an involution. 181. Theoeem. The three pairs of opposite sides of a complete quadrangle are cut by any trans- versal in three pairs of conjugate points of an involution *. Let QRST (Fig. 85) be a complete quadrangle, of which the pairs of opposite sides RT and QS y ST and QR, QT and RS are cut by any transversal in A and A', B and B\ C and C respec- Fig. 85. tively. If P is the point of intersection of QS and RT, then ATPR is a projection of ACA'B* from Q as centre, and ATPR is also a projection of ABA'C from S as centre ; therefore the- group 4C4'£'isprojective vnth.ABA'C\ and therefore (Art. 45) with A' CAB. And since A and A' correspond doubly to one another in the projective groups ACA'B' * DES ARGUES, Correlative Theorem. The straight lines which connect any point with the three pairs of oppo- site vertices of a complete quadri- lateral are three pairs of conjugate rays of an involution. Let qrst (Fig. 86) be a com- plete quadrilateral, of which the pairs of opposite vertices rt and qs } st and qr, qt and rs are projected from any centre by the rays a and a', b and b', c and c' respectively. Fig. 86. Let p be the straight line which joins the points qs and rt. The pencils atpr and aca'b' are in per- spective (their corresponding rays intersect in pairs on q) ; similarly atpr and aba'c' are in perspec- tive (their corresponding rays intersect in pairs on s). The pencil atpr is therefore of course projective with each of the pencils aca'b' and aba'c* ', and loc cit. p. 171. 108 INVOLUTION. [132 and A' CAB, it follows (Art. 122) therefore aca'b' is projective with that AA',BB', CC are three con- aba'c or (Art. 45) with aYab. jugate pairs of an involution. And since a and a? correspond doubly to one another in the pencils aca'b' and a' cab, it follows (Art. 122) that aa', bb', cc' are three pairs of conjugate rays of an involution. The theorem just proved may The theorem just proved may also be stated in the following also be stated in the following form : form : If a complete quadrangle move If a complete quadrilateral in such a way that jive of its sides move in such a way that Jive of its pass each through one of five fixed vertices slide each on one of five collinear points, then its sixth fixed concurrent straight lines, then side will also /;ass through a fixed its sixth vertex will also move on a point collinear with the other five, fixed straight line, concurrent with and forming an involution with the other five, and forming an in- them. volution with them. 132. By combining the preceding theorem (left) with that of Art. 130, we see that If a transversal be cut by the three pairs of opposite sides of a com- plete quadrangle in A and A', B and B' , C and C respectively, these determine upon it segments which are connected by the relation AB\ BC. CA' + A'B.B'C. C'A = o *. 133. In the theorem of Art. 131 (right) let U and V, Pand 7\ W and W denote respectively the opposite vertices rt and qs, st and qr, qt and rs of the quadrilateral qrst, and let AA ', BB f , CC denote respectively the points of intersection of the rays aa' , bb' ', cc' with an * arbitrary transversal. With the help of Art. 124 the following proposition may be enunciated : If the three pairs UTJ', VV, WW of opposite vertices of a complete quadrilateral be projected from any centre upon any straight line, the six points AA', BB', CC so obtained will form an involution. Suppose now, as a particular case of this, that the centre of pro- jection G is taken at one of the two points of inter section of the circles described on UU', W respectively as diameters. Then AG A' and BGB' are right angles, and therefore also (Art. 128) CGC is a right angle ; therefore the circle on WW as diameter will also pass through G. Hence the three circles which have for diameters the three diagonals of a complete quadrilateral pass all through the same two * Pappus, loc. cit., book VII. prop. 130. 135] INVOLUTION. 109 points; that is, they have the same radical axis. The centres of these circles lie in a straight line ; hence The middle points of the three diagonals of a complete quadrilateral are collinear *. The proposition of Art. 131 (right) leads immediately to the Construction for the sixth ray c' of an involution of which jive rays a, a', b,b',c are given. For take on c (Fig. 86) an arbi- trary point, through which draw wnicn taKe any two points v ana any two straight lines q and t, T, and join AT, BT, A'Q, B'Q ; and join the point ta to qb' , and the point tb to qa ; if the joining lines be called r , s respectively, then the straight line connecting the centre of the pencil with the yS point rs is the required ray c'. If, in the preceding problem (left), the point C lies at infinity, its conjugate is the centre of the involution. In order then to find the centre of an involution of i hich two pairs AA', BB f of conjugate points are given, we construct (Fig. 87) a complete quadrangle QSTR of which one pair of opposite sides pass respectively through A and A ', another such pair through B and B f , and which has a fifth side parallel to the base ; the sixth side will then pass through the centre 0. 134. The proposition of Art. 131 (left) leads immediately to the Construction for the sixth point C r of an involution of which jive points A,A',B,B', C are given. For draw through C (Fig. 85) an arbitrary straight line, on [hich take any two points Q and \ and join AT, BT, A'Q, B'Q; if AT, B'Q meet in R, and BT, A'Q in S, the straight line RS will cut the base of the involu- tion in the required point C Fig. 87. The sixth point C which, together with five given points AA'BB'C, forms an involution, is completely deter- mined by the construction ; there is only one point C f which possesses the property on which the construction depends (Art. 127). This may be otherwise seen by regarding C f as given by the equation {AA f BC) = (A'AB'C) between anharmonic ratios; for it is known (Art. 65) that there is only one point C f which satisfies this equation. 135. The theorem converse to that of Art. 131 is the fol- lowing : If a transversal cut the sides of a triangle RSQ (Fig. 85) in three points A', B', C r which, when taken together with three other points A , B , C lying on the same transversal, form three conjugate * Chasles, loc, cit., Arts. 344, 345. Gauss, Collected Works, vol. iv. p. 391. 110 INVOLUTION. [136 pairs of an involution, then the three straight lines RA , SB , QCmeet in the same point. To prove it, let RA , SB meet in T, and let TQ meet the transversal in C v Applying the theorem of Art. 131 (left) to the quadrangle QRST, we have (AA'BCJ = (A'AB'C). But by hypothesis (AA'BC) = (A'AB'C); .-. (AA'BC 1 ) = (AA'BC); consequently (Art. 54) C 1 coincides with C, i.e. QC passes through T. The correlative theorem is : If a point S he joined to the vertices of a triangle rsa (Fig. 86) by three rays a', V, c' which, when taken together with three other rays a,b ,c passing also through S } form three conjugate pairs of an involution, then the points ra , qb , sc lie on the same straight line t. 136. Take again the figure of the complete quadrangle QRST whose three pairs of opposite sides are cut by a trans- versal in A and A', B and B', C and C. Let (Fig. 88) SQ and RT meet in R', QR and ST in S', RS and QT in Q'. Fig. 88. Consider the triangle RSQ ; on each of its sides we have a group of four points, viz. SQR'A', QRS'B', RSQ'C. The projections of these from T on the transversal are BCAA', CABB', ABCC'. The product of the anharmonic ratios of these last thiee groups is ( BA ■ BA\ ,CB CB\ f AC § AC\ { CA : CA') \AB '' AB') ^BC '' BC'J ' 138] INVOLUTION. Ill CA'.AB'.BC 0r BA'. CB'. AC ' which (Art. 130) is equal to — i. Therefore : If any transversal meet the sides of a triangle, and if moreover from as centre each vertex he projected upon the side opposite to it, is of four points this obtained on each of the sides of the triangle will he such that the product of their anharmonic ratios is equal to — i. Conversely, if three pairs of points R'A', S'B', Q'C be taken, one on each of the sides of a triangle RSQ, such that the product of the anharmo?iicratios(SQR'A'),(QRS'B'), (RSQ' C) is equal to —1 ; then, if the straight lines RR', SS', QQ' are concurrent, the points A', B', C will be collinear; and conversely, if the points A\B\C are collinear, the straight lines RR f , SS', QQ' will be concurrent. 137. Suppose now the transversal to lie altogether at infinity; then the anharmonic ratios (SQR'A'), QRS'B'), and (RSQ'C) become (Art. 64) respectively equal to SR' : QR', QS'iRS', and RQ'\ SQ' ; so that the preceding proposition re- duces to the following * : If the straight lines connecting the three vertices of a triangle RSQ with any given point T meet the respectively opposite sides in R f , S', Q', the segments which they determine on the sides will be connected by the relation SR/.QS'.RQ' _ _ QR'.RS'.SQ' ~ I; and conversely : If on the sides SQ , QR , RS respectively of a triangle RSQ points R\ S', Q f be taken such that the above relation holds, then will the straight lines RR f , SS', QQ f meet in one point T. 138. Kepeating this last theorem for two points T' and T", we obtain the following : If the two sets of three straight lines which connect the vertices of a triangle RSQ with any two given points T f and T ff meet the respectively opposite sides in R', S', Q f and R'\ S", Q", then will the product of the anharmonic ratios (SQR'R"), (QRS'S"), and {RSQ' Q") be equal to + i. [For each of the expressions SR'.QS'.RQ' SR". QS" . RQ" QR'.RS'.SQ' ' QR".RS".SQ" * Ceva's theorem. See his book, Be lineis rectis se invicem secantibus statica construct™ (Mediolani, 1678), i. 2. Cf. M6BIUS, Baryc. Calc. § 198. 112 INVOLUTION. [139 is equal to — i ; and the required result follows on dividing one of them by the other.] 139. Considering again the triangle QBS (Fig. 88), and taking the transversal to be entirely arbitrary, let ST , QT be taken so as to be parallel to QB , BS respectively. Then the figure QBST becomes a parallelogram; the points S' and Q' pass to infinity, and B' (being the point of intersection of the diagonals QS , BT) becomes the middle point of SQ. Conse- quently (Art. 64) the anharmonic ratios {SQB'A'), {QBS'B'), {BSQ 'C') become equal respectively to - ( QA ' : SA '), {BB' :QB'), and (SC : RC). Thus*: If a transversal cut the sides of a triangle RSQ in A', B', C respectively, it determines upon them segments which are connected by the relation QA'.BB'.SC _ SA'.QB'.BC'~ I; and conversely: If on the sides SQ , QB , BS respectively of a triangle A', B\ C be taken such that the above relation holds, then these three points be collinear. 140. Repeating the last theorem of the preceding Article for two transversals, we obtain the following : If the sides of a triangle BSQ are cut by two transversals in A', B\ C and in A'\ B", C" respectively, the product of the anharmonic ratios (SQA'A"), (QBB'B'% and (BSC'C") will be equal to + 1. [For each of the expressions QA'. BB'. 8C QA"; RB" . SC" SA'.QB'.BC ' SA". QB". BC" is equal to i ; dividing one by the other, the required result follows.] Reciprocally, if on the sides of a triangle BSQ three pairs of points A' A", B'B", C'C" be taken such that the product of the anharmonic ratios {SQA'A"), (QBB'B"), {BSC'C") may be equal to + i ; then, if the points A', B' , C are collinear, the points A", B", C" will also be collinear, and if the lines BA' , SB', QC are concurrent , the lines BA", SB", QC" will also be concurrent. 141. It has been shown (Art. 122) that if two projective ranges * Theorem of Menelaus ; Sphaerica, iii. I. Cf. Mobius, loc. cit. 142] INVOLUTION. 113 (ABC.) and (A'B'C '...), lying in the same plane, are projected from the point of intersection of a pair of lines such as AB' and A'B, AC and A'C,...ov BC / and B'C., the projecting rays form an involution. The theorems correlative to this are as follows : Given two projective", but not concentric, flat pencils (abc.) and (a'b'c' ...) lying in the same plane ; if they be cut by the straight line which joins a pair of points such as ab' and afb, ac' and a'c,... or he' and b'c., the points so obtained form an involution. Given two projective axial pencils (a/3y...) and (a'Ay* ...) whose axes meet one another ; if they be cut by the plane which is deter- mined by passing through a pair of lines such as a/3' and a'/3, ay and ay,... or fry and /3'y..., the rays so obtained form an involution. Given two projective flat pencils (abc.) and (a'b'c' ...) which are concentric, but lie in different planes ; if they be projected from the point of intersection of a pair of planes such as ab' and a'b, ac and a'c,.. or be' and Kg..., the projecting planes form an involution. 142. Particular Cases. All points of a straight line which lie in pairs at equal distances on opposite sides of a fixed point on the line, form an involution, since every pair is divided harmonically by the fixed point and the point at infinity. Conversely, if the point at infinity is one of the double points of an involution of points, then the other double point bisects the distance between any point and its conjugate. If in such an involution the segments A A' ', BB f formed by any two pairs of conjugate points have a common middle point, then will this point bisect also the segment CC formed by any other pair of conjugates. All rectilineal angles which have a common vertex, lie in the same plane, and have the same fixed straight line as a bisector, form an in- volution, since the arms of every angle are harmonically conjugate with regard to the common bisector and the ray perpendicular to it through the common vertex. Conversely, if the double rays of a pencil in involution include a right angle, then any ray and its conjugate make equal angles with either of the double rays. If in such an involution the angles included by two pairs of conjugate rays aa' and bb f have common bisectors, these will be the bisectors also of the angle included by any other pair of conjugate rays cc' '. All dihedral angles which have a common edge and which have the same fixed plane as a bisector, form an involution ; for the faces of every angle are harmonically conjugate with regard to the fixed plane and the plane drawn perpendicular to it through the common edge. Conversely, if the double planes of an axial pencil in involution are at right angles to one another, then any plane and its conjugate make equal angles with either of the double planes. I CHAPTEE XIII. PROJECTIVE FORMS IN RELATION TO THE CIRCLE. *. 143. Consider (Fig. 89) two directly equal pencils abed... and a'b'c'd'... in a plane, having their centres at and 0' respectively. The angle contained by a pair of corresponding rays aa', bb\ cc', ... is constant (Art. 106) ; the locus of the inter- section of pairs of corresponding rays is therefore (Euc. III. 21) a circle passing • through and 0'. The tangent to this circle at makes with 00' an angle equal to any of the angles OAO\ OBO', 0C0\ &c. ; but this is just the angle which O'O considered as a ray of the second pencil should make with the ray Yi g . 89. corresponding to it in the first pencil ; therefore to O'O or q' considered as a ray of the second pencil corresponds in the first pencil the tangent a to the circle at 0. Imagine the circumference of the circle to be described by a moving point A ; the rays AO , AO' or a, a' will trace out the two pencils. As A approaches 0, the ray AO' will approach 00' or q' and the ray AO will approach q\ and in the limit when A is indefinitely near to 0, the ray AO will coincide with q or the tangent at 0. This agrees with the definition of the tangent at 0, as the straight line which joins two indefinitely near points of the circumference. Similarly, to the ray 00' or j? considered as belonging to the first pencil corresponds the ray p of the second pencil, the tangent to the circle at 0'. 144. Conversely, if any number of points A , B, C, D, . . . on a circle be joined to two points and 0' lying on the same 146] PROJECTIVE FORMS IN RELATION TO THE CIRCLE. 115 circle, the pencils (A, B, C, £,...) and 0' (A, B, C, D,...)so formed will be directly equal, since the angle AOB is equal to AO'B, AOC to AO'C,... BOC to BO'C, &c. But two equal pencils are always projective with one another (Art. 104). If then the points A, B, C, ... remain fixed, while the centre of the pencil moves and assumes different positions on the cir- cumference of the circle, the pencils so formed are all equal to one another, and consequently all projective with one another. The tangent at is by definition the straight line which joins to the point indefinitely near to it on the circle. It follows that in the projective pencils (A,B, C, ...)and 0' (A, B, C,...), the ray of the first which corresponds to the ray O'O of the second is the tangent at 0. 145. It has been seen (Art. 73) that in two projective forms four harmonic elements of the one correspond to four harmonic elements of the other. If then the four rays & (A, B, C, D) form a harmonic pencil, the same is the case with regard to the four rays 0' (A, B, C, D), whatever be the position of the point 0' on the circle. By taking 0' indefinitely near to A, we see that the pencil composed of the tangent at A and the chords AB, AC, AD will also be harmonic; so again the pencil composed of the chord BA, the tangent at B, and the chords BC, BD will be harmonic, &c. When this is the case, the four points A,B,C,I)ofthe circle are said to he harmonic *. 146. The tangents to a circle determine upon any pair of fixed tangents ttvo ranges which are projective with one another. Let M (Fig. 90) be the centre of the circle, PQ and P'Q' a pair of fixed tangents, and A A' a variable tangent. The part AA ' of the variable tangent intercepted between the fixed tangents subtends a constant angle at 31; for if Q, P', T are the points of contact of the tangents respectively, Fig. 90. * Steiner, loc. cit., p. 57. § 43. Collected Works, vol. i. p. 345. I 2 116 PROJECTIVE FORMS IN RELATION TO THE CIRCLE. [147 angle AMA' = AMT + TMA' = \QMT+\TMP' = hQMP'*. Accordingly, as the tangent AA' moves, the rays MA, MA' will generate two projective pencils (Art. 108), and the points A, A' will trace out two projective ranges. Since the angle AM A' is equal to the half of QMP f , it is equal to either of the angles QMQ', PMP f (denoting by P and Q ' the same point, according as it is regarded as belonging to the first or to the second tangent). Consequently Q and Q', P and P' are pairs of corresponding points of the two pro- jective ranges; i.e. the points of contact of the two fixed tangents correspond respectively to the point of intersection of the tangents. Imagine the circle to be generated, as an envelope, by the motion of the variable tangent ; the points A, A' will trace out the two projective ranges. As the variable tangent approaches the position PQ, the point A f approaches Q' , and A ap- proaches the point which corresponds to Q', viz. Q ; and in the limit when the variable tangent is indefinitely near to PQ, the point A will be indefinitely near to Q or the point of contact of the tangent PQ. The point of contact of a tangent must therefore be regarded as the point of intersection of the tangent with an indefinitely near tangent. 147. The preceding proposition shows that four tangents a, b, o, d to a circle are cut by a fifth in four points A, B, C, D whose anharmonic ratio is constant whatever be the position of the fifth tangent. This tangent may be taken indefinitely near to one of the four fixed tangents, to a for example ; in this case A will be the point of contact of a, and B,C ,D the points of intersection ab, ac, ad respectively. As a particular case, if a , b, c, d meet the tangent PQ in four harmonic points, they will meet every tangent in four har- monic points. The group constituted by the point of contact of a and the points of intersection ab, ac, ad will also be har- monic. In this case, the four tangents a, b, c, d are said to be harmonic j\ * Poncelet, Propr. proj., Art. 462. t Steiner, loc. cit., p. 157. § 43. Collected Works, vol. i. p. 345. 148] PROJECTIVE FORMS IN RELATION TO THE CIRCLE. 117 148. The range determined upon any given tangent to a circle by any number of fixed tangents is projective with the pencil formed ly joining their points of contact to any arbitrary point on the circle. Let A, B, C,...X (Fig. 91) be points on the circle, and a, b, c, ...cc the tangents at these points respectively. If the points A f y B\ C", ... in which the tangent x is cut by the tangents a,b,c,... be joined to the centre of the circle, the joining lines will be perpendicular re- spectively to the chords XA,XB, XC,... and will Fig. 91. therefore (Art. 108) form a pencil equal to the pencil X (A , B, C, ...). The range A'B'C ..., is therefore projective with the pencil X (A, B, C,...). Corollary. If four points on a circle are harmonic , then the tangents also at these points are harmonic ; and conversely. For if, in what precedes, X (ABCIJ) is a harmonic pencil, A'B'C'B' will be a harmonic range; and conversely. CHAPTER XIV. PROJECTIVE FORMS IN RELATION TO THE CONIC SECTIONS. 149. Let the figures be constructed which are homological with those of Arts. 144, 146, 148. To the points and tangents of the circle will correspond the points and tangents of a conic section (Art. 23). A tangent to a conic is therefore a straight line which meets the curve in two points which are inde- finitely near to one another; a point on the curve is the point of intersection of two tangents which are indefinitely near to one another. To two equal and therefore projective pencils will correspond two projective pencils, and to two projective ranges will correspond two projective ranges ; for two pencils or ranges which correspond to one another in two homological figures are in perspective. We deduce therefore the following propositions : (l). If any numher of points A, B, 0, D, ... on a conic are joined to two fixed points and 0' lying on the same conic (Fig. 92), the (A,B, C, If,...) and 0' (A, £,0,1),...) so formed are projective with one another. To the ray 00' of the first pencil corresponds the tangent at 0', and to the ray O'O of the » second pencil corresponds the tangent at 0. (2). Any number of tangents a,b,c,d, ...to a conic determine on a pair of fixed tangents and 0' (Fig. 93) two projective ranges. To the point 00' or Q of the first range corresponds the point of contact Q of 0', and to the same point o'o or I" of the second range corre- Fig. 92. sponds the point of contact P of 0*. * Steiner, he. cit., p. 139. § 38. Collected Works, vol. i. pp. 332, 333. 150] PROJECTIVE FORMS IN RELATION TO CONIC SECTIONS. 119 (3). The range which a variable tangent to a conic determines upon a fixed tangent is projective with the pencil formed by joining the S •93- point of contact of the variable tangent to any fixed point of the conic. (Fig. 94.) 150. We proceed now to the theorems converse to those of Art. 149. The proofs here given are due to M. Ed. Dewulf. I. If two {non-concentric) pencils lying in the same plane are pro- jective with one another {but not in perspective), the locus of the points of intersection of pairs of corresponding rays is a conic passing through the centres of the two pencils ; and the tangents to the locus at these points are the rays which correspond in the two pencils respectively to the straight line which joins the two centres. Let and A (Fig. 95) be the respective centres of the two pencils, and let 0M 1 and AM 1 , 0M 2 and AM 2 , 03I 3 and A M z ,... be pairs of corresponding rays. The locus of the points M x } M 2 ,3I 3i ... will pass through 0, since this point is the inter- section of the ray A of the pencil A with the corresponding ray of the pencil 0. Similarly A will be a point on the locus. Fig. 94. 120 PROJECTIVE FORMS IN RELATION [150 Let o be that ray of the pencil which corresponds to the ray AO of the pencil A. Describe a circle touching o at 0, Fig- 95- and let this circle cut OA in A\ and 0M 1} 0M 2 , 0M 3 , ... in the points 1//, J/"/, Jf 8 ',... respectively. The pencils (Jf/ Jf/ M 3 f ... ) and A' (M{ M 2 ' M 3 f ...) are directly equal to one another; and since by hypothesis the pencil (if/ M 2 if/ ... ) or (M x M 2 M 3 ... ) is projective with the pencil A (M 1 M 2 M 3 ...), therefore the pencils A' (1// M 2 ' M z f ... ) and A (M x M 2 M 3 ... ) are projective. But they are in perspective, since the ray A'O in the one corresponds to the ray AO in the other (Artr 80) ; therefore pairs of corresponding rays will intersect in points S^, S 2i # 3 , ... lying on a straight line s. In order, then, to find that point of the locus which lies on any given ray m of the pencil A, it is only necessary to produce m to meet s in S, to join SA' cutting the circle in M\ and to join 0M f ; this last line will cut m in the required point M. But this construction is pre- cisely the same as that employed in Art. 23 (Fig. n) in order to draw the curve homological with a circle, having given the axis s and centre of homology, and a pair of corresponding points A and A'. The locus of the points M is therefore a conic section. II. If two (non-cottinear) ranges lying in the same plane are pro- jective with one another {but not in perspective), the envelope of the straight lines joining pairs of corresponding points is a conic, i.e. the straight lines all touch a conic. This conic touches the bases of the 150] TO THE CONIC SECTIONS. 121 two ranges at the points which correspond in these respectively to the point of intersection of their bases. Let s and / (Fig. 96) be the bases of the two ranges, and let A and A\ B and B\C and C\ , .. be pairs of corresponding Fig. 96. points. The curve enveloped by the straight lines AA\ BB\ CC, ... will touch s, since this is the straight line joining the point ss r or S' of the second range with the corresponding point S of the first. Similarly, / will be a tangent to the envelope. Describe a circle touching s at S. and draw to it tangents a , 9" from the points A,B,C, ... S' respectively. The tangents a'\ b'\ c", . . . will determine on /' a range which is projective with s and therefore also with /. But the point S' corresponds to itself in the two ranges t? and /'; these are therefore in perspective (Art. 80), and the straight lines A!' A\ B" B\ C" C, ... will meet in one point 0. In order then to draw a tangent to the envelope from any given point M lying on the line s, it is only necessary to draw from M a tangent m to the circle, meeting 9" in M ", and to join OM " ; this last line will cut / in that point M f of the range *' which corresponds to the point M of the range s, and MM' will be the required 122 PROJECTIVE FORMS IN RELATION [151 tangent to the envelope. But this construction is precisely the same as that made use of in Art. 23 (Fig. 12) in order to draw the curve homological with a circle, taking a given tan- gent to the circle as axis of homology, any given point as centre of homology, and /, /' as a pair of corresponding straight lines. The envelope of the lines MM' is therefore a conic section. The theorems (I) and (II) of the present Article are correlative (Art. 33), since the figure formed by the points of intersection of corresponding rays of two projective pencils is correlative to that formed by the straight lines joining corresponding points of two projective ranges. Thus in two figures which are correlative to one another [according to the law of duality in a plane), to points lying on a conic in one correspond tangents to a conic in the other. 151. Having regard to Arts. 73 and 79, the propositions of Arts. 149, 150 may be enunciated as follows : The anharmonic ratio of the four straight lines which connect four fixed points on a conic with a variable point on the same is constant. The a?iharmonic ratio of the four points in which four fixed tan- gents to a conic are cut by a variable tangent to the same is constant *. The anharmonic ratio of four joints A, B, C, D lying on a conic is the anharmonic ratio of the pencil (A, B , C , D) formed by joining them to any point on the conic. The anharmonic ratio of four tangents a, b, c, d to a conic is that of the four points (a, b, c, d), where is an arbitrary tangent to the conic. If this anharmonic ratio is equal to — 1, the group of four points or tangents is termed harmonic. The anharmonic ratio of four tangents to a conic is equal to that of their points of contact \. Consequently the tangents at four harmonic points are harmonic, and vice versa. The locus of a point such that the rays joining it to four given points ABC I) form a pencil having a given anharmonic ratio is a conic passi?ig through the given points. * Steiner, he. cit., p. 156. § 43. Collected Works, vol. i. p. 344. t Chasles, Geometrie Supeneure, Art. 663. 152] TO THE CONIC SECTIONS. 123 The tangent to the locus at one of these points, at A for example, is the straight line which forms with AB, AC, AD a pencil whose anharmonic ratio is equal to the given one. The curve enveloped by a straight line which is cut by four give?i straight lines in four points whose anharmonic ratio is given is a conic touching the given straight lines. The point of contact of one, of these straight lines, a for example, forms with the points ah, ac, ad a range whose anhar- monic ratio is equal to the given one *. 152. Through five given ])oints 0, 0' ,A,B ,C in a plane (Fig. 92), no three of which lie in a straight line, a conic can be described. For we have only to construct the two projective pencils which have their centres at two of the given points, and 0' for example, and in which three pairs of corre- sponding rays OA and O'A, OB and O'B, OC and O'C intersect n the three other points. Any other pair OB and O'D of corre- sponding rays will give a new point D of the curve. To construct the tangent at any one of the given points, at for example, we have only to deter- mine that ray of the pencil which corresponds to the ray O'O of the pencil '. Through five given points only one conic can be drawn ; for if there could be two such, they would have an infinite number of other points in common (the intersections of all the pairs of corresponding rays of the projec- tive pencils) ; which is impossible. Given five straight lines o, o', a, b, c in a plane (Fig. 93), no three of which meet in a point, a conic can be described to touch them. For we have only to con- struct the two projective ranges which are determined upon two of the given lines, o and 0' for ex- ample, by the three others a, b,c, and of which three pairs of cor- responding points oa and o'a, ob and o'b, oc and o'c are given. The straight lin« d which joins any other pair of corresponding points of the two ranges will be a new tangent to the curve. To construct the point of con- tact of any one of the given straight lines, that of for ex- ample, we have only to determine that point of the range which corresponds to the point o'o of the range o f . Only one conic can be drawn to touch five given straight lines ; for if there could be two such, they would have an infinite num- ber of common tangents (all the straight lines which join pairs of corresponding points of the pro- jective ranges) ; which is im- * Steiner, loc. cit. f pp. 156, 157, §43. Collected Works, vol. i. pp. 344, 345. 124 PROJECTIVE FORMS IN RELATION [153 From this we see also that : Through four given points can be drawn an infinite number of conies ; and two such conies have no common points beyond these four. There can be drawn an infinite number of conies to touch four given straight lines ; and two such conies have no common tangents beyond these four. 153. The theorems of Art. 88 may now be enunciated in the following manner : If a hexagon is circumscribed to a conic (Figs. 97 and 61), tlie If a hexagon is inscribed in a conic (Figs. 98 and 6o) 5 the three Fig. 97. straight lines which join the three pairs of opposite vertices are con- current. This is known as Bkianchon's theorem *. Fig. 98. pairs of opposite sides intersect one another in three collinear points. This is known as Pascal's theorem f. ' 154. Pascal's theorem has reference to six points of a conic, Brianchon's theorem to six tangents ; these six points or tan- gents may be chosen arbitrarily from among all the points on the curve and all the tangents to it. Now a conic is deter- mined by five points or five tangents; in other words, five points or five tangents may be chosen at will from among all the points or lines of the plane, but as soon as these five * This theorem was published for the first time by Brianchon in 1806, and afterwards repeated in his Memoir e sur les lignes da second ordre (Paris, 181 7: P- 34)- f This theorem was given in Pascal's E. second edition, Paris, 1875). If B lies at infinity, the theorem becomes identical with lemma 20. book i. of Newton's Principia. t Apollonii Pekgaei Conicorum lib. iii. 41. 128 PROJECTIVE FORMS IN RELATION [158 which the various tangents to the parabola meet the two fixed tangents (Fig. ioo), and let P and Q' be the respective points of contact of the latter. The point of intersection of Fig. ioo. the two fixed tangents will be denoted by Q or P' according as it is regarded as a point of the first or of the second tan- gent. We have then AB AC_ BG_ AP_ AQ^ PQ^ TB f ~ A'C' B'C ~ A'P' ~ A'Q' " ■" ~ P'Q f ' (4). Conversely, given two straight lines in a plane, on which lie two similar ranges (which are not in perspective), the straight lines connecting pairs of corresponding points will envelope a parabola which touches the given straight lines at t/ie points which correspond in the two ranges respectively to their point of intersection. For the points at infinity on the given straight lines being corresponding points (Art. 99), the straight line which joins them will be a tangent to the envelope ; thus the envelope is a conic (Art. 150 (II)) which has the line at infinity for a tangent, i.e. it is a parabola. 158. In theorem I of Art. 150 (Fig. 95) suppose that the point A lies at infinity, or, in other words, that the pencil A consists of parallel rays. To the straight line OA, considered as a ray a' of the pencil (viz. that ray which is parallel to the rays of the other pencil), corresponds that ray a of the pencil A which is the tangent at the point A. This ray a may be at a finite, or it may be at an infinite distance. In the first case (Fig. 101) the straight line at infinity is a ray/ of the pencil A, and to it corresponds in the pencil a ray/' different from a' and consequently not passing through 159] TO THE CONIC SECTIONS. 129 A ; the conic will therefore be a hyperbola (Art. 23) having A(z=zaa!) &ndjj' for its points at infinity; the straight line a is one asymptote and/ 7 is parallel to the other. Fig. ioi. Fig-. 102. In the second case (Fig. 102) the line at infinity is the tangent at A to the conic, which is therefore a parabola. 159. If in this same theorem of Art. 150 the points A and are supposed both to lie at infinity (Fig. 103), the two pro- jective pencils will each consist of parallel rays ; and since the conic which these pencils generate must pass through A and it is a hyperbola (Art. 23). The asymptotes of the hyperbola are the tangents to the curve at its infinitely distant points*; they will therefore be the rays a and oint of given i)oint H^ lying on one> of intersection of the curve with any these tangents a, another tangent given straight liner drawn through' to the curve (Fig. 104). one of these points A (Fig. 105). Fig. 104. It c / be the required tangent, ^ab'ca'bc' is a hexagon to which Brianchon's theorem applies. Let r be the diagonal connecting one pair ay and a f b of opposite ver- tices, and let q be the diagonal connecting another such pair ca' and c / a (where ca is the given pointZT) ; then the diagonal which connects the remaining pair hd and b'c must pass through the point qr . If then p be the straight Fig. 105. If C be the required point, AB'CA'BC is a hexagon to which Pascal's theorem applies. Let R be the point of intersection of one pair ^LS' and A'B of oppo- site sides, ancTlet ♦ be the point of intersection of another such pair CA' and r\ then QR must pass through the point of inter- section of the remaining pair BC and B'C. If then PB be joined, .it will cut the given K % 132 CONSTRUCTIONS AND EXERCISES. [162 line joining the points qr and Vc, the straight line which joins plane of a conic two tangents at most can be drawn to the curve (Art. 23) ; so that from a point lying on a given tangent only one other tangent can be drawn. If then the conic is a parabola, it cannot have a pair of parallel tangents. (This has already been seen in Art. 157 (l).) II. Suppose the straight line b to lie at infinity; the problem is then: 163] CONSTRUCTIONS AND EXERCISES. 135 Fig. 112. Given four tangents a,b',c,a' to a parabola, to draw from a given point H lying on one of them, a, another tangent to the curve (Fig. 1 1 %). ■ Solution. Through the point ah' draw the straight line r parallel to a! '; join the points H and a'c by the straight line q, and the points qr and Vc by the straight line p. The straight line drawn through H parallel to p will be the required tan- gent. III. If the straight line a lies at infinity, we have the problem : Given four tangents b', c,a',b to a parabola, to draw the tangent which is parallel to a given straight line (Fig. 113). Solution. Through a'b draw .the straight line r parallel to V, and through a'c draw the straight line q parallel to the given direction ; join the points qr, b'c by the straight line p. The straight line through pb parallel to the. given direction is the tangent required. IV. If in problem II the point H assume different positions on a, or if in III the given straight line assume different directions, we arrive at the solution of the problem : To construct by means of its tangents the parabola which is deter- mined by four given tangents, u^ Fig. 113 CHAPTEE XVI. DEDUCTIONS FROM THE THEOREMS OF PASCAL AND BRIANCHON. 164. We have already given some propositions and con- structions (Arts. 161-163) which follow immediately from the theorems of Pascal and Brianchon, by supposing some of the elements to pass to infinity. Other corollaries may be deduced by assuming two of the six points or six tangents to approach indefinitely near to one another *. If AB'C A' ' BC are six points on a conic, Pascal's theorem asserts that the pencils A(A'B'CC) and B(A'B'CC% for example, are projective with one another. To the ray AB of the first pencil corresponds in the second the tangent at B, so that we may say that the group of four lines AA\ AB\ AC , AB is projective with the group BA\ BB\ BC, tangent at B. But this amounts evidently to saying that the point C\ which was at first taken to have any arbitrary position on the curve, has come to be indefinitely near to the point B. Instead then of the inscribed hexagon we have now the figure made up of the inscribed pentagon AB f CA r B and the tan- gent h at the vertex B (Fig. 114); and Pascal's theorem be- comes the following : 114. . If a pentagon is inscribed in a conic, the points of i?itersection R , Q of two pairs of non-consecutive sides (AB' and A'B, AB and CA f ), and the * Caenot, loc. cit., pp. 455, 456. 165] THEOREMS OF PASCAL AND BRIANCHON j DEDUCTIONS. 137 point P where the fifth side (B'C) meets the tangent aU the opposite vertex, are collinear. { ^* This corollary may also be deduced from the construction (Art. 84, right) for two projective pencils. Three pairs of corresponding rays are here given, viz. AA' and BA', AC'und BC, AB' and BB' . We cut the two pencils by the transversals CA', CB' respectively; if R be the point of intersection of A'B and AB', then any pair of corre- sponding rays of the two pencils must cut the transversals CA', CB' respectively in two points which are collinear with R. In order then to obtain that ray of the second pencil which corresponds to AB, viz. the tangent at B, we join R to the point of intersection Q of CA' and AB, and join QR meeting CB' in P; then BP is the required ray b. But this construction agrees exactly with the corol- lary enunciated above. 165. By help of this corollary the two following problems can be solved : (1). Given five points A , B' , C , A',B of a conic, to draw the tangent at one of them B (Fig. 114). Solution. Join Q, the point of intersection of AB and CA', to R, the point of intersection of AB' and A'B ; if P is the point where QR meets B'C, then BP will be~the required tangent *. Particular cases. Given four points of a hyperbola and the direction of one asymptote, to draw the tangent at one of the given points. (This is obtained by taking one of the points A , B', C , A' to lie at infinity.) Given four points of a hyperbola and the direction of one asymptote, to draw that asymptote. (1> at infinity.) Given three points of a hyperbola and the directions of both asymptotes, to draw the tangent at one of the given points. (Two of the four points A, B', C , A' at infinity.) Given three points of a hyperbola and the directions of both asymptotes, to draw one of the asymptotes. (B and one of the other points at infinity.) - ^ u. (2). Given four points A, B , A', C of a conic and the tangent at one of them B, to construct the conic by points ; for example, to find the point of the curve which lies on a given straight line r drawn through A (Fig. 114). Solution. Let R be the point where A'B meets r, and Q the point where AB meets CA' ; and let QR cut the given tangent in P. The point B' where CPcuts the given straight line r will be the one required. By supposing one or more of the elements of the figure to lie at * Maclaurin, loc. cit., § 40. 11 138 DEDUCTIONS FKOM THE THEOREMS [166 infinity, e. g. one of the points A , A', C ; or two of these points ; or the point A and the line r ; or the point B ; or the point B and one of the other points ; or the point B and the given tangent ; we obtain the following particular cases : To construct by points a hyperbola, having given three points of the curve, the tangent at one of these points, and the direction of one asymptote ; or : two points, the tangent at one of them, and the directions of both asymptotes ; or : three points and an asymptote ; or : two points, one asymptote, and the direction of the other asymptote. Given three points of a hyperbola, the tangent at one of them, and the direction of an asymptote, to find the direction of the other asymptote. To construct by points a parabola, having given three points of the curve (lying at a finite distance) and the direction of the point at infinity on it. 166. Returning to the hexagon AB'CA'BC inscribed in a conic, let not only C be taken in- definitely near to B, but also C indefinitely near to B f . The figure will then be that of an inscribed quadrangle AB'A'B together with the tangents at B and B f (Fig. 115), and Pascal's theorem becomes the follow- ing: If a quadrangle is inscribed in a conic, the points of intersection of the two pairs of opposite sides, and the point of intersection of the tangents at a pair of opposite vertices, are three collinear This property coincides with one already obtained elsewhere (Art. 85, right). For considering the projective pencils of which BA and B'A, BA ' and B'A ', . . . are corresponding rays, it is seen that the straight line which joins the point of intersection Q of BA and B'A' to the point of intersection R of B'A and BA' must pass through the point of intersection P of the rays which correspond in the two pencils respectively to the straight line joining their centres B and B\ 167. By help of the foregoing corollary the following problems can be solved : 167] • OF PASCAL AND BRIANCHON. 139 (1). Given four points A ,B' ,A',B of a conic and the tangent BP at one of them B, to draw the tangent at another of the points B / (Fig. 115). Solution. Let AB and A'B' meet in Q, and AB' and A f B in R ; and let QR meet the given tangent in P. Then B'P will be the required tangent *. By supposing one of the given points, or the given tangent, to lie at infinity, the solutions of the following particular cases are obtained : To draw the tangent at a given point of a hyperbola, having given in addition two other points on the curve, the tangent at one of them, and the direction of one asymptote ; or, one other point, the tangent at' this, and the directions of both asymptotes ; or, one other point, one asymptote, and the direction of the other asymptote. To draw the asymptote of a hyperbola when its direction is known, having given in addition three points on the curve and the tangent at one of them ; or, two points on the curve, the tangent at one of them, and the direction of the second asymptote ; or, two points on the curve and the second asymptote. To draw the tangent at a given point of a parabola, having given two other finite points on the curve, and the direction of the point at infinity on it. (2). To construct a conic by points, having given three points A ,B, B f on the curve and the tangents BP , B'P at two of them ; i. e. to determine, for example, the point A' in which an arbitrary straight line r drawn through B is cut by the conic (Fig. 116). Solution. Join the point of intersection P of the given tangents to the point R where r cuts AB ' ; and let K PR cut AB in Q. If B 'Q be joined, it will 1 cut r in the required point A f . By supposing one of the points A , B , B' or one of the lines BP , B'P, r to lie at infinity, we shall obtain the solutions of the following particular cases : To construct by points a hyperbola, having given two points on the curve, the tangents at these, and the direction of one • asymptote ; or, one point on the curve, the tangent there, one asymptote and the direction of the second asymptote ; or, one point on the curve and both asymptotes. To construct by points a parabola^ having given two points on the * Maclaurin, he. cit., § 38. A* 140 DEDUCTIONS FROM THE THEOREMS [168 curve, the tangent at one of them, and the direction of the point at infinity on the curve. 168. The tangents at the other vertices A and A f of the quadrangle ABA'B' (Fig. 1 16) will also intersect on the straight line joining the points (AH, A'B') and (AB', A f B). Hence the theorem of Art. 166 may be enunciated in the following, its complete form : If a quadrangle is inscribed in a conic, the points of intersection of the two pairs of opposite sides, and the points of intersection of the tangents at the two pairs of opposite vertices, are four collinear points. If two opposite vertices of the quadrangle be taken to lie at infinity, this becomes the following : If on a chord of a hyperbola, as diagonal, a parallelogram be constructed so as to have its sides parallel to the asymptotes, the other diagonal will pass through the point of intersection of the asymptotes. 169. Theorem. The complete quadrilateral formed by four to a conic, and the complete quadrangle formed by their four of contact, have the same diagonal triangle. In the last two figures write C , JD , B , G in pladfe of A f ,B',H, Q respectively. In the inscribed quadrangle ABCD (Fig. 117) the point of intersection of the tangents at A and C, 170] OF PASCAL AND BRIANCHON. 141 that of the tangents at B and D, the point of intersection of the sides AD, BC, and that of the sides AB , CD all lie on one straight line EG. If the same points A,B,C,D are taken in a different order, two other inscribed quadrangles ACDB and ACBD are obtained, to each of which the theorem of Art. 168 may be applied. Taking the quadrangle ACDB, it is seen that the point of intersection of the tangents at A and D, that of the tangents at C and B, the point of intersection of the sides AB , CD, and that of the sides AC , BD all lie on one straight line FG. So too the quadrangle ACBD gives four points lying on one straight line EF; viz. the points of intersection of the tangents at A and B, of the tangents at C and D, of the sides AD , CB, and of the sides AC , BD*. The three straight lines EG , GF , FE thus obtained are the sides of the diagonal triangle EFG (Art. 36, [2] ) of the complete quadrangle whose vertices are the points A,B,C,D; and since the same straight lines contain also the points in which intersect two and two the tangents a,b, c, d at these points, they are also the diagonals of the complete quadrilateral formed by these four tangents. The theorem is therefore proved. 170. In the complete quadrilateral abed the diagonal /, whose extremities are the points ac , bd, cuts the other two diagonals g and e in E and G respectively ; these two points are therefore harmonically conjugate with regard to ac and bd (Art. 56). The correlative theorem is : The two opposite sides of the complete quadrangle ABCD which meet in i^are har- monically conjugate with regard to the straight lines which connect F with the two other diagonal points E and G (Art. 57). Summing up the preceding, we may enunciate the following proposition (Fig. 117): If at the vertices of a {simple) quadrangle ABCD, inscribed in a conic, tangents a,b,c, d be drawn, so as to form a (simple) quadrila- teral circumscribed to the conic, then this quadrilateral possesses the following properties with regard to the quadrangle: (i)the diagonals of the two pass through one point (F) and form a harmonic pencil ; (2) the points of intersection of the pairs of opposite sides of the two lie on one straight line (EG) and form a harmonic range ; (3) the * Maclauiun, loc. cit., § 50 ; Carnot, loc. cit., pp. 453, 454. 142 DEDUCTIONS FROM THE THEOREMS [171 diagonals of the quadrilateral pass through the points of intersection of the pairs of opposite sides of the quadrangle *, 171. By help of the theorem of Art. 169, when we are given four tangents a ,b ,c ,d to a conic and the point of contact A of one of them, we can at once find the points of contact of the three others ; and when we are given four points A , B, G, D on a conic and the tangent a at one of them, we can draw the tangents at the three other points t. Solution. Draw the diagonal triangle EFG of the complete quadrilateral abed ; then AG, AF, AE will cut b, c, d respec- tively in the required points of contact B,G, D. Draw the diagonal triangle EFG of the complete quadrangle ABCD; then the straight lines joining ag,af,ae to B, G, D re- spectively will be the required tangents. 172. The theorem of Art. 169 may be enunciated with re? gard to the (simple) quadrilateral formed by the four straight lines a, b, c,d; it then takes the following form, under which it is seen to be already included in the theorem of Art. 170 J : In a quadrilateral circumscribed to a conic, the straight lines which join the points of contact of the pairs of opposite sides pass through the point of intersection of the diagonals (Fig. 118). This property coincides with one already proved with regard to two projective ranges (Art. 85, left). For consider the projective ranges on a and c as bases, in which ab and cb, ad and cd, ... are corresponding points ; the straight lines which connect the pairs of points ab and cd, cb and ad respectively, must intersect on the straight line which connects the points corresponding in the two ranges respectively to ac ; but this is the straight line joining the points of contact of a and c. If the conic is a hyperbola, and we consider the quadrilateral which is formed by the asymp- totes and any pair of tangents, the foregoing theorem expresses that the diagonals of such a quadrilateral are parallel to the chord which joins the points of contact of the two tangents §. * Chasles, Sections coniques, Art. 121. t Maclaurin, loc. cit., §§ 38, 39. % Newton, loc. cit., Cor. ii. to lemma xxiv. § Apollonius, loc. cit, iii. 44. Fig. 11 175] OF PASCAL AND BEIANCHON. 143 173. The theorem of Art. 172 gives the solution of the problem : To construct a conic by tangents, having given three tangents a,b,c and the points of contact A and G of two of them; to draw, for example, through a given point H lying on a a second tangent to the curve (Fig. 118), Solution. Join the point ab to the point of intersection of AG and H(bc) ; the joining line will meet c in a point which when joined to H gives the required tangent d. If one of the points A , G or one of the given tangents be supposed to lie at infinity, the solution of the following particular cases is obtained : To construct by tangents a hyperbola, having given one asymptote, two tangents to the curve, and the point of contact of one of them ; or, both asymptotes and one tangent. To construct by tangents a parabola, having given the point at infinity on the curve, two tangents, and the point of contact of one of them ; or, two tangents and the points of contact of both. Given four tangents to a conic and the point of contact of one of them, to find the points of contact of the others. 174. If in Pascal's theorem the points A,'B', C' be taken to lie indefinitely near to A , B , C re- * spectively, the figure becomes \vy^n that of an inscribed triangle lew''/ ABC together with the tangents ^^S^^rK at its vert As (Mg. 119); and /^^M\ the theorem reduces to the / ^^y~^^\ * 11 • / /^-• the tangents at the vertices meet the V__^^ respectively opposite sides in three Fi S- ll 9- 175. This gives the solution of the problem : Given three points A, B,G of a conic and the tangents at two of them A and B, to draw the tangent at the third point G (Fig. 119). Solution. Let P, Q be the points where the given tangents at A , B cut BG, CA respectively ; if PQ cut AB in B, then GR is the tangent required. The following are particular cases : Given two points on a hyperbola, the tangents at these points, and the direction of one asymptote, to construct the asymptote itself. Given one asymptote of a hyperbola, one point on the curve, the 144 DEDUCTIONS FROM THE THEOREMS [176 tangent at this point, and the direction of the second asymptote, to construct this second asymptote. Given both asymptotes of a hyperbola and one point on the curve, to draw the tangent at this point. (From the solution of this problem, it follows that the segment determined on any tangent by the asymptotes is bisected at the point of contact). Given two points on a parabola, the direction of the point at infinity on the curve, and the tangent at one of the given points, to draw the tangent at the other given point. 176. The inscribed triangle ABC and the triangle BEF formed by the tangents (Fig. 119) possess the property that their respective sides BC and EF, CA and FD, AB and BE in- tersect in pairs in three collinear points. The triangles are therefore homological, and consequently (Art. 18) the straight lines AB ,BF, CF which connect their respective vertices pass through one point 0. Thus we have the proposition : In a triangle circumscribed to a conic, the straight lines which join the vertices to the points of contact of the respectively opposite sides are concurrent. 177. By help of this proposition the following problem can be solved : Given three tangents to a conic and the points of contact of two of them, to determine the jtoint of contact of the third. Solution. Let DEF (Fig. 119) be the triangle formed by the three tangents, and let A,B be the points of contact of EF, FB re- spectively. If AB and BE intersect in 0, then FQ will cut the tangent BE in the required point of contact C. Particular cases. Given one asymptote of a hyperbola, two tangents, and the point of contact of one of them, to determine the point of contact of the other. Given both asymptotes of a hyperbola, and one tangent, to deter- mine the point of contact of the latter. Given two tangents to a parabola and their points of contact, to determine the direction of the point at infinity on the curve. Given two tangents to a parabola, the point of contact of one of them, and the direction of the point at infinity on the curve, to deter- mine the point of contact of the other given tangent. 178. As a particular case of the theorem of Art. 176, consider a parabola and the circumscribing triangle formed by the tangents at any two points A, B, and the straight line at infinity, which is also 179] OF PASCAL AND BRIANCHON. 145 Fig. 1 20. a tangent. If the tangents at A and B meet in C (Fig. 120), the straight line joining C to the middle point D of the chord AB will be parallel to the direction in which lies the point at infinity on the curve. Again, if any point M be taken on AB, and parallels MP , MQ be drawn to BC , AC respectively to meet AC , BC in P, Q ; and if MP be drawn parallel to DC to meet PQ in R ; then PQ will be a tangent to the parabola, and R its point of contact. 179. Just as from Pascal's theorem a series of special theorems have been derived, relating to the inscribed pen- tagon, quadrangle, and triangle, so also from Brianchon's theorem can be deduced a series of correlative theorems re- lating to the circumscribed pentagon, quadrilateral, and triangle. Suppose e.g. that two of the six tangents a,b , ,c,a / ,b,c / which form the circumscribed hexagon, (Art. 153, left), b and d for example, lie indefinitely near to one another. Since a tangent intersects a tangent indefinitely near to it in its point of contact (Arts. 146, 149), the hexagon will be replaced by the figure made up of the circumscribed pentagon aVca'b together with the point of contact of the side b (Fig. 121). Brianchon's theorem will then become the following : If a pentagon is circumscribed to a conic, the two diagonals which connect any two pairs of opposite vertices, and the straight line join- ing the fifth vertex to the point of contact of the opposite side, meet in the same point. This theorem expresses a property of projective ranges which has already (Art. 85, left) been noticed. For consider the two projective ranges determined by the other tangents on a and b as bases. Three pairs of corresponding points are given, viz. those determined by a', b', and c. Project the first range from the point caf and the second from cb' ; this gives two pencils in perspective of which corresponding, pairs of rays intersect L Fig. 121. 146 DEDUCTIONS FROM THE THEOREMS [180 on the straight line r which joins the points ah', ha' . In order then to obtain that point of the second range which corresponds to the point ah of the first, viz. the point of contact of the tangent h, we draw the straight line q which joins the points ca' and ah, and then the straight line p which joins cb f and qr ; then ph is the point required. But this construction agrees exactly with the theorem in question. 180. By means of the property of the circumscribed pentagon just established the following problems can be solved : (1). Given Jive tangents to a conic, to determine the point of contact of any one of them *. Particular case. Given four tangents to a parabola, to determine their points of contact, and also the direction of the point at infinity on the curve. (2). To construct hy tangents a conic, having given four tangents and the point of contact of one of them. Particular cases. ►To construct by tangents a hyperbola of which three tangents and r one asymptote are given. To construct by tangents a parabola, having given three tangents and the direction of the point at infinity on the curve ; or three tangents and the point of contact of one of them. 181. The corollaries of Brianchon's theorem which relate to the circumscribed quadrilateral and triangle have already been given (they are the propositions of Arts. 172 and 176) ; they are correlative to the theorems of Arts. 166 and 174, just as those of Arts. 164 and 179 are correlative to one another. It will be a very useful exercise for the student to solve for himself the problems enunciated in the present chapter : the constructions all depend upon two fundamental ones, correlative to one another, and following immediately from Pascal's and Brianchon's theorems. 182. The corollaries to the theorems of Pascal and Brianchon show that just as a conic is uniquely determined by five points or five tangents, so al^o it is uniquely determined by four points and the tangent at one of them, by four tangents and the point of contact of one of them, by three points and the tangents at two of them, or by three tangents and the points of contact of two of them. It follows that (1). An infinite number of conies can be drawn to pass through three given points and to touch a given straight line at one of these points ; or to pass through two given points and to touch at them two given straight lines ; but no two of these conies can have another point in common. * Maclaubin, loc. cit, § 41 . 182] OF PASCAL AND BRIANCHON. 147 (2). An infinite number of conies can be drawn to touch a given straight line at a given point, and to touch two other given straight lines ; or to touch two given straight lines at two given points ; but no two of these conies can have another tangent in common. If then two conies touch a given straight line at the same point (i.e. if the conies touch one another at this point), they cannot have in addition more than two common tangents or two common points ; and if two conies touch two given straight lines at two given points {i.e. if two conies touch one another at two points) they cannot have any other common point or tangent. Thus if two conies touch a straight line a at a point A , this point is equivalent to two points of intersection, and the straight line a is equivalent to two common tangents. ^ L 2 CHAPTEK XVII. DESARGUES THEOREM. 183. Theorem. Any transversal whatever meets a conic and the op- posite sides of an inscribed quad- rangle in three conjugate pairs of points of an involution. This is known as Des argues' theorem *. Let QEST (Fig. 122) be a quadrangle inscribed in a conic, Correlative Theorem. The tangents from an arbitrary point to a conic and the straight lines which join the same point to the opposite vertices of any circumscribed quad- rilateral form three conjugate pairs of rays of an involution. Let qrst (Fig. 123) be a quad- rilateral circumscribed about 8 Fig. 122. and let s be any transversal cut- ting the conic in P and P', and the sides QT, ES, QE, TS of the x Fig. 123. any point S let ' be drawn to the conic ; from tangents p, p conic, and let the straight lines * Desargues, loc. cit., pp. 171, 176. 185] DESARGUES THEOREM, 149 quadrangle in A , A ', B, B' re- spectively. The two pencils which join the points P, R, P', T of the conic to Q and S respectively are projective with one another (Art. 149), and the same is therefore true of the groups of points in which these pencils are cut by the transversal. That is, the group of points PBP' A is pro- jective with the group PA'P'B', and therefore (Art. 45) with P'B'PA'\ consequently (Art. 123) the three pairs of points PP', AA\ BB f are in involution. 184. This theorem, like that of Pascal (Art. 153, right), enables us to construct by points a conic of which five points P,Q,R,S,T are given. For if (Fig. 122) an arbitrary transversal s be drawn through P, cutting QT, RS, QR, TS in A , A', B, B' respectively ; *and if (as in Art. 134) the point P' be found, conjugate to P in the involution determined by the pairs of points A, A' andi?, B' \ then will P' be another point on the conic to be constructed. 185. The pair of points C, C in which the transversal cuts the diagonals QS and RT of the inscribed quadrangle belong also (Art. 131, left) to the involution determined by the points A , A f and B, B'. Moreover, since the points A, A' and B, B f suffice to deter- mine the involution, the points a, a',b,b' be drawn which join S to the vertices qt, rs, qr, ts of the quadrilateral respectively. The two groups of points in which q and s are cut by the tangents p, f M p' , t are pro- jective with one another (Art. 149), and the same is therefore true of the pencils formed by joining these points to S. That is, the group of rays pbp'a is projective with the group^a / jp / 6', and therefore (Art. 45) with 2>'b'pa' ; consequently (Art. 123) the three pairs of rays pp', aa', bb' are in involution. This theorem, like that of Brianchon (Art. 153, left), en- ables us to construct by tangents a conic of which five tangents p, q, r, s, t are given. For if (Fig. 123) an arbitrary point S be taken on p, and this point be joined to the points qt, rs, qr, ts respectively by the rays a,a',b,b'\ and if (Art. 134) the ray p' be constructed, conjugate to p in the involution determined by the pairs of rays a, a' and b,V ' ; then will p f be another tangent to the conic to be constructed. The pair of rays c, c' which connect S with the points of intersection qs and rt of the opposite sides of the circum- scribed quadrilateral belong also (Art. 131, right) to the involu- tion determined by the rays a, a' and b , V. Moreover, since the rays a, a' and b, b' suffice to determine the involution, the rays p, p' are a 150 DESARGUES THEOREM. [186 P, P f are a conjugate pair of this involution for every conic, whatever be its nature, which circumscribes the quadrangle QEST. Thus: Any transversal meets the conies circumscribed about a given quad- rangle in pairs of points forming an involution. If the involution has double points, each of these is equivalent to two points of intersection P and P f lying indefinitely near to one another ; and will therefore be the point of contact of the transversal with some conic cir- cumscribing the quadrangle. There are therefore either two conies which pass through four given points Q, P, S, T and touch a given straight line s (not passing through any of the given points), or there is no conic which satisfies these con- ditions. 186. If, from among the six points AA', BB', PP' of an involution, five are given, the sixth is determined (Art. 1 34). If ,( ! then in Fig. 122 it is supposed that the conic is given, and that the quadrangle varies in such a way that the points A, A', B remain fixed, then also the point B' will remain invariable ; consequently : If a variable quadrangle move in such a way as to remain always inscribed in a given conic, while three of its sides turn each round one of three fixed collinear points, then the fourth side will turn round a fourth fixed point, conjugate pair of this involution for every conic, whatever be its nature, which is inscribed in the quadrilateral qrst. Thus : The pairs of tangents drawn from any point to the conies inscribed in a given quadrilateral form an involution. If the involution has double rays, each of these is equivalent to two tangents p and p' lying indefinitely near to one another ; and will therefore be the tangent at S to some conic inscribed in the quadrilateral. There are therefore either two conies which touch four given straight lines q, r, s, t and pass through a given point S (not lying on any of the given lines), or there is no conic which satis- fies these conditions. If, from among the six rays aa', bb f , pp r of an involution, five are given, the sixth is deter- mined (Art. 134). If then in Fig. 123 it is supposed that the conic is given, and that the quadrilateral varies in such a way that the rays a, a', b remain fixed, then also the ray b f will remain invariable ; consequently: If a variable quadrilateral move in such a way as to remain always circumscribed to a given conic, while three of its vertices slide each along one of three fixed con- current straight lines, then the fourth vertex will slide along a 187] DESARGUES THEOREM. 151 Fig. 124. collinear with the three given fourth fixed straight line, concur- ones. rent with the three given ones. 187. The theorem of the preceding Art. (left) may be ex- tended to the case of any inscribed polygon having an even number of sides. Suppose such a polygon to have 2 n sides, and to move in such a way that in—i of these pass respec- tively through as many fixed points all lying on a straight line s (Fig. 124). Draw the diagonals connecting the first of its vertices with the 4 «S 6 th , 8 th , ... 2 (n - i) th vertex, thus dividing the polygon into n — 1 simple quadrangles. In the first of these quadrangles the first three sides (which are the first three sides of the polygon) pass respectively through three fixed points on s ; therefore also the fourth side (which is the first diagonal of the polygon) will pass through a fixed point on s. In the second quadrangle the first three sides (the first diagonal and the fourth and fifth side of the polygon) pass re- spectively through three fixed points on * ; therefore the fourth side (the second diagonal of the polygon) will pass through a fixed point on s. Continuing in the same manner, we arrive at the last quadrangle and find that the fourth side of this (i.e. the 2n th side of the polygon) passes through a fixed point on s. We may therefore enunciate the general theorem : If a variable polygon of an even number of sides move in such a way as to remain aUvays inscribed in a given conic, while all its sides but one pass respectively through as many fixed points lying on a straight line, then the last side also will pass through a fixed point collinear with the others *, If tangents can be drawn to the conic from the fixed point round which the last side turns, and if each of these tangents is considered as a position of the last side, the two vertices • which lie on this side will coincide and the polygon will have only 2 n — 1 vertices. The point of contact of each of the two * Poncelet, loc. cit, Art. 5T3. 152 DESARGUES THEOREM. [188 remain tangents will therefore be one position of one of the vertices of a polygon of zn—i sides inscribed in the conic so that its sides pass respectively through the zn — i given collinear points. 188. The solution of the correlative theorem is left as an exercise to the student ; the enunciation is as follows : If a variable polygon of an even number (zn) of sides moves so as to to a given conic, while all its vertices but one slide along as many fixed straight lines radiating from a centre, then the last vertex also loill slide a fixed straight line passing the same centre (Fig. 125). If the straight line on which this last vertex slides cut the conic in two points, and if the tangents at these be drawn, each of them will be one position of a side of a polygon of zn — i sides circumscribed about the conic so that its vertices lie each on one of the z n — i given con- Fig. 125. current straight lines. 189. If in Fig. 122 it be sup- posed that the points S and T lie indefinitely near to one another on the conic, or in other words that ST is the tangent at S, then the quadrangle QRST reduces to the inscribed triangle QRS and the tangent at S (Fig. 126), so that Desargues' theorem becomes the following : If a triangle QRS is inscribed in a conic, and if a transversal s meet two of its sides in A and A', the third side and the tangent at the opposite vertex in B and B', and the conic itself in P and P', If in Fig. 123 the tangents s and t be supposed to lie indefi- nitely near to one another, so that st becomes the point of contact of the tangent s, then the quadri- lateral qrst reduces to the circum- scribed triangle qrs and the point of contact of s (Fig. 127), so that the theorem correlative to that of Desargues becomes the following : If a triangle qrs is circum- scribed about a conic, and if from any point S there be drawn the straight lines a, a' to two of its vertices, the straight lines b,b' to the third vertex and tfie point of 191] DESARGUES THEOREM. 153 these three pairs of points are in involution. 190. This theorem gives a solution of the problem : Given Jive Fig. 126. points P,P',Q,B ,S on a conic, to dravj the tangent at any one of them S. Forif^ 5 ^ / 5J B(Fig.i26) are the points in which the straight line PP' cuts the straight lines QS, SB, BQ respectively, we construct (as in Art. 1 34) the point B ' conjugate to B in the involution determined by the two pairs of points A , A' and P, P'; then B'S will be the required tangent. |0 -191. If in Fig. 126 it be now supposed in addition that the points Q and B also lie inde- finitely near to one another on the conic, i. e. that QB is the tangent at Q, then the inscribed quadrangle QBST is replaced by the two tangents at Q and S and their chord of contact QS counted twice (Fig. 128). Since, the straight lines QT, BS now coincide, A and A' will contact of the opposite side, and the tangents p, p f to the conic, then these three jyairs of rays are in involution. This theorem gives a solution of the problem : Given five tangents Fig. 127. p, p',q, r, 8 to a conic, to find the point of contact of any one of them s. For if a,a',b (Fig. 1 2 7) are the rays joining the point pp' to the points qs,sr,rq respectively, we construct (as in Art. 134) the ray b' conjugate to b in the involu-. tion determined by the two pairs of rays a, a f and p ,//; then b's will be the required point of con- tact. If in Fig. 127 it be now sup- posed in addition that the tan- gents q and r lie indefinitely near to one another, i.e. that qr is the point of contact of the tangent q, then the circumscribed quadri- lateral qrst is replaced by the points of contact of the tangents q and s and the point of intersec- tion qs of these tangents counted twice (Fig. 129). Since the points qt, rs now co- incide in a single point qs, the 154 DESABGUES THEOREM. [191 also coincide in one point, which is consequently one of the double points of the involution deter- mined by the pairs of conjugate rays a and a' will also coincide in a single ray a, which is conse- quently one of the double rays of the involution determined by the Fig. 128. points P, P' and B, B'. case, then, Desargues' becomes the following : In this theorem If a transversal cut two tan- gents to a conic in B and B\ tlieir chord of contact in A, and the conic itself in P and P\ then the point A is a double point of the involution determined by the pairs of points P, P' and B, B'. Or, differently stated : If a variable conic 2?ass through two given points P and P' and touch two given straight lines, the chord which joins the points of contact of these two straight lines will always pass through a fixed point on PP'. If the tangents QU, SU vary at the same time with the conic, while the points P,P',B,B' re- main fixed, the chord of contact Fig. 129. pairs of conjugate rays p, p f and b , b' . The theorem correlative to that of Desargues then becomes the following : If a given point 8 be joined to two points on a conic by the straight lines b, b', and to the point of intersection of the tan- gents at these points by the straight line a; and if from the same point S tliere be drawn the two tan- gents p, p' to the conic ; then a is a double ray of the involution de- termined by the pairs of raysp,p' and b, V ' . Or, differently stated : If a variable conic touch two given straight lines p and p' and pass through two given points, the tangents at these two points will always intersect on a straight line passing through pp/. . If the points of contact of q and s vary at the same time with the conic, while the straight lines p,p', b, b f remain fixed, the point 193] DESARGUES THEOREM. 155 QS must still always pass through one or other of the double points of the involution determined by the pairs of points P, P ' and B,B'. If then four collinear points P, P ', B, B' are given and any conic is drawn through P and P\ and then the pairs of tangents from B and B' to this conic ; then if each tangent from B is taken to- gether with each tangent from B\ four chords of contact will be obtained, which intersect one another two and two in the double points of the involution determined byP, />' and £ ,"£'*, 192. From the theorem of the last Article (left) is derived a solution of the problem : Given four points P,P f >Q,S on a conic and the tangent at one of them Q, to draw the tangent at any other of the given points S (Fig. 128). For if A , B are the points in which PP' cuts QS and the given tangent respectively, and we con- struct the point B' conjugate to B in the involution determined by the pair of points P, P' and the double point A ; then the straight line SB' will be the tan- gent required. of intersection qs must still always lie on one or other of the double rays of the involution determined by the pairs of rays p,y/ and 6, b'. If then four concurrent straight lines p , p\ b , b f are given and any conic is drawn touching p and 7/, and then the two pairs of tan- gents to this conic at the points where it is cut by b and &'; then if the tangents at the two points on b are combined with the tangents at the two points on &', each with each, four points of intersection will be obtained, which lie two and two on the double rays of the involution de- termined by p, p f and b, b' '. From the theorem of the last Article (right) is derived a solu- tion of the problem : Given four tangents %), p' , q, s to a conic and the point of contact of one of them q, to determine the point of contact of any other of the given tangents s (Fig. 129). For if a , b are the rays which connect pp r with qs and with the given point of contact respec- tively, and we construct the ray b' conjugate to b in the involu- tion determined by the pair of rays p , p' and the double ray a ; then sb f will be the required point of contact. 193. Consider again the theorem of Art. 191 ; and suppose that the conic is a hyperbola, and that its asymptotes are the tangents given (Fig. 1 30). The chord of contact QS lies in this case entirely at infinity; so that the involution (PP\ BB', ...) has one double point at infinity, and therefore (Arts. 59, 125) the other double point * Brianchon, loc. eft., pp. 20, 21, 156 DESARGUES THEOREM. [194 is the common point of bisection of the segments PP[, BB', ... We conclude that : If a hyperbola and its asymptotes be cut by a transversal, the seg- ments intercepted by the curve and by the asymptotes respectively have the same middle point. Fig. 130. From this it follows that PB = B'P' and PB' = BP' *, which gives a rule for the construction of a hyperbola when the two asymptotes and a point on the curve are given t. 194. Consider once more the Consider once more the theorem theorem of Art. 191 (left), and of Art. 191 (right), and suppose suppose now that the points P and P' are indefinitely near to one another, i. e. let the transversal be a tangent to the conic (Fig. 131). Its point of contact P will now that the tangents p and p' lie indefinitely near to one another, i.e. let the point S lie on the conic itself (Fig. 132). The tan- gent to the conic at S will be the a Fig. 131. be the second double point of the involution determined by the pair of points B,B f and the double points ; consequently (Art. 125) P and A are harmonic conjugates Fig. 132. second double ray of the involu- tion determined by the pair of rays b , V and the double ray a ; consequently (Art. 125) p and a are harmonic conjugates with re- * Apollonius, loe. cit., ii. 8, 16. t Ibid., ii. 4. 196] DES ARGUES THEOREM. 157 with regard to B and B' ; and we conclude that : In a triangle UBB' circum- scribed to a conic, any side BB f is divided harmonically by its point of contact P and the point where it meets the chord QS joining the points of contact of the other two sides. 195. From A a second tangent can be drawn to the conic ; let its point of contact be 0. Since the four points P,A,B,B', which have been shown to be harmonic, are respectively the point of contact of the tangent AB, and the three points where this tangent cuts three other tangents OA, QB, SB' respectively, it follows that the tangents AB, OA, QB, SB' will be cut by every other tangent in four harmonic points (Art. 149); i.e. they are four harmonic tan- gents (Art. 151). And since the chord of contact QS of the con- jugate tangents QB, SB' passes through A the point of intersec- tion of the tangents at P and 0, we have the theorem : If the chord of contact of one pair of tangents to a conic pass through the point of intersection of another pair of tangents, then each pair is harmonically conjugate with regard to the other. And conversely : If four tangents to a conic are harmonic, the chord of contact of each pair of conjugate tangents passes through the point of inter- section of the other pair. gard to b and b'; and we con- clude that : In a triangle ubb' inscribed in a conic, any two sides b and b' are harmonic conjugates with re- gard to the tangent p at the vertex in which they meet and the straight line joining this vertex to the point of intersection of the tangents q and s at the other two vertices. The straight line a cuts the conic in a second point ; let the tangent at this be o. Since the four rays p ,a,b ,b' , which have been shown to be harmonic, are respectively the tangent at S, and the straight lines which join S to three other points on the conic (the points of contact of o , q, and s) it follows that the straight lines connecting these four points with any other point on the conic will form a harmonic pencil (Art. 149); i.e. the four points are harmonic (Art. 151). And since the point of intersection of the tangents q and s lies on the chord of contact of the tangents p and o, we have the theorem : If the point of intersection of the tangents at one pair of points on a conic lie on the chord join- ing another such pair of points, then each pair is harmonically conjugate with regard to the other. And conversely : If four points on a conic are harmonic, the point of intersection of the tangents at each pair of con- jugate points lies on the chord joining the other pair. 196. These two correlative propositions can be combined into one 158 DESAKGUES' THEOREM. [197 by virtue of the property already established (Arts. 148, 149) that the tangents at four harmonic points on a conic are themselves har- monic, and conversely. We may then enunciate as follows : If a pair of tangents to a conic meet in a point lying on the chord of contact of another ptair^ then also the second pair will meet in a point lying on the chord of contact of the first ; and the four tangents (and likewise their points of contact) will form a harmonic system*. Thus in Fig. 131 QS passes through A , the point of intersection of PA and OA, and similarly OP passes through £7, the point of inter- section of QB and SB'; and the pencil U (QSPA) is harmonic, and likewise the pencil A (OPQ U). In Fig. 132 the point qs lies on a, the chord of contact of and j), and similarly the point op lies on the straight line u which joins the points of contact of q and $ ; and the range u (qsap) is harmonic, and the range a (o2)qu) also. 197. Example. Suppose the conic to be a hyperbola (Fig. 133). Its asymptotes are a pair of tangents whose chord of contact QS is the straight line at infinity ; consequently the chord joining the points of contact of a pair of parallel tangents will pass through the point of intersection U of the asymptotes ; and conversely, if through U a transversal be drawn, the tangents at the points P and 0, where it cuts the curve, will Fig. 133. De parallel. The point U will lie midway between P and 0, since in general UVPO (Fig. 131) is a harmonic range, and in this case V lies at in- finity. Any tangent to the curve cuts the asymptotes in two points B and B' which are harmonically conjugate with regard to the point of con- tact P and the point where the tangent meets the chord of contact of the asymptotes ; but this last lies at infinity ; therefore P is the middle point of BB\ Thus The part of a tangent to a hyperbola which is intercepted between the asymptotes is bisected at its point of contact t. This proposition is a particular case of that of Art. 193. 198. Theoeem J. If a quadrangle is inscribed in a conic, the rectangle contained by the distances of any i^oint on the curve from * De la Hire, loc. cit., book i. prop. 30. Steiner, loc. cit., p. 1 59, § 43 ; Collected Works, vol. i. p. 346. + Apollonius, loc. cit., ii. 319. t To this Chasles has given the name of Pappus' theorem, since it corresponds to the celebrated 'problema ad quatuor linens' of this ancient geometer. Cf. Apercu historique, pp. 37, 338. 199] DESARGUES' THEOREM. 159 one pair of opposite sides is to the rectangle contained by its distances from the other pair in a constant ratio. In Fig. 122, the pairs of points P and P', A and A', B and B' being, by Desargues' theorem, in involution, the anharmonic ratios (PP'AB) and (P'PA'B') are equal to one another, or PA PB_ _ P\A/ P'B' P'A : P'B ~ PA '' PB' PB' PA' P'B'' P'A' But PA : P'A is equal to the ratio of the distances (measured in any the same direction) of the points P and P' from the straight line QT, and the other ratios in the foregoing equation may be interpreted similarly ; we have therefore (A±.{B)__(&).M (A)'-{B)' {B'f(A')" (B).(B')-(By.(B'Y' where (A), (A'), (B), (B') denote the distances of the point P from the sides QT, R/S, QR, ST respectively of the inscribed quadrangle QRST, and (A)', (A')', (B)', (B')' denote similarly the distances of the point P' from these sides respectively. (These distances may be measured either perpendicularly or obliquely, so long as they are all measured parallel to one another.) The ratio (A) (A') (B)(B') is therefore constant for all points P on the conic ; which proves the theorem. 199. Theoeem. If a quadrilateral is circumscribed about a conic, the rectangle contained by the distances of one pair of opposite vertices from any tangent is to the rectangle contained by the distances of the other pair from the same tangent in a constant ratio *. In Fig. 123 let the vertices qr, qt, st, sr of the circumscribed quadrilateral qrst be denoted by R, T, T x , R x respectively ; let the points where the tangents p, p' meet the side q be called P, P' respectively t, and let the points where these same tangents meet the side s be called P x , P x ' respectively. Since by the theorem corre- lative to that of Desargues, the pairs of rays p and p', a and a', b and b', are in involution, the anharmonic ratios (bapp') and (b'a'p'p) are equal to one another. Hence by theorem (2) of Art. 149, * Chasles, Sections coniques, Art. 26. f P' is not shown in the figure. 160 DESARGUES' THEOREM. [199 (RTPP') = (T&PfPJ = (2?^^/) by Art. 45; #P &£* _ R^P X R X P( •*• Tp • Tp , - T ^ ' T ^, » RP.T.P. RP'.T.P/ Whence TP^P^ TP'.RSf But RP : TP is equal to the ratio of the distances (measured in any the same direction) of the points R and T from the straight line p ; so also T 1 P 1 : R X P X is the ratio of the distances of the points T x and R x from the same straight line p. The foregoing equation therefore expresses that the ratio RP.T 1 P 1 TP.R.P, is constant for every tangent p to the conic ; which proves the theorem. CHAPTEK XVIII. SELF-CORRESPONDING ELEMENTS AND DOUBLE ELEMENTS. 200. Consider two projective flat pencils, concentric or non- concentric. Through their common centre or through their two centres and 0' draw a conic or a circle, and let this cut the rays of the first pencil in A,B,C, ... and those of the second in A', B\ C", ... . Project these two series of points from two new points 1 , 0/ (or from the same point) lying on the conic ; the two projecting pencils Y [ABC ... ) and 1 / (A / B / C / ... ) are by Art. 149 projective with the two given pencils (ABC ... ) and Cf (A' B' C ... ) respectively ; and are therefore projective with one another. The two series of points ABC ... and A'B'C ... are said to form two projective ranges on the conic *. I. Now project these two ranges (Fig. 134) from two of then- corresponding points, say from A' and A. The projecting pencils A'(A,£,C,...) and A(A\B\C\...) will be projective with one another ; and since they have the self-corresponding ray AA ', they are A in perspective. Corresponding pairs n/f~/^^\ of rays will therefore (Art. 80) inter- ^n^V„: A^>\ sect on a fixed straight line, so that ^j ^^n T^., 4 \ r AB' and A'JB, AC and A'C, AD' and \AT p?T?k^ A' I) . . . , will meet on one straight line s. ^I' or AM.A'N = AN.A'M. But by Ptolemy's theorem (Euc. vi. D), AA'. MN = AM. A'N+AN. A'M. If then M , N , A , A ' are four harmonic points on a circle, \ AA'. MN = AM . A'N = AN . A'M. 206. The properties established in Art. 200 and the following Articles lead at once to the solution of the important problem : To construct the self- corresponding elements of two superposed pro- jective forms, and the double elements of an involution. $t5/ I. Let two concentric projective pencil's be given, which are deter- "IX mined by three pairs of corresponding rays (Fig. 142); it is required to construct their self-corresponding rays. Through the common centre describe any circle, cutting the three given pairs of rays in A and A', B and B', C and C re- spectively. Let AB', A'B meet in E, and AC, A'C in Q ; if the straight line QR cut the circle in two points M and N, then OM, ON will be the required self-corre- sponding rays. Fig. 142. 170 SELF-CORltESPONDING ELEMENTS [206 II. Let A and A', B and B\C and C (Fig. 143) be three pairs of corresponding points of two collinear ranges; it is construct the self-corresponding points. to fr Fig. 143. Describe any circle touching the common base of the two ranges, and to this circle draw from the given points the tangents a and a 7 , b and b', cand c' . Let r be the straight line which joins the points* ab\ a'b, and q that which joins the points ac\ a'c. If the point qr lies outside the circle and from it the tangents m and n be drawn to the circle, then the points om , on in which these meet the base will be the required self-corresponding points of the two ranges. Fig. 144. Otherwise (Fig. 144): Draw any circle whatever in the plane and take on it any point - UNTVEKi 206] AND DOUBLE ELEMENTS. 171 0. From project the given points upon the circumference of the circle, and let A[ and A x ', B x and B x ', C x and C x be the projec- tions of A and A', B and B', C and C respectively. Join A X B X , A X B X meeting in R, and A X C X , A X 'C X meeting in Q (or B X C X , B X 'C X meet- ing in P). If the straight line PQR cut the circle in two points M x , N x , and these be projected from the- point back upon the given base o, then their projections M , N will be the required self-corresponding points of the given ranges *. III. In (I) let the two pencils be in involution (Fig. 145), and let it be required to find the double rays. Two pairs of conjugate rays . suffice now to determine the pen- cils. Draw through the centre any circle cutting the given rays in A and A\ B and B' respectively. Let AB', A'B meet in R, and AB , A'B' in Q; if the straight line QR cut the circle in two points M and N, then OM , ON will be the required double rays of the involution., IV. Let A and A\B and B' be two given pairs of conjugates of an Fig- 145- involution of points on a straight line ; it is required to find the double points (Fig. 146). Draw any circle in the plane and take on it any point 0. From project the given points upon the circumference of the circle, and let A x and A/, B X and B( be the projections of A and A', B and B' respectively. Let A x B x ' i A l 'B 1 meet in R, and A X B X , A(B X ' in Q. If QR cuts the circle in M x , N lt and these points be projected from back upon the given straight line, then their projections M , N will be the required double points. * Steiner, loc. cit.y pp. 68 and 174, §§17 and 46; Collected Works, vol. i. PP- 285, 356. 172 SELF-CORRESPONDING ELEMENTS [207 , Otherwise : Describe a circle touching the base AB... (Fig. 147), and draw to this circle from the points A and A', B and B', the tangents a and a', Kg. 147. b and b\ respectively. Let r be the straight line which joins the points ab', a'b, and q that which joins the points ab, afb'. If the point qr lies outside the circle, the tangents m and n from this point to the circle will cut the base line of the involution in the required double points. 207. Theorem. A pencil in involution is either such that every ray is at right angles to its conjugate, or else it contains one and only one pair of conjugate rays including a right angle. Consider again Art. 206, III ; if the point of intersection S of the straight lines AA\BB\ ... is the centre of the circle (Fig. 148) then AA\ BB', ... are all diameters, and therefore Fig. 148. Fig. 149. each ray OA, OB, ... will be at right angles to its conjugate 0A\ 0B\ ... In this case then the involution is formed by a series of right angles which have their common vertex at 0. 208] AND DOUBLE ELEMENTS. 173 But if 8 is not the centre of the circle (Fig. 149), draw ;he diameter through it ; if C and C are the extremities of this diameter, the rays OC, OC will include a right angle. But these will be the only pair of conjugate rays which possess this property, since through S only one diameter can be drawn. 208. This proposition is only a particular case of the following one : Too superposed involutions (or such as are contained in the same one-dimensional form) have always a pair of conjugate eleynents in common, except in the case where the involutions have double elements and the double elements of the one overlap those of the other. Take two involutions of rays having a common centre 0, and let a circle drawn through cut the pairs of con- jugate rays of the first involution in the pairs of points (AA',BB\...) and those of the second in (GG',HH' ',...)• Let 8 be the point of intersection of AA',BB' f .,. and T that of GG\HTL\.... If the straight line ST cut the circle in two points i?and E', these will be a conjugate pair of each involu- tion, since they are collinear with 8 and with T also. Let us now examine in what cases ST will cut the circle. * h Fig. 150. Fig. 151. In the first place, it will certainly do so if one at least of the points S, T lies within the circle (Art. 203, VIII), i.e. if one at least of the involutions has no double elements (Figs. 150, 151). Secondly, if both the points 8, T lie outside the circle, i. e. if both the involutions have double elements, then the straight line ST may or may not cut the circle. If OMflN are the double elements of the first involution, 0JJ y 0V those of the second, the rays OE, OW must be harmonically conjugate both with regard to 0M,0N and with regard to 0U,0F; but (Art. 70) in order that there should exist a pair of elements which 174 SELF-CORRESPONDING ELEMENTS [209 are at the same time harmonically conjugate with regard to each of the two pairs OM , Oi\^and 0U,0V, it is necessary and sufficient that these two pairs should not overlap. If then these pairs do not overlap, ST will cut the circle (Fig. 152) ; Fig. 152 whereas if they do overlap, ST will not cut the circle (Fig. 153). The two involutions have therefore a common pair of conjugate elements in all cases except this last, viz. when they both have double elements and these overlap. [In Figs. 150, 151 and 152, are shown cases of two involutions having a common pair of conjugate elements E and E' ; Fig. 153 on the other hand illustrates the case where no such pair exists.] 209. The preceding problem, viz. that of determining the common pair of conjugate elements of two involutions superposed one upon the other, depends upon the following, viz. to determine (in a range, in a pencil, or on a conic) a pair of elements which are harmonically conjugate with regard to each of two given pairs. This problem has already been solved, for the case of a range, in Art. 70 ; the following is another solution : Suppose that we have to deal with a range of points lying on a straight line. Take any circle and a point on it, and project the given points from upon the circumference ; let M, N and U, V be their projections (Fig. 152). Let the tangents at M and N to the circle meet in S, and the tangents at U and V in T. If the pair MN does not overlap the pair UV, then ST will cut the circle in two points E and US', which when projected back from.0 upon the given straight line will give the points required. 210. The double points of the involution determined by the pairs A 9 A' and B, B / are the common pair of conjugate elements of two other involutions ; one of these is determined by the pairs A , B 21lj AND DOUBLE ELEMENTS. 175 and A', B', the other by the pairs A, B' and A', B (Art. 203, VII.) From this follows a construction for the double points of an involution of collinear joints which is determined by the pairs A, A' and B, B'. Take any point G outside the base of the involution and describe the circles GAB, GA'B' ; they will meet in another point, say in H. Similarly let K be the second point of intersection of the circles GAB', GA f B. Every circle passing through G and H meets the base in a pair of conjugate points of the involution AB, A'B' (Art. 127) ; so too every circle passing through G and K gives a pair of conjugate points of the involution AB', A'B. If then the circle GHK be described and it meet the base, the two points of intersection will be the double elements of the involution AA', BB' *. 211. It follows from the foregoing that the determination of the self-corresponding points of two projective ranges ABC ... and A'B'C ... on a conic (and consequently of the self-corresponding points of any two superposed projective forms) reduces to the con- struction of the straight line s on which intersect the pairs of straight lines AB' and A'B, AG' and A'G, BC' and B'C, .... Simi- larly the determination of the double points of an involution AA', BB', ... depends on the construction of the straight line s on which intersect the pairs of straight lines AB and A'B', AB' and A'B, ... or the pairs of tangents at A and A', B and B', ... . Conversely, if any straight line s (which does not touch the conic) is given, an involution of points on the conic is thereby determined ; for it is only necessary to draw, from different points of s, pairs of tangents to the conic, and the points of contact will be pairs of conjugate points of an involution. But, on the other hand, in order that two projective ranges of points ABC ... and A'B'C ... may be determined, there must be given, in addition to the straight line s, a pair of conjugate points A and A' also ; then the straight lines joining A and A' to any point on s will cut the conic in a pair of corresponding points B' and B. Two projective ranges of points determine an involution ; for they determine the straight line s, which determines the involution. If the two ranges have two self-corresponding points, there will also be the double points of the involution. * Chasles, Giomilrie, suptrieure, Art. 263. CHAPTEK XIX. PROBLEMS OP THE SECOND DEGREE. 212. Problem. Given five points 0, 0\ A, B, C on a conic, to determine the points of intersection of the curve with a given straight line s. Solution. Join any two of the points 0, 0' to each of the others A, B, C (Fig. 154); the Problem. Given five tangents 0, o', a, b,c to a conic, to draw a pair of tangents to the curve from a given point S. Consider the points where two of the tangents 0, o' are met by the others a, b, c (Fig. 155) ; the *%■ i54 pencils (A, B, C, ... ) and O f (A,B,C,...) will be projective, and will cut the transversal s in points forming two collinear projective ranges. A point M which corresponds to itself in these two ranges will ranges (a, b, c, ... ) and o / (a, b, c, ... ) will be projective, and if projected from JS as centre will give two concentric projec- tive pencils. Any ray m which corresponds to itself in these two pencils will 213] PROBLEMS OF THE SECOND DEGREE. 177 also be a point on the conic, since a pair of corresponding rays of the two pencils must meet in M. . The points of intersection of the conic with the straight line s are, therefore found as the self-corre- sponding points of the two colli-, near ranges which are determined on s by the three pairs of corre- sponding rays OA and O'A, OB and O'B, OG and O'C. There may be two such self-correspond- ing points, or only one, or none at all ; consequently the straight line s may cut the conic in two points, or it may touch it, or it may not meet it at all. The construction of the self-corresponding points themselves may be effected by either of the methods explained in Art. 206, II. 213. In a similar manner the problem may be solved if there be given four points , 0', A, B on a conic and the tangent o at one of them ; or three points 0, 0', A and the tangents o and o' at two of them and 0'. In the first case the two pencils are determined by the three pairs of rays o and O'O, OA and O'A, OB and O'B ; and in the second case by the three pairs o and O'O, 00 f and o', OA and O'A. If however there be given five tangents, or four tangents and the point of contact of one of them, or three tangents and the points of contact of two of them, we may begin by first con- structing such of the points of contact of the tangents as are not also be a tangent to the conic, since a pair of corresponding points of the two ranges o and o' must lie on m. The tangents from S to the conic are therefore found as the self-corresponding rays of the two concentric pencils which are determined by the rays joining JS to the three pairs of corresponding points oa and o'a, ob and o'b, oc and o'c. There may be two such self-correspond- ing rays, or only one, or none at all ; consequently there can either be drawn from the point £ two tangents to the conic, or JS is a point on the conic, or else from S no tangent at all can be drawn. The construction of the self-cor- responding rays themselves may be effected by the method ex- plained in Art. 206, I. In a similar manner the pro- blem may be solved if there be given four tangents o, o', a, b to a conic and the point of contact of one of them o ; or three tan- gents o, o', a and the points of contact and 0' of two of them o and o'. In the former case the three pairs of points which deter- mine the two ranges are and o'o , oa and o'a , ob and o'b ; in the latter case they are and o'o, oo' and 0', oa and o'a. If however there be given five points on the conic, or four points and the tangent at one of them, or three points and the tangents at two of them, we may begin by first constructing such of the tan- gents at the points as are not already given (Arts. 165, 171, N 178 PROBLEMS OF THE SECOND DEGREE. [214 already given (Arts. 180, 171, 177); the problem will then reduce to one of the cases given above. 175); the problem will then re- duce to one of the cases given above. 214. In the construction given in Art. 212 (left) suppose that the conic is a hyperbola and that the given straight line s is one of the asymptotes (Fig. 156). The collinear projective ranges de- termined on s by the pencils (A, B, C, ... ) and 0' (A, B, C, ... ) will have in this case one self-corresponding point, and this (being the point of contact of the hyperbola and the asymptote) will lie at an infinite dis- tance. But in two collinear ranges whose self-corresponding points coincide in a single one at infinity, the segment intercepted between any pair of corre- sponding points is of constant length (Art. 103). We therefore conclude that If from two fixed points and / on a hy2>erbola there be drawn two rays to cut one another on the curve, the segment PP' which these intercept on either of the asymptotes is of constant length *. Fig. 156. Fig. 157- 215. If in Art. 212 (left) the straight line s be taken to lie at infinity, the problem becomes the following : Given five ptoints 0, 0', A, B, C on a conic, to determine the joints at infinity on it (Fig. 157). * Brianchon, loc. cit., p. 36. 216] PROBLEMS OF THE SECOND DEGREE. 179 Consider again the projective pencils (A , B , C , . . . ) and 0' (A, B, C, ... ), which determine on the straight line at infinity s two collinear ranges whose self-corresponding points are the required points at infinity on the conic. Since each of these self-corresponding points must lie not only at the intersection of a pair of corresponding rays of the two pencils but also on the line at infinity s, the corre- sponding rays which meet in such a point must be parallel to one another ; the problem therefore reduces to the determination of the pairs of corresponding rays of the two pencils which are parallel to one another. In order then to solve the problem we draw through the parallels OA', OB', OC to O'A, O'B, O'C respectively, and then tonstruct (Art. 206, I) the self-corresponding rays of the two concentric pencils which are determined by the three corresponding pairs OA and OA', OB and OB', OC and OC. If there are two self-corre- sponding rays OM and ON, the conic determined by the five given points is a hyperbola whose points at infinity lie in the directions OM, ON ; i.e. whose asymptotes are parallel to OM and ON respectively. If there is only one self-corresponding ray OM, the conic deter- mined by the five given points is a parabola whose point at infinity lies in the direction OM. If there is no self-corresponding ray, the conic determined by the five given points is an ellipse, since it does not cut the straight line at infinity. If in the first case (Fig. 157) it is desired to construct the asymp- totes themselves of the hyperbola, we consider this latter as determined by the two points at infinity and three other points, say A, B, and C ; in other words, we regard the hyperbola as generated by the two projective pencils, one of which consists of rays all parallel to OM, and the other of rays all parallel to ON, and which are such that one pair of corresponding rays meet in A, a second pair in B, and a third pair in C. The rays which correspond in the two pencils respectively to the straight line at infinity (the line joining the centres of the pencils) will be the asymptotes required. Let then a, b, c (Fig. 157) be the rays parallel to OM which pass through A, B, C respectively, and let a', b', c' be the rays parallel to ON which pass through the same points respectively. Join the points ab' and a'b and the points bo' and b'c, and let K be the point of intersection of the joining lines ; the straight lines drawn through K parallel to OM and ON will be the .required asymptotes. 216. Problem. Given Jive points A, B,C, D, E on a conic, to draw the tangents from a given point S to the conic. This problem also can be made to depend on that of Art. 212 N % 180 PROBLEMS OP THE SECOND DEGREE. [217 Fig. 158. (left), by making use of the properties of the involution (Art. 203) obtained by cutting the conic by transversals drawn through S. Join SA,SB (Fig. 158); these straight lines will cut the conic again in two new points A' and B', which can be determined (making use of the ruler only, and without drawing the curve) by means of Pascal's theorem (Art. 161, right). (In the figure the points A' and B / have been constructed by means of the hexagons ADCBEA' and BECADB' respectively). Now let the point of intersection of A B and A'B' be joined to that of AB' and A'B ; the joining line s will pass through the points of contact of the tangents from S (Art. 203). The problem therefore reduces to that of determining the points of intersection of the conic and the straight line s (Art. 212, left). 217. The problem, To find the points of intersection of a given straight line s and a conic which is determined by Jive given tangents, may similarly be made to depend on that of Art. 212 (right), by making a construction (Fig. 159) analogous to the foregoing one. And the problem, To draw through a given i^oint a straight line which shall divide a given triangle into two parts having to one another a given ratio, may be solved by reducing it to the follow- ing construction : To draw from the given point a tangent to a hyperbola of which the asymptotes and a tangent are known. These are left as exercises to the student. Fig. 159- 218. Problem. To construct a conic which shall pass through four given points Q , R , £ , T, and shall touch a given straight line s which does not pass through any of the given p>oints. Solution. Let A « A' , B , B' To construct a conic which shall touch four given straight lines q, r , s, t , and shall p>ass through a given point S which does not lie on any of the given lines. Let a , a' , b , b' be the rays 219] PROBLEMS OF THE SECOND DEGREE. 181 be the points where the sides joining the point Sto the vertices QT , RS , QR , ST respectively qt , rs , qr , st respectively of the of the quadrangle QRST cut quadrilateral qrst (Fig. 161). the straight line s (Fig. 160). Construct the double rays (if Fig. 1 60. Fig. 161. Construct the double points (if such exist) of the involution de- such exist) of the involution de- termined by the pairs of rays a and a', b and &'. If there are two double rays m and n, each of them will be (Art. 185, right) a tangent at S to some conic inscribed in the quadrilateral qrst. Each of the conies qrstm , qrstn therefore gives a solution of the problem; and these conies can be con- structed by tangents by help of Brianchon's theorem (Art. 161, left). If however there are no double rays, there is no conic which satisfies the conditions of the problem. 219. If in the foregoing Art. (left) the straight line s be taken to lie at infinity, the problem becomes the following : To construct a -parabola which shall ^;ass through four given points ft, '*»*.>. To solve it, take any point (Fig. 162), and through it draw the rays a , a' , b , b' parallel respectively to the straight lines QT , RS , QR , ST ; and construct the double rays (if such exist) of the involution determined by the pairs of rays a and a' } b and b'. termined by the pairs of points A and A', B and B'. If there are two double points M and N, each of them will be (Art. 185, left) the point of con- tact with s of some conic cir- cumscribed about the quadrangle QRST. Each of the conies QRSTM, QRSTN therefore gives a solution of the problem ; and these conies can be constructed by points by help of Pascal's theorem (Art. 161, right). If however there are no double points, there is no conic which satisfies the conditions of the problem. 182 PROBLEMS OF THE SECOND DEGREE. [220 Each of these double rays will determine the direction in which lies the point at infinity on a parabola passing through the four given points ; the problem therefore reduces to the last problem of Art. 165. If however the involution has no double rays, no parabola can be found which satisfies the conditions of the problem. Through four given points therefore can be drawn either two parabolas or none ; in the first case the other conies which pass through the given points are ellipses and hyperbolas ; in the second case they are all hyperbolas. The first case occurs when each of the four points lies outside the triangle formed by the other three (i. e. when the quadrangle formed by the four points is non-reentrant) ; the second case when one of the four points lies within the triangle formed by the other three (i.e. when the quadrangle formed by the four points is reentrant). 220. If in Art. 218 (right) one of the straight lines q ,r , s ,t lies at infinity, the problem becomes the following : To construct a parabola which shall touch three given straight lines and shall ptass through a given point. 221. Pkoblem. To construct a conic which shall pass through three given points P , P\ P" and shall touch two given straight lines q and s, neither of which p>asses through any of the given points. Solution. This depends on the theorem of Art. 191 (left). Join PP\ and consider it as a trans- versal which cuts the conic in P and P', and the pair of tan- gents q and s in the two points B and B' (Fig. 163). If A and A x are the double points of the in- volution determined by the two pairs of points P and P\B and B', the chord of contact of the conic and the tangents q and s must pass through one of these points, by the theorem quoted above. To construct a conic which shall touch three given straight lines p , p', p" an ^ shall pass through two given points Q and S, neither of which lies on any of the given straight lines. The solution depends on the theorem of Art. 191 (right). Con- sider pp f as a point from which the tangents p and p' have been drawn to the conic, and the rays b and b f to the two points Q and S (Fig. 164). If a and a x are the double rays of the involution determined by the two pairs of rays p and p', b and b\ the point of intersection of the tangents at Q and S to the conic must lie on one of these rays, by the theorem quoted above. Repeat the same 221] PROBLEMS OF THE SECOND DEGREE. 183 Repeat the same reasoning for the case of the transversal PP", which cuts q and s in D and D" \ reasoning for the case of the point pp'\ from which are drawn the rays d and d" to the points Fig. 163. if C and C x are the double points of the involution determined by the two pairs of points P and P" ', D and D'\ the chord of contact must similarly pass through C or C x . The problem admits therefore of four solutions ; viz. when the two involutions (PP'iBB') and {PP" , DD") both have double points, there are four conies which satisfy the given conditions. If the double points are A , A x and C , C : respectively, the chords of con- tact of the four conies and the tangents q and s are AC, A X C , AC X , and A 1 G 1 . Of each of these conies five points are known, viz. P , P / i P", and the two points of intersection of AC (or of A X C ', or AC li or A x C xi as the case may be) with q and s ; they can ac- cordingly be constructed by points by means of Pascal's theorem (Art. 161, right). Fig. 164. Q and S; if c and c x are the double rays of the involution de- termined by the two pairs of rays 2) and ;;", d and d", the point of intersection of the tangents must similarly lie on c or c v The problem admits therefore of four solutions ; viz. when the two in- volutions (ppf, bb') and (pp", dd") both have double rays, there are four conies which satisfy the given conditions. If the double rays are a , a x and c , c x respectively, the points of intersection of the tangents at Q and S to the four conies are ac , a x c , ac 3 , and a x c x . Of each of these conies five tangents are known, viz. p ,p',p", and the two straight lines which join ac (or a 2 c, or ac 1} or a y c x , as the case may be) to Q and S ; they can accordingly be con- structed by tangents by means of Brianchon's theorem (Art. 161, left).. 184 PROBLEMS OF THE SECOND DEGREE. 222. Pboblem. To construct a polygon whose vertices shall lie on given straight lines (each on each), and whose sides shall pass through given points (each through each *). Solution. For the sake of simplicity suppose that it is required to construct a quadrilateral, whose vertices 1,2,3,4 shall lie respectively on four given straight lines s x , s 2 , s 3 , s 4 , and whose sides 12 , 23 , 34 , 41 shall pass respectively through four given points #12 5 ^23 5 #34 J #41 ( Fi g* l6 5)« The method and reasoning will be the Fig. 165. same as for a polygon of any number of sides. Take any points A li B li C lf ... on«j and project them from S 12 as centre upon s 2 ; and let A 2 , B 2 5 ^2 »••• be their projections. Project A 2 , B 2 , C 2 , ... from S 23 as centre upon s 3i and let A 3 , B 3 , C 3i ... be their projections. Project A 3 .B 3 ,C 3 ,... from S u as centre upon s 4 , and let A 4 , B i , C 4 ,. . . be their projections. Finally project A 4 , 2? 4 , 4 , ... from # 4J as centre upon s 15 and let A , 5 , C , ... be their projections. The points S 12 , S n , S 3i , £ 41 are the centres of four projectively related pencils ; for the first and second are in perspective (since their pairs of corresponding rays A X A 2 , BJB 2 , ... and A 2 A 3 , B 2 B 3 , ... intersect on s 2 ), the second and third are in perspective (pairs of corresponding rays intersect on s 3 ), and similarly the third and fourth are in perspective (pairs of corresponding rays intersect on s 4 ). Con- sequently (Art. 150) pairs of corresponding rays of the first and fourth pencils (such as A X A 2 and A 4 A) will intersect on a conic ; or in other words the locus of the first vertex of the variable quadri- lateral whose second, third, and fourth vertices (A 2 , A 3 , A 4 ) slide respectively on three given straight lines (s 2 , s 3 , s 4 ) and whose sides (A X A 2 , A 2 A 3 , A S A 4 , A 4 A) pass respectively through four given points, * Poncelet, he. cit., p. 345. 223] PROBLEMS OF THE SECOND DEGREE. 185 is a conic *. This conic passes through the points S 12 , aS"^ , the centres of the pencils which generate it ; in order therefore to deter- mine it, three other points on it must be known ; the intersections of the three pairs of corresponding rays A X A 2 and A±A , B 1 B 2 and BJB , C X C 2 and Cfi will suffice. It is then only necessary further to con- struct (Art. 212) the points of intersection if and N of the straight line s x with the conic determined by these five points ; either M or N can then be taken as the first vertex of the required quadrilateral. This construction may be looked at from another point of view. The broken lines A X A 2 A 3 A A A , B X B 2 B 3 BJ$ , and C x C 2 C z Cfi may be regarded as the results of so many attempts made to construct the required quad- rilateral ; these attempts however give polygons which are not closed, for A does not in general coincide with A lt nor B with JB lt nor C with G x . These attempts and all other conceivable ones which might similarly be made, but which it is not necessary to perform, give on the straight line s x two ranges A t B 1 C x ... and ABC.) one being traced out by the first vertex and the other by the last vertex of the open polygon. These ranges are projective with one another, since the second has been derived from the first by means of projections from /S^ , S 2Z , S 34 , S A1 as centres, and sections by the transversals 8 2 ,s 3 , s 4 , s x . Each of the self-corresponding points therefore of the two ranges will give a solution of the problem ; for, if the first vertex of the polygon be taken there, the last vertex will also fall on the same point, and the polygon will be closed. In the following examples also the method remains the same whatever be the number of sides of Cz M the polygon which it is required to k M^-ff7''^k\ — "^/f ' construct. A'^vS&Z >K' 'V 223. Peoblem. To inscribe in a / ^^S{^^^>J^/ given t conic a polygon whose sides /'Z^i^ / f\ pass respectively through given points. ^xi/l \ \V\ / / / Solution. Suppose that it is re- C of the conies with HK * 228. The solution just given of problem II holds good equally when the points A and B lie indefinitely near to one another, i.e. when the two conies touch a given straight line at the same given point. In this case two conies are given which touch one another at a point A, and the straight line HK is constructed which joins their remaining points of intersection C and C'. If HK passes through A, one of the points C or C' must coincide with A, since a conic cannot cut a straight line in three points. When this is the case, three of the four points of intersection of the conies lie indefinitely near to one another, and may be said to coincide in the point A ; and the conies are said to osculate at the point A. The construction gives a point H of the chord which joins A to the fourth point of intersection C of the conies. It may happen that this chord coin- cides with the tangent at A ; in this case A represents four coincident points of intersection of the two conies (or rather, four such points lying indefinitely near to one another). * Gaskin, The geometrical construction of a conic section , &c. (Cambridge, 1852), pp. 26, 40. 190 PROBLEMS OF THE SECOND DEGREE. [229 229. Let now the lemma of Art. 226 be applied to the case of a conic and a circle touching it at a point A. At A draw the normal to the conic (the perpendicular to the tangent at A), and let it cut the conic again in F and the circle again in F'. On AF as diameter describe a circle ; this circle, which touches the conic at A and cuts it at F, will cut it again at another point G such that AG F is a right angle. Join AG and let G f be the point where it cuts the first circle. Join FG,F f G'\ by the lemma they will intersect on the chord UK; but they are parallel to one another, since AG'F' also is a right angle. Thus for any circle whatever which touches the conic at A> the chord of intersection HK with the conic has a constant direction, viz. that parallel to FG. If HK passes through A, the conic and the circle osculate at this point. If then a parallel through A to FG cut the conic again in C, the circle which touches the conic at A and cuts it at C will be the osculating circle (circle of curvature) at A *. [In the particular case where A is a vertex (Art. 297) of the conic, F will be the other vertex, FG the tangent at F, AC the tangent at A, and C will coincide with A . It is seen then that the osculating circle at a vertex of a conic has not only three but four indefinitely near points in common with the conic] Conversely, the conic can be constructed which passes through three given points A, P, Q and has a given circle for its osculating circle at one of these points A. For join AP, AQ, and let them cut the given circle in P f , Q' re- spectively; and join PQ, P'Q f , meeting in U. If AU be joined and cut the circle again in C, the required conic will pass through C. It is therefore determined by the four points A ,P,Q , G and the tangent at A (which is the same as the tangent to the circle there). 230. The proposition correlative to the lemma of Art. 226 may be enunciated as follows : If a and b are a pair of common tangents to two conies, and if from two points taken on a and b respectively the tangents f,gbe drawn to the first conic and the tangents f , g' to the second, then the points fg and fg r will be collinear with the point of intersection of the second pair of common tangents to the conies. This proposition enables us to solve the problems which are corre- lative to I and II of Art. 227 ; viz. given three (or two) of the com- mon tangents to two conies, and in addition two (or three) tangents to the first and two (or three) tangents to the second, to determine the remaining common tangent (or the two remaining common tan- gents) to the conies. 231. Pkoblem. Given eleven points A, B,C,D,E; A 1 ,B 1 ,C v D 1 ,F i ; P; * Poncelet, loc. cit., Arts. 334-337. 234] PROBLEMS OF THE SECOND DEGREE. 191 to construct by 2>oints the conic which passes through P and through the four points of intersection of the two conies which are determined by the points A ,B ,0 ,D , E and A 1 ,B 1 ,0 1 ,D 1 , E x respectively. The conies are supposed not to be traced, nor are their points of intersection given *. Solution. Draw through P any transversal, and construct (Art. 212, left) the points M and M' in which it cuts the conic ABC BE and the points N and N f in which it cuts the conic A v B x O l D 1 E v Since these two conies and the required one all pass through the same four points, Desargues' theorem may be applied to them. If therefore (Art. 134, left) the point P r be constructed, conjugate to P in the involution determined by the pairs of points M and M ', N and N', this point P' will lie on the required conic. By causing the trans- versal to turn about the point P, other points on the required conic may be obtained. 232. Pkoblem. Given ten points A,B,C,D,E; A 1 ,B 1 ,C l ,D 1 ,E 1 and a straight line s ; to construct a conic which shall touch s and shall pass through the four points of intersection of the two conies which are determined by the points A,B,0 ,D,E and A 1 ,B 1 ,0 1 ,D 1 ,E 1 respectively. The conies are supposed not to be traced, nor are their points of intersection given. Solution. Construct (Art. 212) the points of intersection M and M' of s with the conic ABODE, and the points of intersection N and N' of s with the conic A l B x C 1 D 1 E 1 , and then (Art. 134) the double points of the involution determined by the two pairs of points M and M' ', N and N'. If P is one of these double points, it will be the point of contact (Art. 185) of s with a conic drawn through the four points of intersection of the conies ABODE and A 1 B 1 C 1 D 1 E l to touch s. The problem thus reduces to that of the preceding Article. 233. The correlative constructions give the solutions of the corre- lative problems : viz. to construct a conic which passes through a given point (or which touches a given straight line), and which is inscribed in the quadrilateral formed by the four common tangents to two conies ; the conies being supposed each to be determined by five given tangents, but not to be completely traced ; and their four common tangents being supposed not to be given. 234. Pkoblem. Through a given point S to draw a straight line which shall be cut by four given straight lines a ,b ,c ,d in four 2>oints having a given anharmonic ratio. Solution. It has been seen (Art. 151) that the straight lines which are cut by four given straight lines in four points having a given anharmonic ratio are all tangents to one and the same conic * Poncelet, loc. cit., Art. 389. 192 PROBLEMS OF THE SECOND DEGREE. [.235 touching the given straight lines ; and that if A , B , C are the points where d cuts a ,b ,c respectively, and D is the point of contact of d, the anharmonic ratio (ABCD) is equal to that of the four points in which the straight lines a ,b , c ,d are cut by any other tangent to the conic. Accordingly, if on the straight line d that point D be constructed (Art. 65) which gives with the points ad(=A),bd(=B),cd(=C) an anharmonic ratio {ABCD) equal to the given one, and if then the straight lines be constructed (Art. 213, right) which pass through S and touch the conic determined by the four tangents a,b ,c,d and the point of contact D of d, each of these straight lines will give a solution of the proposed problem. If one of the straight lines a , b , e , d lie at infinity, the problem becomes the following : Given three straight lines a ,b , c and a point S, to draw through S a straight line such that the segment intercepted on it between a and b may be to that intercepted on it between a and c in a given ratio. To solve this, construct on the straight line a that point A which is so related to the points ab (=B) and ac (==C) that the ratio AB : AC has the given value ; and draw from S the tangents to the parabola which is determined by the tangents a ,b , c and the point of contact A of a. The correlative construction gives the solution of the following problem : On a given straight line s to find a point such that the rays joining it to four given points A , B ,C ,D form a pencil having a given anharmonic ratio. 235. Problem. Given two projective ranges of points lying on the straight lines u , u' respectively ; to find two corresponding segments MP, M'P' such that the angles MOP , M / / P / which they subtend at two fixed points , r respectively may be given in sign and mag- nitude. Solution. Take on u' two points A' and D' such that the angle A'O'D' may be equal to the second of the given angles ; let A and D be the points on u which correspond respectively to A' and D' t and let A x be a point on u such that the angle A x OD is equal to the first of the given angles. The problem would evidently be solved if OA x coincided with OA, since in this case the angles AOD and A'O'D' would be equal to the given angles respectively. If the rays O'A', OA , 0'D',OD , OA x be made to vary simultaneously, they will trace out pencils which are projectively related. For those traced out by O'A' and O'D' respectively are projective, and similarly those traced out by OA x and OD respectively, since the angles A'O'D' and A x OD are constant (Art. 108); and the pencils traced out by 237] PROBLEMS OF THE SECOND DEGREE. 193 OA and O'A' respectively, and by OD , 0' ' D' respectively, are pro- jective since the given ranges on u and u' are so. Consequently the pencils generated by OA and OA x respectively are projective, and their self-corresponding rays give the solutions of the problem. If three trials be made of a similar kind to the foregoing one, three pairs of corresponding rays OA and OA xi OB and OB x , OG and 0G X will be obtained ; let the self-corresponding rays of the concentric projective pencils determined by these three pairs be constructed (Art. 206, 1). If one of these self-corresponding rays meets u in M, and if the point P be taken on u such that the angle MOP is equal to the first of the given ones, and if then on u' the points M', P f be found which correspond to M,P respectively, the angle M'O'P' will be equal to the second of the given angles, and the problem will be solved. 236. Problem. Given two projective ranges of points A ,B ,G , ... and A' ,B f ,G f ... lying on the straight lines u and u' respectively, to find two corresponding segments which shall be equal, in sign and magnitude, to two given segments. Solution. Take on u' a segment A'])' equal to the second of the given ones, and let AD be the segment on u which corresponds to A'D'. Take on u the point A x such that A X D is equal to the first of the given segments ; then the problem would be solved if A x coincided with A. If the points A , A', D', D ,A X be made to vary simulta- neously, the ranges traced out by A and A' respectively will be pro- jective with one another, as also those traced out by D and !)' respectively (by reason of the projective relation existing between ABG... and A'B'C' ...} ; and the ranges traced out by A and D respectively, and similarly those traced out by A' and D' respectively, will be projective with one another, since they are generated by segments of constant length sliding along straight lines (Art. 103). Consequently also the ranges traced out by A and A x are protectively related, and their self- corresponding points give the solutions of the problem. It is there- fore only necessary to obtain three pairs of corresponding points A and A',B and B', G and C', by making three trials, and then to construct the self- corresponding points of the ranges determined by these three pairs (Art. 206, II). 237. The student cannot have failed to remark that the method employed in the solution of the preceding problems has been in all cases substantially the same. This method is general, uniform, and direct ; and it may be applied in a more or less simple manner to all problems of the second degree, i.e. to all questions which when treated algebraically would depend on a quadratic equation. It consists in making three trials, which give three pairs of corresponding elements of two superposed projective forms ; the self-corresponding elements of these systems give the solutions of the problem. This method is 194 PROBLEMS OF THE SECOND DEGREE. [238 precisely analogous to that known in Arithmetic as the ' rule of false position/ and it has on that account been termed a geometric method of false position*. 238. Problems of the second degree (and those which are reducible to such) are solved, like all those occurring in elementary Geometry, by means of the ruler and compasses only, that is to say by means of the intersections of straight lines and circles t. But again, the solu- tion of any such problem can be made to depend on the determination of the self-corresponding elements of two superposed projective forms, which determination depends (Art. 206) on the construction of the self -corresponding points of two projective ranges lying on a circle whose position and size is entirely arbitrary. It follows that a single circle, described once for all, will enable us to solve all problems of the second degree which can be proposed with reference to any given elements lying in one plane (the plane in which the circle is drawn) J. This circle once described, any such problem will reduce to that of constructing three pairs of points of the two pro- jective systems whose self-corresponding elements give the solution of the problem. This done, we proceed to transfer to the circumference of the circle, by means of projections and sections, these three pairs of points. This will give three pairs of points on the circle ; taking these as the pairs of opposite vertices of an inscribed hexagon, we have only further to draw the straight line which joins the points of intersection of the three pairs of opposite sides (the Pascal line) of this hexagon. It is hardly necessary to remark that instead of the solution of such a problem being made to depend on the common elements of two superposed pro- jective forms, it may always be reduced to the determination of the double elements of an involution (Art. 211). The following Articles (239 to 249) *& contain examples of problems solved by Fig. 169. means of the method just explained. 239. Pkoblem. Given (Fig. 169) two projective ranges of points lying on the straight lines u and u respectively, and two other projective ranges of points lying on the straight lines v * Chasles, Geom. sup., p. 212. + A problem is said to be of the first degree when it can be solved with help of the ruler only, i. e. by the intersections of straight lines. See Lambert, loc. cit., p. 161 ; Brianchon, loc. cit, p. 6; Poncelet, loc. cit., p. 76. X Poncelet, loc. cit., p. 187; Steiner, Die geometrischen Constructionen aus- gefiihrt mittelst der geraden Linie und eines festen Kreises (Berlin, 1833), p. 67 ; Collected Works, vol. i. pp. 461-522 ; Staudt, Geometric der Lage (Niirnberg, !847)> § 23- 241] PROBLEMS OF THE SECOND DEGREE. 195 and v respectively ; it is required to draw through a given point two straight lines 8 and s', which shall cut u and u' in a pair of corre- sponding points and also v and v in a pair of corresponding p>oints. Through draw any straight line cutting u f , i/in A f ,P' respectively; let A be the point on u which corresponds to A', and let P be the point on v which corresponds to P' '. The problem would be solved if the straight lines OA and OP coincided with one another. If these straight lines be made to change their positions simultaneously, they will trace out two concentric projective pencils (determined by three trials of a similar kind to the one just made) ; and the self-corre- sponding rays of these pencils will give the solutions of the problem. 240. In the preceding problem the straight lines u and u f might be taken to coincide, and similarly v and v'. If all four straight lines coincided with one another, the problem would become the following : Given two projective ranges u, u' and two other projective ranges v, v all lying on one straight line, to find a pair of points which shall correspond to one another when regarded as points of the ranges u, u respectively, and likewise when regarded as points of the ranges v, v* respectively. 241. Peoblem. Between twio given straight lines u and u x to place a segment such that it shall subtend given angles at two given points and S (Fig. 170). Fig. 170. Draw any ray SA to meet u in A; draw SA X to meet u x in A 1 so that ASA X may be equal to the second of the given angles ; join 0A lt and draw 0A r to meet u in A' so that A f OA x may be equal to the first of the given angles. Then the problem would be solved if OA coincided with 0A\ Three trials of a similar kind to the one just made will give three pairs of corresponding rays (OA and 0A ( \ OB and 0B / i OC and 0C) of the two projective pencils which would be traced out by causing OA and OA' to change their positions simultaneously; the self-corresponding rays OM and ON of these pencils will give the solutions (MM x and NN^) of the problem. O % 196 PROBLEMS OF THE SECOND DEGREE. [242 242. Pkoblem. Given two projective ranges u and u; if a pair of corresponding points A and A' of these ranges be taken, it is required to find another pair of corresponding points M and M' such that the ratio of the length of the segment AM to that of the segment A'M' may be equal to a given number X. Let A and A',B and B', C and C be three pairs of corresponding points of the two ranges. On u take two new points B", C" such that AB"=\.A'B' said AC"=\.A'C. The points A , B", C" determine a range which is similar (Art. 99) to the range A', B',C, ... and therefore projective with A , B , C , ... . The collinear ranges A , B", C" , . . . and A ,B ,C , ... have already one self-corresponding . point in A ; their other self-corresponding point M (Art. 90) will give the solution of the problem, since AM=AM"=\.A'M'. This problem is therefore of the first degree. 243. Problem. Given two collinear projective ranges ABC ... and A'B 'C . . . , to find a pair of corresponding points M and M' such that the segment MM f shall be bisected at a given point 0. Take three points A?', B", C" such that is the middle point of each of the segments AA",BB", CC" '; the points A",B", C" determine a range which is equal to the range ABC ... , and therefore projective with the range A'B'C .... Construct the self- corresponding points of the collinear projective ranges A'B'C ... and A"B"C" ... ; if M' or M ff is one of them, then MM ' will have its middle point at 0, and will be a segment such as is required. 244. Pkoblem. Given a straight line and two points E ,F on it ; to determine on the straight line two points M and M' such that the segment MM' may be equal in length to a given segment, and the anharmonic ratio (EFMM / ) equal to a given number. Take on the given straight line any three points A,B, C ', then find on it three points A', B f ,C such that the anharmonic ratios {EFAA') , (EFBB') , (EFCC) may each be equal to the given number; and again three points A", B", C" such that the segments AA", BB h ', CC" may each be equal in length to the given segment. The ranges ABC ... and A'B'C ... will be projectively related (Arts. 79, 109), and the same will be the case with regard to the ranges ABC ... and A"B"G" . . .(Art. 103); therefore A'B'C ... and A"B"C" ... will be projective with one another. If these ranges have self-corresponding points, and if M ' or M" is one of them, the segment MM' and the anharmonic ratio (EFMM') will have the given values, and the problem is solved. 245. Problem. To inscribe in a given triangle PQR.a rectangle of given area (Fig. 171). Suppose MSTU to be the rectangle required ; if MS' be drawn parallel to PR, a parallelogram MSPS' will be formed which is equal 245] PROBLEMS OF THE SECOND DEGREE. 197 Fig. 171. in area to the rectangle ; so that for the given problem may be substituted the following equivalent one : To find on the base QB of a given triangle PQB a point M such that if MS , MS' be drawn parallel to the sides PQ , PR to meet PB, PQ in S, S' respectively, the rectangle contained by PS and PS' shall be equal to a given square k*. Take any point A on QB, draw AD parallel to PQ to meet PB in D, and take on PQ a point D' such that the rectangle contained by PD and PD' maybe equal to k* ; then draw D'A' parallel to PB to meet QB in A '. If the points A and A' coincided with one another, the problem would be solved. Now let the points A , D , D', A' be made to vary simultaneously ; they will trace out ranges which are all projective with one another. For since D is the projection of A made from the point at infinity on PQ, and A' the projection of D' made from the point at infinity on PB , the first and second ranges are in perspective, and the third and fourth likewise. But the second and third ranges are projective with one another, since the relation PD . PD'=k 2 shows (Art. 74) that the points D and D', in moving simultaneously, describe two projective ranges such that the point P , regarded as belonging to either range, corresponds to the point at infinity regarded as belonging to the other*. Three similar trials give three pairs of points similar to A and A'; if the self-corresponding points of the ranges determined by these pairs be constructed, they will give the solutions of the problem. Instead of taking the point A quite arbitrarily in the three trials, any particular positions may be chosen for it, and by this means the construction may often be simplified. This remark applies to all the problems which we have discussed. With regard to the present one, it is clear that if A be taken at infinity, its projection D will also lie at infinity; consequently D' will coincide with P, and therefore A' with B. Again, if A be taken coincident with Q, its projection D will coincide with P , and consequently D ', and therefore also A', will pass off to infinity. We have thus two trials, neither of which requires * If the two ranges be called u and vf, and the construction of Art. 85 (left) be referred back to, it will be seen that the auxiliary range u" lies in this case entirely at infinity. If then a pair of corresponding points D and D' have been found, and we wish to find the point E' which corresponds to any other point E of PR (=w), we have only to join D'E, and to draw BE' parallel to D'E to meet PQ (=«,') in £'. 198 PROBLEMS OF THE SECOND DEGREE. [240 any construction ; the pairs which result from them are composed respectively of the point at infinity and R , and of Q and the point at infinity. If the pair given by the third trial be called B ,B' } and if A , A f stand for any pair whatever, we have (Art. 74) QA.BA'szQB.MB', and therefore, if M is a self-corresponding point, QM.RM=QB.RB\ from which the self- corresponding points could be found. But it is better in all cases to go back to the general construction of Art. 206. In this case the three pairs of conjugate points of the two ranges which are given are : B and B' ; the point at infinity and R; Q and the point at infinity. Let then any circle be taken, and a point on its circumference ; from draw the straight lines OB , OB', OR , OQ, and a parallel to QR, and let these cut thecircle again inB 1 , 2?/, R v Q ti and i" respectively*. Join the point of intersection of B x R 1 and B(I with that of B X I and B{Q X ; if the joining line cut the circle in two points M x and N % , the straight lines which join these to will meet QR in the self-corresponding points M and N, and these give the solutions of the problem. 246. Problem. To construct a polygon, whose sides shall pass respectively through given points, and all whose vertices except one shall lie respectively on given straight lines ; and which shall be such that the angle included by the sides which meet in the last vertex is equal to a given angle. Suppose, for example, that it is required to construct a triangle LMN (Fig. 172) whose sides MN ,NL ,LM shall pass through the given points , V, U respectively, and whose vertices M , N shall lie on the given straight lines u , v re- spectively; and which shall be such that the angle MLN is equal to a given angle. Through draw any straight line to cut u in A and v in B; join BV, and through U draw the Fig. 172. straight line TJX making with BV an angle equal to the given one. Let TJX meet u in A' \ the problem would be solved if the point A f coincided with A. If the rays OA, UA' be made to vary simultaneously, they will determine on u two projective ranges ; the solutions of the problem will be found by constructing the self-corre- sponding points of these ranges. * Of these points only I is marked in the figure. 248] PROBLEMS OF THE SECOND DEGREE. 199 247. The following problem is included in the foregoing one : A ray of light emanating from a given point is reflected from n given straight lines in succession ; to determine the original direction which the ray must have, in order that this may make with its direction after the last reflexion a given angle. Let u x , u 2 , ... u n be the given straight lines (Fig. 173). If the ray OA x strike u x at A x , then by the law of reflexion the incident and re- flected rays will make equal angles with u x ; but the incident ray passes through the fixed point ; therefore the reflected ray will always pass through the point O x which is symmetrical to with regard to Wj*. So again, if the ray after one reflexion strikes u 2 at A 2 , it will be reflected according to the same law; consequently the ray after two reflexions will pass through a fixed point 2 which is symmetrical to x with regard to u 2 ; and so on. The paths of the ray before reflexion, and after one, two, ... n reflexions form therefore a polygon OA x A 2 A 3 ... , whose n+ 1 sides pass respectively through n + 1 fixed points , O x , 2 , ... O n , and which is such that n of its vertices lie respectively on n given straight lines u x ,u 2 , ... u n ; while the angle included by the sides which meet in the last vertex is to be equal to a given angle. Thus the problem reduces, as was stated, to that of Art. 246. 248. Pkoblem. To construct a polygon whose vertices shall lie respectively on given straight lines, and whose sides shall subtend given angles at given points respectively. Suppose it required to construct a triangle whose vertices 1,2,3 shall lie on the given straight lines u x , u 2 , u 3 respectively, and whose sides 23 , 31, 12 shall subtend at the given points S x , S 2 , S 3 respec- tively the angles a> x , o> 2 , o> 3 which are given in sign and magnitude (Fig. 174). On u x take any point A ; join AS Z , and make the angle ASJB equal to a> 3 ; let S 3 B cut u 2 in B. Join BS X ; make the angle BS X C equal to a> x , and let S X C cut u 3 in C. Join CS 2 ; make the angle CS 2 A ' equal to © 2 , and let S^' cut u x in A'. The problem would be solved if S 2 A' coincided Fig. 174. * i.e. a, point 0^ such that 00 t is bisected at right angles by u x . 200 PROBLEMS OF THE SECOND DEGREE. [240 with S 2 A. If S 2 A be made to turn about S 2 , the other rays S 3 A , S S B , S t B , Sfi, S 2 C, and S 2 A' will change their positions simul- taneously, and will trace out pencils which are all protectively related. For the ranges traced out by S d A and S a B respectively will be projective (Art. 108) since the angle AS 3 B is constant; the ranges traced out by S Z B and S t B respectively are projective since they are in perspective ; and so on. The solutions of the problem will therefore be given by the self-corresponding rays of the concentric projective pencils which are generated by S 2 A and S 2 A f respectively. In the same manner is solved the more general problem in which the straight lines joining S li S ii ... to the vertices of the polygon are no longer to include given angles, but are to be such that together with pairs of given straight lines meeting in S t , S 2 , . . . respectively they form at each of these points a pencil of four rays having a given anharmonic ratio. If at each of the points the pencil is to be harmonic, and the given straight lines such as to include a right angle, the problem can be enunciated as follows (Art. 60) : To construct a polygon whose vertices shall lie respectively on given straight lines, and whose sides shall subtend at given points angles whose bisectors are given, 249. The same method gives the solution of the problem : To construct a polygon whose sides shall pass respectively through given jwints, and which shall be such that the pairs of adjacent sides divide given segments respectively in given anharmonic ratios*. Particular cases of this problem may be obtained by supposing that each pair of adjacent sides is to intercept on a given straight line a segment given in magnitude and direction ; or a segment which is divided by a given point into two parts having a given ratio to one another t. * That is to say, two adjacent sides are to cut a given straight line, on which are two given points A, JB, in two other points C, D such that the anharmonic ratio (ABCD) may be equal to a given number. f Chasles, Geom. sup., pp. 219-223; andTowNSEND, Modern Geometry (Dublin, 1865), vol. ii. pp. 257-275. CHAPTEE XX. POLE AND POLAR. 250. Let any point 8 be taken in the plane of a conic (Fig. 175), and through it let any number of transversals be drawn to cut the conic in pairs of points A and A\B and B\ C and C\ ... . The tangents a and a' , b and b' } c and c' at these points will, by Arts. 203, 204, intersect in pairs on a fixed straight line s, on which lie also the points of contact of the tan- gents from S to the conic (when the position of 8 is such that tangents can be drawn). Further, the pairs of chords AB'*nd-A'B,AC'm&A' i Fig. 178. Join SA,SB, and find the points A',B' where these cut the conic again respectively (Art. 161, right). The straight line s which joins the point of intersection of AB' and A'B to that of AB and Fig. 179. From the points sq, , sb draw the second tangents a' ,b f respec- tively to the conic (Art. 161, left). The point S in which the diagonals of the quadrangle aba'b' intersect one another will be * Desargues, loc. cit, p. 191. f Poncelet, loc. cit., Art. 195. 206 POLE AND POLAR. [258 A'B' will be the polar of the the pole of the given straight given point (Art. 250 [2] ). line. II. Let the conic be determined II. Let the conic be deter- by five tangents a,6,c,d,e mined by five points A, B,C,D,E (Fig. 180). ( ri g- lSl )- Fig. 180. Through S draw two trans- versals u and v, and construct their poles U and V (as on the right hand side above) ; U V will be the polar of JS (Art. 256). To simplify matters the transversal u may be drawn through the point ab ; if then the second tangent c' be drawn to the conic (Art. 161) from the point uc, U will be the point of intersection of the diagonals of the quadri- lateral acbc'. So too if the transversal v be drawn through the point ac for example, and the second tangent b' be drawn to the conic from the point vb, then V will be the point of intersection of the diagonals of the quadri- lateral abcb'. .. Fig. 181. On s take two points U and V, and construct their polars u and v (as on the left hand side above) ; the point uv will be the pole of s (Art. 256). To simplify matters the point £7 may be taken on the straight line AB ; if then UC be joined, and the second point C in which it meets the conic be constructed, u will be the straight line joining the points of intersection of the pairs of oppo- site sides of the quadrangle ACBC. So too if V be taken on the straight line AC for example, and VB be joined, and its second point of intersection B' with the conic be constructed, then v will be the straight line joining the points of intersection of the pairs of opposite sides of the quadrangle ABCB'. 258, Let E and F (Fig. 182) be a pair of conjugate points 259] POLE AND POLAB. 207 and let G be the pole of EF; then G will be conjugate both to E and to F, so that the three points E, F, G are conjugate to one another two and two. Every side therefore of the triangle EFG is the pole of the opposite vertex, and the three sides are conjugate lines two and two. A triangle such as EFG, in which each vertex is the pole of the opposite side with regard to a given conic is called a self-conjugate or self-polar triangle with regard to the conic. 259. To construct a triangle self-conjugate with regard to a given conic. One vertex E (Fig. 182) maybe taken arbitrarily; construct its Fig. 182. polar, take on this polar any point F, and construct the polar of F. This last will pass through E, since E and F are conjugate points ; if G be the point where it cuts the polar of E, then E and G , F and G, will be pairs of conjugate points ; and therefore EFG is a self- conjugate triangle. In other words : take any point E and draw through it any two transversals to cut the conic in A and D, B and C respectively ; join AC, BD, meeting in F, and AB , CD meeting in G ; then EFG is a self- conjugate triangle. Or again, one side e may be taken arbitrarily, and its pole E con- structed ; if through E any straight line / be drawn, and its pole 208 POLE AND POLAK. [260 (which will lie on e) be constructed and joined to the pole of e by the straight line g, then efg will be a triangle such as is required ; for the straight lines e ,/, g are conjugate two and two. Thus, after having taken the side e arbitrarily, we may proceed as follows : take two points on e and from them draw pairs of tangents a and d, b and c, to the conic ; join the points ac , bd by the straight line/, and the points ab,cd by the straight line g ; then will efg be a self-conjugate triangle. 260. From what has been said above the following property is evident : The diagonal points of the complete quadrangle formed by any four points on a conic are the vertices of a triangle which is self- conjugate with regard to the conic. And the diagonals of the complete quadrilateral formed by any four tangents to a conic are the sides of a triangle which is self-conjugate with regard to the conic *. Or, in other words : The triangle whose vertices are the diagonal points of a complete quadrangle is self-conjugate with regard to any conic circumscribing the quadrangle. And the triangle whose sides are the diagonals of a complete quadrilateral is self-conjugate with regard to any conic inscribed in the quadrilateral. 261. From the properties of the circumscribed quadrilateral and the inscribed quadrangle (Arts. 166 to 172) it follows moreover that : If EFG (Fig. 182) is a triangle self-conjugate with regard to a given conic, and ABC is a triangle inscribed in the conic, such that two of its sides CA, AB pass through two of the vertices F, G respectively of the other triangle, then will the re- maining side BC pass through the remaining vertex E, and every side of the inscribed triangle will be divided harmonically 'by the corresponding vertex of the self- conjugate triangle and the side which joins the other two vertices of it. The three straight lines EA,FB , GC meet in one point D on the conic ; the two triangles are therefore in perspective, and the three pairs of corresponding sides FG and BC, GE and CA, EF and AB, will meet in three collinear points. Hence it follows that a self-conjugate triangle EFG and a point A of a conic determine an inscribed quadrangle ABCD, whose diagonal * Desargues, he. cit., p. 186. 263] POLE AND POLAR. 209 triangle is EFG. The points B ,C ,D are those in which the straight lines AG, AF, AE cut the conic again. The enunciation of the correlative property is left to the student *. 262. Of the three vertices of the triangle EFG, one always lies inside the conic, and the two others outside it. For if E is an internal point, its polar does not cut the conic, and con- sequently F and G are both external to the conic. If, on the other hand, E is an external point, its polar cuts the conic, and F and G are harmonic conjugates with regard to the two points of intersection ; of the two points F and G therefore, one must be internal and the other external to the conic. From this property and that of Art. 254, 1, we conclude that of the three sides of any self-conjugate triangle, two always cut the curve, and the third does not. 263. (1). On every straight line there are an infinite number of pairs of points which are conjugate to one another with respect to a given conic, and these form an involution f. (2). Through every point pass an infinite number of pairs of straight lines which are conjugate to one another with respect to a given conic, and these form an involution f. (3). If a point describes a range, its polar with respect to a given conic will trace out a pencil which is projective with the given range. And, conversely, if a straight line describes a pencil, its pole with respect to a given conic will trace out a range which is projective with the given pencil %. To prove these theorems, consider Fig. 183, and suppose in it the conic and the three points A,B ,G to be given. Let the point C be supposed to move along the conic. Then the rays AC, EC will trace out two pencils which are projective with one another (Art. 149 [l]) ; and therefore the ranges in which these pencils cut the polar of G will be pro- jective also ; that is to say, the conjugate points i^and E will describe two collinear projective ranges. In these ranges the points F and E correspond to one another doubly, since the polar of E passes through F, and the polar of F passes through E ; consequently the ranges in question are in in- volution. From what has been said it follows also that the pairs of * Poncelet, loe. cit., p. 104. + Desargues, loc. cit., pp. 192, 193. X Mobius, Baryc. Cole, § 290. P 210 POLE AND POLAR. [264 conjugate lines GF, GE in like manner form an involution, and that the range of poles E,F,... is projective with the pencil of polars GF, GE,.... 264. If the straight line incuts the conic, the two points of Fig. 183. intersection are the double points of the involution formed by the pairs of conjugate poles. The centre of the involution lies on the diameter which passes through the pole G of the given straight line (Art. 290). If the point G is external to the conic, the tangents from G to the conic are the double rays of the involution formed by the pairs of conjugate polars. Consequently (Art. 125): A chord of a conic is harmonically divided by any pair of points lying on it which are conjugate with respect to the conic ; and The pair of tangents drawn from any point to a conic are har- monic conjugates with respect to any pair of straight lines meeting in the given point which are conjugate with respect to the conic. If the point G lies at infinity, the pairs of conjugate straight lines form an involution of parallel rays, the central ray of which is a diameter of the conic (Arts. 129, 276). 265. Theorem. If tivo complete quadrangles have the same diagonal points, their eight vertices lie either four and four on two straight lines or else they all lie on a conic. 265 POLE AND POLAR. 211 Fig. 184. Let ABCD and A'B 'CD' (Fig. 184) be two quadrangles having the same diagonal points E ,F,G; so that BC,AD, B'C, A'D' all meet in E, CA , BB , C'A' , B'D' „ „ F, AB, CD, A'B', CD' „ „ G. (1). In the first place let the eight vertices be such that some three of them are collinear. Suppose for example that A ' lies on AB. Since AB and A'B' meet in G, therefore B' also must lie on AB; and since the straight lines GE , GF are har- monically conjugate with regard both to AB, CD and to A'B' ,CD', and AB coincides with A'B', therefore also CB coincides with CD'. Thus the four points C,D,C,D' are collinear, and the eight points A,B,C,D, A',B',C,D' lie four and four on two straight lines. (2). But if this case be excluded, *. e. if no three of the eight vertices He in a straight line, then a conic can be drawn through any five of them. Let a conic be drawn through A,B,C,D,A' (Fig. 185); then shall B',C,D' lie on the same conic. For since E,F,G are the diagonal points of the inscribed quadrangle ABCD, G is the pole of EF, and therefore G and the point where its polar EF meets the transversal GB'A' are harmonically conjugate with regard to the points where this transversal cuts the conic. But one of these last points is A', therefore the other is B' '; for since E,F,G are also the diagonal points of the quadrangle A'B' CI/, the points A' and B' are harmonically conjugate with regard to G and the point where EF cuts A'B'. In a similar manner it can be shown that C and D' also lie on the same conic. The eight vertices A,B,C ,D,A',B',C,D' therefore lie on a conic, and the proposition is proved. p % Fig. 185. 212 POLE AND POLAK. [266 Since the straight lines AB and A'B' meet in G, therefore A A' and BB', as also AB' and A'B, will meet on EF, the polar of G. This property gives the means of constructing the point B f when the points A, B,C ,B,A' are given. The point C will then be found as the point of intersection of A'F and B'E, and the point B' as that of B'F, A'F, and C'G. 266. Suppose now that two conies are given which are inscribed in the same quadrilateral. Let the four common tangents which form this quadrilateral be a, b, c, d, and let their points of contact with the conies heA,B,C,B and A', B',C',B f respectively. By the theorem of Art. 169, the triangle formed by the diagonals of the circumscribed quadrilateral abed has for its vertices the diagonal points of the inscribed quadrangle ABCB and also those of the inscribed quadrangle A'B'C'B'; thus ABCB and A'B'C'B' have the same diagonal points. Accordingly, by the theorem of Art. 265, the eight points A,B, C , B , A',B', C',B' lie either four and four on two straight lines, or they lie all on a conic. 267. By writing, as usual, line for point, and point for line, the propositions correlative to those of Arts. 265 and 266 can be proved, viz. If two complete quadrilaterals have the same three diagonals, their eight sides either pass four and four through two points, or else they all touch a conic. If two conies intersect in four points, the eight tangents to them at these points either pass four and four through two points, or they all touch a conic *. 268. If there be given the diagonal points E, F, G and one vertex A of a quadrangle ABCB, the quadrangle is completely determined, and can be constructed. For B is that point on AE which is harmonically conjugate to A with respect to E and the point where FG cuts AE ; so C is that point on AF which is harmonically conjugate to A with respect to jF'and the point where GE cuts AF; and B is that point on AG which is harmonically conjugate to A with respect to G and the point where incuts AG. But if there be given the diagonal points E, F, G of a quadrangle ABCB and the conic with respect to which EFG is a self-conjugate triangle, the quadrangle is not completely * Staudt, loc. cit., p. 293. 270] POLE AND POLAR. 213 determined. For we may take arbitrarily on the conic a point A as one vertex of the quadrangle ABCB ; then the other vertices B, C, B are the second points of intersection of the conic with the straight lines AG, AF, AE respectively. Hence it follows that : All conies with respect to which a given triangle EFG is self- conjugate, and which pass through a fixed point A, pass also through three other fixed points B ,C ,D. 269. Peoblem. To construct a conic passing through two given points A and A', and with respect to which a given triangle EFG shall be self -conjugate. Solution. Construct, in the manner just shown, the three points B, C, D which form with A a complete quadrangle having E , F, and G for its diagonal points. Five points A, A', B , C, D on the conic are then known, and by means of Pascal's theorem any number of other points on it may be found. Or we may construct the three points B', C / ,D / which form with A' a complete quadrangle having E, F, and G for its diagonal points ; the eight points A, B,C, D , A', B' ', C', D f will then all lie on the conic required. 270. Consider again the problem (Art. 218) of describing a conic to touch four given straight lines a,b,c,d and to pass through a given point S (Fig. 186). The diagonals of the quadrilateral abed form a triangle EFG which is self- conju- gate with regard to the conic; consequently, if the three points P , Q , R be constructed which together with S form a quadrangle having 2? , F, and G for its diagonal points, the three points so con- structed will lie also on the required conic. Now it may happen that there is no conic which satisfies the problem, or again there may be two conies which satisfy it (Art. 218, right); in the second case, since the construction for the points P, Q, R is linear, the two conies will both pass through these points. Thus : If two conies inscribed in the same quadrilateral abed %)ass through the same point S, they will intersect in three other points P, Q, R ; and the triangle formed by the diagonals of tlie circumscribed quadrilateral abed will coincide with that formed by the diagonal points of the inscribed quadrangle PQRS. In order to find a construction for the points P, Q, R, consider Fig. 1 86. 214 POLE AND POLAR. [271 the point P for example which lies on ES (Fig. 186). It is seen that the segment SP must be divided harmonically by E and its polar FG (Art. 250) ; but the diagonal (ab) (cd) which passes through E is also divided harmonically, at E and F. We have therefore two harmonic ranges, which are of course projective (Art. 51) and which are in perspective since they have a self-corresponding point at E ; therefore the straight lines P (ab), S (cd), and FG, which join the other pairs of corresponding points, will meet in a point (Art. 80). We must therefore join S to one extremity of one of the diagonals passing through E, for example to the point cd, and take the point where the joining line meets FG. This point, when joined to the other extremity ab of the diagonal, will give a straight line which will meet ES in the required point P*. 271. The propositions and constructions correlative to those of the last three Articles, and which will form useful exercises for the student, are the following : All conies with respect to which a given triangle is self-conjugate, and which touch a fixed straight line, touch three other fixed straight lines. To construct a conic to touch two given straight lines, and with respect to which a given triangle shall be self-conjugate. If two conies circumscribing the same quadrangle have a common tangent, they have three other common tangents. To construct the three remaining common tangents to two conies which pass through four given points and touch a given straight line (Art. 218, left). Fig. 187. 272. Let ABCD (Fig. 187) be a complete quadrilateral whose diagonal points are E, F, and G. Let also L and P be the points where FG meets AD and BG respectively. J/" and Q „ „ GE „ ED and CA -tfandP „ „ EF „ GD and AB The six points so obtained are the vertices of a complete quad- rilateral. For the triangle EFG is in perspective with each of the * Bbianchon, loc. cit., p. 45 ; Maclaurin, De tin. Georn., § 43. 274] POLE AND POLAR. 215 triangles ABC, DCB, CD A, BAD, the centres of perspective being D, A, B, C respectively ; whence it follows that the four triads of points PQR, PMN, LQN, and LMR lie on four straight lines (the axes of perspective). These four axes form a quadrilateral whose diagonals LP, MQ, NR form the triangle EFG. Accordingly, a conic inscribed in the quadrangle ABCD and passing through L will pass also through N , P, and R (Art. 270) ; similarly a conic can be inscribed in the quadrangle ABDC to pass through R, M, N, and Q', and a conic can be inscribed in the quadrangle ACBD to pass through Q, P, M, and L. It will be seen that for each of these conies the four tangents shown in the figure (the four sides of the complete quadrangle ABCD) are harmonic, and that the same will therefore be the case with regard to their points of contact (Arts. 148, 204). For take one of the sides of the quadrangle, for example AB ; a consideration of the complete quadrangle CDEF shows that this side is harmonically divided in R and G. Now the points A, B, G are the points of intersection of the tangent AB with the other three tangents, and R is the point of contact of AB ; therefore the four tangents are cut by any other tangent to the conic in four harmonic points *. 273. If ABCD is a parallelogram, the points E, G, M, Q pass off to infinity, and LNPR also becomes a parallelogram. Of the three conies considered above the first will in this case be an ellipse which touches the sides of the parallelogram ABCD at their middle points ; the second a hyperbola which touches the sides AB and CD at their middle points and has AC and BD for asymptotes ; and the third a hyperbola having the same asymptotes and touching the sides AD and BC at their middle points. 274. From that corollary to Brianchon's theorem which has reference to a quadrilateral circumscribed about a conic (Art. 172) we have already, in Art. 173, deduced a method for the construction of tangents to a conic when we are given three tangents a , b , c and the points of contact B , C of two of them (Fig. 183). We take any point E on BC and join it to the points ab,ac by the straight lines g, f, respectively; if the point in which g meets c be joined to that in which /meets b, the joining line d will be a tangent to the conic. The four tangents a , b , , d form a complete quadrilateral two of whose diagonals # = (ah) (cd) and/= (ac) (bd) intersect * Steiner, loc. cit., p. 160, § 43, 4; Collected Works, vol. i. p. 347; Staudt, Beitrage zur Geometrie der Laye, Art. 329. 216 POLE AND POLAK. [275 in E\ therefore also (Art. 172) the chords of contact AD and BC of the tangents a and d, b and c respectively will intersect in E. The straight lines joining E to the points ab and ac, being two of the diagonals of the quadrilateral abed, are con- jugate lines with respect td the conic ; consequently : If a triangle abe is circumscribed about a conic, the straight lines which join two of its vertices ab and ac to any point E on the polar of the third vertex be are conjugate to one another with respect to the conic. And conversely : If two straight lines (c and b) touch a conic, any two conjugate straight lines (fandg) drawn from any point (E) on their chord of contact will cut the two given tangents in points such that the straight line (a) joining them touches the conic. 275. Let us now investigate the correlative property. Sup- pose three points A , B , C on a conic to be given, and the tangents b , c at two of these points (Fig. 183). If a straight line e drawn arbitrarily through the point be cut AB in G and AC in F; then if GC and FB be joined they will intersect in a point D lying on the conic. The four points A,B,B,C form a complete quadrangle two of whose diagonal points lie on e\ therefore (Art. 166) the point be and the point of intersection of the tangents at A and B will lie on e. The points G and F, being two of the diagonal points of the quadrila- teral ABCB, are conjugate with respect to the conic; consequently Fi l8g If a triangle ABC (Fig. 188) is inscribed in a conic, the points F and G in which two of the sides are cut by any straight line drawn through the pole S of the third side are conjugate to one another with respect to the conic. And conversely : If two given points (B , C) on a conic be joined to two conjugate points (G ,F) which are collinear with the pole (S) of the chord (BC) joining the given points, then the joining lines will intersect in a point (A) lying on the conic. CHAPTEE XXL THE CENTRE AND DIAMETERS OF A CONIC. Fig. 189. 276. Let an infinitely distant point be taken as pole, and through it let a transversal be drawn (Fig. 1 89) to cut the conic in two points A and A'. The segment A A' will be harmonically divided by the pole and the point where it is cut by the polar (Art. 250); this point will there- fore be the middle point of AA' (Art. 59). That is to say: If any number of parallel chords of a conic be drawn, the locus of their middle points is a straight line ; and this straight line is the polar of the point at infinity in which the chords intersect *. 277. This straight line is termed the diameter of the chords which it bisects. If the diameter meets the conic in two points, these will be the points of contact of the tangents drawn to the conic from the pole, i. e. of those tangents which are parallel to the bisected chords. If the tangents at the ex- tremities A and A r of one of these chords be drawn, they will meet in a point on the diameter. If A A ' and BB r are two of the bisected chords, the straight lines AB and A f B\ AB / and A'B will intersect in pairs on the diameter (Art. 250). If, conversely, from a point on the diameter can be drawn a pair of tangents a and a' to the conic, their chord of contact A A' will be bisected by the diameter; and if through the same point there be drawn the straight line which is harmoni- cally conjugate to the diameter with respect to the two * Apollonius, Conic., lib. i. 46, 47, 48 ; lib. ii. 5, 6, 7, 28-31, 34-37. 218 THE CENTRE AND DIAMETERS OP A CONIC. [278 tangents, this straight line will be parallel to the bisected chords. If from two points on the diameter there be drawn two pairs of tangents a and a\ b and b', the straight line join- ing the points ab and a'U and that joining the points ab K and a'b will both be parallel to the bisected chords (Art. 252). 278. To each point at infinity, that is, to each pencil of parallel rays, corresponds a diameter. The diameters all pass through one point ; for they are the polars of points lying on one straight line, viz. the straight line at infinity; the point in which the diameters intersect is the pole of the straight line at infinity (Art. 256). 279. Since every parabola is touched by the straight line at infinity, and the point of contact is the pole of this straight line (Art. 254, II), it follows (Art. 278) that all diameters of a parabola are parallel to one another (they all pass through the point at infinity on the curve) ; and conversely, every straight line which cuts a parabola at infinity is a diameter of the curve. 280. If 8 is any point from which a pair of tangents a and a' can be drawn to the conic (Fig. 189), the chord of contact AA', the polar of 8, will be bisected at B by the diameter which passes through 8 ; for 8 and the point at infinity on AA' are conjugate points with respect to the conic. If the diameter cuts the curve in M and M' ? the tangents at these points are parallel to AA', and MM' is divided harmonically by the pole 8 and the polar AA' (Art. 250). If then the conic is a parabola (Fig. 190) the point 31' moves off to infinity, and therefore M is the middle point of the segment SR ;' thus The straight line which joins the middle point of a chord of a parabola to the pole of the chord is bisected by the curve *. 281. When the conic is not a parabola, Fi x the straight line at infinity is no longer a tangent to the curve, and consequently the pole of this straight line, or the point of intersection of the diameters, is a point lying at a finite distance. Since any two points on the conic which are collinear with the pole are separated harmonically by the pole and the polar (Art. 250), the pole will lie midway between the two points on the curve * Apollonius, loc. cit., lib. i. 35. 284] THE CENTRE AND DIAMETERS OF A CONIC. when the polar lies at infinity. Every chord of the conic therefore which passes through the pole of the straight line at infinity is bisected at this point. On account of this property the pole of the straight line at infinity or the point in which all the diameters intersect is called the centre of the conic. 282. Applying the properties of poles and polars in general (Arts. 250—253) to the case of the centre and the straight line at in- finity, it is seen (Fig. 191) that : If A and A' are any pair of points on the conic collinear with the centre, the tangents at A and A'~ are parallel. rig -g If A and A',B and B' are any two pairs of points on the conic which are collinear with the centre, the pairs of chords AB and A'B', AB' and A' B are parallel, so that the figure ABA'B' is a parallelogram. If a and V are any pair of parallel tangents, their chord of contact passes through the centre, as also does the straight line lying midway between a and a' and parallel to both. If a and a', b and V are any two pairs of parallel tangents, the straight line joining the points ab and a'b' and that joining the points ab' and a'b both pass through the centre ; in other words, if aba'b' is a parallelogram circumscribed to the conic, its diagonals intersect in the centre. 283. If the conic is a hyperbola, the straight line at in- finity cuts the curve ; consequently the centre is a point exterior to the curve (Art. 254, I) in which intersect the tan- gents at the infinitely distant points, i.e. the asymptotes (Fig. 197). If the conic is an ellipse, the straight line at infinity does not cut the curve ; consequently the centre is a point inside the curve (Figs. 191, 192). 284. Two diameters of a central conic (ellipse or hyper- bola*) are termed conjugate when they are conjugate straight * In the case of the parabola there are no pairs of conjugate diameters ; for since the centre lies at infinity, the diameter drawn parallel to the chords which are bisected by a given diameter must coincide always with the straight line at infinity. 220 THE CENTRE AND DIAMETERS OP A CONIC. [285 lines with respect to the conic, i. e. when each passes through the pole of the other (Art. 255). Since the pole of a diameter is the point at infinity on any of the chords which the diameter bisects, it follows that the diameter V conjugate to a given diameter b is parallel to the chords bisected by b ; conversely, V bisects the chords which are parallel to b*. Fig. 192. * Any two conjugate diameters form with the straight line at infinity a self-conjugate triangle (Art. 258), of which one vertex is the centre of the conic and the other two are at infinity. Since in a self-conjugate triangle two of the sides cut the conic and the third side does not (Art. 262), and since the straight line at infinity cuts a hyperbola but does not cut an ellipse, it follows that of every two conjugate diameters of a hyperbola one only cuts the curve, while an ellipse is cut by all its diameters. 285. Pkoblbm. Given jive points A ,B ,C ,D ,E on a conic, to determine its centre. Solution. We have only to repeat the construction given in Art. 257 5 II (right), assuming the straight line s to lie in this case at infinity. Draw through G a parallel to AB, and determine the point C in which this parallel meets the conic again ; draw also through B a parallel to AC, and determine the point B f in which this parallel meets the conic again. The straight line u which joins the points of intersection of the pairs of opposite sides of the quadrangle ACBG\ and the straight line v which joins the points of intersection of the pairs of opposite sides of the quadrangle ABGB' , will meet in the required point 0, which is the pole of the straight line at infinity and therefore the centre of the conic t. The straight lines u and v are the diameters conjugate respec- tively to AB and AG', if through there be drawn the straight lines v! , v' parallel to AB , AG respectively, then u and u', v and v' will be two pairs of conjugate diameters. If the conic is determined by five tangents, its centre may be found by a method which will be explained further on (Art. 319). * Apollonius, loc. cit., lib. ii. 20. t If w and v should be parallel, the conic is a parabola, whose diameters are parallel to u and v. 288] THE CENTRE AND DIAMETERS OF A CONIC. 221 286. Four tangents to a conic form a complete quadrilateral whose diagonals are the sides of a self-conjugate triangle (Art. 260). Suppose the four tangents to be parallel in pairs (Fig. 191); then one diagonal will pass to infinity, and con- sequently the other two will be conjugate diameters (Art. 284); thus: The diagonals of any parallelogram circumscribed to a conic are conjugate diameters. The points of contact of the four tangents form a complete quadrangle whose diagonal points are the vertices of the self- conjugate triangle (Arts. 169, 260). In the case where the four tangents are parallel in pairs one of these diagonal points is the centre of the conic, and the other two lie at infinity. That is to say, the six sides of the quadrangle are the sides and diagonals of an inscribed parallelogram ; its sides are parallel in pairs to the diagonals of the circumscribed paral- lelogram, and its diagonals intersect in the centre of the conic. 287. Conversely, let ABA!B' (Fig. 191) be any inscribed parallelogram, and consider it as a complete quadrangle. Since its three diagonal points must be the vertices of a self- conjugate triangle, one of them, will be the centre of the conic, and the other two will be the points at infinity on two conju- gate diameters ; thus: In any parallelogram inscribed in a conic, the sides are parallel to two conjugate diameters and the diagonals intersect in the centre. Or again : The chords which join a variable point A on a conic to the ex- tremities B and B f of a fixed diameter are always parallel to two conjugate diameters. 288. The following conclusions can be drawn at once, from Art. 286. Any two parallel tangents (a and a') are cut by any pair of conjugate diameters in two pairs of points, the straight lines connecting which give two other parallel tangents (b and b'). If from the extremities (A and A') of any diameter straight lines be drawn parallel to any two conjugate diameters, they will meet in two points on the curve, and the chord joining these will be a diameter. Given any two parallel tangents a and a' whose points of 222 THE CENTRE AND DIAMETERS OF A CONIC. [289 contact are A and A' respectively, and any third tangent b ; if from A a parallel be drawn to the diameter passing through a'b this parallel will meet the tangent b at its point of con- tact B. Given any two parallel tangents a and a' whose points of contact are A and A' respectively, and another point B on the conic ; the tangent at B will meet the tangent a in a point lying on that diameter which is parallel to A'B, and it will meet the tangent a' in a point lying on that diameter which is parallel to AB. 289. Suppose now that the conic is a circle (Fig. 193), i.e. the locus of the vertex of a right angle AMB whose arms AM and BM turn round fixed points A and B respectively. These arms in moving generate two equal and consequently projective pencils ; therefore the tangent at A will be the ray of the first pencil which corresponds to the ray Flg ' m ' BA of the second (Art. 143). The tangent at A must therefore make a right angle with BA ; and simi- larly the tangent at B will be perpendicular to AB. The tangents at A and B are therefore parallel, and consequently AB is a diameter, and the middle point of AB is the centre of the circle (Art. 282). I. Since AB is a diameter, the straight lines AM and BM will be parallel to a pair of conjugate diameters, whatever be the position of M (Art. 287) ; therefore : Every pair of conjugate diameters of a circle are at right angles to one another. II. Since the diagonals of any parallelogram circumscribed about the circle are conjugate diameters, they will inter- sect at right angles ; thus any parallelogram which circumscribes a circle must be a rhombus. III. In a rhombus, the distance between one pair of opposite sides is equal to the distance between the other pair ; thus by allowing one pair of opposite sides of the circumscribed rhom- bus to vary while the other pair remain fixed, we see that the distance between two parallel tangents is constant. This distance is the length of the straight line joining the points of contact of the tangents, for this straight line, which is a 290] THE CENTKE AND DIAMETERS OF A CONIC. 223 diameter, cuts at right angles the conjugate diameter and the tangents parallel to it ; therefore all diameters of a circle are equal in length. IV. The diagonals of any inscribed parallelogram are diameters ; but all diameters are equal in length ; therefore any parallelogram inscribed in a circle must be a rectangle. 290. Returning to the general case where the conic is any whatever (Fig. 189), let s be any straight line and 8 its pole. All chords parallel to s will be bisected by the diameter passing through 8 ; for since 8 and the point at infinity on s are conjugate points with respect to the conic, the polar of the second point will pass through the first. We may also say that : If a diameter pass through a fixed point, the conjugate diameter will be parallel to the polar of this point. I. If the diameter passing through 8 cuts the conic in two points M and M', then MM' is divided harmonically by the pole 8 and the polar s*\ thus if is the middle point of MM', that is, the centre of the conic, and R the point where MM' is cut by the polar s, we have (Art. 69) OS.OR= OM 2 . II. From this follows a construction for the semi-diameter conjugate to a chord AA' of a conic, having given the extremities A and A' of the chord and three other points on the conic. We determine (Art. 285) the centre 0, and join it to the middle point R of AA' ; we then construct the tangent at A and take its point of intersection S with OR. If now a point M be taken on OR such that OM is the mean propor- tional between OR and OS, then OM will be the required semi-diameter. If lie between R and S, so that OR and OS have opposite signs, the diameter OR will not cut the conic ; but in this case also the length OM, the mean proportional between OR and OS, is called the magnitude of the semidiameter conjugate to the chord AA'. An analogous definition can be given for the case of any straight line (Art. 294). III. If the conic is a circle, the perpendicularity of the conjugate diameters in this case gives the theorem : * Apollonius, loc. cit„ i. 34, 36 ; ii. 29, 30. 224 THE CENTRE AND DIAMETERS OF A CONIC. [291 The polar of any point with respect to a circle is perpendicular to the diameter which passes through the pole. 291. From this last property can be derived a second de- monstration of the very important theorem of Art. 263 (3), viz. The range formed by any number of collinear points, and the pencil formed by their polars with respect to any given conic, are two projective forms. Consider as poles the points A,B ,C ,... lying on a straight line, s (Fig. 194); the diameters (A,B,C ,...) obtained by joining them to the centre of the conic will form a pencil which is in perspective with the range A,B ,C , Another pencil will be formed by the polars a,b,c,... of the points A, B, C , ... since these polars all pass through Fi a point S (Art. 256), the pole of s. If now the conic is a circle, then by the property proved in Art. 290, III, the straight lines 0(A,B,C,...) are perpendicular respectively to a, b,c , . . . ; and the two pencils are in this case equal. The range of poles A,B,C,... is therefore projective with the pencil of polars a , b , c , . . . with regard to a circle. This result may now be extended and shown to hold not only for a circle but for any conic. For any given conic may be regarded as the projection of a circle (Arts. 149, 150). In the projection, to harmonic forms correspond harmonic forms (Art. 51); consequently to a point and its polar with regard to the conic will correspond a point and its polar with regard to the circle, and to a range of poles and the pencil formed by their polars with regard to the conic will correspond a range of poles and the pencil formed by their polars with regard to the circle. But it has been seen that this range and pencil are projective in the case of the circle ; therefore the same is true with regard to the range and pencil in the case of the conic, and the theorem is proved. 292. Theoeem. A quadrangle is inscribed in a conic, and a point is taken on the straight line which joins the points of intersection of the pairs of opposite sides. If from, this point be drawn tlie straight lines connecting it with the two pairs of opposite vertices, and also a pair of 292] THE CENTRE AND DIAMETERS OP A CONIC. 225 Fig. 195. tangents to the conic , these straight lines will be three conjugate pairs of an involution. Let ABCD be a simple quadrangle inscribed in a conic (Fig. 195); let the diagonals AC , BD meet in F, and the pairs of opposite sides BC ,AD and AB ,CD in E and G respectively ; the points E,F,G will then be conjugate two and two with respect to the conic (Art. 259). Take any point / on EG and join it to the vertices of the quadrangle, and draw also the tangents IP, IQ to the conic. The two tangents are harmonically separated by IE, IF (Art. 264), since these are conjugate straight lines, F being the pole of IE. But the rays IE , IF are harmonically conjugate also with regard to I A , IC ; for the diagonal A C of the complete quadrilateral formed by AB , BC , CD, and DA is divided harmonically by the other two diagonals BD and EG, and the two pairs of rays in question are formed by joining / to the four harmonic points on AC. For a similar reason the rays IE , IF are harmonically conjugate with regard to IB , ID. The pair of tangents, the rays I A , IC, and the rays IB , ID are therefore three conjugate pairs of an involution, of which IE, IF are the double rays (Art. 125). I. By virtue of the theorem correlative to that of Desargues (Art. 183, right), a conic can be inscribed in the quadrilateral ABCD so as to touch the straight lines IF and IQ. II. The theorem correlative to the one proved above may be thus enunciated : If a simple quadrilateral ABCD (Fig. 196) is circumscribed about a conic, and if through the point of intersection of its diagonals any transversal be drawn, this will cut the conic and the pairs of opposite sides AB and CD, BC and AD, in three pairs of conjugate points of an invo- lution. III. By virtue of Desargues' theorem (Art. 183, left), a conic can be described to pass through the four vertices of the quadrilateral and through the two points where the conic is cut by the transversal *. * Chasles, Sections coniyues, Arts. 122, 126. Q Fig. 196. 226 THE CENTRE AND DIAMETERS OF A CONIC. [293 293. The theory of conjugate points with regard to a conic gives a solution of the problem : To construct the points of intersection of a given straight line s with a conic which is determined by five points or by five tangents. Take on s any two points XI and V, construct their polars u and v (Art. 257), and let U' and V be the points where these meet s. If the involution determined by the two pairs of reciprocal points U and TJ\ V and V, has two double points M and N, these will be the required points of intersection of the conic with s. If V and V' should coincide, the conic touches s at the point in which they coincide. If the involution has no double points, the conic does not cut s *. By a correlative method may be solved the problem : to draw from a given point S a pair of tangents to a conic which is determined by five tangents or by five points. 294'. Let A and A' be a pair of points lying on a straight line s which are conjugate with respect to the conic, and let be the point where s meets the diameter passing through its pole S (the diameter bisecting chords parallel to s). Then will be the centre of the involution formed on s by the pairs of conjugate points such as A and A% and therefore (Art. 125) OA . OA f = constant. If s cuts the conic in two points M and if, these will be the double points of the involution, and OA.OA'=OM 2 =01V 2 . If 8 does not cut the conic, the constant value of OA . OA' will be negative (Art. 125); in this case there exists a pair H and H' of conjugate points of the involution, or of conjugate points with regard to the conic, such that lies midway between them, and OA . OA' = OH . OH' =-OH*=- OH' 2 . The segment HH' has been called an ideal chord f of the conic, just as MJ¥ in the first case is a real chord. Accepting this defini- tion we may say that a diameter contains the middle points of all chords, real and ideal, which are parallel to the conjugate diameter. When two conies are said to have a real common chord MN, it is meant that they both pass through the points M and N. When two conies are said to have an ideal common chord HH', this signifies that II and H ' are conjugate points with regard to both conies, and that the diameters of the two conies which pass through the respective poles of HH' both pass through the middle point of IIII'. * Staudt, Geomelrie der Lage, Art. 305. t PONCELET, IOC. Clt., p. 29. 297] THE CENTRE AND DIAMETERS OF A CONIC. 227 295. A pencil of rays in involution has in general (Art. 207) one pair of conjugate rays which include a right angle. Therefore Through a given point can always be drawn one pair of straight lines which are conjugate with respect to a given conic and which include a right angle ; and these are the internal and external bisec- tors of the angle made with one another by the tangents drawn from the given point, when this is exterior to the conic. 296. In Art. 263 (Fig. 183) let the point G be taken to co- incide with the centre of the conic (hyperbola or ellipse) ; two conjugate lines such as GF, GF will then become conju- gate diameters, and we see that the pairs of conjugate diameters of a conic form an involution. If the conic is a hyperbola, the asymptotes are the double rays of the involution (Arts. 264, 283); thus any two conjugate diameters of a hyperbola are har- monically conjugate with regard to the asymptotes*. If the conic is an ellipse, the involution has no double rays. Consider two pairs of conjugate elements of an involution ; the one pair either overlaps or does not overlap the other, and according as the first or the second is the case, the involution has not, or it has, double points (Art. 128) ; thus: Of any two pairs of conjugate diameters of an ellipse, the one aa' is always separated by the other W (Fig. 192) ; Of any two pairs of conjugate diameters of a hyperbola, the one aa' is never separated by the other bb f (Fig. 197). 297. The involution of conjugate diameters will have one pair of con- jugate diameters including a right angle (Art. 295). If there were a second such pair, every diameter would be perpendicular to its con- jugate (Art. 207), and in that case the angle subtended at any point F] * on the curve by a fixed diameter would be a right angle (Art. 287), and consequently the conic would be a circle. Every conic therefore which is not a para- bola or a circle has a single pair of conjugate diameters which are at right angles to one another. These two diameters a and a' are called the axes of the conic (Figs. 192, 197). In the * De la Hike, loc. cit., book ii. prop. 13, Cor. 4. Q2 228 THE CENTRE AND DIAMETERS OF A CONIC. [298 hyperbola (Fig. 197) the axes are the bisectors of the angle between the asymptotes m and n (Arts. 296, 60). In the ellipse both axes cut the curve (Art. 284); the greater (a f ) is called the major, the smaller (a) the minor axis. In the hyperbola only one of the axes cuts the curve ; this one (a') is called the transverse axis, the other (a) the conjugate axis. The points in which the conic is cut by the axis a' in either case are called the vertices. Regarding an axis as a diameter which bisects all chords perpendicular to itself, it is seen that the parabola also has an axis. For since all chords at right angles to the common direction of the diameters are parallel to one another, their middle points lie on one straight line, which is the axis a of the parabola (Fig. 190). The parabola has one vertex at infinity ; the other, the finite point in which the axis a cuts the curve, is generally called the vertex of the parabola. 298. Since each of the orthogonal conjugate diameters of a central conic (ellipse or hyperbola) bisects all chords perpen- dicular to itself, it follows that the conic is symmetrical with re- spect to each of the diameters in question (Art. 76). The ellipse and the hyperbola have therefore each two axes of symmetry ; the parabola, on the other hand, has only one such axis. The ellipse and hyperbola are also symmetrical with respect to a point ; the centre of symmetry being in each case the pole of the straight line at infinity. In general, given a conic, a point 8, and s the polar of S with / , respect to the conic; if S be r*zz™-~/ft~~-~zsdx, taken as centre and s as axis of / m^^tOv^/ harmonic homology (Art. 76), the x rJ^§^p^™%7 conic is homological with itself / yS L 299. In the theorem of Art. I/' 275 suppose the inscribed triangle ^ Fig I98 to be AA X M (Fig. 198); that is, let two of its vertices A and A x be collinear with the centre of the conic, which is taken to be an ellipse or hyperbola. The pole of the side AA X will be the point at infinity common to the chords bisected by the diameter AA X , and the theorem will become the following : * See also Art. 396, below. 302] THE CENTRE AND DIAMETERS OF A CONIC. 229 The straight lines which join two conjugate points P and P' to the extremities A and A x of that diameter whose conjugate is parallel to FT?' intersect on the conic. 300. The pairs of conjugate points taken, similarly to P and P\ on the diameter conjugate to AA X form an involution (Art. 263) whose centre is the centre of the conic. If this involution has two double points B and B lt these lie on the curve, which is therefore an ellipse. If the involution has no double points, the conic is a hyperbola (Art. 284) ; in this case two points B and B 1 can be found which are conjugate in the involution and consequently conjugate with respect to the conic, and which lie at equal distances on opposite sides of (Art. 125). In both cases the length of the diameter conjugate to AA ± is interpreted as being the segment BB 1 (Arts. 290, 294). In the ellipse we have (Art. 294) OP . OF = constant = OB 2 = OB •, and in the hyperbola OP. 0P'= constant = OS. 0# 1= - 0B 2 = - OB 2 . 301. The foregoing theorem enables us to solve the problem : To construct by points a conic, having given a fair of conjugate diameters AA X and BB X in magnitude and fwsition. Fig. 199. In the case of the ellipse (Fig. 198) the four points A,A l ,B,B 1 all lie on the curve ; in the case of the hyperbola (Fig. 199) let AA X be that one of the two given diameters which meets the conic. Construct on the diameter BB X several pairs of conjugate points P and P f of the involution determined by having as centre and B and B x in the first case as double points, in the second case as conjugate points. The straight lines AP and A V P' (as also A X P and AP') will intersect on the curve. 302. The straight lines OX, OX' drawn parallel to AP,A Y P f respectively are a pair of conjugate diameters (Art. 287). The 230 THE CENTRE AND DIAMETERS OF A CONIC. [303 pairs of conjugate diameters form an involution (Art. 296) ; consequently the pairs of points analogous to X, X' (in which the diameters cut the tangent at A) also form an involution, the centre of which is A, since OA and the diameter OB parallel to AX are a pair of conjugate diameters. If the conic is a hyperbola, the involution of conjugate diameters has two double rays, which are the asymptotes ; therefore the points K and K x , in which AX meets the asymptotes, are the double points of the involution XX\ ...*. 303. Since OP AX is a parallelogram, AX= — OP; and from the similar and equal triangles 0P'A t and AX'O, AX'= OP'f. But OP. 0P'= ± OB 2 (Art. 125) ; therefore AX.AX f = + 0B 2 ; or The rectangle contained by the segments intercepted on a fixed tangent to a conic between its point of contact and the points where it is cut by any two conjugate diameters is equal to the square ( + OB 2 ) on the semi-diameter drawn parallel to the tangent. 304. We have seen (Art. 302) that in the case of the hyper- bola K and K x are the double points of the involution of which A is the centre and X,X' a pair of conjugate points ; thus AX.AX'=AK 2 =0B 2 . Therefore AK= OB, and OAKB is a parallelogram. Accord- ingly: If a parallelogram be described so as to have a pair of conjugate semi-diameters of a hyperbola as adjacent sides, one of its diagonals will coincide with an asymptote %. Further, the other diagonal AB is parallel to the second asymptote. For consider the harmonic pencil (Art. 296) formed by the two asymptotes and the two conjugate diameters OA , OB. The four points in which this pencil cuts AB will be harmonic ; but one of the asymptotes OK meets AB in its middle point, therefore the other will meet it at infinity (Art. 59). , 305. Let X 1 be the point where the diameter OX meets the tangent at A v Since OX' and 0X X are a pair of conjugate lines which meet in a point on the chord of contact AA X of * In Fig. 199 only one of the points jK, K x is shown. + In order to account for the signs, it need only be observed that in the case of the ellipse OP and OP' are similar, but AX and AX' opposite to one another in direction ; while in the case of the hyperbola OP and OP' are opposite, but AX and AX' similar as regards direction. X Apollonius, loc. cit., book ii. 1. 309] THE CENTRE AND DIAMETERS OF A CONIC. 231 the tangents AX and A X X X , the straight line X / X 1 (Art. 274) ■will be a tangent to the conic. The point of contact of this tangent is M, the point of inter- section of AP and A X P' (Art. 299). 306. It is seen moreover that X / X 1 is one diagonal of the parallelogram formed by the tangents at A and A 1 and the parallels to AA X drawn through P and P / '; this may also be proved in the following manner. All points of a diameter have for their polars straight lines which are parallel to the conjugate diameter (Art. 284) ; if then through the conjugate points P and P' parallels be drawn to AA X , the first will be the polar of P'and the second the polar of P; consequently these parallels are conjugate lines. If now the theorem of Art. 274 be applied to these conjugate lines and the two tan- gents at A and A 1 , we obtain the following proposition : If a parallelogram is such that one pair of its opposite sides are tangents to a conic, and the other pair are straight lines, conjugate with regard to the conic and drawn parallel to the chord of contact of the two tangents, then its diagonals also will be tangents to the conic. 307. This gives the following solution of the problem : To construct a conic by tangents, having given a pair of conjugate diameters AA X and BB X in magnitude and direction. Suppose BB X to be that diameter which meets the conic in the case where the latter is a hyperbola. On BB X determine a pair of con- jugate points P and P / of the involution which has the centre of the conic as centre and the points B , B x either as double points or as conjugate points, according as the conic to be drawn is an ellipse or a hyperbola. Draw through A and A 1 parallels to BB X , and through P and P / parallels to AA X ; the diagonals of the parallelo- gram so obtained will be tangents to the required conic. 308. The segments AX and A X X^ are equal in magnitude and opposite in sign ; and it has been seen that AX.AX'= + OB 2 ; therefore AX'. A.X^ ± OB 2 ; or The rectangle contained by the segments intercepted upon two parallel fixed tangents between their points of contact and the points where they are cut by a variable tangent (X'Xj) is equal to the square (+ OB 2 ) on the semi-diameter parallel to the fixed tangents*. 309. Since the straight line OB is parallel to AX and A Y X X and half-way between them, the segments determined by AM * See Art. 160. 232 THE CENTRE AND DIAMETERS OF A CONIC. [310 and A X M respectively on A^X X and AX (measured from A L and A respectively) are double of OP and 0P' : , but by the theorem of Art. 300 the rectangle OP. OP' is constant ; thus The straight lines connecting the extremities of a given diameter with any point on the conic meet the tangents at these extremities in two points such that the rectangle contained by the segments of the tangents intercepted between these points and the points of contact is 310. Since X is (Art. 288) the point of intersection of the tangent at A and the tangent parallel to X'X Y , the proposition of Art. 303 may also be expressed as follows : The rectangle contained by the segments (AX, AX') determined by two variable parallel tangents upon any -fixed tangent is equal to the square (+ OB 2 ) on the semi-diameter parallel to the fixed tangent. 311. From the theorems of Arts. 299, 300 is derived the solution of the following problem : Given the two extremities A and A t of a diameter of a conic, a third point M on the conic, and tlie direction of the diameter conjugate to AA J , to determine the length of the latter diameter (Fig. 199). Through 0, the middle point of AA X , draw the diameter whose direction is given ; let it be cut by AM and A X M in P and P f respec- tively, and take OB the mean proportional between OP and OP'; then OB will be the half of the length required. 312. The proposition of Art. 303 gives a construction for pairs of conjugate diameters, and in par- ticular for the axes, of an ellipse of which two conjugate semi-diameters OA and OB are given in magnitude and direction (Fig. 200). Through A draw a parallel to OB ; this will be the tangent at A, and will be cut by any two conju- D gate diameters in two points X and X' such that AX.AX'=-OB 2 . If now there be taken on the normal at A two segments AC and AD each equal to OB, every circle passing through C and D will cut this tangent in two points X and X' which possess the property ex- pressed by the above equation ; these points are therefore such that the straight lines joining them to the centre will give the direc- tions of a pair of conjugate diameters. If the circle be drawn * Apollonius, loc. cit., lib. iii. 53. I ° \x' I / ^s. \ / w\ \^^ j X \ X } / 314] THE CENTRE AND DIAMETERS OF A CONIC. 233 through the angle XOX' becomes a right angle, and consequently OX , OX' will be the directions of the axes. Since the circular arcs CX', X f B are equal, the angles COX', X'OD are equal ; consequently OX' , OX are the internal and external bisectors of the angle which 00 , OD make with one another. In order then to construct the semi-axes OP , OQ in magnitude, let fall perpendiculars AX X , AX/ on OX, OX' respectively. Then XandX 15 X f and X( are pairs of conjugate points; therefore OP will be the geometric mean between OX and 0X X , and OQ the geometric mean between OX' and OX/ *. 313. Through the extremities A and A! (Fig. 201) of two conjugate semi-diameters OA and OA! of a conic draw any two parallel chords AB and A'B'. To find the points B and B' we have only to join the poles of these chords; this will give the diameter OX' which passes through their middle points. Let OX be the diameter conjugate to OX', i.e. that diameter which is Fi 20I parallel to the chords AB, A'B'. The pencils O(XX'AB) and 0(X'XA'B') are each harmonic (Art. 59), and are therefore projective with one another ; consequently the pairs of rays 0(XX', AA', BB') are in involution (Art. 123). But the two pairs (XX', AA') determine the involution of conjugate diameters (Arts. 1 27, 296); therefore also OB and OB' are conjugate diameters. Thus If through the extremities A and A! of two conjugate semi-diameters parallel chords AB , A'B' be drawn, the points B and B' will be the extremities of two other conjugate semi-dia?neters. Two diameters AA and BB determine four chords AB which form a parallelogram (Arts. 260, 287). The diameters conjugate respectively to them form in the same way another parallelogram, which has its sides parallel to those of the first ; that is, every chord AB is parallel to two chords A'B', and not parallel to two other chords A'B'. 314. Let E,Kbe the points where AB is cut by OA', OB' respectively. The diameter OX' which bisects A'B' will also bisect UK ; therefore AB and IIK have the same middle point ; thus AH=KB and AK=HB. The triangles OAK and OBH * Chasles, Aperg>j, historique, pp. 45, 362; Sections coniques, Art. 205. 234 THE CENTRE AND DIAMETERS OP A CONIC. [315 are therefore equal in area (Euc. I. 37), as also AKB' and BRA', and therefore also OAB' and OA!B are equal. Accord- ingly : The parallelogram described on two semi-diameters (OA , OB') as adjacent sides is equal in area to the parallelogram described similarly on the two conjugate semi-diameters. In the same way the triangles OAB and OA'B' can be proved equal. The triangles AHA', BKB' are equal for the same reason ; and OAR, OBK &re equal, and therefore also OAA' and OBB'. Therefore The parallelogram described on a pair of conjugate semi-diameters as adjacent sides is of constant area *. 315. Let M and N be the middle points of the non-parallel chords AB and A'B'. Since AB and A'B' are parallel to a pair of conjugate diameters (Art. 287) and since ON is the diameter conjugate to the chord A'B', therefore ON will be parallel to AB; so also OM will be parallel to A'B'. The angles OMA and ON A! are therefore equal or supplementary ; and since the triangles OMA and ON A' are equal in area (being halves of the equal triangles OAB and OA'B'), we have (Euc. VL 15), OM.AM=±ON.NA'f. Now project (Fig. 202) the points A,M,B,A', N, B' from the point at infinity on OB as centre upon the straight line B'B'. The ratio of the parallel segments AM and ON, OM and NA' is equal to that of their projections ; we con- clude therefore from the equality just proved that the rectangle contained by the projections of OM and AM is equal to that Fi S- 202 - contained by the projections of ON and NA'. As the projecting rays are parallel to OB, the projections of OM and MA are * Apollonius, loc. cit., lib. vii. 31, 32. t The signs + and — caused by the relative direction of the segments OM, NA' and ON, AM correspond respectively to the case of the ellipse (Fig. 201) and to that of the hyperbola (Fig. 202). 316] THE CENTRE AND DIAMETERS OP A CONIC. 235 each equal to half the projection of BA or of OA. Since N is the middle point of A'B' } the projection of ON will be equal to half the sum of the projections of OA' and OB', and the projection of NA' will be equal to half the projection of B'A', that is, to half the difference between the projections of OA and OB'. We have therefore (proj. OA) 2 = + proj. (OA' + OB') xproj. {OB'-OA'), or (proj. OA') 2 ± (proj. OA) 2 = (proj. OB') 2 . In the same manner, by projecting the same points on OB by means of rays parallel to OB' (Fig. 203), we should obtain . (proj. OA) 2 ± (proj. OA') 2 = (proj. OB) 2 . This proves the following proposition : If any pair of conjugate diameters are projected upon a fixed diameter by means of parallels to the diameter conjugate to this last, then the sum (in the ellipse) or difference (in the hyperbola) of the squares on the projections is equal to the square on the fixed diameter. By the Pythagorean theorem (Euc. I. 47) the sum of the squares on the orthogonal pro- jections of a segment on two Fi straight lines at right angles to one another is equal to the square on the segment itself. If then a pair of conjugate diameters are projected orthogonally on one of the axes of a conic and the squares on the pro- jections of each diameter on the two axes are added together, the following proposition will be obtained : The sum, (for the ellipse) or difference (for the hyperbola) of the squares on any pair of conjugate diameters is constant, and is equal to the sum or the difference of the squares on the axes *. 316. If five points on a conic are given, then by the method explained in Art. 285 the centre and two pairs of conjugate diameters u and u', v and v f can be constructed. If these pairs overlap one another, the conic is an ellipse ; in the contrary case it * Apollonius, loc. cit., lib. vii. 12, 13, 22, 25. 236 THE CENTRE AND DIAMETERS OF A CONIC. [317 is a hyperbola (Art. 296). If in this second case the double rays of the involution determined by the two pairs u and ?/, v and v' be constructed, they will be the asymptotes of the hyperbola. If in either case the orthogonal pair of conjugate rays of the in- volution be constructed, they will be the axes of the conic. The direction of the axes can be found without first constructing the centre and two pairs of conjugate diameters*. Let A, B, C, F, G be the five given points (Fig. 168) ; describe a circle round three of them ABC, and construct (Art. 227, I) the fourth point of intersection C" of this circle with the conic determined by the five given points. Any transversal will cut the two curves and the two pairs of opposite sides of the common inscribed quadrangle ABCC in pairs of points forming an involution (Art. 183). The double points P and Q (if such exist) of this involution will be conjugate with regard to each of the curves (Arts. 125, 263); i.e. they will be the pair common (Art. 208) to the two involutions which are formed on the transversal by the pairs of points conjugate with regard to the circle and by the pairs of points conjugate with regard to the conic (Art. 263). Suppose that the straight line at infinity is taken as the transversal. As this straight line does not meet the circle, one at least of these two involutions will have no double points, and consequently (Art. 208) the points P and Q do really exist. Since these points are infinitely distant and are conjugate with regard to both curves they will be (Arts. 276, 284) the poles of two conjugate diameters of the circle and also of two conjugate diameters of the conic ; but conjugate diameters of the circle are perpendicular to one another (Art. 289) ; therefore P and Q are the poles of the axes of the conic. Further, the segment PQ is harmonically divided by either pair of opposite sides of the quadrangle ABCC ; consequently P and Q are the points at infinity on the bisectors of the angles included by the pairs of opposite sides (Art. 60). In order then to find the required directions of the axes, we have only to draw the bisectors t of the angle included by a pair of opposite sides of the quadrangle ABCC, for example by AB and CC (Fig. 168). 317. Let qrst (Fig. 161) be a complete quadrilateral, and S any point. It has already been seen (Art. 185, right) that the pairs of rays a and a', b and b', which join S to two pairs of opposite vertices, belong to an involution of which the tangents drawn from S to any conic inscribed in the quadrilateral are a pair of conjugate rays. Suppose the involution to have two double rays m and?*; they will be harmonically conjugate * Poncelet, loc. cit., Art. 394. t See also the note to Art. 387. 318] THE CENTRE AND DIAMETEES OF A CONIC ■ 237 with regard to such a pair of tangents (Art. 125), and will consequently be conjugate lines with respect to the conic. But (Art. 218, right) m and n are the tangents at 8 to the two conies which can be inscribed in the quadrilateral qrst so as to pass through S. Therefore If tioo conies which are inscribed in a given quadrilateral pass through a given point, their tangents at this point are conjugate lines with respect to any conic inscribed in the quadrilateral. Instead of taking an arbitrary point S, let m be supposed given. If this straight line does not pass through any of the vertices of the quadrilateral, there will be one conic, and only one, which touches the five straight lines m,q,r,s,t (Art. 152). Let S be the point where this conic touches m; there will be a second conic which is inscribed in the quadrilateral and which passes through S ; let the tangent to this at S be n. The straight lines m and n will then be conjugate to one another with respect to all conies inscribed in the quadrilateral ; and therefore (Art. 255), The poles of any straight line m with respect to all conies inscribed hi the same quadrilateral lie on another straight line n. Moreover, since the straight lines m and n are the double rays of the involution of which the rays drawn from S to two opposite vertices are a conjugate pair, therefore m and n divide harmonically each diagonal of the quadrilateral. 318. I. The correlative propositions to those of Art. 317 are the following : If a straight line touches two conies which circumscribe the same quadrangle, the two points of contact are conjugate to one another with respect to all conies circumscribing the quadrangle. The polars of any given point M with respect to all the conies circumscribing the same quadrangle meet in a fixed point N. The segment MN is divided harmonically at the two points where it is cut by any pair of opposite sides of the complete quadrangle. II. Suppose in the second theorem of Art. 317 that the straight line m lies at infinity ; then the poles of m will be the centres of the conies (Art. 281), and n will bisect each of the diagonals of the quadri- to u Fig. 204. lateral (Art. 59); therefore: The centres of all conies inscribed in the same quadrilateral lie 238 THE CENTRE AND DIAMETERS OF A CONIC. [319 on the straight line (Fig. 204) which passes through the middle points of the diagonals of tlhe quadrilateral*. III. Suppose similarly in theorem I of the present Article that the point 31 lies at infinity ; the polars of M will become the diameters conjugate to those which have M as their common point at infinity ; thus : In any conic circumscribing a given quadrangle, the diameter which is conjugate to one drawn in a given fixed direction will pass through a fixed point. 319. Newton's theorem (Art. 318, II) gives a simple method for finding the centre of a conic deter- mined by five tangents a,b, c, d, e (Fig. 205). The four tangents a,b,c,d form a quadrilateral; join the middle points of its diagonals. Let the same be done with regard to the quadrilateral abce ; the two straight lines thus obtained will meet in the required centre 0. The five tangents, taken four and four together, form five quad- rilaterals ; the five straight lines which join the middle points of the diagonals of each of the quadri- laterals will therefore all meet in the centre of the conic inscribed Fig. 205. in the pentagon abcde. The same theorem enables us to find the direction of the diameters of a parabola which is determined by four tangents a, b,c, d. For each point on the straight line joining the middle points of the diagonals of the quadrilateral abed is the pole of the straight line at infinity with regard to some conic inscribed in the quadrilateral (Art. 318, II); therefore the point at infinity on the line will be the pole with regard to the inscribed parabola (Arts. 254 III, and 23). The straight line therefore which joins the middle points of the diagonals is itself a diameter of the parabola (Fig. 204). * Newton, Principia, book i. lemma 25. Cor. 3. CHAPTEE XXII. POLAR RECIPROCAL FIGURES. 320. An auxiliary conic K being given, it has been seen (Art. 256) that if a variable pole describes a fixed straight line its polar turns round a fixed point, and reciprocally, that if a straight line considered as polar turns round a fixed point, its pole describes a fixed straight line. Consider now as polars all the tangents of a given curve C, or in other words suppose the polar to move, and to envelope the given curve. Its pole will describe another curve, which may be denoted by C'. Thus the points of C' are the poles of the tangents of C. But it is also true that, reciprocally, the points of C are the poles of the tangents of C'. For let M' and N' be two points on C' (Fig. 206) • their polars m and n will be two tangents to C and the point mn where they meet will be the pole of the chord M'N' (Art. 256). Now suppose the point N'to approach M' indefinitely; the chord M'N' will ap- proach more and more nearly to the position of the tangent at M' to the curve C'; the Fi 6 straight line n will at the same time ap- proach more and more nearly to coincidence with m, and the point mn will tend more and more to the point where m touches C. In the limit, when the distance M'N' becomes indefinitely small, the tangent to C' at M' will become the polar of the point of contact of m with C. Just then as the tangents of C are the polars of the points of C', so also are the tangents of C' the polars of the points of C ; if a straight line m touches the curve C at M, the pole M ' of m 240 POLAR RECIPROCAL FIGURES. [321 is a point of the curve C ' and the polar m ' of 31 is a tangent to the curve C' at M f . Two curves C and C' such that each is the locus of the poles of the tangents of the other, and at the same time also the envelope of the polars of the points of the other, are said to be polar reciprocals * one of the other with respect to the auxiliary conic K. 321. An arbitrary straight line r meets one of the reciprocal curves in n points say ; the polars of these points are n tan- gents to the other curve all passing through the pole R f of r. To the second curve therefore can be drawn from any given point R f the same number of tangents as the first curve has points of intersection with the straight line r, the polar of R' \ and vice versa. In other words, the degree and class of a curve are equal to the class and degree respectively of its polar reciprocal tvith respect to a conic. 322. Now suppose the curve C to be a conic, and a , b two tangents to it; they will be cut by all the other tangents c,d,e,... in corresponding points of two projective ranges (Art. 149). In other words, C may be regarded as the curve enveloped by the straight lines c,d ,e , ... which connect the pairs of corresponding points of two projective ranges lying on a and b respectively (Art. 150). The curve C' will pass through the poles A\ B\C\D\W,... of the tangents a, b, c, d, e, ... of C. The straight lines A'(C' S D',I!',...) will be the polars of the points a(c,d,e,...) and will form a pencil projective with the range of poles lying on the straight line a (Art. 291); so too the straight lines i?'(C", i)', J5", ...) will be the polars of the points b(c ,d ,e,...) and will form a pencil projective with the range of poles lying on I. But the ranges a(c,d,e,...) and b(c ,d,e,...) are projective; therefore also the pencils A\C\B\E\...) and £'(C D',E', ...) are projective. Consequently C' is the locus of the points of intersection of corresponding rays of two projective pencils ; that is (Art. 150) a conic. Accordingly : The polar reciprocal of a conic with respect to another conic is a conic f. 323. When an auxiliary conic K is given and another conic * PONCELET, IOC. tit., Art. 232. t Ibid., Art. 231. 324] POLAR RECIPROCAL FIGURES. 241 C whose polar reciprocal C ' is to be determined, the question arises whether C' is an ellipse, a hyperbola, or a parabola. The straight line at infinity is the polar of the centre of K ; there- fore the points at infinity on C ' correspond to the tangents of C which pass through 0. It follows that the conic C ' will be an ellipse or a hyperbola according as the point is interior or exterior to the conic C, and O f will be a parabola when lies upon C. If A is the pole of a straight line a with respect to C, and a' the polar of A and A! the pole of a with respect to K, then will A' be the pole of a' with respect to C', since to four poles forming a harmonic range correspond four polars forming a harmonic pencil (Art. 291) and vice versa. Therefore the centre M' of C' will be the pole with respect to K of the straight line m which is the polar of with respect to C. To two conjugate diameters of C' will correspond two points of m which are conjugate with respect to C, &c. 324. Let there be given in the plane of the auxiliary conic a -figure (Art. 1) or complex of any kind composed of points, straight lines, and curves ; and let the polar of every point, the pole of every line, and the polar reciprocal of every curve, be constructed. In this way a new figure will be obtained ; the two figures are said to be polar reciprocals one of the other, since each of them contains the poles of the straight lines of the other, the polars of its points, and the curves which are the polar reciprocals of its curves. To the method whereby the second figure has been derived from the first the name of polar reciprocation is given. Two figures which are polar reciprocals one of the other are correlative figures in accordance with the law of duality in plane Geometry (Art. 33) ; for to every point of the one corresponds a straight line of the other, and to every range in the one corre- sponds a pencil in the other. They lie moreover in the same plane ; their positions in this plane are determinate, but may be interchanged, since every point in the one figure and the corresponding straight line in the other are connected by the relation that they are pole and polar with respect to a fixed conic. Thus two polar reciprocal figures are correlative figures which are coplanar, and which have a special relation to one another with respect to their positions in the plane in which they lie. On the other hand, if two figures are merely 242 POLAB RECIPROCAL FIGURES. [325 correlative in accordance with the law of duality, there is no relation of any kind between them as regards their position *. 325. If one of the reciprocal figures contains a range (of poles) the other contains a pencil (of polars), and these two corresponding forms are projective (Art. 291). If then the points of the range are in involution, the rays of the corre- sponding pencil will also be in involution, and to the double points of the first involution will correspond the double rays of the second (Art. 124). If there is a conic in one of the figures there will also be one in the other figure (Art. 322); to the points of the first conic will correspond the tangents of the second, and to the tangents of the first will correspond the points of the second; to an inscribed polygon in the first figure will correspond a circumscribed polygon in the second (Art. 320). If the first figure exhibits the proof of a theorem or the solution of a problem, the second will show the proof of the correlative theorem or the solution of the correlative problem ; that namely which is obtained by interchanging the elements ' point ' and ' line.' 326. Theoeem. If two triangles are both self-conjugate with regard to a given conic, their six vertices lie on a conic, and their six sides touch another conic f. Let ABC and DEF be two triangles (Fig. 207) each of which is self-conjugate (Art. 258) with regard to a given conic K. Let DE and DF cut BC in B x and C x respectively, and let AB and AC cut EF in E x and F x respectively. The point B is the pole of CA, and C is the pole of AB ; B x is the pole of the straight line joining the poles of BC and DE, i.e. of AF; and C x is the pole of the straight line joining the poles of BC and DF, i. e. of AE. The range of poles BCB X C X is therefore (Art. 291) projective with the pencil of polars A(CBFE), and therefore with the range of points F X E X FE in which this pencil is cut by the transversal EF. Thus (BCB X C X ) = (F X E X FE) = (E X F X EF) by Art. 45, which shows that the two ranges in which the straight lines BC and ^respectively are cut by AB, CA, DE, FD are protectively related. * Steiner, loc. cit., p. vii of the preface; Collected Works, vol. i. p. 234. t Steiner, loc. cit., p. 308, § 60, Ex. 46 ; Collected Works, vol. i. p. 448 ; Chasles, Sections coniqnes, Art. 215. 329] POLAR RECIPROCAL FIGURES. 243 These six straight lines therefore, the six sides of the given triangles, all touch a conic C (Art. 150, II). The poles of these six sides are the six vertices of the triangles ; these vertices therefore all lie on another conic C' which is the polar reciprocal of C with regard to the conic K *. 327. The proposition of the preceding Article may also be expressed as follows : Given two triangles which are self-conjugate with regard to the same conic K ; if a conic C touch five of the six sides it will touch the sixth side also, and if a conic pass through five of the six vertices it will pass through the sixth vertex also. It follows that if a conic C touch the sides of a triangle abc which is self-conjugate with regard to another conic K , there are an infinite number of other triangles which are self-conjugate with regard to the second conic and which circumscribe the first. For let d be any tangent to C ; from D, its pole with regard to K, draw a tangent e to C, and let / be the polar with regard to K of the point de ; then the triangle def will be self- conjugate with regard to K (Art. 259). But C touches five sides a, b, c, d, e of two triangles which are both self-conjugate with respect to K; therefore it must also touch the sixth side / ; which proves the proposition. 328. If the point D is such that from it a pair of tangents e' and f can be drawn to K, the four straight lines e,f, ef, f will form a harmonic pencil (Art. 264), since e and / are conjugate straight lines with respect to the conic K ; consequently the straight lines e' and f f are conjugate to one another with respect to C. The locus of D is the conic C ' which is the polar reciprocal of C with regard to K ; therefore : If a conic C is inscribed in a triangle which is self-conjugate with respect to another conic K, the locus of a point such that the j)airs of tangents drawn from it to the conies C and K form a harmonic pencil is a third conic C' which is the polar reciprocal of C with respect to K. 329. Correlatively : If a conic C' circumscribes a triangle which is self-conjugate with respect to another conic K, there are an infinite number of other triangles which are inscribed in C and are self -con- jugate with respect to K ; and the straight lines which cut C f and "&. in two pairs of joints which are harmonically conjugate to one another all touch a third conic C which is the polar reciprocal of C f with regard to K. * "We may show independently that the six vertices lie on a conic as follows. It has been seen that the pencil of polars A {CBFE) is projective with the range of poles BCB^Px ; it is therefore projective with the pencil D {BCB X CC) formed by joining these to the point D. Therefore A {CBFE) = D (BCB&) = D {BCEF) = D{CBFE) by Art. 45, which shows (Art. 150, I) that A, B, 0, D, E, F lie on a conic. R % 244 POLAR RECIPROCAL FIGURES. [330 330. Theorem. If two triangles circumscribe the same conic, their six vertices lie on another conic. Let OQ'R' and O'PS be two triangles each circum- scribing a given conic C (Fig. 208). The two tangents PS and Q'R' are cut by the four other tangents O'P, OQ', OP', O'S in two groups of corresponding points PQPS and P'Q'R'S' of two projective ranges u and u' (Art. 149); consequently the. pencils O(PQPS) and O'(P'Q'P'S') formed by connecting these points with and 0' respectively are projective. Therefore the points P, Q',B',S, in which their pairs of corresponding ray s intersect, lie on a conic C' (Art. 150,1) passing through the centres and 0' '; which proves the theorem. 331. The theorem correlative and converse to the foregoing one is the following : If two triangles are inscribed in the same conic, their six sides touch another conic*. This may be proved by considering the triangles OQ'R' and O'PS as both inscribed in the conic C', and by reasoning in a manner exactly analogous, but correlative, to that above. 332. It follows at once that: If two triangles circumscribe If two triangles are inscribed the same conic, the conic which in the same conic, the conic which passes through five of their ver- touches five of their sides touches tices passes through the sixth the sixth side also, vertex also. Or: If two conies are such that a triangle can be inscribed in the one so as to circumscribe the other, then there exist an infinite number of other triangles which possess the same property f. 333. There are in the figure (Fig. 208) four projective forms : the two ranges u and u f , which determine the tangents to the conic C, and the two pencils and ', which determine the points of C'; the pencil is in perspective with the range u * Brianchon, loc. cit., p. 35; Steiner, he. cit, p. 173, § 46, II; Collected Works, vol. i. p. 356. f PONCELET, loc. Cit., Art. 565. 334 POLAR RECIPROCAL FIGURES. 245 and the pencil O f is in perspective with the range u'. If then any tangent to C cut the bases u and u' of the two ranges in A and A' respectively, the rays OA and O'A' will meet in a point M lying onC'; and, conversely, if any point M on C' be joined to the centres and 0', the joining lines will cut u and u' respectively in two points A and A! such that the straight line joining them is a tangent to C. Therefore : If a variable triangle AA'M is such that two of its sides pass respectively through two fixed points O f and lying on a given conic, and the vertices opposite to them lie respectively on two fixed straight lines u and u f , while the third vertex lies always on the given conic, then the third side will touch a fixed conic which touches the straight lines u and u' '. If a variable triangle AA'M is such that two of its vertices lie respectively on two fixed tangents u and u f to a given conic, and the sides opposite to them pass respectively through two fixed points O f and 0, while the third side always touches the given conic, then the third vertex will lie on a fixed conic which passes through the points and 0' '. 334. Theorem. If the extremities of each of two diagonals of a complete quadrilateral are conjugate points with respect to a given conic, the extremities of the third diagonal also will be conjugate points with respect to the same conic *. Let ABXY (Fig. 209) be a complete quadrilateral such that A is conjugate to X, and B to Y, with respect to a given conic K (not shown in the figure). Let the sides AB , XT meet in C, and the sides AY,BX in Z; then shall C and Z be conjugate points with respect to the conic K. Suppose the polars of the points A,B,C (with respect to K ) to cut the straight line ABC in A', B',C r respectively. The three pairs of conjugate points A and A!, B and B f , C and C are in involution; consequently, considering XYZ as a triangle cut by a transversal A!B'C\ it follows by Art. 135 that the * Hesse, De octo punctis intersectionis trium superficiei'um secundi ordinis (Dissertatio pro venia legendi, Regiomonti, 1 840), p. 1 7. 246 POLAR RECIPROCAL FIGURES. [335 straight lines XA', YB', ZC meet in one point Q. Since evidently X^'is the polar of A and YB' the polar of B with respect to K, their point of intersection Q is the pole of AB. Since then C is a point on AB and is conjugate to C, its f>olar will be QC ; but QC passes through Z; therefore Cand i^are conjugate points, which was to be proved. 335. The proof of the following, the correlative theorem, is left as an exercise to the student : If two pairs of opposite sides of a complete quadrangle are conju- gate lines with respect to a conic, the two remaining sides also are conjugate lines with respect to the same conic. In order to obtain such a complete quadrangle, it is only necessary to take the polar reciprocal of the quadrilateral con- sidered in Hesse's theorem, i. e. the figure which is formed by the polars of the six points A and X, B and Y, C and Z. 336. The following proposition is a corollary to that of Art. 334: Two triangles which are reciprocal with respect to a conic are in *. Let ABC (Fig. 210) be any triangle; the polars of its vertices with respect to a given conic form another triangle A!B'C f reciprocal to the first, that is, such that the sides of the first triangle are also the polars of the vertices of the second. Let the sides CA and C'A' meet in E, and the sides AB and Fig. 210. A'B' in F. The points B and E are conjugate with respect to the conic, since E lies on C'A\ the polar of B ; similarly C and .Fare conjugate points. Thus in the quadri- lateral formed by BC, CA, AB, and EF, two pairs of opposite vertices B and E, C,and F are conjugate; therefore the third pair are conjugate also, viz. A and the point B where BC meets EF. The polar B'C of A therefore passes through B\ thus BC and B'C meet in a point I) lying on EF. Since then the pairs of opposite sides of the two triangles meet one another in three collinear points, the triangles are in homology, and the straight lines AA', BB', CC which join * Chasles, loc. cit., Art. 135, 339] POLAR RECIPROCAL FIGURES. 247 the pairs of vertices meet (Art. 1 7) in a point 0, the pole of the straight line DEF. 337. By combining this theorem with that of Art. 155 the following property may be enunciated : If two triangles are reciprocals with respect to a given conic K, the six points in which the sides of the one intersect the non- corresponding* sides of the other lie on a conic C, and the six straight lines which connect the vertices of the one with the non-corresponding vertices of the other touch another conic C r , the polar reciprocal of C with respect to K (Art. 322); these straight lines are in fact the polars with regard to K of the six points just mentioned. If one of the triangles A'B'C is inscribed in the other ABC, the three conies C, C', and K coincide in one which is circumscribed about the former triangle and inscribed in the latter (Arts. 174, 176). •» 338. Problem. Given two triangles ABC, A'B'C which are in homology ; to construct (when it exists) the conic with regard to which they are reciprocal. Take one of the sides, BG for example ; the points in which it is cut by G'A' a,n&A'B' are conjugate to the points .Sand G respectively, and these two pairs of conjugate points determine an involution (Art. 263), the double points of which (if they exist) are the points where BG is cut by the conic in question. In order then to find the points in which this conic cuts BG, it is only necessary to construct these double points. In this way the points in which the sides of the triangles meet the conic can be found, and the latter is determined. Since A' and B are the poles of BG and G'A', these points and that in which C'A' meets BG will be the vertices of a self-conjugate triangle (Art. 258). If then, in finding the points of intersection of the conic and the straight lines BG and C'A' in the manner just explained, it should hajypen that the two involutions found have neither of them double points, the conclusion is that no conic exists such as is required ; for if it did exist, it must be cut by two of the side^ of the self-conjugate triangle (Art. 262). 339. The centre of homology of the given triangles (Fig. 210) is the pole of the axis of homology DBF; and the projective corre- spondence (Art. 291) between the points (poles) lying on the axis and the straight lines (polars) radiating from the centre of homology is determined by the three pairs of corresponding elements D and * Two sides BG and WC of the triangles may be termed corresponding, when each lies opposite to the pole of the other. And two vertices A and A' may be termed corresponding, when each lies opposite to the polar of the other. 248 POLAR RECIPROCAL FIGURES. - [339 A A', E and BB' , F and CC. Consequently it is possible to construct with the ruler only (Art. 84) the polar of any other point on the axis, and the pole of any other ray passing through the centre 0. What has just been said with regard to the point and the axis of homology may also be said with regard to any vertex of one of the triangles and its polar (the corresponding side of the other triangle). For if e. g. the vertex A' and the side BG be considered, the projective correspondence between the straight lines radiating from A' and the points lying on BC is determined by the three pairs of corresponding elements A'B' and C, A'C and B, A f O and D. This being premised, it will be seen that the polar of any point P and the pole of any straight line p can be constructed with the help of the ruler only. For suppose P to be given ; it has been shown that the pole^s of the straight lines PO, PA, PB, PC, PA', ... can be constructed, and these all lie on a straight line X which is the required polar of P. So again i f th e straight line p is given, the polars of the points in which it meeVtfC, CA, ... can be constructed, and will meet in a point which is the pole of p. It will be noticed that all these determinations of poles and polars are linear (i. e. of the first degree) and independent of the construction (Art. 338) of the auxiliary conic, which is of the second degree, since it depends on finding the double elements of an involution. The construction of the poles and polars is therefore always possible, even when the auxiliary conic does not exist. In other words : the two given triangles in homology determine between the points and the straight lines of the plane a reciprocal correspondence such that to every point corresponds a straight line and to every straight line a point, to the rays of a pencil the points of a range projective with the pencil, and vice versa. Any point and the straight line corre- sponding to it may be called pole and polar, and this assemblage of poles and polars, which possesses all the properties of that determined by an auxiliary conic (Art. 254), may be called a polar system. Two triangles in homology accordingly determine* a polar system. If an auxiliary conic exists, it is the locus of the points which lie on the polars respectively corresponding to them, and it is at the same time the envelope of the straight lines which pass through the poles respectively corresponding to them. If no auxiliary conic exists, there is no point which lies on its own polar *. * Staudt, loc. cit, Art. 241. CHAPTEE XXIII. FOCI *, 340. It has been seen (Art. 263) that the pairs of straight lines passing through a given point 8 and conjugate to one another with respect to a given conic form an involution. Let a plane figure be given, containing a conic C ; and let the figure homological with it be constru^lfcd, taking 8 as centre of homo- logy ; let C ' be the conic corresponding to C in the new figure. Since in two homological figures a harmonic pencil corre- sponds to a harmonic pencil, any pair of straight lines through 8 which are conjugate with respect to C will be conjugate also with respect to C'. The polars of S with respect to the two conies will be corresponding straight lines ; if then the polar of S with respect to C be taken as the vanishing line in the first figure, the polar of 8 with respect to C' will lie at infinity; i. e. the point 8 will be the centre of the conic C'. In this case therefore any two straight lines through 8 which are conjugate with respect to C will be a pair of conju- gate diameters of C'. If 8 is external to C, the double rays of the involution formed by the conjugate lines through 8 are the tangents from S to C, and therefore the asymptotes of C', which is in this case a hyperbola. If S is internal to C, the involution has no double rays, and therefore C' is an ellipse. We conclude then that to every point S in the plane of a given conic C corresponds a conic G r homological with C and having its centre at S; which conic C' is a hyperbola or an ellipse according as 8 is external or internal to the given conic C. * Steineb, Vorlesungenuber synthetische Geometrie (ed. Schroter), II ter Abschnitt, § 35 ; Zech, Hohere Geometrie (Stuttgart, 1857), § 7 > R EYE > Geometrie der Lage (2nd ed., Hannover, 1877), Vortrag 13. 250 FOCI. [341 341. For certain positions of the point S the conic C ' will be a circle. When S has one of these positions it is called a focus* of the conic C. Since all pairs of conjugate diameters of a circle cut one another orthogonally the involution at S of conjugate lines with respect to C will in this case consist entirely of orthogonal pairs. If C is a circle, its centre is a focus ; for every pair of conjugate lines which meet in 0, i.e. every pair of conjugate diameters of C, cut orthogonally. And a circle C has no other focus but its centre 0. For let any points be taken (Fig. 211) distinct from and a straight line SQ be drawn not passing through 0; and let P be the pole of $Q. Then since PO must be perpendicular to SQ, the conjugate lines SP, SQ cannot be orthogonal, and there- fore 8 cannot be a focus of C. The foci of a conic C which is not a circle are of necessity internal points ; this follows from what has been said above (Art. 340). Further, they lie on the axes. For if F is a focus and the centre of the conic, the pole of the diameter FO will lie on the perpendicular drawn through F to FO; therefore FO is perpendicular to its conjugate diameter, i. e. FO is an axis of the conic. Again, the straight line connecting two foci F and F x is an axis. For if straight lines perpendicular to FF X be drawn through F and F 1 these will both be conjugate to FF X , and their point of intersection will therefore be the pole of FF X ; but this point lies at infinity ; therefore FF t is an axis. 342. Let a point P be taken arbitrarily on an axis a of a conic ; through P draw a straight line r, and from It, the pole of r, draw the straight line / perpendicular to r ; let P' be the point where / meets the axis. The straight lines passing through P and those passing through P' and conjugate to them respectively form two projective pencils ; for the second pencil is composed of rays which project from P f the range * Db la Hire, Sectiones conicae (Parisiis, 1685), lib. viii. prop. 23 ; Poncelet, Proprtitis projectives, Art. 457 et seqq. 343] FOCI, 251 formed by the poles of the rays of the first pencil, which range is (Art. 291) projective with the first pencil itself. The two pencils in question have three pairs of corresponding rays which are mutually perpendicular ; for if A be the point at infinity which is the pole of the axis a, the rays PA , PP\ r of the first pencil correspond to the rays P'P,P'A, / of the second, and the three former rays are severally perpendicular to the three latter. The two pencils therefore by the inter- section of corresponding rays generate a circle of which PP' is a diameter ; and therefore every pair of corresponding rays of the two pencils P and P f intersect at right angles. Thus: To every point P lying on an axis of the conic corresponds a point P f on the same axis such thai any two conjugate straight lines which pass one through P and the other through P r are perpendicular to one another. ■flie pairs of points analogous to P, P f form an involution. For let the ray r move parallel to itself ; the corresponding rays r' (which are all perpendicular to r) will all be parallel to each other. The pencil of parallels r is projective (Art. 291) with the range which the poles R of the rays r determine upon the diameter conjugate to that drawn parallel to r\ and the pencil of parallels r ' is in perspective with this same range. Therefore the pencils r,r / are projective, and consequently the points P, P f in which a pair of corresponding rays ;•, r f of the pencils cut the axis a trace out two projective ranges. To the straight line at infinity regarded as a ray r corresponds in the second pencil the diameter parallel to the rays r' '; and similarly, to the line at infinity regarded as a ray r' corre- sponds in the first pencil the diameter parallel to the rays r. Therefore the point at infinity on the axis has the same corre- spondent whether it be regarded as a point P or as a point P': viz. the centre of the conic. We conclude that the pairs of points P \P' constitute an involution of which the centre is the centre of the conic, 343. If the involution formed by the points P, P r on the axis a has double points, each of them will be a focus of the conic, since every straight line through such a double point will be conjugate to the perpendicular drawn to it through the point itself. 252 FOCI. [344 If the involution has no double points, each of the two points (Art. 128) at which the pairs PP' subtend a right angle will be a focus of the conic. For every pair of mutually perpendicular straight lines which meet in such a point will pass through two points P, P', and will therefore be conju- gate lines with respect to the conic. From this it follows that one at least of the two axes of a conic contains two foci. Further, a conic has only two foci ; for every straight line which joins two foci is an axis (Art. 341), and no conic (except it be a circle) has more than two axes. Consequently a central conic {ellipse or hyperbola) has two foci, which are the double points of the involution PP f on an axis and are also the points at which the pairs of points PP'ofthe involution on the other axis subtend a right angle. The axis which contains the foci may on this account be called the focal axis. Since the foci are internal points, it is seen that in the hyperbola the focal axis is that one which cuts the curve (the transverse axis). Since the centre of the conic is the centre of the involution PP', it bisects the distance between the two foci. From what has been said it follows that two perpendicular straight lines which are conjugate with respect to a conic meet the focal axis in two points which are harmonically conjugate with respect to the foci ; and they determine upon the other axis a segment which subtends a right angle at either focus. 344. The normal at any point on a curve is the perpen- dicular at this point to the tangent. Since the tangent and normal at any point on a conic are conjugate lines at right angles, they meet the focal axis in a pair of points harmoni- cally conjugate with respect to the foci ; and they determine on the other axis a segment which subtends a right angle at either focus (Art. 343). Accordingly: If a circle be draivn to pass through the two foci and through any point on the conic, it will have the two points in which the non-focal axis is cut by the tangent and normal at that point as extremities of a diameter. And again (Art. 60): The tangent and normal at any point on a conic are the bisectors 347] FOCI. 253 of the angle made with one another ly the rays which join that point to the foci*. These rays are called the focal radii of the given poinlw 345. A pair of conjugate lines which intersect at ight angles in a point 8 external to the conic are harmonically conjugate with respect to the tangents from 8 to the conic (Art. 264) as well as with respect to the rays joining 8 to the foci (Art. 343); therefore: The angle between two tangents and that included ly the straight lines which join the point of intersection of the tangents to the foci have the same bisectors \. 346. In the parabola, the point at infinity on the axis, re- garded as a point P, coincides with its correspondent P'; for the straight line at infinity, being a tangent to the conic at the said point P, passes through iti~own pole. Accordingly one of the double points of the involution determinftrbn the axis by the pairs, of conjugate orthogonal rays, i.e. one of the foci, is at infinity. The other double point lies at a finite distance, and is generally spoken of as the focus of the parabola ; Since in the case of the parabola one focus is at infinity, the theorems proved" above (Arts. 343-345) become the follow- ing: Two conftigate orthogonal rays, and in particular the tangent and normal at any point on the parabola, meet the axis in tioo points which are equidistant from the focus. The tangent and normal at a point on a parabola are the bisectors of the angle which the focal radius of the point makes with the diameter passing through the point %. The straight line which connects the focus with the point of inter- section of two tangents to a parabola makes with 'either of the tange?its the same angle that the axis makes with the other tangent. 347. From the last of these may be immediately deduced the following theorem : The circle circumscribing a tria?igle formed by three tangents to a parabola passes through the focus. Let P QB (Fig. 2 1 z) be a triangle formed by three tangents * Apollonius, loc. cit.y iii. 48. f Ibid., iii. 46. % De la Hire, loc. cif., lib. viii. prop. 2. 254 FOCI. [348 Fig. 212. to a parabola, and let F be the focus. Considering the tangents which meet in P, the angle FPQ is equal to that made by PR with the axis ; and considering the tangents which meet in P, the angle FPQ is equal to that made by PP with the axis. Hence the angles FPQ , FPQ are equal, and there- fore P, Q , P , F lie on the same circle. Corollary. The locus of the foci of all parabolas which touch the three sides of a given triangle is the circumscribing circle of the triangle. This corollary gives the construction for the focus of a parabola which touches four given straight lines. And since only one such parabola can be drawn (Art. 157), we conclude that: Given four straight lines, the circles circumscribing the four triangles which can be formed by taking the lines three and three together all pass through the same point. 348. The polar of a focus is called a directrix. The two directrices are straight lines perpendicular to the transverse axis and external to the conic, since t^e foci lie on the transverse axis and are internal to the conic (Art. 343). In the case of the parabola, the straight line at infinity is one directrix ; the other lies at a finite distance, and is generally spoken of as the directrix of the parabola. If F be a focus, and if the tangent at any point X on a conic cut the corresponding directrix in Y, this point I will be the pole of the focal radius FX. Therefore FX, FY are conjugate lines with respect to the conic, and since they meet in a focus, they will be at right angles: consequently: Fig. 213. 349] FOCI. 255 The part of a tangent to a conic intercepted between its point of contact and a directrix subtends a right angle at the corresponding focus. 349. Let the tangent and normal at any point M on a conic meet the focal axis in P ,P f respectively, and let them meet the other axis in Q , Q' respectively (Fig. 213). From M let perpendiculars MP", MQ" be drawn to the axes. From the similar triangles OPQ , Q"MQ OP:OQ=Q"M:Q"Q,\ and from the right-angled triangle Q'MQ Q"M:Q"Q = Q'Q":Q"M; .-. OP:OQ = Q'Q":Q"M = Q'Q":OP", or OP.OP"=OQ.Q'Q" = OQ(Q'0+OQ"), so that OP.OP"-OQ.OQ"=OQ.Q'0 (l) But P and P" are a pair of conjugate points, since MP" is the polar of P; similarly Q and Q" are conjugate points. Therefore (Art. 294) OP . OP" = OA 2 and OQ . OQ" = ± OP 2 , where OA , OP are the lengths of the semiaxes, and the double sign refers to the two cases of the ellipse and the hyperbola. Again, the points Q , Q' subtend a right angle at either of the two foci F,F' (Art. 343) so that OQ . Q'O = OF 2 . Substituting, (l) becomes OF 2 = OA 2 + OP 2 . This shows that in the ellipse OA > OP ; so that the focal axis is the axis major. Eeferring now to Figs. 214 and 215, FA = FO+OA, FA'= FO + OA' =FO-OA; .: FA.FA' = F0 2 -OA 2 = + OP 2 . If F> be the point in which a directrix cuts the focal axis, the vertices A and A f of the conic will be harmonically conju- gate with respect to F and the point B where the polar of F cuts AA' (Art. 264); therefore, since bisects AA f , OA 2 = OF.OD. 256 FOCI. [350 The parabola has one vertex at infinity; consequently the other lies midway between the focus and the directrix (Fig. 218). 350. If a focus F of a conic C be taken as centre of homo- logy, and a conic C' be constructed homological with C and Fig. 214. Fig. 215. having its centre at F, it has been seen (Arts. 340, 341) that C' is a circle. But by what has been proved in Art. 77, if if and M' are a pair of corresponding points of C and C', FM : MP = constant, FM' or FM MP FM' x constant, where MP (Figs. 214, 215) is the distance of M from the vanishing line, that is from the polar of F, i. e. the correspond- ing directrix. Now FM' is constant, because C' is a circle ; therefore The distance of any point on a conic from a focus bears a constant ratio to its distance from the corresponding directrix. Moreover, this ratio is the same for the two foci. For let (Figs. 214, 215) be the centre of the conic, F, F' the foci, A , A' the vertices lying on the focal axis, D,l)' the points in which this axis is cut by the directrices ; then (Art. 294) OA 2 = OA' 2 = 0F.0D = OF'. OB'. But OF' = - OF, so that A'D' = - AD and F'A'= - FA, and therefore FA:AD= F'A' : A'D', which shows that the ratio is the same for F and for F'. In the case of the parabola the ratio in question is unity, 352] FOCI. 257 because (Art. 349) the vertex of a parabola is equally distant from the focus and the directrix. Therefore The distance of any point on a parabola from the focus is equal to its distance from the directrix. 351. Conversely, the locus of a point M which is such that its distance from a fixed point F bears a constant ratio e to its distance from a fixed straight line d is a conic of which F is a focus and d the corresponding directrix *. For let MP (Figs. 214, 215) be drawn perpendicular to d; then by hypothesis FM _ MP ~ e * Let now the figure be constructed which is homological with the locus of M ; F being taken as centre of homology, and d as vanishing line. If M' be the point corresponding to M s then (Art. 77) FM -j^f-. : MP = constant. FM' These two equations show that FM f is constant ; thus the locus of M ' is a circle, centre F. The locus of M is there- fore a conic (Art. 23) having one focus at F (Art. 341). And since the straight line at infinity is the polar of F with respect to the circle, the straight line d is the polar of F with respect to the conic ; i.e. it is the directrix corresponding to F. 352. The length of a chord of a conic drawn through a focus perpendicular to the focal axis is called the latus rectum or the parameter of the conic. Let MFM' (Fig. 216) be a chord of a conic drawn through a focus F, and let iV be the point where it cuts the corresponding directrix. Let LFL f be the latus rectum drawn through F. Then since the directrix is the polar of the focus, N and F are harmonic conjugates with regard to M and M f . There- fore 211 + NF~ NM^ NM' and if perpendiculars MK , FD , M'K' be let fall on the directrix, 2 7 + FD ~ M'K' T MK * Pappus, Math. Collect., lib. vii. prop. 238. 258 But by Art. 350 FOCI. [353 'K'\ FD : MK = M'F: FL : FM: " FL" M f F^ FM' that is to say : In any conic, half the latus the segments of any focal chord. is a harmonic mean between - "X X ■M Tu 0' \ JP K' -7^ N ID 3 L' A.' Fig. 2 1 6. Corollary. If 31, M f be taken at A', A respectively, FL = i (~AF + FA') - i _^1_ ~ 2 AF.FA' OA = Tm ( b y Art - 349 )> so that FL= + — — , — 0^4 which gives the length of the semi-latus rectum in terms of the semi-axes. In the parabola r—j = o , so that FL = 2 FA. 353. Theorem. In the ellipse the sum, and in the hyperbola the difference, of the focal radii of any point on the curve is constant*. Let M be any point on a central conic (Figs. 314, 215) whose * Apolloxius, he. cit., iii. 51, 52. 356] FOCI. foci are F, F' and directrices d, d'; and let (if, d) &c. denote as usual the distance of M from d, &c. By Art. 351 FM F'M (M,d)~(M,d')- € ' FM±F'M •*' (M,d)±(M i d / )~ € ' But (Fig. 214) in the ellipse (Jf, d) + (M t d'), and (Fig. 215) in the hyperbola (M, d) — (M, d') is equal to the distance DJ)' between the two directrices ; therefore FM±F'M=*.JDD', which proves the proposition. Conversely : The locus of a point the sum (difference) of whose- distances from two fixed points is constant is an ellipse (a hyperbola) ofzvhich the given points are the foci. 354. If in the proposition of the last Article the point M be taken at a vertex A, e.DD'=FA±F'A = zOA = AA' ) so that the length of the focal axis is the constant value of the sum or difference of the focal radii. It is seen also that the constant e is equal to the ratio of the length of the focal axis to the distance between the directrices. 355. Since by Art. 294 OA 2 = OF.OD, or AA /2 = FF'.DI>', _ AA' _ FF A € -DD'-AA' ; so that the constant e is equal to the ratio of the distance between the foci to the length of the focal axis. Now in the ellipse FF' AA\ in the parabola FF'=AA'= 00, in the circle FF'=o. Therefore the conic is an ellipse, a hyperbola, a parabola, or a circle, according as ei, e = i, or e = o. This constant e is called the eccen- tricity of the conic. 356. Theoeem. The locus of the feet of perpendiculars let fall from a focus upon the tangents to an ellipse or hyperbola is the circle described on the focal axis as diameter *. * Apollonius, loc. tit., iii. 49, 50. S 2 260 FOCI. [357 Take the case of the ellipse (Fig. 217). If F, F' are the foci, and M is any point on the curve, join F'M and produce it to G making MG equal to MF. Then F'G will (Art. 354) be equal to AA ' whatever be the posi- tion- of M ; thus the locus of G is a circle, centre F' and radius equal to AA'. If FG be joined, it will cut the tangent at M perpendicu- larly, since this tangent (Art. 344) bisects the angle F3IG; and the point U where the two lines intersect will be the mid- dle point of FG because FMG is an isosceles triangle. There- fore OU is parallel to F'G and equal to \F'G, that is, to OA ; i. e. the locus of U is the circle on AA ' as diameter. A similar proof holds good for the hyperbola, except that from the greater of the two MF, MF' must be cut off a, part MG equal to the less. 357. If FU, FU' (Fig. 217) are the perpendiculars let fall from a focus F on a pair of parallel tangents, U,F,U' will evidently be collinear. And since U and U' both lie on the circle described oh A A ' as diameter, FU . FU' = FA . FA' = + OW (Art. 349), according as the conic is an ellipse or a hyperbola. Thus the product of the distances of a pair of parallel tangents from a focus is constant. Since the perpendicular let fall from the other focus F' on the tangent at M is equal to FU', it follows that The product of the distances of any tangent to an ellipse (hyper- bola) from the two foci is constant, and equal to the square of half the minor (conjugate) axis. Conversely : The envelope of a straight line which moves in such a way that the ptroduct of its distances from two fixed points is constant is a conic ; an ellipse if the value of the constant is positive, a hyperbola if it is negative. 358. Let F (Fig. 218) be the focus of a parabola, A the vertex, M any point on the curve, N the point of intersection of the tangents at M and A. If NF' be drawn to the infinitely 360] FOCI. 26 L distant focus I' (i. e. if NF' be drawn parallel to the axis), the angles ANF f , FNM will be equal (Art. 346). But ANF' is a right angle ,'theref ore FNM is a right angle also. Thus The foot of the perpen- dicular let fall from the focus of a parabola on any tangent lies on the tangent at the vertex. Coeollaby. Since any point on the circumscribing circle of a triangle may be regarded (Art. 347) as the focus of a parabola inscribed in the triangle, it follows at once from the theorem just proved that if from any jwint on the circumscribing circle of a triangle perpendiculars be let fall on the three sides, their feet will be collinear *. 359. The theorem of Art. 356 may be put into the following form: If a right angle move in its plane in such a way that its vertex describes a fixed circle, while one of its arms passes always through a fixed point, the envelope of its other arm will be a conic concentric with the given circle, and having one focus at the fixed point. The conic is an ellipse or a hyperbola according as the given point lies within or without the given circle t. So too the corresponding theorem (Art. 358) for the parabola may be expressed in a similar form as follows : If a right angle move in its plane in such a way that its vertex describes a fixed straight line, while one of its arms passes always through a fixed point, the other arm will en- velope a parabola having the fixed point for focus and the fixed straight line for tangent at its vertex. 360. I. Let the tangents at the vertices of a central conic be cut in P, P r by the tangent at any point M (Fig. 219). The three tangents form a triangle circumscribed about, the conic, two of the vertices of which Fi s- 2I 9- are P and P r , the third (at infinity) being the pole of the * For other proofs of this see Art. 416. f Maclauein, Geometria Organica, pars II a . prop. xi. 262 FOCI. [361 axis AA'. Therefore (Art. 274) the straight lines drawn from P and P' to any point on the axis will be conjugate to one another with respect to the conic. Thus, in particular, the straight lines joining P and P' to a focus will be conjugate to one another; but conjugate lines which meet in a focus are mutually perpendicular (Art. 343); consequently the circle on PP' as diameter will cut the axis A A' at the foci*. II. Let the tangent PMP' cut the axis A A' at N; then J^is the harmonic conjugate of if with respect to P, P f (Art. 194). Consider now the complete quadrilateral formed by the lines FP,F'P,FP\F'P'. Two of its diagonals are FF' and PP'; the third diagonal must then cut FF' and PP' in points which are harmonically conjugate to N with regard to F, F' and P , P' respectively. It must therefore be the normal at M to the conic t- 361. Let TM , TN (Fig. 220) be a pair of tangents to a conic, M and N their points of contact, F a focus, d the corresponding directrix. If the chord MN cut d in P, this point is the pole of TF; therefore TFP is a right angle (Art. 343) J. But MN is divided harmonically by FT and its pole P ; thus F(MNTP) is a harmonic pencil, and consequently FT,FP are the bisectors of the angle MFN. Accordingly : One of the bisectors of the angle which a chord of a conic subtends at a focus passes through the pole of the chord. The other bisector meets the chord at its point of intersection with the directrix corre- sponding to the focus. . Or the same thing may be stated in a different manner, thus : The straight line which joins a focus to the point of intersection of a pair of tangents to a conic makes equal (or supplementary) angles with the focal radii of their points of contact §. * Apollonius, loc. tit., iii. 45. Desargues, (Euvr'es, i. pp. 209, 210. t Apollonius, loc. tit, iii. 47. X If the points M and N are taken indefinitely near to one another, this reduces to the theorem already proved in Art. 348. § De la Hire, loc. tit., lib. viii. prop. 24. Fig. 220. 362] FOCI. 263 362. Let the tangents TM , TN be cut by any third tangent in M' ' , N' respectively (Figs. 221, 222); let L be the point of contact of this third tangent. The following relations will hold among the angles of the figures : N'FL = NFN' = I NFL, LFM' = M'FM = I LFM, whence by addition N'FL + LFM' = $ (NFL + LFM), or N'FM' = \ NFM = NFT = TFM* Let now the tangents TM, TN be fixed, while the tangent M'N' is supposed to vary. By what has just been proved, the angle subtended at the focus by the part M'N' of the Fig. 221, Fig. 222. variable tangent intercepted between the two fixed ones is constant. As the variable tangent moves, the points M ' , N' describe two projective ranges (Art. 149), and the arms FM ', FN' of the constant angle M'FN' trace out two con- centric and directly equal pencils (Art. 108). Accordingly: * In this reasoning it is supposed that FM', FN', FT are all internal bisectors ; 4. e. that either the conic is an ellipse or a parabola, or that if it is a hyperbola, the three tangents all touch the same branch (Fig. 221). If on the contrary two of the tangents, for example TM and TN, touch one branch and the third M'N' the other branch (Fig. 222), then FM' and FN' will be external bisectors. In that case, N'FL = ± NFL- - LFM' = I LFM + - 2 a (the angles being measured all in the same direction) ; .-. N'FM' = \ NFM, just as in the case above. 264 FOCI. [363 The ranges which a variable tangent to a conic fixed tangents are projected from either focus by ' pencils. on two of two This theorem clearly holds good for the cases of the parabola and its infinitely distant focus, and the circle and its centre. For the parabola it becomes the following : Two fixed tangents to a parabola intercept on any variable tangent to the same a segment whose projection on a line perpendicular to the axis is of constant length. The general theorem may also be put into the following form : One vertex F of a variable triangle M'FN' is fixed, and the angle M'FN f is constant, while the other vertices M', N' move respectively on fixed straight lines TM, TN. The envelope of the side M'N' is a conic of which F is a focus, and which touches the given lines TM, TN. 363. The problem, Given the two foci F, F' of a conic and a tangent t, to construct the conic, is determinate, and admits of a single solution, as follows. Join FF' (Figs. 223, 224) and let it cut t in P; take P' the harmonic conjugate of P with respect to F and F'. If a straight line P'M be drawn perpendicular to t, it will be the normal corresponding to the tangent t (Art. 344), i.e. if will be the point of contact of t. Draw MP" perpendicular to FF' ; it will be the polar of P, and P , P" will be conjugate points with respect to the Fig. 223. Fig. 224. conic. If then FF' be bisected at 0, and on FF' there be taken two points A, A' such that OA* = OA" 1 = OP. OP", A and A' will 364] FOCI. 265 be the vertices of the conic. The conic is therefore completely deter- mined ; for three points on it are known (M, A , A') and the tangents at these three points (t and the straight lines AC, A'C drawn through A , A' at right angles to AA'). An easy method of constructing the conic by tangents is to describe any circle through i^and F', cutting AC, A / C / in H and K, II' and K f respectively (Fig. 224). Then if the chords UK', H'K be drawn which intersect crosswise in the centre of the circle (which lies on the non-focal axis), these will be tangents to the conic (Art. 360). Every circle through F and F' which cuts AC and A'C thus deter- mines two tangents to the conic. The conic is an ellipse or a hyperbola according as t cuts the segment FF r externally or internally. The conic is a parabola when F f is at infinity (Fig. 225). In this case produce the axis FF to P' making FP' equal to PF, and draw P'M perpendicular to t ; then M will be the point of contact of the given tangent t. Draw MP" perpendicular to the axis ; then P and P" will be conjugate points with respect to the parabola. And since the involution of conjugate points on the axis has one double point at infinity, the middle point A of PP" will be the other double point, i.e. the vertex of the parabola. The parabola is therefore com- pletely determined, since two points on it are known (M and A), and the tangents at these points (t and the straight line drawn through A at right angles to the axis). 364. On the other hand, the problem, To construct the conic which has its foci at two given points F, F f and which passes through a given point M, which is also a determinate one, admits of two solutions. For if the locus of a point be sought the sum of whose distances from F and F' is equal to the constant value FM+F'M, an ellipse is arrived at ; but if the locus of a point be sought the difference of whose distances from i^and i^'is equal to FM^FM', a hyperbola is found. This may also be seen from the theorem of Art. 344, which shows that if the straight lines t, t' be drawn bisecting the angle FMF' (Fig. 223) each of these lines will be a tangent at M to a conic which satisfies the problem, the other line being the corresponding normal to this conic. The finite segment FF' is cut or not by the tangents according as the conic is a hyperbola or an ellipse. There will consequently be two conies which have F, F' for foci and which pass Fig. 225. 266 FOCI. [365 through M\ a hyperbola having for tangent at M that bisector t r which cuts the segment FF f , and for normal the other bisector t ; and an ellipse having t for tangent at M and t f for normal. These two conies, having the same foci, are concentric and have their axes parallel. They will cut one another in three other points besides M ; and their four points of intersection will form a rectangle inscribed in the circle of centre and radius OM ; in other words, the three other points will be symmetrical to M with respect to the two axes and the centre. This is evident from the fact that a conic is symmetrical with respect to each of its axes. 365. Through every point M in the plane then pass two conies, an ellipse and a hyperbola, having their foci at F and F'. In other words, the system of confocal conies having their foci at F and F' is composed of an infinity of ellipses and an infinity of hyperbolas ; and through every point in the plane pass one ellipse and one hyperbola, which cut one another there orthogonally and intersect in three other points. Two conies of the system which are of the same kind (both ellipses or both hyperbolas) clearly do not intersect at all. Two conies of the system however which are of opposite kinds (one an ellipse, the other a hyperbola) always intersect in four points, and cut one another orthogonally at each of them. This may be seen by observing that the vertices of the hyperbola are points lying within the segment FF', and therefore within the ellipse. On the other hand, there must be points on the hyperbola which lie outside the ellipse ; for the latter is a closed curve which has all its points at a finite distance, while the former extends in two directions to infinity. The hyperbola therefore, in passing from the inside to the outside of the ellipse, must necessarily cut it. No two conies of the system can have a common tangent ; because (Art. 363) only one conic can be drawn to have its foci at given points and to touch a given straight line. Any straight line in the plane will touch a determinate conic of the system, and will be normal, at the same point, to another conic of the system, belonging to the opposite kind. The first of these conies is a hyperbola or an ellipse according as the given straight line does or does not cut the finite segment FF' . 366. If first point F' lies at infinity, the problem of Art. 364 becomes the following : Given the axis of a parabola, the focus F, and a point M on the curve, to construct the parabola. Just as in Art. 364, there are two solutions (Fig. 226). The tangents at M to the two parabolas which satisfy the problem are the bisectors of the angle made by MF with the diameter passing through If; therefore the parabolas cut orthogonally at M and 367] FOCI. 267 consequently intersect at another point, symmetrical to M with respect to the axis. The parabolas cannot intersect in any other finite point, since they touch one another at infinity *. The tangents to the two parabolas at M cut the axis in two points P, P' which lie at equal distances on opposite sides of F; and if P" is the foot of the perpendicular let fall rom M on the axis, the vertices A , A' of the parabolas are the middle points respec- Fig. 226. of the segments PP", P'P' tively. Suppose A and P " to fall on the same side of F. Then since P'P" < P'P, and P'A' is the half of P'P", and P'F the half of P'P, therefore. P'A' OB. If OA = OB, i.e. if the hyperbola is equilateral (Art. 395), the di- rector circle reduces simply to the centre ; that is, the asymptotes are the only pair of tangents which cut at right angles. If OA < OB, the director circle has no real existence ; the hyperbola has no pair of mutually perpendicular tangents. 1/ Fig. 228. Fig. 229. 369. Consider now the case of the parabola (Fig. 229). Let F be the focus, A the vertex, TH and TK a pair of mutually perpendicular tangents. If these meet the tangent at the vertex in i^Tand irrespectively, the angles FHT , FKT will be right angles (Art. 358), so that the figure THFK is a rectangle. Therefore TH=KF; and since the triangles TEH, FAK are evidently similar, TF=AF. The locus of the point T is * De la Hire, loc. cit., lib. viii. props. 27, 28. + Gaskin, The geometrical construction of a conic section, 1852), chap. iii. prop. 10 et seqq. (Cambridge, 270 FOCI. [370 therefore a straight line parallel to HK, and lying at the same distance from HK (on the opposite side) that F does. That is to say : The locus of the point of intersection of two tangents to a para- bola which cut at right angles is the directrix*. Since the director circle of a conic is concentric with the latter, it must in the case of the parabola have an infinitely great radius. In other words, it must break up into the line at infinity and a finite straight line. And we have just seen that this finite straight line is the directrix. 370. The director circle possesses a property in relation to the self-conjugate triangles of the conic which we will now proceed to investigate. Let XYZ (Fig. 230) be a triangle which is self-conjugate with respect to a conic whose centre is 0. Join OX and let it cut YZ in X' and the conic in A'. Draw OB* parallel to YZ-, let it cut XY in L and the conic in B'\ and draw ZL' parallel to OX to meet OB' in L'. Then A' and OB ' are evidently conjugate semi-diameters; alsoXand X', L and L' are pairs of conjugate points with respect to the conic. Therefore OX. OX' = + OA' 2 , and 01 .OL' = ± 03'* t where the positive or the negative signs are to be taken according as the semidiameters OA' , 0B f are real or ideal (Art. 294). Thus for the ellipse OX . 0X F +OL.OL'= OA' 2 + OB' 2 = OA 2 + OB 2 , and for the hyperbola OX . 0X'+ OL . OL' = ±(0A /2 - OB' 2 ) = ±(OA 2 -OB 2 ), so that in both cases (Art. 368) OX.OX'+OL.OL'^OT 2 , (1) where OT is the radius of the director circle. * De la Hire, he. cit., lib. viii. prop. 26. Fig. 230. 371] FOCI. 271 Now let a circle be described round the triangle XYZ, and let U be the point where it cuts OX again ; then X'Y.X'Z=X'X.X'U; .-. ru = ^-rz ~ox xz ' (from the similar triangles OLX, X'YX) 01 or = ox 0L ■ Therefore equation (l) gives OT 2 = OX.OX'+OX.X'U = OX.OU, that is to say : The centre of a conic has with respect to the circum- scribing circle of any triangle self-conjugate to the conic a constant power, which is equal to the square of the radius of the director circle. Or in other words : The circle circumscribing any triangle which is self-conjugate with regard to a conic is cut orthogonally by the director circle *, The following particular cases of this theorem are of interest : I. The centre of the circle circumscribing any triangle which is self- conjugate with respect to a parabola lies on the directrix. II. The circle circumscribing any triangle which is self-conjugate with respect to an equilateral hyperbola passes through the centre of the conic. 371. Consider a quadrilateral circumscribed about a conic. Since each of its diagonals is cut harmonically by the other two, the circle described on any one of the diagonals as diameter is cut orthogonally by the circle which circumscribes the diagonal triangle (Art. 69). But the diagonal triangle is self-conjugate with respect to the conic (Art. 260), and therefore its circumscribing circle cuts orthogonally the director circle (Art. 370). Consequently the director circle and the three circles described on the diagonals as diameters all cut orthogonally the circle circumscribing the diagonal triangle. Now by Newton's theorem (Art. 318) the centres of the four first-named circles are collinear; and circles whose centres are collinear and which all cut the same circle orthogonally have a common radical axis. Therefore : The director circle of a conic, and the three circles described on * Gaskin, he. cit., p. 33. 272 FOCI. [372 the diagonals of any circumscribed quadrilateral as diameters, are coaxial. In the parabola the director circle reduces to the directrix and the straight line at infinity; in this case then the above theorem becomes the following : If a quadrilateral is circumscribed about a imrabola, the three circles described on the diagonals of the quadi-ilateral as diameters have the directrix for their common radical axis. 372. If in the theorem of Art. 371 the quadrilateral be supposed to be given, and the conic to vary, we arrive at the following theorem : The director circles of all the conies inscribed in a given quadri- lateral form a coaxial system, to which belong the three circles having as diameters the diagonals of the quadrilateral. There is one circle of such a system which breaks up into two straight lines : that namely which degenerates into the radical axis and the straight line at infinity. Now the director circle breaks up into two straight lines — viz. the directrix and the line at infinity — in the case of a parabola (Art. 369). Therefore the common radical axis of the system of coaxial director circles is the directrix of the parabola which can be inscribed in the quadrilateral. If the circles of the system do not intersect, there are two of them which degenerate into point-circles (the limiting points). Now the director circle degenerates into a point in the case of the equilateral hyperbola (Art. 368). Therefore when the circles do not cut one another, the two limiting points of the system are the centres of the two equilateral hyperbolas which can in this case be inscribed in the quadrilateral. If the circles do intersect, the system has no real limiting points ; and in this case no equilateral hyperbola can be inscribed in the quadrilateral. The circles which cut orthogonally the circles of a coaxial system form another coaxial system ; if the first system has real limit- ing points, the second system has not, and vice versa. In order then to inscribe an equilateral hyperbola in a given quadrilateral, it is only necessary to describe circles on two of the diagonals of the quadrilateral as diameters, and then to draw two circles cutting the former two orthogonally. When the problem is possible, these two orthogonal circles will intersect ; and their two points of intersection are the centres of the two equilateral hyperbolas which satisfy the conditions of the problem. 373. If five points are taken on a conic, five quadrangles may be formed by taking these points four and four together ; and the diagonal triangles of these five quadrangles are each of them self- conjugate with respect to the conic. If the circumscribing circles of 376] FOCI. 273 these five diagonal triangles be drawn, they will give, when taken together in pairs, ten radical axes. These ten radical axes will all meet in the same point, viz. the centre of the conic. 374. Consider again a quadrilateral circumscribing a conic ; let P and P\ Q and Q\ R and IV be its three pairs of opposite vertices. If these be joined to any arbitrary point S, and if moreover from this point S the tangents t , t f are drawn to the conic, it is known by the theorem correlative to that of Desargues (Art. 183, right) that t and t', SP and SP', SQ and SQ', SR and SR' are in involution. Now let one of the sides of the quadrilateral (say P'Q'R') be taken to be the straight line at infinity, so that the inscribed conic is a parabola ; and let 8 be taken at the orthocentre (centre of perpen- diculars) of the triangle PQR formed by the other three sides of the quadrilateral. Then each of the three pairs of rays SP and SP', SQ and SQ', SR and SR' cut orthogonally; therefore the same will be the case with the fourth pair t and t'. But tangents to a parabola which cut orthogonally intersect on the directrix (Art. 369) ; therefore : The orthocentre of any triangle circumscribing a parabola lies on the directrix. 375. If in the theorem of the last Article the triangle be supposed to be fixed, and the parabola to vary, we obtain the theorem : The directrices of all parabolas inscribed in a given triangle meet in the same point, viz. the orthocentre of the triangle. Given a quadrilateral, one parabola (and only one) can always be inscribed in it. By taking the sides of the quadrilateral three and three together, four triangles are obtained ; and the four ortho- centres of these triangles must all lie on the directrix of the parabola. It follows that Given four straight lines, the orthocentres of the four triangles formed by taking them three and three together are collinear. 376. Let C be any given conic, and let C r be its polar reciprocal with respect to an auxiliary conic K. The particular case in which K is a circle whose centre coincides with a focus F of the conic C is of great interest ; we shall now proceed to consider it. If r, / be any two straight lines which are conjugate with respect to C, and if R , R' be their poles with respect to K, it is known (Art. 323) that R , R r will be conjugate points with respect to C '. Consider now two such lines r , / which pass through F; they will be at right angles since every pair of conjugate lines through a focus cut one another orthogonally. T 274 FOCI. [377 They will therefore be perpendicular diameters of the circle K, and their poles B , R' with respect to K will be the points at infinity on /, r respectively. These points are conjugate with respect to C', and the straight lines joining them to the centre of this conic are therefore a pair of conjugate diameters of C'; consequently two conjugate diameters of C' are always mutually perpendicular. This proves that C' is a circle; i.e. the polar reciprocal of a conic, with respect to a circle which has its centre at one of the foci, is a circle. By taking the steps of the above reasoning in the opposite order, the converse proposition may be proved, viz. The polar reciprocal of a circle with respect to an auxiliary circle is a conic having one focus at the centre of the auxiliary circle. As in Art. 323, it is seen that the conic is an ellipse, a hyperbola, or a parabola, according as the centre of the auxiliary circle lies within, without, or upon the other circle. 377. If d be the directrix of the conic C corresponding to the focus F, and if its pole be taken with respect to the circle K, this point will evidently be the centre of the circle C' (Art. 323). The radius of the circle C' may also easily be found. For in Fig. i\6 let two points X, X' be taken in the latus rectum LFL' such that FX.FL = FX'.FL'=k 2 , where k denotes the radius of the circle K ; and let straight lines be drawn through X and X' perpendicular to XFX'. These straight lines are evidently parallel tangents of the circle C', and the distance XX' between them is therefore equal in length to the diameter of C'. But \X , X=FX=^ i FL so that the radius of the circle C' is equal to -^ • The eccentricity e of the conic C may be expressed in a simple manner in terms of quantities depending upon the two circles K and C'. For if 0' be the centre and p the 379] FOCI. 275 radius of the latter circle, it has been seen that the directrix is the polar of 0' with respect to K ; therefore (Fig. 216) FD.FO'=k 2 . But it has just been proved that FL . p = k 2 ; FL FO' therefore (Art. 351), € = ^tft = x tit p 378. The proposition of Art. 376 may be proved in a different manner, so as to lead at once to the position and size of the circle C'. Take any point if on the (central) conic C (Fig. 217); from the focus F draw FU perpendicular to the tangent at 31, and on FU take a point Z such that FZ . FU= k 2 , k being as before the radius of the circle K. Then the locus of Z is the polar reciprocal of C with respect to K. Now it is known (Arts. 356, 357) that CHies on the circle on A A ' as diameter, and that if UF cut this circle again at U' FU.FU'= + OB 2 . Therefore FZ :FU' = k 2 : + OB 2 ; which proves (Art. 23 [6]) that the locus of Z is a circle whose centre 0' lies on FO and divides it so that FO f :FO = k 2 : OB 2 , OA and whose radius p is equal to k 2 . jp^ , that is, (Art. 352 Cor.) to ~ And again, since OF. 0D= OA 2 and FD = FO+ 02, (Figs. 214, 215), .-. F£.FO=OF 2 -OA 2 = + OB 2 =Jc 2 ~, by what has just been proved. .-. FO'.FD = Jc 2 ; i.e. 0' is the pole of the directrix d with respect to K. In the particular case where h = OB, p = OA ; that is to say : The polar reci])rocal of an ellipse (hyperbola) with respect to a circle having its centre at a focus and its radius equal to half the minor {conjugate) axis is the circle described on the major {transverse) axis as diameter. 379. In the case where C is a parabola, let M be any point on the curve (Fig. 218); let fall FN perpendicular to the tangent at 11, and take on FN a point Z such that FZ. FN=& 2 . Then, T 2 276 FOCI. [379 as before, the locus of Z will be the polar reciprocal of C with respect to K. Draw ZQ perpendicular to ZF to cut the axis of the parabola in Q. Then a circle will evidently go round QANZ, so that FA.FQ = FN.FZ=k 2 ; therefore Q is a fixed point, and the locus of Z is the circle on QF as diameter. If 0' be the centre, p the radius of this circle, In the particular case where k is equal to half the latus rectum, that is, to 2 FA , we have p = k ; that is to say : The polar reciprocal of a parabola with respect to a circle having its centre at the focus and its radius equal to half the latus rectum is a circle of the same radius, having its centre at the point of intersection of the axis with the directrix. CHAPTEE XXIV. COROLLARIES AND CONSTRUCTIONS. 380. In the theorem of Art. 275 suppose the vertices B and C of the inscribed triangle ABC (Fig. 188) to be the points at infinity on a hyperbola ; then S will be the centre of the curve, and the theorem will become the following : If from any point ion a hyperbola parallels be drawn to the asymptotes, they will meet any given diameter in two points #and G which are conjugate to one another with regard to the curve. Or : If through two points lying on a diameter of a hyperbola, which are conjugate to one another with regard to the curve, parallels be drawn to the asymptotes, they will intersect on the curve. From this follows a method for the construction of a hyperbola by points, having given the asymptotes and a point M on the curve. On the straight line SM, which joins if to the point of inter- section S of the asymptotes, take two conjugate points of the in- volution determined by having S for centre and M for a double point. These points will be conjugate to one another with respect to the conic (Art. 263) ; if then parallels to the asymptotes be drawn through them, the two vertices of the parallelogram so formed will be points on the hyperbola which is to be constructed. 381. Let similarly the theorem of Art. 274 be applied to the hyperbola, taking the sides b and c of the circumscribed triangle abc to be the asymptotes ; it will then become the following : If through the points where the asymptotes are cut by any tangent to a hyperbola any two parallel straight lines be drawn, these will be conjugate to one another with respect to the conic. Or : Two parallel straight lines which are conjugate to one another with respect to a hyperbola cut the asymptotes in points, the straight lines joining which are tangents to the curve. From this we deduce a method for the construction, by means of its tangents, of a hyperbola, having given the asymptotes b and c and one tangent m. Draw parallel to m two conjugate rays of the involution (Art. 129) determined by having m for a double ray and the parallel diameter for central ray. The two straight lines so drawn will be conjugate to one another with respect to the conic ; if then the points where they cut the asymptotes be joined to one another, we shall have two tangents to the curve. 278 COROLLARIES AND CONSTRUCTIONS. [382 382. Let B and C be any two points on a parabola, and A the point where the curve is cut by the diameter which bisects the chord BC. Let F and G be two points lying on this diameter which are conjugate with respect to the parabola, i.e. two points equidistant from A (Art. 142) ; by the theorem of Art. 275, BF and CG, and likewise BG and CF, will meet on the curve. This enables us to construct by points a parabola which circum- scribes a given triangle ABC and has the straight line joining A to the middle 'point of BG as a diameter. Or we may proceed according to the following method : On BC take two points H and H / which shall be conjugate to one another with regard to the parabola, i.e. any two points dividing BC harmonically. Since H and H' are collinear with the pole of the diameter passing through A, therefore by the theorem of Art. 275, a point on the parabola will be found by constructing the point of intersection of AH with the diameter passing through W , and another will be found as the point where AW meets the diameter passing through H. 383. In the theorem of Art. 274 suppose the tangent c to lie at infinity ; then we see that If a and b are two tangents to a parabola, and if from any point on the diameter passing through the point of contact of a there be drawn two straight lines, one passing through the point ab and the other parallel to b, these wiH be conjugate to one another with regard to the parabola. This enables us to construct by tangents a parabola, having given two tangents a and t, the point of contact A of one of them a. and the direction of the diameters. Draw the diameter through A and let it meet t in ; the second tangent t' from will be the straight line which is harmonically conjugate to t with respect to the diameter OA (the polar of the point at infinity on a) and the parallel through to a. If now two straight lines /* and h' be drawn through which shall be conjugate to one another with regard to the parabola, i.e. two straight lines which are harmonic conjugates with regard to t and t f , the parallel to $ drawn from the point ha and the parallel to h drawn from the point h r a will both be tangents to the required parabola. 384. If in the theorem of Art. 274 the straight line a be supposed to lie at infinity, and b and c to be two tangents to a parabola, we obtain the following : The parallels drawn to two tangents to a parabola, from any point on their chord of contact, are conjugate lines with regard to tJie conic. By another application of the same theorem we deduce a result already proved in Art. 178, viz. that 385] COROLLARIES AND CONSTRUCTIONS. 279 If, from a point on the chord of contact of a pair of tangents b and c to a parabola, two straight lines h and h' be drawn parallel to b and c respectively, the straight line joining the points he and h'b will be a tangent to the curve *. From this may be deduced a construction for the tangents to a parabola determined by two tangents and their points of contact. 385. Theorem. If a conic cut the sides BC , CA , AB of a triangle ABC in the points B and B', E and E', F and F' re- spectively, then will BB.BB' CE.CE' AF.AF' _ cb.cb'' ae.aw"bf7bF~ y r\ This celebrated theorem is due to Carnot f. Consider the sides of the triangle ABC (Fig. 231) as Fig. 231. cut by the transversals BE and B' E' in the points B and B', i?and E', G and G'; by the theorem of Menelaus (Art. 139) BBCJ^AG_ CB' AE'BG ~ T ' ( ' BB' CE' AG' and = 1. CB' AE' BG' Again, BEE'B' is a quadrangle inscribed in the conic, and by Desargues' theorem (Art. 183) the transversal^^ meets the opposite sides and the conic in three pairs of points in involu- tion; therefore (Art. 130) the anharmonic ratios (ABFG) and (BAF'G') are equal; thus (Art. 45) (ABFG) = (ABG'F'), or (ABFG): (ABG'F')=i, which gives AF.AF' AG. AG' .... (4) BF.BF'' BG. De la Hire, loc. cit., lib. iii. prop. 21. = I. BG' + G&omitrie de position, p. 437. 280 COROLLARIES AND CONSTRUCTIONS. [386 Multiplying together (2), (3), and (4), we obtain the relation stated in the enunciation*. 386. Conversely, if on the sides BC \CA,AB respectively of a triangle ABC there he taken three pairs of points B and B\ B and B', F and F' such that the segments determined by them and the vertices of the triangle satisfy the relation (1) of Art. 385, these six 'points lie on a conic. For let the conic be drawn which passes through the five points B,B', E, B\ F, and let F" be the point where it cuts AB again. By Carnot's theorem a relation holds which differs only from (l) in that it has F" in the place of F'. This relation, combined with (1), gives AF':BF'=AF":BF", whence (ABF'F") = i ; and therefore (Art. 72, VII) F" coincides with F f . * Carnot's theorem, being evidently true for the circle (since in this case BD . BD' = CD . CD'> &c), may be proved without making use of involution properties as follows : Let I , J, K be the points at infinity on BO , CA , AB respectively, and sup- pose Fig. 231 to have been derived by projecting from any vertex on any plane a triangle A X B X C X whose sides are cut by a circle in D x and D x ', E x and E x ', F x and F' respectively. Let J, , /„ K x be the points on the sides B X C X , C X A X , A V B 1 which project into I, J, Irrespectively; they will of course be collinear. Then BD So CD m (BCDI) (Art. 64) = (B l C l DJ 1 ) (Art. 63) B X D X B X I X ~~C 1 D 1 ''C 1 I l ' BD BD' CD' .BD' B X D X ' ~ C X D{ B X D X B 1 I 1 ■ B X D X ' B X I X 2 CD, CD' ~ C X D X . CJ? " BJ*' C x D x ' (Euc, ' c x i x * i". 35, 36.) CE AE AF .CE' .AE' .AF' A X J X * B X K* Similarly, and BF.BF'-AKr Multiplying these three equations together, and remembering that by the theorem of Menelaus the product on the right-hand side is equal to unity, we have the result required. Carnot's theorem is true not only for a triangle but for a polygon of any num- ber of sides ; the proof just given can clearly be extended so as to show this, the theorem of Menelaus being capable of extension to the case of a polygon. Menelaus' theorem is included in that of Carnot. It is what the latter reduces to when the conic degenerates into two straight lines of which one lies at infinity. 387] COROLLARIES AND CONSTRUCTIONS. 281 387. If the point A pass off to infinity (Fig. 232) the ratios AF: AE and AF' : AE / become in the limit each equal to unity, and the equation (1) of Art. 385 accordingly reduces to BD . BD' CF.CF' _ CD.CD'' BF.BF'" T * • W Draw parallel to BC a straight line to cut CFE' in Q and the conic in P and P'\ the preceding equation, applied to the triangle whose vertices are C, Q, and the point at infinity where PP' and BC meet, gives QE.QE' CD . CD' _ CE.CE'' QP.QP'~ *' Fig. 232. Multiplying together these last two equations, we obtain BD.BD' _ QP.QP\ BF.BF f ~ QE.QE'' that is to say : If through any point Q there be drawn in given directions two transversals to cut a conic in P,P' and E,E' respectively, then the rectangles QP . QP' and QE . QE' are to one another in a constant ratio* f. * Apollonius, loc. city lib. iii. 16-23. Desargues, loc. cit., p. 202. De la Hibe, loc. cit., bk. v. props. 10, 12. f From this follows at once the result already proved in a different manner in Art. 316, viz. that if a conic is cut by a circle, the chords of intersection make equal angles with the axes. For let P , P', E, E' be the points of intersection of a circle with the conic ; then (Euc. iii. 35) QP . QP' = QE . QE'. But if MOM', NCN' be the diameters of the conic parallel respectively to QPP' and QEE', we have, by the theorem in the text, QP.QP'-.QE. QE' = CM. CM': CN.CN' = CM* : CN*. Therefore CM = CN, and consequently CM and CN (and therefore also QPP' and QEE') make equal angles with the axes. 282 COROLLARIES AND CONSTRUCTIONS. [388 388. Suppose in equation (5) of Art. 387 that the conic is a hyperbola and that in place of BC is taken an asymptote RK of the curve; then the ratio RB .RB' : KB .KB' becomes equal to unity, and therefore rf.hf'Lke.ke', that is to say : If through any point R [or R') lying on an asymptote there be drawn, parallel to a given straight line, a transversal to cut a hyper- bola in two points F and F' (B and B'), then the rectangle HF.RF' (R'B . R'B') contained by the intercepts will be constant. If the diameter parallel to the given direction R'B meets the curve, then if S and 8' are the points where it meets it, and if is the centre, R'B . R'B' = 08.08' = - OS 2 . If the diameter OT parallel to the given direction RF does not meet the curve, a tangent can be drawn which shall be parallel to it. The square on the portion of this tangent intercepted between its point of contact and the asymptote will be equal to the rectangle RF. RF' by the theorem now under consideration; but this portion is (Art. 303) equal to the parallel semidiameter OT; therefore RF. RF'= OT 2 , or : If a transversal, cut a hyperbola in F and F' (in B and B') and an asymptote in H (in R'), the rectangle HF.RF' (R'B .R'B') is equal to + the square on the parallel semidiameter OT (08); the positive or negative sign being taken according as the curve has or has not tangents parallel to the transversal. 389. If the transversal cuts the other asymptote in L (ml'), then by Art. 193 HF'=FL or H'B'=BL', and consequently FH. FL = - OT 2 or BR'. BL' = OS 2 ; therefore : If a transversal drawn from any point F (B) on a hyperbola cut the asymptotes in H and L (in H' and L'), the rectangle FH . FL (BR'. BL') is equal to + the square on the parallel semidiameter ; the negative or positive sign being taken according as the curve has or has not tangents parallel to the transversal. 391] COROLLARIES AND CONSTRUCTIONS. 283 390. From the proposition of the last Article may be deduced a construction for the axes of a hyperbola, having given a pair of conjugate semidiameters OF and OT in magnitude and direction (Fig. 233). We first construct the asymptotes. Of the two given semidiameters, let OF be the one which cuts the curve. Draw through F a parallel to OT', this will be the tangent at F. Take on this parallel FP and FQ each equal to OT ; then OP and OQ will be the asymptotes (Art. 304). In order now to obtain the directions of the axes, we have only to find the bisectors of the angle included by the asymptotes, or, in other words, the two perpendicular rays OX, Y which are conjugate to one another in the in- volution of which OP and OQ are the double rays (Arts. 296, 297). To determine the lengths of the axes, draw through F a parallel to OX, and let it cut the asymptotes in B and B'\ and on OX take OS the mean proportional between FB and FB'. Then will OS be the length of the semiaxis in the direction OX; and OX will or will not cut the curve according as the segments FB , FB f have or have not the same direction. Again, construct the parallelogram of which OS is one side, which has an adjacent side along OY, and one diagonal along an asymptote ; its side OR will be the length of the semiaxis in the direction OY (Art. 304). 391. In the plane of a triangle ABC take any two points and 0'; if OA , OB , 00 meet the respectively opposite sides BO, CA, AB of the triangle in D, E, F, Ceva's theorem (Art. 137) gives BD CE AF_ _ cd'ae'bf~ l ' Similarly, if (/A , O'B , O'C meet the respectively opposite sides in D',E',F', then BD' CE' AF' _ CD'' AE''BF' ~ *' If these equations be multiplied together, equation (1) of Art. 385 is obtained ; therefore : If from any two points the vertices of a triangle are projected upon the respectively opposite sides, the six points so obtained lie on a conic. For example, the middle points of the sides of a triangle and the feet of the perpendiculars from the vertices on the opposite sides are six points on a conic *. * This conic is a circle (the nine-point circle). See Steiner, Annales de Mathe- matiques (Montpellier, 1828), vol. xix. p. 42 ; or his Collected Works, vol. i. p. 195. 284 COROLLARIES AND CONSTRUCTIONS. 392. Pkoblem. To construct a conic which shall jmss through three given points A , B , G, and with regard to which the pairs of corre- sponding points of an involution lying on a given straight line u shall he conjugate points. Let AB and AC (Fig. 234) be joined, and let them meet u in D and E. Let the points corresponding in the involution to D and E respectively be D' and E'\ let D" be the harmonic conjugate of D Fig. 234. with respect to A and B, and let E" be the harmonic conjugate of E with respect to A and C. Thus D will be conjugate (with respect to the required conic) both to D' and to D", and therefore D'D" will be the polar of D s . So too E'E" vti\\ be the polar of E. Join BE , CD, and let them cut E'E" and D'D" in E and D respectively ; then E will be conjugate to E and D to D. If then two points B', C be found such that the ranges BB'EE and CC'DD are harmonic, they will both belong to the required conic. In the figure, F and F', G and G' are the pairs of points which determine on u the involution of conjugate points. 393. Pkoblem. To construct a conic which shall pass through four given points Q , R , S , T and shall divide harmonically a given seg- ment MN (Fig. 235). Let the pairs of ^opposite sides of the quadrangle QRST meet the straight line MN in A and A', B and B'. If the required conic cuts MN, the two points of inter- section will be a pair of the invo- lution determined by A and A', B and B' (Art. 183). If then the involution of which M and N are the double points and the involution determined by the pairs of points A and A', B and B' have a pair P and P' in common, the required conic will pass through each of the points P and P' (Arts. 125, 208). 395] COROLLARIES AND CONSTRUCTIONS. 285 In order to construct these points, describe any circle (Art. 208) and from any point on it project the points A , A ', B , B', M , N upon the circumference, and let A x , A x , B 1 , B\ , M x , N x be their respective projections. If the chords A X A X and B X B( meet in V, and the tangents at M x and N x meet in U, all straight lines passing through U determine on the circumference, and consequently (by projection from 0) on the straight line MN, pairs of conjugate points of the first involution, and the same is true, with regard to the second involution, of straight lines passing through V. If the straight line UV meets the circle in two points P x and P x , let these be joined to 0; the joining lines will cut MN in the required points P and P f . Let W be the pole of UV with respect to the circle. Every straight line passing through W and cutting the circle determines on it two points which are harmonically conjugate with regard to P x and P x ; and these points, when projected from on MN, will give two points which are harmonically conjugate with regard to P and P', and which are therefore conjugate to one another with respect to the required conic. If then UV does not cut the circle, so that the points P and P f cannot be constructed, draw through W two straight lines cutting the circle, and project the points of inter- section from the centre upon the straight line MN ; this will give two pairs of points which will determine the involution on MN of conjugate points with respect to the conic. The problem therefore reduces to that treated of in the preceding Article. 394. Problem. To construct a conic which shall pass through four given points Q , R , S , T, and through two conjugate points [which are not given) of a known involution lying on a straight line u. This problem is similar to the preceding one ; since it amounts to constructing the pair of conjugate points common to the given involution and to that determined on u by the pairs of opposite sides of the quadrangle QRST (Art. 183). Such a common pair will always exist when the given involution has no double points ; and the two points composing it will both lie on the required conic. If the given involution has two double points M and N, the present problem becomes identical with that of Art. 393. The problem clearly admits of only one solution, and the same is the case with regard to those of the two preceding Articles. 395. Consider a hyperbola whose asymptotes are perpen- dicular to one another, and to which, on this account, is given the name of rectangular hyperbola (Fig. 236). Since the asymptotes are harmonically conjugate with regard to any pair of conjugate diameters (Art. 296), they will in 286 COROLLARIES AND CONSTRUCTIONS. [395 this case be the bisectors of the angle included between any such pair (Art. 60). But the parallelogram described on two conjugate semidiameters as adjacent sides has its diagonals parallel to the asymptotes (Art. 304); in this case therefore every such parallelogram is a rhombus ; that is, every diameter is equal in length to its conjugate. On account of this property the rectangular hyper- bola is also called equilateral*. I. Since the chords joining the extremities P and P' of any diameter to any point M on the curve are parallel to a pair of conjugate diameters (Art. 287), the angles made by PM and Fig. 236. P'M with either asymptote are equal in magnitude and of opposite sign. If the points P and P' remain fixed, while M moves along the curve, the rays PM and P'M trace out two pencils which are oppositely equal to one another (Art. 106). II. Conversely, the locus of the points of intersection of pairs of corresponding rays of two oppositely equal pencils is an equilateral hyperhola. For, in the first place, the locus is a conic, since the two pencils are projective (Art. 1 50). Further, the two pencils have each a pair of rays which include a right angle, and which are parallel respectively to the corresponding rays of the other pencil (Art. 106); the conic has thus two points at infinity lying in directions at right angles to one another, and is there- fore an equilateral hyperbola. It will be seen moreover that the centres P and P f of the two pencils are the extremities of a diameter. For the tangent p at P is the ray corresponding to P'P regarded as a ray p 'of the second pencil, and the tangent q' at P' is the ray corresponding to PP' regarded as a ray q of the first pencil (Art. 150); but the angles^ andj/^' must be equal and opposite; therefore, since p' and q coincide, p and q' must be parallel to one another. III. The angular points of a triangle ABC and its ortho- centre (centre of perpendiculars) B are the vertices of a * Apollonius, loc. cit., vii. 21. De la Hire, loc. cif., book v. prop. 13. 395] COROLLARIES AND CONSTRUCTIONS. 287 complete quadrangle in which each side is perpendicular to the one opposite to it, and whose six sides determine on the straight line at infinity three pairs of points subtending each a right angle at any arbitrary point 8. The three pairs of rays formed by joining these points to 8 belong therefore to an in- volution in which every ray is perpendicular to its conjugate (Arts. 131 left, 124, 207). But this involution of rays projects from S the involution of points which, in accordance with Desargues' theorem, is determined on the straight line at infinity by the pairs of opposite sides of the quadrangle and by the conies (hyper- bolas*) circumscribed about it. The pairs of conjugate rays therefore of the first involution give the directions of the asymptotes of these conies ; thus : If a conic pass through the angular points of a triangle and through the orthocentre, it must be an equilateral hyperbola-^. IV. Conversely, if an equilateral hyperbola be drawn to pass through the vertices A,B ,C of a triangle, it will pass also through the orthocentre D. For imagine another hyper- bola which is determined (Art. 162, I) by the four points A, B , C , D and by one of the points at infinity on the given hyperbola. This new hyperbola will be an equilateral one by the foregoing theorem, and will consequently pass through the second point at infinity on the given curve ; and since the two hyperbolas thus have five points in common (A,B,C, and two at infinity) they must be identical; which proves the proposition. Therefore : If a triangle be inscribed in an equilateral hyperbola, its ortho- centre is a point on the curve. V. If the point B approach indefinitely near to A, i. e. if BAC becomes a right angle, we have the following proposition : If EFG (Fig. 236) is a triangle, right-angled at B, which is * No ellipse or parabola can be circumscribed about the quadrangle here con- sidered (Art. 219). f This may be deduced directly from Pascal's theorem. For let a conic be drawn through A,B,C,D, and let I x and I 2 be the points where it meets the line at infinity. Since ABCD^Iiis a hexagon inscribed in a conic, the inter- sections of AB and DI lt of BC and I X I 2 , and of CD and I 2 A, are three collinear points. Therefore the straight line joining the point in which D/j meets AB to that in which AI 2 meets CD must be parallel to BC. Thus AI 2 must be at right angles to Dl x , and as these lines are parallel to the asymptotes of the conic the latter is a rectangular hyperbola. 288 COROLLARIES AND CONSTRUCTIONS. . [396 inscribed in an equilateral hyperbola, the tangent at E is perpendicular to the hypotenuse FG. VI. Through four given points Q,R,S,T can be drawn only one equilateral hyperbola (Art. 394). The orthocentre of each of the triangles QRS , BST ', STQ , QRT lies on the curve * VII. Given four tangents to an equilateral hyperbola, to construct the curve. Since the diagonal triangle of the quadrilateral formed by the four tangents is self-conjugate with respect to the hyperbola, the centre of the latter will lie on the circle circumscribing this triangle (Art. 370, II). But the centre of the hyperbola lies also on the straight line which joins the middle points of the diagonals of the quadrilateral (Art. 318, II). Either of the points of intersection of this straight line with the circle will therefore give the centre of an equilateral hyperbola satisfying the problem ; there are therefore two solutions. For another method of solution see Art. 372. VIII. The polar reciprocal of any conic with respect to a circle K having its centre on the director circle is an equilateral hyperbola. For since the tangents to the conic from the centre of the circle K are mutually perpendicular, the conic which is the polar reciprocal of the given one must cut the straight line at infinity in two points subtending a right angle at 0. That is to say, it must be an equilateral hyperbola. 396. Suppose given a conic, a point JS, and its polar s ; and let a straight line passing through JS cut the conic in A and A'. Let the figure he constructed which is homological with the given conic, JS being taken as centre of homology, s as axis of homology, and A , A' as a pair of corresponding points. Then every other point B r which corresponds to a point B on the conic will lie on the conic itself. For if AB meets the axis s in P, then B', the point of intersection of SB and A'P, is likewise a point on the conic (Art. 250). The curve homological with the given conic will therefore be the conic itself. Any two corresponding points (or straight lines) are separated har- monically by JS and s ; this is, in fact, the case of harmonic homology (Arts. 76, 298). To the straight line at infinity will therefore correspond the * These theorems are due to Brianchon and Poncelet ; they were enunciated in a memoir published in vol. xi. of the Annales de Mathe'maUques (Montpellier, 1 821), and were given again in vol. ii. (p. 504) of Poncelet's Applications a" Analyse et de G6ometrie (Paris, 1864). 397] COROLLARIES AND CONSTRUCTIONS. 289 ib) ls: straight line j which is parallel to s and which lies midway between S and s ; and the points in which j meets the conic will correspond to the points at infinity on the same conic. From this may be derived a very simple method of determining whether a given arc of a conic, however small, belongs to an ellipse, a parabola, or a hyperbola. Draw a chord s joining any two points in the arc ; construct its pole S, and draw a straight line j parallel to s and equidistant from S and s. If j does not cut the arc, the latter is part of an ellipse (Fig. 237 a). If j touches the arc at a point J, the arc belongs to a parabola of which SJ is a diameter (Fig. 237 b). If, finally, j cuts the are in two points J x , J 2 (Fig. 237c), the arc will be part of a hyperbola whose asymptotes are parallel to 8J X and SJ 2 *. 397. Pkoblem. Given a tangent to a conic, its point of contact, and the position (but not the magnitude) of a pair of conjugate diameters; to construct the conic (Fig. 238). Suppose the point of intersection of the given diameters, and P and Q the points in which they are cut by the given tangent. Through the point of contact M of this tangent draw parallels to OQ , OP to meet OP, OQ in P' and Q' respectively. Since the polar of M (the tangent) passes through P, the polar of P will pass (C) / Fig. 237. Fig. 238. through M ; and since the polar of P is parallel to OQ, it must be MP') therefore P and P' are conjugate points. If now points A and A' be taken on OP such that OA and OA' may each be equal to the mean proportional between OP and OP' } then AA f will be equal in length to the diameter in the direction OP (Art. 290). In the same way the length of the other diameter BB' will be found by making OB and OB f each equal to the mean pro- portional between OQ and OQ'. * Poncelet, loc. cit., Arts. 225, 226. u 290 COROLLARIES AND CONSTRUCTIONS. If the points P and P' fall on the same side of 0, the involution of conjugate points has a pair of double points A and A' (Art. 128); that is to say, the diameter OP meets the curve. If, on the other hand, P and P' lie on opposite sides of 0, the involution has no double points, and the diameter OP does not meet the curve. In this case A and 4' are two conjugate points lying at equal distances from 0. The figure shows two cases : that of the ellipse (a) and that of the hyperbola (6). 398. Pboblem. Given a point M on a conic and the positions of two pairs of conjugate diameters a and RN\ this case the conic does not cut any of the sides of the self- conjugate triangle ; therefore (Art. 262) it does not exist. If the point P lies outside the triangle, one only of the three points A', B', C lies on the corresponding side; the two others lie on the respective sides produced. If these two other sides are cut by p , none of the involutions possesses double points, and the conic does not exist. If on the other hand p cuts the first side, or if p lies entirely outside the triangle, the conic exists, and may be constructed as above. In all cases, whether the conic has a real existence or not, the polar system (Art. 339) exists. It is determined by the self-conjugate triangle EFG, the point P, and the straight line p. To construct this system is a problem of the first degree, while the construction of the conic is a problem of the second degree. 404. Pkoblem. Given a pentagon ABODE, to describe a conic with regard to which each vertex shall be the pole of the opposite side *. Let F be the point of intersection of AB and CD . If the conic K be constructed (Art. 403) with regard to which ADF is a self- conjugate triangle and E the pole of BC, then the points B and C in which BC is cut by AF and DF respectively will be the poles of ED and EA, the straight lines which join E to the points D and A respectively. Every vertex of the pentagon will therefore be the pole of the opposite side ; that is, K will be the conic required. If the conic C be constructed which passes through the five vertices of the pentagon, and also the conic C ' which touches the five sides of the pentagon (Art. 152), these two conies will be polar reciprocals one of the other with respect to K (Art. 322). 405. Pkoblem. Given Jive points A , B , C , D , E (no three of which are collinear), to determine a point M such that the pencil MiABCDE) shall be projective with a given pencil abede (Fig. 244). Through D draw two straight lines DD', DE' such that the pencil D (ABCD'E') is projective with abede (Art. 84, right). Construct the point E' in which DE' meets the conic which passes through the four points ABCD and touches DD' at D (Art. 165) ; then construct the point M in which the same conic meets EE'. M will be the point required. For since M , A , B , C , D , E' lie on the same conic, the pencil M(ABCDE') is projective with the pencil D (ABCD'E'), Fig. 244. * Staudt, loc. cit., Arts. 238, 258. 294 COROLLARIES AND CONSTRUCTIONS. [406 which by construction is projective with the given pencil abcde. Since then ME' and ME are the same ray, the problem is solved. As an exercise may be solved the correlative problem, viz. Given Jive straight lines a , b , c , d , e, no three of which are con- current, to draw a straight line m to meet them in Jive points forming a range projective with a given range ABCDE *. 406. Peoblem. To trisect a given arc AB of a circled. On the given arc take (Fig. 245) a point N, and from B measure in the opposite direction to AN an arc BN' equal to twice the arc AN. If BT be the tangent at B, and if be the centre of the circle rig- 245. of which the arc AB is a part, the angles AON and TBN f are equal and opposite. If N and N' vary their positions simultaneously, the rays ON and BN' will describe two oppositely equal pencils, and the locus of their point of intersection M will therefore (Art. 395, II) be an equilateral hyperbola passing through and B. The asymptotes of this hyperbola are parallel to the bisectors of the angle made by AO and i^with one another ; for these straight lines are correspond- ing rays (being the positions of the variable rays ON and BN' for which the arcs AN and BN' are each zero). The centre of the hyperbola is the middle point of the straight line OB which joins the centres of the two pencils. The hyperbola having been constructed by help of Pascal's theorem, the point P will have been found in which it cuts the arc AB. Two corresponding points N and N' coalesce in this point ; therefore the arc AP is half of the arc PB, and P is that point of trisection of the arc AB which is the nearer to A. The hyperbola meets the circle in two other points E and Q. The point R is one of the points of trisection of the arc which together * Staudt, loc. cit., Art. 263. t Staudt, Beitrdge, Art. 432. Chasles, Sections coniques, Art. 37. 408] COROLLARIES AND CONSTRUCTIONS. 295 with AB makes up a semicircle ; and the point Q is one of the points of tri section of the arc which together with AB makes up the cir- cumference of the circle. 407. It has been seen (Art. 191) that if P', P", Q\ Q" (Fig. 246) are four given collinear points, and if any conic be described to pass through P r and P", and then a tangent be drawn to this conic from Q' and another from Q", the chord joining the points of contact of these tangents passes through one of the double points M ', N' of the involution which is determined by the two pairs of points P f and P", Q' and Q". The two tangents which can be drawn from Q', combined with the two from Q", give four such chords of contact, of which two pass through M* and two through N'. From this may Fig. 246. be deduced a construction for the double points of the involution P'P", Q'Q", or, what is the same thing (Art. 125), for the two points M', N' which divide each of the two given segments P'P" and Q'Q" harmonically. Describe any circle to pass through P f and P", and draw to it from Q' the tangents f and u', and from Q" the tangents f and u". The chord of contact of the tangents t' and t fr and that of the tangents u r and u" will cut the straight line P'P" in the two required points M' and N'. 408. This construction has been applied by Brianchon * to the solution of the two problems considered in Art. 221, viz. I. To construct a conic of which two points P', P y/ and three tangents q , q', q" are given. Join P'P", and let it cut the three given tangents in Q , #', Q" respectively (Fig. 246). Describe any circle through P', P" and draw to it tangents from Q, Q', Q /f . ' The chords which join the points of contact of the tangents from Q" to the points of contact of the tangents from Q meet P'P" in two points M and N\ and simi- larly the tangents from Q" combined with those from- Q' determine two points M ' and N'. The chord of contact of the tangents q\ q" to the required conic will therefore pass through one of the points M , N, and that of the * Brianchon, loc. cit., pp. 47, 51. 296 COROLLARIES AND CONSTRUCTIONS. [409 tangents q f , q" will pass through one of the points M', N\ The four combinations MM ', MN\ NM', NN' give the four solutions of the problem. The problem therefore reduces to the following : To describe a conic which shall touch three given straight lines q , q', q" in such a way that the chords of contact of the two pairs of tangents q , q" and q\ q" shall pass respectively through two given points M and M' . Let QQ'Q" (Fig. 247) denote the triangle formed by the three given tangents, and let A , A', A" be the points of contact to be deter- mined. By a corollary to Desargues' theorem (Art. 194), the side q = Q'Q" is divided harmonically at the point of contact A and at the point where it is cut by the chord A' A". If these four harmonic points be projected on MQ" from A" as centre, it follows that the segment RQ" intercepted on MQ" between q" and qf is divided harmonically by M and the chord A' A". Let then MQ" be joined ; it will cut q" in some point R \ and let the point V be determined which is harmonically conjugate to M with regard to R and Q". In order to do this, draw through M any straight line to cut q" and q' in S and T respectively ; join SQ" and TR, meeting in TJ\ and join QU, meeting RQ" in V. Join VM'\ it will meet q f and q" in A' and A" ; and finally if MA" be joined, it will cut Q'Q" in A. II. To construct a conic qf which three points P , P\ P" and two tangents q , qf are given. Join PP', and let it meet q and q' in Q and Q' respectively; join PP", and let it meet q and q / in R and R' respectively. Describe a circle round PP'P'\ and to it draw tangents from Q and Q'; the chords of contact will meet PP' in two points M and N. Similarly draw the tangents from R and R / ; the chords of contact will meet PP" in two other points M / and N'. Then each of the straight lines MN\ NN\ M'N ', MM' will meet the tangents q and q' in two of the points of contact of these two tangents with a conic circumscribing the triangle PP'P". This construction differs from that given in Art. 221 (left) only in the method of finding the double points M and N, M f and N'. 409. Theorem. If two angles AOS and AO'S of given magnitude turn about their respective vertices and 0' in such a way that the point of intersection S qf one pair of arms lies always on a fixed straight line u, the point of intersection of the other pair of arms will describe a conic (Fig. 248). 411] COROLLARIES AND CONSTRUCTIONS. 297 The proof follows at once from the property that the pencils traced out by the variable rays OA and OS, OS and O'S , O'^and O'A are projective two and two (Arts. 42, 108), and that consequently the pencils traced out by OA and O'A are projective. This theorem is due to Newton, and was given by him under the title of The Organic De- scription of a conic *. 410. The following, which depend on the foregoing theorem, may serve as exercises to the student : — 1. Deduce a construction for a conic passing through five given points , 0', A , B , C. 2. Given these five points, determine the magnitude of the angles AOS , AO'S and the position of the straight line u in order that the conic generated may pass through the five given points. 3. On the straight line 00 ' which joins the vertices of the two given angles a segment of a circle is described containing an angle equal to the difference between four right angles and the sum of the given angles. Show that according as the circle of which this seg- ment is a part cuts, does not cut, or touches the straight line u, so the conic generated will be a hyperbola, an ellipse, or a parabola. ' 4. Determine the asymptotes of the conic, supposing it to be a hyperbola ; or its axis, in the case where it is a parabola. 5. When is the conic (a) a circle, (b) an equilateral hyperbola, (c) a pair of straight lines ? 6. Examine the cases in which the two given angles are directly equal, or oppositely equal, or supple- mentary t. 411. Theokem. If a variable triangle AM A' move in such a way that its sides turn severally round three given points , ' , S (Fig. 249) while two of its vertices A , A' slide along two fixed straight lines u , vf respec- tively, the locus of the third vertex M is a conic passing through the following five points, viz. , 0' ', uu', and the intersections B and C of u and u' with O'S and OS respectively %. * Principia, lib. i. lemma xxi; Enumeratio linearum tertii ordinis (OpUcks, 1704), p. 158, § xxxi. f Maclaurin, Geometria Organica (London, 1720), sect. i. prop. 2. X See Art. 156. COROLLARIES AND CONSTRUCTIONS. [412 412. this). 413. Theoeem. (The theorem of Art. 411 is a particular case of If a variable polygon move in such a way that its n sides turn severally round n fixed points O x , 0% , . . . O n (Fig. 250) while n — 1 of its vertices slide respectively along n — 1 fixed straight lines u x ,u 2 , ... u n _ 1 , then the last vertex will describe a conic; and the locus of the point of intersection of any pair of non-adjacent sides will also be a conic *. The proof of this theorem and its cor- ■Fig. 250. relative is left to the student t. Theoeem. From two given points A and A' tangents AB , AG and A'B',A'C are drawn to a conic; then will the four points oj contact B , C , B',C', and the two given points A , A' all lie on a conic (Fig. 251 J). Let A'C, A'B f meet BC in D and E respectively; these points will evidently be the poles of AC, AB' respectively. The pencil A(BCB'C) is projective with the range of poles BCED (Art. 291), and therefore with the pencil A'(BCED) or A'(BCB;C)>, which proves the theorem. 414. Theoeem (correlative to that of Art. 413). From two given points A and A' tangents AB, AC and A'B', A'C are drawn to a conic; then will the four tangents and the two chords of contact all touch a conic J. For (Fig. 251) the range of points BC (AB, AC, A'B', A'C) or BCED is projective with the pencil A (BC B'C) formed by their polars; but this pencil is projective with the range B'C(AB,AC,A'B', A'C); therefore the six lines AB, AC, A'B', A'C, BC, B'C all touch a conic. 415. Theoeem. On each diagonal of a complete quadrilateral is taken a pair of points dividing it harmonically ; if of these six points three (one from each diagonal) lie in a straight line, the other three will also lie in a straight line. Coeollaey. The middle points of the three diagonals of a complete quadrilateral are collinear. * This theorem is due to Maclaurin and Braikenridge (Phil. Trans., London, 1735). t Poncblbt, loc. cit., Art. 502. X Chasles, Sections coniques, Arts. 213, 214. 418] COBOLLABIES AND CONSTRUCTIONS. 299 416. Theorem. If from any point on the circle circumscribing a triangle ABC straight lines OA' , OB', OC be inflected to meet the sides BC, CA, AB in A', B' , C respectively, and to make with them equal angles (both as regards sign c » and magnitude) ; then the three /^^"""^n" points A',B',C will be collinear /^\\\^s^^L f (Fig. 252). (// y\\ \r*- Through draw OA", OB", OC" c *U1 \ \\ /\ parallel to BC, CA, AB respec- Xi^^^-W- ^s. tively; then it is easily seen that b\^^^J^a "ft the angles AOA", BOB", COC" Fig> 2g2 have the same bisectors. The same will therefore be true with regard to the angles AOA', BOB', COC; consequently (Art. 142) the arms of these last three angles will form an involution, and therefore (Art. 135) the points A', B', C will be collinear * t. 417. Theorem. If from the vertices of a triangle circumscribed about a circle straight lines be inflected to meet any tangent to the circle, so that the angles they subtend at the centre may be equal (in sign and magnitude), then the three straight lines will meet in a point J. The proof is similar to that of the theorem in the preceding Article. 418. Problems. (1). Given three collinear segments AA\ BB', CC; to find a point at which they all subtend equal angles (Art. 109). In what case can these angles be right angles % (See Art. 128). (2). Given two projective ranges lying on the same straight line ; to find a point which is harmonically conjugate to a given point on the line, with respect to the two self-corresponding points of the two ranges (which last two points are not given) §. (3). Given two pairs of points lying on a straight line ; to deter- mine on the line a fifth point such that the rectangle contained by its distances from the points of the first pair shall be to that contained * Chasles, loc. cit, Art. 386. f Otherwise : Since the triangles BOC ', COB' are similar, BC : CB' m OB : OC. So also CA' : AC = OC : OA, an d AB'\BA' = OA:OB; whence by multiplication, paying attention to the signs of the segments, BC .CA'.AB' = - CA . B'C . A'B, which shows (Art. 139) that A', B' , C are collinear. % Chasles, loc. cit, Art. 387. § Chasles, Giom. sup., Art. 269. 300 COROLLAKIES AND CONSTKUCTIONS. [419 by its distances from the points of the second pair in a given ratio *. (4). Through a given point to draw a transversal which shall cut off from two given straight lines two segments (measured from a fixed point on each line) which shall have a given ratio to one another ; or, the rectangle contained by which shall be equal to a given one f. 419. It will be a useful exercise for the student to apply the theory of pole and polar to the solution of problems of the first and second degree, supposing given a ruler, and a fixed circle and its centre. We give some examples of problems treated in this manner : I. To draw through a given point P a straight Une parallel to a given straight line q. The pole Q of q and the polar p of P (with respect to the given circle) must be found ; if A be the point where p is cut by the straight line OQ joining Q to the centre of the circle, then the polar a of A will be the straight line required. II. To draw from a given point P a perpendicular to a given straight line q. Draw through P a straight line parallel to OQ ; it will be the perpendicular required. III. To bisect a given segment AB. Let a and b be the polars of A and B respectively, and c that diameter of the given circle which passes through ab ; if d be the harmonic conjugate of c with respect to a and b, the pole of d will be the middle point of AB. IV. To bisect a given arc MN of a circle. Construct the pole S of the chord MN ; the diameter passing through S will cut MN in the middle point of the latter. V. To bisect a given angle. If from a point on the circle parallels be drawn to the arms of the given angle, the problem reduces to the preceding one. VI. Given a segment AC ; to produce it to B so that AB may be double of AC. Let 1 a and c be the polars of A and C respectively, d the diameter of the given circle which passes through ac, and b the ray which makes the pencil abed harmonic ; the pole of b will be the required point B. * This is the problem ' de sectione determinata ' of Apollonius. See Chasles, G6om. sup., Art. 281. t These are the problems ' de sectione rationis ' and ' de sectione spatii ' of Apollonius. See Chasles, Giom. sup., Arts. 296, 298. 420] COROLLARIES AND CONSTRUCTIONS. 301 VII. To construct the circle whose centre is at a given point U and whose radius is equal to a given straight line UA. Produce AU to B, making UB equal to AU (by VI), and draw perpendiculars at A and B to AB (by II). Bisect the right angles at A and B (by V) ; and let the bisecting lines meet in C and D. We have then only to construct the conic of which AB and CD are a pair of conjugate diameters (Art. 301). 420. The following problems* depend for their solution on the theorem of Art. 376. I. Given three points A, B, C on a conic and one focus F, to construct the conic. With centre F and any radius describe a circle K, and let the polars of A, B, C with respect to this circle be a, b, c respectively. Describe a circle touching a, b, c and take its polar reciprocal with respect to K; this will be the conic required. Since there can be drawn four circles touching a, b, c (the inscribed circle of the triangle abc and the three escribed circles), there are four conies which satisfy the problem. II. Given two points A, B on a conic, one tangent t, and a focus F> to construct the conic. Describe a circle K as in the last problem, and let a, b be the polars of A, B, and I 1 the pole of t, with respect to K. Draw a circle to pass through T and to touch a and b ; the polar reciprocal of this circle with respect to K will be the conic required. Since four circles can be drawn to pass through a given point and touch two given straight lines, this problem also admits of four solutions. III. Given one point A on a conic, two tangents b, c, and a focus F, to construct the conic. Describe a circle K as in the last two problems ; let a be the polar of A, and let B, C be the poles of 6, c respectively with regard to this circle. Draw a circle to pass through B and C and to touch a ; its polar reciprocal with respect to K will be the conic required. Since two circles can be described through two given points to touch a given straight line, this problem admits of two solutions. IV. Given three tangents a, b, c to a conic and one focus F, to construct the conic. Describe a circle K as in the last three problems, and let A, B, C be the poles of a, b, c respectively with regard to this circle. Draw the circle through A, B, C and take its polar reciprocal with respect to K ; this will be the conic required. This problem clearly admits of only one solution. * Solutions of these problems were given by De la Hire (see Chasles, Aperfu historique, p. 125), and by Newton (Principia, lib. i. props. 19, 20, 21). 302 COKOLLAKIES AND CONSTBUCTIONS. 421. Problem. Given the axes of a conic in position (not in magnitude) and a pair of conjugate straight lines which cut one anotlter orthogonally, to construct the foci. If be the centre of the conic, and P, P / and Q, Q f the points in which the two conjugate lines respectively cut the axes, then of the two products OP . OP f and OQ . OQ', one will be positive and the other negative. This determines which of the two given axes is the one containing the foci. If now a circle be circumscribed about the triangle formed by the two given conjugate lines and the non- focal axis, it will cut the focal axis at the foci (Art. 343). 422. The following are left as exercises to the student. 1. Given the axes of a conic in position, and also a tangent and \j its point of contact, construct the foci, and determine the lengths of the axes (Art. 344). 2. Given the focal axis of a conic, the vertices, and one tangent, construct the foci (Art. 360). (V^ 3. Given the tangent at the vertex of a parabola, and two other tangents, find the focus (Art. 358). ^^—-4. Given the axis of a parabola, and a tangent and its point of contact, find the focus (Art. 346). 5. Given the axis and the focus of a parabola, and one tangent, construct the parabola by tangents (Arts. 346, 349, 358). 6. The locus of the pole of a given straight line r with respect to any conic having its foci at two given points is a straight line «*' perpendicular to r. The two lines r, / are harmonically separated by the two foci. 7. The locus of the centre of a circle touching two given circles consists of two conies having the centres of the given circles for foci. ^^ 8. The locus of a point whose distance from a given straight line is equal to its tangential distance from a given circle consists of two parabolas. 9. In a central conic any focal chord is proportional to the square of the parallel diameter. 10. In a parabola, twice the distance of any focal chord from its pole is a mean proportional between the chord and the parameter. INDEX. Affinity, pp. 18, 19. Angle of constant magnitude turning round its vertex traces out two directly equal pencils, 91. bisection of an, 300. trisection of an, 294. Angles, two, of given magnitude ; gene- ration of a conic by means of, 297. Anharmonic ratio defined, 54, 57. unaltered by projection, 54. of a harmonic form is — 1, 57. cannot have the values + I, o, or 00, 62. of four points or tangents of a conic, 122. Anharmonic ratios, the six, 60, 61. Apollonius, x, xi, xii. on the parabola, 127, 218. on the hyperbola, 130, 142, 156, 158, 286. on the diameters of a conic, 217, 223, 230, 232, 234, 235. on focal properties of a conic, 253, 258, 259, 262. section-problems, 300. Arc of a conic, determination of kind of conic to which it belongs, 289. of a circle, trisection of, 294. of a circle, bisection of, 300. Asymptotes, tangents at infinity, 1 6, 1 2 9 . meet in the centre of the conic, 219. determination of the, given five points on the conic, 178, 179. Auxiliary conic, 203, 239, 240. circle of a conic, 260. Axes of a conic defined, 227, 228. case of the parabola, 228. focal and non-focal, 252. bisectors of the angle between its chords of intersection with any circle, 236, 281. Axes of a conic, construction of the, given a pair of conjugate diameters, 232, 283. given five points, 236, 292. Axis of perspective or homology, 10. of affinity, 18. of symmetry, 64. Bellavitis, xi, 64, 161. Bisection of a given segment or angle by means of the ruler only, 300. Brianchon, x, xi, xii, 124, 125. Brianchon's theorem, xi, 124. points, the sixty, 125. Carnot's theorem, xi, 279, 280. Centre of projection, 1, 3. of perspective or homology, 10, 12, 98. of similitude, 18. of symmetry, 64. of an involution, 102. Centre of a conic, the pole of the line at infinity, 218. bisects all chords, 219. the point of intersection of the asymptotes, 219. when external and when internal to the conic, 219. locus of, given four tangents, 237. construction of the, given five points, 220. construction of the, given five tan- gents, 238. Ceva, theorem of six segments, ill. Chasles, xi, xii. on homography, 34. method of generating conies, 127. correlative to the theorem ■ ad qua- tuor lineas,' 159. on the geometric method of false position, 194. solutions of problems of the second degree, 200. Circle, curve homological with a, 14, 15. generated by the intersection of two directly equal pencils, 114. harmonic points and tangents of a, 115, 116. fundamental projective properties of points and tangents of a, 115. of curvature at a point on a conic, 190. cutting a conic ; the chords of inter- section make equal angles with the axes, 236, 281. circumscribing triangle formed by three tangents to a parabola, 253. auxiliary, of a conic, 260. Class of a curve, 4. is equal to the degree of its polar reciprocal with regard to a conic, 240. 304 INDEX. Coefficient of homology, 63. Collinear projective ranges, 68. their self- corresponding points, 78, 9 1 , 9 2 . 93- construction for these, 1 70. Complementary operations, 33. Concentric pencils, 69. construction for their self-correspond- ing rays, 169. Cone, sections of the, 14, 18. Confocal conies, 266. Congruent figures, 64. Conic, homological with a circle, 15, 16. generated by two projective pencils, 119. generated as an envelope from two projective ranges, 120. determined by five points or five tangents, 123. fundamental projective property of points and tangents, 118. projective ranges of points and series of tangents of a, 161. homological with itself, 228, 288. polar reciprocal of a, 240. homological with a given conic, and having its centre at a given point, 249. confocal with a given conic, and passing through a given point, 266. Conic, construction of a, having given five points or tangents, 131, 149, 176, 179, 180, 297. four points and the tangent at one of them, 137, 177. three points and the tangents at two of them, 139, 177. three tangents and the points of con- tact of two of them, 143, 177. four tangents and the point of con- tact of one of them, 146, 177. four points and a tangent, 180. four tangents and a point, 180. three points and two tangents, 182, 296. three tangents and two points, 182, 295. the asymptotes and. one point or tangent, 277. the two foci and one tangent, 264. the two foci and one point, 265. one focus and three tangents, 268, 301. one focus and three points, 301. ne focus, two points and a tangent, 3d. one focus, two tangents and a point, 301. a pair of conjugate diameters, 229, 231. a pair of conjugate diameters in posi- tion, and two points or tangents, 291, a pair of conjugate diameters in posi- tion, and a tangent and its point of contact, 289. two pairs of conjugate diameters in position, and one point or tan- gent, 290, 291. two reciprocal triangles, 247. v a self- conjugate triangle, and appoint and its polar, 292. a self-conjugate pentagon, 293. three points and the osculating circle at one of them, 190. Conic, construction of a, homological with itself, 228, 288. passing through three points and determining a known involution on a given line, 284. passing through four points and di- viding a given segment harmoni- cally, 284. passing through four points and through a pair of conjugate points of a given involution, 285. Conies, osculating, 189. having a common self-conjugate triangle, 213, 214. circumscribing the same quadrangle, 150, 214, 237. inscribed in the same quadrilateral, 150, 213, 214, 237. Conjugate axis of a hyperbola, 228. Conjugate diameters, defined, 219. of a circle cut orthogonally, 222. form an involution, 227. parallelogram described on a pair as adjacent sides is of constant area, 234- sum or difference of squares is con- stant, 235. construction of, given two pairs, 232. construction of, given five points on the conic, 236. ' including a given angle, construction of, 292. Conjugate lines meeting in a point, one orthogonal pair can be drawn, 227. orthogonal, the involution determined by them on an axis of the conic, 251. orthogonal, with respect to a para- bola, 253. Conjugate points and lines with regard to a conic, 204. involution-properties of, 209. Conjugates, harmonic, 46. in an involution, 101. Construction of a figure homological with a given one, 13. for the fourth element of a harmonic form, 47. for the fourth point of a range whose anharmonic ratio is given, 55, INDEX. 305 / of pairs of corresponding elements of two projective forms, when three are given, 70. for the self-corresponding elements of two superposed projective forms, 169. for the sixth element of an involu- tion, 109. of pairs of elements of an involution, given two, 104. for the centre of an involution, 109. for the double elements of an in- volution, 169, 175, 295. for the common pair of two super- posed involutions, 1 73. for the pole of a line or polar of a point, 2*05, 206. of a triangle self-conjugate to a conic, 207. of the centre and axes of a conic, 220, 236, 238, 283, 292. of conjugate diameters, 232, 236, 292. for diameters of a parabola, having given four tangents, 238. for the focus of a parabola, given four tangents, 254. for the foci of a conic, given the axes and a pair of orthogonal conjugate lines, 302. Copolar and coaxial triangles, 7, 8. Correlative figures, 26, 85, 241. Curvature, circle of, 190. Degree of a curve, 4. is equal to the class of its polar reci- procal with respect to a conic, 240. De la Hire, x, xii. Desargues, ix, x, xii, 101, 102, 107, 148. Desargues' theorem, 148. Descriptive, the term, as distinguished from metrical, 50. Diagonal triangle, of a quadrangle or quadrilateral, 30. common to the complete quadri- lateral formed by four tangents to a conic, and the complete quad- rangle formed by their points of contact, 140. Diagonals of a complete quadrilateral, each is cut harmonically by the other two, 46. their middle points are collinear, 109, 299. form a triangle self-conjugate to any conic inscribed in the quadri- lateral, 208. if the extremities of two are conju- gate points with regard to a conic, those of the third are so too, 245. Diameters of a conic defined, 217. of a parabola, 218. conjugate, 219. ideal, 223. of a parabola, construction for, given four tangents, 238. Dimension of a geometric form, 25. Directly equal ranges, defined, 88. generated by the motion of a seg- ment of constant length, 89. Directly equal pencils, defined, 90. two, the projection of two concentric projective pencils, 89. two, generate a circle by their inter- section, 114. subtended at a focus of a conic by the points in which a variable tangent cuts two fixed ones, 264. Director circle, defined, 269. the locus of the intersection of or- thogonal tangents, 269. cuts orthogonally the circumscribing circle of any self-conjugate tri- angle, 270. Directrix, defined, 254., property of focus and, 256. Directrix of a parabola, the locus of the intersection of orthogonal tan- gents, 270. the locus of the centre of the cir- cumscribing circle of a self-con- jugate triangle, 271. the locus of the orthocentre of a circumscribing triangle, 273. Division of a given bisected segment into n equal parts, by means of the ruler only, 97. Double elements of an involution, they separate harmonically 'of conjugates, 103, construction for the, i6Qj_£9&. Duality, the principle of, 26-32. tion, ifllX any pair J Eccentricity, 259. of the polar reciprocal of a circle with respect to another circle, 274. Ellipse, 16. its centre an internal point, 219. is cut by all its diameters, 220. is symmetrical in figure, 228. Envelope of connectors of correspond- ing points 'of two projective is a conic, 120. if the ranges are similar, it is a para- bola, 128. of a straight line the product of whose distances from two given points is constant, 260. Equal ranges and pencils, 86-90. Equianharmonic forms and figures are projective, and vice versa, 54, 56, 62,66. Equilateral hyperbola, why so called 286. triangles self-conjugate with regard to a, 271. inscribed in a quadrilateral, 272. 306 INDEX. circumscribing a triangle passes through the orthocentre, 287. is the polar reciprocal of a conic with regard to a point on the director circle, 288. construction of, given four tangents, 272, 288. Euclid, porisms of, x, 96. External and internal points with re- gard to a conic, 203. False position, geometrical method of, 194. Focal axis of a conic, 252. radii of a point on a conic, 253. radii, their sum or difference is con- stant, 258. Foci, denned, 250. are points such that conjugate lines meeting in them cut orthogonally, 250. are internal points lying on an axis, 250. are the double points of the involu- tion determined on an axis by pairs of orthogonal conjugate lines, 251 . of a parabola, one at infinity, 253. of parabolas inscribed in a given triangle, locus of, 254. properties of, with regard to tangent and normal, 259-264. reciprocation with respect to the, 2 74, 275. construction of, under various con- ditions, 302. Focus of a parabola, 253. inscribed in a given triangle, locus of, 254. reciprocal of the curve with regard to, 275. Forms, geometric, defined, 22, 164. elements of, 23, 164. prime, of one, two, three dimensions, 24. dual generation of, 23, 24, 26. projective, 34-38. 'larmonic, 39-49. projective, when in perspective, 67. projective, superposed, 68, 69. Gaskin, 189, 269, 271. Gergonne, x. Harmonic forms defined, 39, 40. forms are projective, 41, 43. pairs of points necessarily alternate, 45. conjugates, 46. point or ray, construction for the fourth, 47. forms, metrical relations, 57, 58. homology, 64, 228, 288. points and tangents of a circle, 115, 116, 169. and of a conic, 122, 157, 168. Hesse, theorem, relating to the ex- tremities of the diagonals of a complete quadrilateral, 245. Hexagon, inscribed in a line-pair, 76. circumscribed to a point-pair, 76. inscribed in a conic, 124. circumscribed to a conic, 124. complete, contains sixty simple hexa- gons, 125. Homographic, the term, 34. figures, construction of, 81. figures may be placed in homology, 84. Homological figures, construction of, 13-20. metrical relations between, 63-65. Homology, defined, 9, 10. • in space, 20. plane of, 20. coefficient or parameter of, 63. harmonic, 64, 228, 288. Homothetic figures, 18. Hyperbola, tangent - properties of a, 129, 130. and asymptotes cut by a transversal, 156, 282. tangent cut off by the asymptotes is bisected at the point of contact, 158.. centre is an external point, 219. is cut by one only of every pair of conjugate diameters, 220. is symmetrical in figure, 228. properties of the asymptotes and conjugate points and lines, 277. equilateral, 285. Ideal diameters and chords, 223, 226. Infinity, points and line at, 5. line at, a tangent to the parabola, 16. plane at, 21. Internal and external points with re- gard to a conic, 203. Intersection of a conic with a straight line; constructions, 176, 177, 180, 226. , of two conies; constructions, 189. /Involution, defined, 101. the two kinds, elliptic and hyper- bolic, 105, 168. construction for the sixth element of an, ipg. determined by two pairs of conju- gates, 104, 165. of points or tangents of a conic, 165.- construction for the double elements of an, 169, 295. formed by cutting a conic by a pencil, 166. of conjugate points or lines with ; regard to a conic, 209. of conjugate diameters of a conic, 227. INDEX. 307 Involution-properties of the complete quadrangle and quadrilateral, 107. of a conic and an inscribed or circum- scribed quadrangle, 148, 225. of a conic and an inscribed or cir- cumscribed triangle, 152, 157. of a conic, two tangents, and their chord of contact, 154. of conjugate points and lines with regard to a conic, 209. Lambert, ix, xi, 96-98. Latus rectum, 257, 258. Locus of the centre of perspective of two figures when one is turned round the axis of perspective, 1 2,98. of the intersection of corresponding rays of two projective pencils is a conic, 119. ad quatuor lineas, 158. of middle points of parallel chords of a conic, 217. of poles of a straight line with regard to conies inscribed in a quadri- lateral, 237. of the centre of a conic, given four tangents, 237. of foot of perpendicular from the focus of a conic on a tangent, 260. of the intersection of orthogonal tangents to a conic, 269. Maclaurin, xi, 127, 141, 185, 297^298^ Major and minor axes of an ellipse, 228. Menelaus, theorem on triangle cut by a transversal, 112, 280. Metrical, the term, distinguished from descriptive, 50. Mobius, theorem on figures in per- spective, 12. on anharmonic ratio, x, 56, 61. Monge, xii. Newton, locus of centre of a conic in- scribed in a quadrilateral, 238. organic description of a conic, xi, 297. Nine-point circle, 283. Normal, 252. Oppositely equal pencils, 90. they generate an equilateral hyper- bola by their intersection, 286. Oppositely equal ranges, 88. Organic description of a conic, 297. Orthocentre of a triangle circumscribing , a parabola lies on the directrix, 273. of a triangle inscribed in an equi- lateral hyperbola lies on the curve, 287. Orthogonal projection, 19. pair of rays in a pencil in involution, 172. pair of conjugate diameters of a conic, 227. conjugate lines with respect to a conic, 251, 252. Osculating conies, 189. circle of a conic, 190. Pappus, x, xii. on a hexagon inscribed in a line- pair, 7$,- porismsof, 95, 96. fundamental property of the an- harmonic ratios, 54. problem 'ad quatuor lineas,' 158. on the focus and directrix property of a conic, 257. Parabola, touches the line at infinity, 16. is determined by four points or tan- gents, 127. two fixed tangents are cut propor- tionally by the other tangents, 127. generated as an envelope from two similar ranges, 128. diameters of a, 218. construction of the diameters, having given four tangents, 238. focal properties of the, 253, 254. focus and directrix property, 257. self-conjugate triangle, property of, 271. inscribed in a triangle, its directrix passes through the orthocentre, 273- Parabola, construction of a, given four points, 181. given four tangents, 135. given three tangents and a point, 182. under various conditions, 138, 139, 143, 146. given the axis, the focus, and one point, 266. given two tangents, the point of con- tact of one of them, and the direction of the axis, 278. given two tangents and their points of contact, 279. Parallel lines meet at infinity, 5. projection, 19. lines, construction of, with the ruler only, 96, 300. Parallelogram, inscribed in or circum- scribed about a conic, 219, 221. described on a pair of conjugate semi- diameters of a conic is of constant area, 234. Parameter of homology, 63. Pascal's theorem, xi,_j_24. lines, the sixty, 125. Pencil, flat, defined, 22. axial, 22. X 2 308 INDEX. harmonic, 40, 42. in involution, 101. in involution, orthogonal pair of rays of a, 172. cut by a conic in pairs of points forming an involution, 166. Pentagon, inscribed in a conic, 136. circumscribed to a conic, 145. self- conjugate with regard to a conic, 293- Perpendiculars, centre of, see Ortho- centre. from a focus on tangents to a conic, the locus of their feet a circle, 259. from the foci of a conic on a tangent, their product constant, 260. from any point of the circumscribing circle of a triangle to the sides, their feet collinear, 261, 299. construction of, with the ruler only, 97, 3°°- Perspective, figures in, 3. triangles in, 7, 8, 246. forms in, 35. plane, 10. relief, 20. Plane of points or lines, 22. Planes, harmonic, 42. involution of, 101. Points, harmonic, on a straight line, 40. harmonic, on a circle, 116. harmonic, on a conic, 122, 157. projective ranges of, on a conic, 161. Polar reciprocal curves and figures, 240, 241. of a conic with respect to a conic is a conic, 240. of a circle with respect to a circle, 274. of a conic with respect to a focus, 274* 275. of a conic with respect to a point on the director circle, 288. Polar system, defined, 248. determined by two triangles in per- spective, 248. determined by a self-conjugate tri- angle and a point and its polar, 293. Pole and polar, defined, 201, 202. reciprocal property of, 204. theory of, applied to the solution of problems, 300. construction of, 205, 206, 248. Poles, range of, projective with the pencil formed by their polars, 209, 224. of a straight line with regard to all conies inscribed in the same quad- lateral lie on a fixed straight line, Polygon, inscribed in a conic, whose sides pass through fixed points, 151, 185/187. circumscribed to a conic, whose ver- tices slide on fixed lines, 152, 186. whose sides pass through fixed points and whose vertices lie on fixed lines, 184. Poncelet, ix, x, xii. on variable polygons inscribed in or circumscribed to a conic, 151, 184- 187. on ideal chords, 2 26. on polar reciprocal figures, 240. on triangles inscribed in one conic and circumscribed about another, . 2 44- Porisms, of Euclid and Pappus, 95, 96. of in- and circumscribed triangle, 94, 244. of the inscribed and self-conjugate triangle, 243. of the circumscribed and self-con- jugate triangle, 243. Power of a point with respect to a circle, 58. Prime-forms, the six, 24. Problems, solved with ruler only, 96-98. of the second degree, 1 76-200. solved by means of the ruler and a fixed circle, 194, 300. solved by polar reciprocation, 301. Projection, operation of, 2, 22, 164. central, 3. . orthogonal, 19. parallel, 19. of a triad of elements into any other given triad, 36. of a quadrangle into any given quad- rangle, 80. of a plane figure into another plane figure, 81. Projective forms and figures, 34. forms, when in perspective, 67 . forms, when harmonic, 69. ranges, metrical relations of, 62. forms, construction of, 70-74. figures, construction of, 81-84. plane figures can be put into homo- logy 84. properties of points and tangents of a circle, 11 4-1 17. properties of points and tangents of a conic, 11 8-1 30. Projectivity of any two forms ABC &nd A' B'C f , 36. of two forms ABCD and BABC, 38. of harmonic forms, 41,-43. of the anharmonic ratio, 54."^" of any two plane quadrangles, 80. of a range of poles and the pencil formed by their polars, 209, 2 24. Quadrangle, complete, defined, 29. two plane quadrangles always pro- jective, 80. INDEX. 309 harmonic properties, 39, 47. involution properties, 107, 225. inscribed in a conic, 138, 140, 208, 225. if two pairs of opposite sides are con- jugate lines with regard to a conic, the third pair is so too, 246. Quadrangles having the same diagonal points ; their eight vertices lie on a conic or a line-pair, 210. Quadrilateral, complete, defined, 29. harmonic properties, 39, 46. involution properties, 107, 225. middle points of diagonals are col- linear, 109, 299. circumscribed to a conic, 142, 208, 225, 272. locus of centres of inscribed conies, 237. theorem of Hesse relating to the ex- tremities of the three diagonals, 245. Quadrilaterals having the same di- agonals ; their eight sides touch a conic or a point-pair, 212. Range, defined, 22. harmonic, 40. Ranges, projective, on a conic, 161. Ratio, of similitude, 18. harmonic, 57. anharmonic, 54-62. Reciprocal figures, 85. points and lines with regard to a conic, 204. triangles, two, are in perspective, 246. Reciprocation, polar, 241. with respect to a circley 274, 275. applied to solution of problems, 301. Rectangular hyperbola, see Equilateral. Ruler only, problems solved with, 96- 98. Ruler and fixed circle, problems solved by help of the, 194, 300. Section, operation of, 2, 22, 164. of a cone, 14, 18. of a cylinder, 19. Segment, dividing 'two given ones har- monically, 58, 103, 295. of constant magnitude sliding along a line generates two directly equal ranges, 89. bisected, its division into n equal parts by aid of the ruler only, 97. Segments of a straight line, metrical relations between, 51, 52. Self-conjugate pentagon with regard +0 a conic, 293. Self-conjugate triangle, 207-209. circumscribing circle of a, its pro- perties, 271. Self-conjugate triangles with regard to a conic, two; properties of, 242. Self-corresponding elements, defined, 67. of two superposed projective forms, 68, 69, 78, 91-93. general construction for these, 169. of two coplanar projective figures, 79. of two projective ranges on or series of tangents to a conic, 162, 163. Sheaf, defined, 22. Signs, rule of, 51. Similar ranges and pencils, 86, 87, 128. and similarly placed figures, 18. Staudt, vi, vii. on the geometric prime-forms, 24. on the principle of duality, 26. on harmonic forms, 39. on the construction of two projective figures, 81. on the polar system, 248. on an involution of points on a conic, 165. Steiner, vii, 3, xii on the sixty Pascal lines and Brian- chon points, 125. on the solution of problems of the second degree by means of a ruler and a fixed circle, 194. Superposed geometric forms, 68, 69. construction of their self-correspond- ing elements, 169. plane figures, if projective, cannot have more than three self-corre- sponding elements, 79. Supplemental chords, 221. Symmetry, a special case of homology, 64. Tangents, harmonic, of a circle, 11 6, in- harmonic, of a conic, 168. to a conic, series of projective, 163, 164. orthogonal, to a conic, 269. to a conic from a given point ; con- structions, 176, 177, 179, 226. common, to two conies; construc- tions, 190. Tetragram and Tetrastigm, 29. Townsend, 200. Transversal, cut by the sides of a tri- angle, 112. cutting a quadrangle or a quadri- lateral, 107, 108. cutting a conic and an inscribed quadrangle, 150. drawn through a point to cut a conic; property of the product of the segments, 281. cutting a hyperbola and its asymp- totes, 156, 282. Transverse axis of a hyperbola, 228. 310 INDEX. Triangle, inscribed in one triangle and circumscribed about another, 94. inscribed in a conic, 143, 216. circumscribed to a conic, 144, 216. inscribed or circumscribed, involu- tion-properties, 152, 157. self-conjugate with regard to a conic, 207, 270. circumscribed to a parabola, 253, 273. self-conjugate with regard to a para- bola, 271. self-conjugate with regard to an equilateral hyperbola, 271. cut by a conic, Carnot's theorem, 279. inscribed in an equilateral hyper- bola, 287. Triangles, two, self-conjugate with re- gard to a conic ; properties of, 242. inscribed in one conic and self-con- jugate to another, 243. circumscribed to one conic and self- conjugate to another, 243. inscribed in one conic and circum- scribed to another, 244. reciprocal, are in perspective, 246. formed by two pairs of tangents to a conic and their chords of contact, 298. Trisection of an arc of a circle, 294. Vanishing points and lines, 5. plane, 21. Vertex of a conic, 228, 256. circle of curvature at a, 19 D. THE END n SELECT WORKS PUBLISHED BY THE CLARENDON PRESS. Baynes. LESSONS ON THERMODYNAMICS. By R. E. Baynes, M.A., Senior Student of Christ Church, Oxford, and Lee's Reader in Physics. Crown 8vo. cloth, ys. 6d. Chambers. A HANDBOOK OF DESCRIPTIVE ASTRONOMY. By G. F. Chambers, F.R.A.S. Third Edition. Demy 8vo. cloth, 2%s. Clarke. GEODESY. By Colonel Alexander Ross Clarke, C.B., R.E. 8vo. cloth, 12s. 6d. Donkin. ACOUSTICS. By W. F. DonKin, M.A., F.R.S., late Savilian Professor of Astronomy, Oxford. Crown 8vo. cloth, ys. 6d. Maxwell. A TREATISE ON ELECTRICITY AND MAGNE- TISM. By J. Clerk Maxwell, M.A., F.R.S. Second Edition. 2 vols. Demy 8vo. cloth, \l. lis. 6d. AN ELEMENTAR Y TREA TISE ON ELECTRICITY. By the same Author. Edited by William Garnett, M.A. Demy 8vo. cloth, js. 6d. Minchin. A TREA TISE ON STA TICS. By G. M. Minchin, M.A., Third Edition, Corrected and Enlarged. Vol. I. Equilibrium of Coplanar Forces. 8vo. cloth, ys. Vol. II. In the Press. UNI PLANAR KINEMATICS OF SOLIDS AND FLUIDS. By the same Author. Crown 8vo. cloth, 7s. 6d. SELECT WORKS PUBLISHED BY THE Ci_,,RENDON PRESS. Price. TREATISE ON INFINITESIMAL CALCULUS. By Bartholomew Price, M.A., F.R.S., Professor of Natural Philo- sophy, Oxford. Vol. I. Differential Calculus. Second Edition, 8vo. cloth, x^s. 6d. Vol. II. Integral Calculus, Calculus of Variations, and Differential Equations. Second Edition. 8vo. cloth, \%s. Vol. III. Statics, including Attractions ; Dynamics of a Material Particle. Second Edition. 8vo. cloth, \(ss. Vol. IV. Dynamics of Material Systems ; together with a Chapter on Theoretical Dynamics, by W. F. Donkin, M.A., F.R.S. 8vo. cloth, 16s. Rigaud. CORRESPONDENCE OF SCIENTIFIC MEN OF THE SEVENTEENTH CENTURY, with Table of Contents by A. de Morgan, and Index by the Rev. J. Rigaud, M.A., Fellow of Magdalen College, Oxford. 2 vols. 8vo. cloth, \%s. 6d. Smyth and Chambers. A CYCLE OF CELESTIAL OBJECTS. Observed, Reduced, and Discussed by Admiral W. H. Smyth, R.N. Revised, condensed, and greatly enlarged by G. F. Chambers, F.R.A.S. 8vo. cloth, 12s. Stewart. A TREATISE ON HEAT, with numerous Woodcuts and Diagrams. By Balfour Stewart, LL.D., F.R.S., Professor of Natural Philosophy in Owens College, Manchester. Fourth Edition. Extra fcap. 8vo. cloth, 7s. 6d. Watson and Burbury. THE MATHEMATICAL THEORY OF EIEC- TRICITY AND MAGNETISM. By H. W. Watson, D.Sc, F.R.S., formerly Fellow of Trinity College, Cambridge, and S. H. BURBURY, M.A., formerly Fellow of St. John's College, Cambridge. Vol. I. Electrostatics. 8vo. cloth, 10s. 6d. Vol. II. Electrodynamics. A TREATISE ON THE APPLICATION OF GENE- RALISED COORDINATES TO THE KINETICS OF A MA TERIAL SYSTEM. By the same Authors. 8vo. cloth, 6s. LONDON : HENRY FROWDE, Oxford University Press Warehouse, Amen Corner, E.C. T 14 DAY USE RETURN TO DESK FROM WHICH BORROWED LOAN DEPT. This book is due on the last date stamped below, or on the date to which renewed. Renewed books are subject to immediate recall. 5 . W^ ? IN STACKS * APR 61964 REC'D LD APR?fe'fi/|-4PI\ | » t r» oi a rn^n l( :o General Library ?n,i,m'i^ University of California (D3279sl0)476B Berkeley y 388 7'/ -. t'1 UNIVERSITY OF CALIFORNIA LIBRARY fii