Moving points tutorial
by yayups, Jan 5, 2019, 9:50 AM
I think a lot of people will really like this. (original link here https://artofproblemsolving.com/community/c6h1763036)
Hi All,
I've been getting many requests lately to explain this technique, so here is a short writeup which explains the method of projective maps from scratch. Any suggestions are welcome, and I hope it helps!
Basic familarity with projective geometry is assumed. Mainly just good intuition with cross ratios is required.
Let
be a conic, a line, or a pencil of lines through a fixed point (note that this last one is often ommited from the definition, but I find it to be quite important). These objects have a structure of cross ratio built into them, specifically, if
, then we can talk about
(note that if
is a pencil of lines, then
are lines). The definition of a projective map is now very simple.
Definition A projective map
from
to
where these are either conics, lines, or a pencil of lines is a function that preserves cross ratio. In particular, if
, then
And that's all!
The reason this is useful is the following theorem.
Theorem If
are projective, then
if
and
coincide on three different input values.
Proof: Suppose
,
,
. Then,
The important equation is
. Since the cross ratio is bijective, we have
, as desired. 
When solving problems, if we can phrase the problem as in the theorem, then actually we only have to check the problem for 3 cases! Before doing examples, let's cover some basic transformations that are projective maps. Your intuition with cross ratios is going to be really helpful here. Firstly, note that the composition of two projective maps is projective, and the inverse of a projective map is projective (Exercise: Prove this from the definition!). Here is a list I could come up with for projective maps (not exhaustive by any means).
You might be wondering where projecting from a line to a line fits in. It turns out this is actually a composition of two of the maps we listed above. In particular, if
is a point and
are lines, then we first map
using the first map, then map
using the inverse of the first map. Their composition is the classic perspectivity map. It's time to do some examples!
![[asy] size(8cm);
pair A = dir(129); pair B = dir(220); pair C = dir(320); filldraw(A--B--C--cycle, invisible, black);
pair M = 0.5*A + 0.5*B; pair N = 0.5*A + 0.5*C; pair O = circumcenter(A,M,N); pair Q = rotate(-90,A)*O; pair X = 4.2*Q - 3.2*A; pair Y = intersectionpoints(circle(X,abs(X-A)),circumcircle(A,M,N))[0]; pair Z = 0.5*A + 0.5*Y;
filldraw(circumcircle(C,Y,N), invisible, blue+opacity(0.5));
draw(A--X,red); pair S = circumcenter(C, N, Y); pair T = foot(S, B, C); pair D = 2*T - C; pair P = extension(S, midpoint(N--C), X, N); draw(N--X);
dot("$A$", A, dir(A)); dot("$B$", B, dir(B)); dot("$C$", C, dir(0)); dot("$M$", M, dir(135)); dot("$N$", N, dir(80)); dot("$X$", X, dir(X)); dot("$D$", D, dir(225)); [/asy]](//latex.artofproblemsolving.com/1/6/6/1663ff6c1db80a29582615c8e7cf1f07a223ee24.png)
Diagram made by v_Enhance. Let
be the tangent at
to
and let
. Let
be the map from
to
given by
. Note that this means
(directed angles).
So consider the map from
to
given by rotation by angle
. This is clearly projective. We have the map from
by
, and we have
by
. Therefore, composing all these projective maps in the order
we have that
is a projective map that maps
.
Similarly, we define a map
that sends
. We have again that its projective. We want to show
, so by the theorem, we just have to check this for three choices of
. This is now a geometry problem that is left to the reader!
Note: This proof is written much shorter like this. Let
. Then,
is projective, so
is projective. Similarly if
, then
is projective. So it suffices to check for three values of
.
The next example will have the proof again written nice and short like the above.
![[asy]
import olympiad;
import cse5;
size(8cm);
defaultpen(fontsize(9pt));
pair A = dir(110);
dot("$A$", A, dir(A));
pair B = dir(210);
dot("$B$", B, dir(B));
pair C = dir(330);
dot("$C$", C, dir(C));
pair I = incenter(A, B, C);
dot("$I$", I, dir(45));
pair O = origin;
pair D = extension(A, I, O, B+C);
dot("$D$", D, dir(225));
pair E = dir(310);
dot("$E$", E, dir(E));
pair F_1 = B*C/E;
pair F = extension(B, C, A, F_1);
dot("$F$", F, dir(F));
pair G = midpoint(I--F);
dot("$G$", G, dir(-15));
pair K = extension(E, I, D, G);
dot("$K$", K, dir(K));
pair I_A = 2*D-I;
pair P = extension(D, K, A, F);
pair Z = extension(I, P, I_A, F);
draw(A--B--C--cycle, red);
draw(unitcircle, blue);
draw(A--E--K--D--cycle, green);
draw(I--F);
[/asy]](//latex.artofproblemsolving.com/0/4/6/046adea104842e9f6d362caf7c8735e7a89e8679.png)
Diagram from v_Enhance. We will animate
on
, which means that we have mappings generated by varying
. Let
and
. We have that
is projective by the third item on our list.
We'll now show that
is projective. Note that
is projective since
is a reflection which clearly preserves cross ratio. Also,
is projective since it is a homothety at
with factor
(here the codomain is the scaling of
by
at
). Thus,
is projective. Projecting through
, we have that
is projective, so
is projective, as deisred.
Therefore, to show
, it suffices to check three values of
. It is now a geometry problem (an easy one at that) to check the problem for
(possible hint: fact 5).
Hi All,
I've been getting many requests lately to explain this technique, so here is a short writeup which explains the method of projective maps from scratch. Any suggestions are welcome, and I hope it helps!
Basic familarity with projective geometry is assumed. Mainly just good intuition with cross ratios is required.
Let





Definition A projective map




![\[(AB;CD) = (f(A)f(B);f(C)f(D)).\]](http://latex.artofproblemsolving.com/e/e/f/eef3e6921bf3e50a0478e5605bee11f297ab1906.png)
The reason this is useful is the following theorem.
Theorem If




Proof: Suppose



![\[(AB;CD)=(f(A)f(B);f(C)f(D))=(g(A)g(B);g(C)g(D))=(f(A)g(B);f(C)g(D)).\]](http://latex.artofproblemsolving.com/4/3/7/43702b598276b26d7f8611e252df0b17cb61c3a7.png)



When solving problems, if we can phrase the problem as in the theorem, then actually we only have to check the problem for 3 cases! Before doing examples, let's cover some basic transformations that are projective maps. Your intuition with cross ratios is going to be really helpful here. Firstly, note that the composition of two projective maps is projective, and the inverse of a projective map is projective (Exercise: Prove this from the definition!). Here is a list I could come up with for projective maps (not exhaustive by any means).
- Given a line
and a point
, the map from
to
(the pencil of lines through
) given by
- Given a conic
and a point
on the conic, the map from
to
given by
- Given a conic
and any point
, the map from
to
by
- This one's a little off beat, but still useful. Given two clines
and
, any inversion (or Mobius transform for that matter) that sends
to
is projective. This is because of the famous fact that inversion preserves cross ratio.
You might be wondering where projecting from a line to a line fits in. It turns out this is actually a composition of two of the maps we listed above. In particular, if




USA Winter TST for IMO 2019 Problem 1 wrote:
Let
be a triangle and let
and
denote the midpoints of
and
, respectively. Let
be a point such that
is tangent to the circumcircle of triangle
. Denote by
the circle through
and
tangent to
, and by
the circle through
and
tangent to
. Show that
and
intersect on line
.
Merlijn Staps



















Merlijn Staps
![[asy] size(8cm);
pair A = dir(129); pair B = dir(220); pair C = dir(320); filldraw(A--B--C--cycle, invisible, black);
pair M = 0.5*A + 0.5*B; pair N = 0.5*A + 0.5*C; pair O = circumcenter(A,M,N); pair Q = rotate(-90,A)*O; pair X = 4.2*Q - 3.2*A; pair Y = intersectionpoints(circle(X,abs(X-A)),circumcircle(A,M,N))[0]; pair Z = 0.5*A + 0.5*Y;
filldraw(circumcircle(C,Y,N), invisible, blue+opacity(0.5));
draw(A--X,red); pair S = circumcenter(C, N, Y); pair T = foot(S, B, C); pair D = 2*T - C; pair P = extension(S, midpoint(N--C), X, N); draw(N--X);
dot("$A$", A, dir(A)); dot("$B$", B, dir(B)); dot("$C$", C, dir(0)); dot("$M$", M, dir(135)); dot("$N$", N, dir(80)); dot("$X$", X, dir(X)); dot("$D$", D, dir(225)); [/asy]](http://latex.artofproblemsolving.com/1/6/6/1663ff6c1db80a29582615c8e7cf1f07a223ee24.png)
Diagram made by v_Enhance. Let









So consider the map from







![\[\ell_1\to\mathcal{C}_N\to\mathcal{C}_N\to\ell_2,\]](http://latex.artofproblemsolving.com/2/d/b/2db26114caa7a4f9bfbc2ab8e648312be613996f.png)


Similarly, we define a map




Note: This proof is written much shorter like this. Let

![\[X\mapsto NX\mapsto ND\mapsto D\]](http://latex.artofproblemsolving.com/6/6/3/663e83070f354db6cc3506378d2f63795acc3a1d.png)




The next example will have the proof again written nice and short like the above.
IMO 2010/2 wrote:
Given a triangle
, with
as its incenter and
as its circumcircle,
intersects
again at
. Let
be a point on the arc
, and
a point on the segment
, such that
. If
is the midpoint of
, prove that the meeting point of the lines
and
lies on
.
















![[asy]
import olympiad;
import cse5;
size(8cm);
defaultpen(fontsize(9pt));
pair A = dir(110);
dot("$A$", A, dir(A));
pair B = dir(210);
dot("$B$", B, dir(B));
pair C = dir(330);
dot("$C$", C, dir(C));
pair I = incenter(A, B, C);
dot("$I$", I, dir(45));
pair O = origin;
pair D = extension(A, I, O, B+C);
dot("$D$", D, dir(225));
pair E = dir(310);
dot("$E$", E, dir(E));
pair F_1 = B*C/E;
pair F = extension(B, C, A, F_1);
dot("$F$", F, dir(F));
pair G = midpoint(I--F);
dot("$G$", G, dir(-15));
pair K = extension(E, I, D, G);
dot("$K$", K, dir(K));
pair I_A = 2*D-I;
pair P = extension(D, K, A, F);
pair Z = extension(I, P, I_A, F);
draw(A--B--C--cycle, red);
draw(unitcircle, blue);
draw(A--E--K--D--cycle, green);
draw(I--F);
[/asy]](http://latex.artofproblemsolving.com/0/4/6/046adea104842e9f6d362caf7c8735e7a89e8679.png)
Diagram from v_Enhance. We will animate






We'll now show that













Therefore, to show



This post has been edited 1 time. Last edited by yayups, Mar 27, 2019, 11:40 PM