CURVES WITH A DIRECTRIX DISSERTATION Submitted to the Board of University Studies of the Johns Hopkins University in conformity with the requirements for the deg;ree of Doctor of Philosophy BY CLYDE SHEPHERD ATCHISON J907 BALTIMORE, MD., V. S. A. 1908 To Professor Morley, without whose helpful suggestions and constant inspiration this paper would have been impossible, and to Dr. Cohen, Dr. Hulbert, and Dr, Coble for their encouragement in his university course, the author desires thus to express his thanks. 186885 Digitized by the Internet Archive in 2008 with funding from IVIicrosoft Corporation http://www.archive.org/details/curveswithdirectOOatchrich / OF THE I UNIVERSITY Curves ivith a Directrix. Br Clyde S. Atchison. Introductory. It is a well-known fact of elementary geometry that the orthocentre of any three lines of a parabola, which is of course uniquely determined by four lines, lies on a line, called the directrix of the parabola. In his Ortliocentric Properties of the Plane n-Line, Professor Morlet* has extended the meaning of the term orthocentre, so that with every odd number of lines there is associated an ortho- centre, and with every even number of lines a directrix. It is my purpose to discuss the class of curves, of which the orthocentre of a certain number of lines shall always lie on a line, the directrix of the curve in question. As the curves to be considered are entirely rational, throughout the discussion, I shall employ almost exclusively the method of vector analysis, worked out by Professor MoRLEY, which is especially convenient for the purpose. The line equation of a curve is here uniformly expressed by means of conjugate coordinatesf, while the point equation is expressed by means of what is commonly called the map equation ; that is, a point of the curve is expressed as a rational algebraic function of a parameter t, which is limited to the unit circle, and is, therefore, always equal in absolute value to unity. Throughout y and h will be considered as conjugates of x and a respectively; that is, if X and Y are rectangular coordi- nates of a point, then x=zX -\-iY, y = X-iY, and similarly for a and h. { * Transactions of the American Mathematical Society, Vol. IV (1903), pp. 1-13. f F. Fbanklin: Some Applications of Circular Coordinates, ^mcWcan Journal of Mathematics, Vol. XII (1890), p. 161; F. Moklby : On the Metric Geometry of the Plane n-Line, Transactions of the American Society, Vol. I (1900), p. 97. t H. A. Cosvbrsb: Annals of Mathematics, Ser. II, Vol. V (1904), pp. 106-109, where is given a fnlle explanation of these methods. 2 Curves with a Directrix. Employing the notation used in the memoirs of Prof. Morley, a line a is written in the form where x^ is the reflection of the origin in the line, and — 1/<„ is the clinant. The definition of the first orthocentre, as extended for 2n — 1 lines, is the point of intersection of the In — 1 perpendiculars, let fall from the centres of the 2n — 1 A^"~"s, one of which is uniquely determined by every 2n — 2 lines, upon the line left out in each case. By a* A^"~^ is meant a deltoid of order 2n — 2 and class 2n — 3, having 2n — 3 cusps, and containing the line at infinity as a special {In — 4)-fold line, being tangent to it n — 2 times at / and n — 2 times at /. Thus, in the case of three lines, the three A"s each of which is determined by two lines of the three, are simply points, the vertices of the triangle of the three lines, and the orthocentre is the intersection of the perpendiculars dropped from the vertices upon the opposite sides. But in the case of five lines, every four lines uniquely determine a A', and the perpendiculars dropped from the centres of these five A^'s upon the line omitted in each case meet in a point, which is the orthocentre of the five given lines. For 2n — 1 lines, the first orthocentre* is of the form i'n-i = «! — ^j ag + A^g ag — + (— )" ( + /S'„_2 a„_i — S^-i a,), where Si stands for symmetric functions of the 2n — 1 <'s, and the a„ are the so-called characteristic constants of the lines, and in the case of 2« — 1 lines 2n-l x,tl 2n-l-a "« - 2 {t, - t,) {t, -t,).... ( ' 7 ' 8 « « 8 " -^, 8 « ^-, 8 « ^^, 4 angles*^. 9 « « 9 « _|_, 9 '. -^, 9 « ^^, 9 " ^. 10 " « 10 " -^, 10 « -4J-' 1^ " -T^'^^ " r^'^^'^Slesf^. 2m " " 27» « ^i^^, 2m « -^^, 2m " -^^^ 2m ' 2m 27« 2 m angles ("^~ ■>'' ,« angles ^. ^ 2m " 2m (2m + l) « " (2m + l) " 3^, (2- + 1) " ^lll' (^'" + 1) " 2-^1' ' (^- + 1) " TOT' ^^^'^+1^^"^'^^ 2^1- § 3. General Discussion of the K^^. In the first section, the equation of the JT^""^ was obtained in trilinear complex line coordinates, and the orthocentre of any 2n — 1 of its lines was proved to lie on the directrix of the curve. In the present section an equation of the curve will be derived by another method, and a general discussion will follow. Some special cases will also be considered. 12 Curves ivith a Directrix. Envelope of a A^"~^ Moving Under Parabolic Translation. Consider the general case of a A^"~^ moving under parabolic translation; that is, a translation in which one and therefore all the points of the A^"~^ describe a parabola, although not the same parabola. A convenient form in which to write the map equation of the general parabola for the present purpose is X r' {r^ — ty where t is a constant turn. This parabola is of unit size, but since the unit of measurement may be taken arbitrarily, it may be considered as any parabola. On dividing the map equation by its conjugate, and substituting the value of t for which x becomes infinite, 1/t^ is found to be the clinant of the vector from the origin to the point at infinity on the parabola. The map equation of the general A^"~^ with its centre at the origin is (-)"[na„_ir-^-(n-l)a„_2r-H- ••• + (-)"(•••• + 2a, ^ + ^J-....) (n-2)&„_2 (n-l)&^i 1 •••• r=i ^ r J' The condition for the envelope of the A^"~^ , moving with its centre on the parabola X X ^ T^ {t^ — ty while its orientation remains constant, is that tDtf , . < A/ — =r-^, = a real quantity = = (....+ 2a.' >. + I -....)... - ('"-ft^--' + C"-^) ^•■- ] . This condition is satisfied, as before, when That is, the equation of the envelope is of the form It is evident, therefore, that when m<^n, the equation reduces simply to and is another K^, with the same directrix as the E^^ over which it was trans- lated, for the equation of the directrix is independent of the constants of the deltoidal term. When m^n, the equation of the envelope reduces to "f j_ A2m-1 {r^—tf n. — ^ _1_ A2m— 1 and is evidently a K^^ with the same directrix as the iT^" over which it was translated, since both curves can be obtained by trao slating the proper deltoid over the same parabola, and, consequently, according to Theorem 5, have the same directrix. Hence, we may state the theorem : Theorem 6. The envelope of a A""'"'^ whose orientation is constant, and whose centre moves on a K^", is another K^" if nK^n, and is a K^'" if my-n, and the directrices of all curves generated thus are the same; namely, the directrix of the K^". Curves voith a Directrix. 19 It should be observed that any curve jK"2"+2 can be obtained by translating a cycloid (n+l)a„P+-^ over a K^^. Now, in the equation of the ^^", if l/r^ be taken equal to unity, the directrix is perpendicular to the axis of reals, and the curve is tangent to the line at infinity n — 1 times at /, n — 1 times at J, and once in the direction of the axis of reals, and is then designated by jff'g", according to the notation of § 1. The equation then takes the form X = J~(f + (-)" ["«"-! *"'' - (»^ - 1) «»-2 «""' + • • • • § 4. Caustics* Any rational curve can be represented by a map equation which of course carries with it its conjugate equation y=m and the general form of its caustic by reflection, where the incident rays are parallel to the real axis, can easily be derived. Letting t" represent the unknown clinant of the line tangent to the curve at the point /(O, the equation of this line can be written x-f{t) = t'\jj-f{t)-], which, for varying t, is the line equation of the curve. On dividing through by t% and eliminating y by partial differentiation with respect to t, the equation or .^f^e>-'m+'nm is obtained, which must be the map equation of the curve. But, by hypothesis, x = f{i) ♦Salmon: Higher Plane Curves. Heath: Oeometrical Optict, F. Morley: On the CauHie of the Epicycloid. 20 Curves with a Directrix. is the map equation of the curve. It follows, therefore, that tnt)^ i^^'f{t) % z ♦ whence ~m' This then is the clinant of the tangent to the curve at any point given by t. Since when the incident rays are parallel to the real axis, the reflected rays make twice the angle with the axis of reals that the tangents to the curve at the points of reflection make with the axis of reals, it is evident that the clinant of any ray after reflection at the point given by t, is K]' -ft. and its equation is which, as t varies, is the line equation of the caustic of the given curve. Since the condition for the cusps of the curve obtained by equating or to zero, is m = 0, which of course carries with it its conjugate equation /'(O = 0, it is evident further that: Theorem 7. The cusps of any rational curve lie on its caustic. Now, consider the case of the K'^^\ The condition for cusps is of degree 2n + 2, showing that the K^^ has 2n + 2 cusps, although not necessarily all real. The line equation of the caustic by reflection, formed as in the general case, is X - 1 A. [ <' ,1 (_Y\^^n-i (n— l)6n_a - t^\y \{\ — tf^^ 'I t^-i - f.-2 ^ • • • • + (»-l)a„_iP]}), Curves with a Directrix. 21 which may be written in the form t'x-y = i-Y ^na„_, <»+» - (n - l) (a„_, + a„_i) T + (« - 2) (a„_3 + a„_2) <"-! -.... + (— )» j . . . . + 2<» (aj + a,) + ^"^ (V. + &„-i) - "^11^- a, t\ Its map equation is + (»-2H"-l) (a„_3 + a„_3) <"-3_. . . . + (_)» I . . . . + 3<(ai + a,) (n— 1)(« — 2),, , , \,n(n—l), ~\ showing that the caustic is a cycloidal curve. It is to be observed, that in forming the equation of the caustic of the K^^, since the parabolic term, j- t^, f appeard on the left side of the equation, and its conjugate, /.. ...g , appeared on the right multiplied by Ijf, these terms canceled each other. Hence, since the ^" is made up only of a parabolic term and a deltoidal term of class 2n — 1, it is evident that the caustic by reflection of the general A'^""^ with centre at the origin is the same as the caustic by reflection of the Kl", and is the cycloidal curve on which lie the 2n — 1 cusps of the A"'~^ and the 2n + 2 cusps of the A'§". In general, given any self-conjugate equation, which is a polynomial in t, thus: G: ajr— as^-^+agr-"— . . . . +(—)*( a;^ <""*+! +»<"-* -a^+a <"-'-'+ • ...)•••• +(-)"-*( b^^.f^' + yf-b.t^-^ + ....)....+(-)"(• ••• + &3<'-&2< + ^) = 0, where a and b are conjugates, it represents a cyclogen, the clinant of any one of whose lines is ( — )"l/<"~^*. Considering the incident rays as coming from the point at infinity on the axis of reals, the clinant of the line equation of the caustic of any (7„ is therefore 1/^201-2*) q^ writing out a few caustic equations 22 Curves vyith a Directrix. for the less complicated cyclogens, it is readily observed, that the caustic equations also always represent cyclogens, and that if y is the coefficient of t'' in the C„, it is also the coefficient of t^ in the caustic. Further, considering the caustic as a polynomial in t, it is observed that the middle term is always absent, and that the class of the caustic is equal to the sum of the order of the C„ and of the order of t in its clinant ; that is, the resulting caustic is of degree 2n — 2^ in t. Hence, the general form of the caustic of the general (7„ may be written, thus: aj«i!n-2*_a2«2"-2'=-i + a3<2"-2*-^- .... +(-)'(n-7^- |)(a;+a,+i) <'""'*+ • • ■ • + 0<"-* + . . . . + {-f{n-h-^{y+^,^,)t'^ + . . . ■+^,t'-^,t + (i, = 0, where a and j8 are conjugates. §5. Discussion of the K\. (a) Cusps. The map equation of the K\ is The condition for cusps, obtained by equating ^ to zero, was found to be (j^.+''-^=0. (C) which is a sextic in t, showing the -ff"* has six cusps, although not necessarily all real. There are just as many real cusps as equation (C) has roots <{ which are real turns. Considering a and h as variables, and calling them x and y, equation (C) may be written - {1—tf ' t' For varying t, this is the line equation of the curve whose map equation is This curve, composed of two loops, is tangent to the line at infinity in the direction of the real axis, has a double point at X =■ 1/4, cuts the axis of reals again at the point a; = — 1/16, and is a sextic in lines. Now a is any point in the plane, and for any point a there are as many f's, which are real turns, satisfying equation (C), as there can be drawn real tangents to the curve (A) from the point a. It is evident geometrically that if "^~(i-0*' Curves with a Directrix. 23 where t may take any value from i to — i (that is, if a lies inside both loops), no real tangents can be drawn to (A), and the K\ has no real cusps. But if — — 1 a-x- (i_^)4, where t may take any value from i to — i (that is, if a lies on the smaller loop), two real coincident tangents can be drawn to (A), and the K^ has two real coin- cident cusps. If ^ «>-(]:=7/' where t may take any value from i to — i ; and if at the same time ''<-(T^' whei'e t may take any value from 1 to i and from — i to 1 (that is, if a lies out- side the smaller loop, but inside the larger), two real distinct tangents can be drawn to (A), and the Kl has two distinct real cusps. If _ _ 1 where t may take any value from 1 to i and from — i to 1 (that is, if a lies on the larger loop), two real distinct tangents and two real coincident tangents can be drawn to (A), and the K\ has two distinct real cusps and two coincident real '"^P'- ^f a = a: =1/4, that is, if a lies on the node of (A), two distinct pairs of coincident tangents can be drawn to a, and the Kl has two distinct pairs of real coincident cusps. If ">-(T:^' where t may take all values (that is, if a lies outside both loops), four real tangents can be drawn to a, and the K\ has four distinct real cusps. Consequently, it is impossible for the K\^ to have more than four real cusps. (b) Conic on the Cusps. The line equation of the /fj may be written as a polynomial in t, thus: at'^ — t^{x->ra) + t'^{x-\-y—\) — t{y-\-h)-\-h = Q. The invariant g.;^ of this quartic in t is ^l (x + a)(y-f fe) ) {x + y-iy ^ whose vanishing expresses the condition that four Vs be self-apolar. Hence, the clinants of the four tangents to the Kq from any point on the conic , {x + a){y + h) {x + y--iy _ ab ^ + ~ The Kq with four real cusps. 24 Curves with a Directrix. 25 are self-apolar. But from a cusp of the K\ three tangents coincide, and the four clinants are self-apolar regardless of the position of the fourth tangent. There- fore, this conic passes through the six cusps of the K\. (c) Nodes. From consideration of PLiJCKER's equations, we know that the Kl has four nodes. Since the equation '—s'-r^, + '-»''=" is of class four, there are in general four tangents to the K^ from any point of the plane. From a node, however, two pairs of tangents coincide, and so there are only two distinct tangents. Hence, if the equation of the Kl be written as a polynomial in t, thus : ^'-^'(^+l) + l'^'' + 2/-l)-~(2^ + &) + ^ = 0, (B) having for roots