[alogo] 1. Conic generated from a line by a projectivity

Let F be a projective transformation. Below I work in the set D of points X such that {X, F(X), F(F(X))} are not collinear. D can be empty (f.e. when F is a translation or homothety) but here I assume that D in non-empty. Under this assumption one can correspond to each line L of the plane a conic in the following way:
[1] The image F(L) = L' is also a line. For every point X of L the image F(X) is a point of L' and line LX which passes through {X, F(X)} is tangent to a conic cL.
The conic cL has the following properties:
[2] Let S be the contact point of line L with the parabola cL. Then S' = F(S) is the intersection point of lines {L, L'=F(L)}.
[3] Point F(F(S)) = S'' is a contact point of cL with line L'.
[4] For every point X of the line L line XX', where X'=F(X) (X on line L') is tangent to cL at a point W which is the harmonic conjugate of V with respect to {X,X'}, where V is the intersection point of XX' with line SS''.
[5] Triangle SS'S'' has the conic cL tangent to its sides {S'S, S'S''} at its vertices {S,S''} and is unique with this property. In other words there is no other point R of the same conic having corresponding triangle RR'R'' with R'=F(R), R''=F(F(R)) and lines {R'R, R'R''} tangent to cL at {R,R''} correspondingly.

[0_0] [0_1] [0_2]
[1_0] [1_1] [1_2]

The first four claims are proved in the same way are proved the corresponding claims for the special case of affinities studied in AffinityGeneratedParabola.html (*). The only difference is in the kind of the conic, which now is a general one and not necessarily a parabola. This follows from the way the conic is generated. One can namely show easily, introducing projective coordinates along lines SS' and S'S'' that the line coordinates between X and F(X) are in a homographic relation x' = (ax+b)/(cx+d). Hence one can apply the Chasles-Steiner argument and conclude that lines XX' envelope a conic (see Chasles_Steiner.html ).

To prove the fifth claim I apply here a modified argument of that used in (*).
In fact, assume that there are two triangles SS'S'' and RR'R'' as required in [5]. By Brianchon's theorem the diagonals of the quadrilateral formed by the tangents {SS', S'S'', RR', R'R''} and the lines joining the contact points pass through a common point O. There results the figure drawn below.
Define then the projectivity G by requiring {S'=G(S), S''=G(S'), R'=G(R), R''=G(R')}.
- Point O is a fixed point of G.
In fact, point Ï, being the intersection of {SR,S'R'} maps to the intersection point of the lines {S'R',S''R''} which is again Ï, hence this is a fixed point of G.

[0_0] [0_1] [0_2] [0_3]
[1_0] [1_1] [1_2] [1_3]
[2_0] [2_1] [2_2] [2_3]

- The polar PO of O with respect to the conic is an invariant line of G.
In fact, consider the intersection point T of SR with the polar PO. The cross ratio (S,R,O,T)=-1 (harmonic) is preserved by G. Besides line SR maps via G to S'R'. If follows that T maps via G to a point T' on PO.
Analogously it is shown that T' maps via G to the intersection point T'' of S''G'' with PO. This completes the argument showing that PO is invariant under G.
- Assuming now that F is a projectivity mapping every point X on SS' to a point X' on S'S'' so that XX' is tangent to the conic and also every point Y on RR' to a point Y' on R'R'' such that YY' is tangent to the conic we come to a contradiction.
In fact, projectivities F and G coincide at the four points {S,S',R,R'} hence they are identical. Taking {X,X',Y,Y'} along a tangent at a point M of the conic we see that this tangent is invariant under F hence its intersection point M' with PO (which is also invariant under F) is a fixed point of PO. It follows that PO consists entirely from fixed points. Hence the contradiction, since OS should coincide with OS' and OS'' which was excluded at the beginning.

[alogo] 2. Correspondence of a conic to a point

Let F be a projectivity. Let also D be the set of the plane for which {×,×',×''} with X'=F(X), X''=F(X') are not collinear. The previous analysis shows that for each × in D there is a conic, which I denote with k(X) (coincides with cL above), tangent at {×,×''} to lines {××', ×''×'}. The analysis shows that the correspondence is 1-1, i.e. to different points in D correspond different conics.
From the fifth property above follows also that no conic in the subset k(D) of all conics can be invariant under the projectivity F. Thus F introduces a permutation F* in k(D) with no fixed points.

See Also

AffinityGeneratedParabola.html
Chasles_Steiner.html

Return to Gallery


Produced with EucliDraw©