[P1] JBMO 2013-2 (Three collinear midpoints)
by MSTang, Jan 3, 2016, 1:59 AM
JBMO 2013-2. Let
be an acute-angled triangle with
and let
be the centre of its circumcircle
. Let
be a point on the line segment
such that
. Let
be the second point of intersection of
and the line
. If
,
and
are the midpoints of the line segments
,
and
, respectively, show that the points
,
and
are collinear.
Solution
The condition
might be familiar! If it does, you might already know what point
really is. (Or you might be able to tell what's about to happen from the right-angle mark above
)
If not, let's take a look. We can calculate
by considering triangle
. This triangle is isosceles, with
, and we know that
by the inscribed angle theorem. Therefore,
and
must both equal
. Then,
. But
, so
. Point
is just the foot of the perpendicular from
to
!
(If you knew that
and
are isogonal conjugates with respect to
,
being the orthocenter, then point
being the altitude would be immediate.)
Well, now we can simplify the problem statement quite a bit:
Now, how can we get to points
,
, and
? How can we connect them to the rest of the diagram?
For one, we know that
and
, since
and
are midpoints of chords on the circle
. Also, we can recall a fact about the midpoint of the hypotenuse of a right triangle, which gives us
. With this in mind, why not draw in some extra segments, like
or
, and see if something pops up?
Wait - are
and
parallel? And equal in length, too?
Now we have a good idea: can we show that
and
are congruent? If so, that would immediately imply that
,
, and
are collinear. This seems promising, since we already know that
- this gives us a way to use point
!
We need to prove three things to establish congruence - so what other two? Maybe we can show that
. We break down
into
, which equals
, a right angle
, and angle
, which we currently have no information on. (And looking at it, it's hard to see a clean way to get that angle - it involves a foot of an altitude and the circumcenter.) So,
We can tease out
by considering quadrilateral
:
This works! Angles
and
sum to
, so
.
Now we're feeling confident about this congruence. Maybe the third thing to prove can be that
, since we know a lot about both lengths. For one, considering
, we have
(since
). Also,
. We can use some trigonometry to also prove that this equals
. A particularly quick way is the extended law of sines, with
:
So indeed,
, and
.
That's all we need, then - success!
by SAS, so
,
,
are collinear (from
). 
Here's a formal, short solution like the type I would write on an olympiad:



















Solution
![[asy]
size(8cm);
pair A=dir(100),B=dir(212),C=dir(-32),D=foot(A,B,C),E=dir(-100),O=(0,0),M=(B+E)/2,P=(A+C)/2,N=(O+D)/2;
draw(A--B--C--cycle^^B--E--A^^D--O);
draw(M--P,dashed+gray);
draw(unitcircle);
dot("$A$",A,dir(90));
dot("$B$",B,SW);
dot("$C$",C,.7*SE);
dot("$D$",D,SE);
dot("$E$",E,S);
dot("$M$",M,.5*SW);
dot("$O$",O,dir(90));
dot("$N$",N,SE);
dot("$P$",P,NE);
markscalefactor /= 6;
draw(rightanglemark(A,D,B));
add(pathticks(B--M,2));
add(pathticks(M--E,2));
add(pathticks(A--P,3));
add(pathticks(P--C,3));
add(pathticks(O--N,1));
add(pathticks(N--D,1));
[/asy]](http://latex.artofproblemsolving.com/1/3/8/138e10c6d467a6aa71c22689191f6e46f539b44e.png)



If not, let's take a look. We can calculate













(If you knew that





Well, now we can simplify the problem statement quite a bit:
Quote:
Let
be an acute-angled triangle with
and let
be the centre of its circumcircle
. Let
be the foot of the perpendicular from
to
. Let
be the second point of intersection of
and the line
. If
,
and
are the midpoints of the line segments
,
and
, respectively, show that the points
,
and
are collinear.



















Now, how can we get to points



For one, we know that








![[asy]
size(8cm);
pair A=dir(100),B=dir(212),C=dir(-32),D=foot(A,B,C),E=dir(-100),O=(0,0),M=(B+E)/2,P=(A+C)/2,N=(O+D)/2;
draw(A--B--C--cycle^^B--E--A^^D--O--P^^M--D);
draw(M--P,dashed+gray);
draw(unitcircle);
dot("$A$",A,dir(90));
dot("$B$",B,SW);
dot("$C$",C,.7*SE);
dot("$D$",D,SE);
dot("$E$",E,S);
dot("$M$",M,.5*SW);
dot("$O$",O,dir(90));
dot("$N$",N,SE);
dot("$P$",P,NE);
markscalefactor /= 6;
draw(rightanglemark(A,D,B));
draw(rightanglemark(O,P,C));
add(pathticks(B--M,2));
add(pathticks(M--E,2));
add(pathticks(A--P,3));
add(pathticks(P--C,3));
add(pathticks(O--N,1));
add(pathticks(N--D,1));
[/asy]](http://latex.artofproblemsolving.com/3/2/b/32bfe0f232b39f7ae369607be5ccb9342b39dd4a.png)


Now we have a good idea: can we show that







We need to prove three things to establish congruence - so what other two? Maybe we can show that






![\[\angle MDN = (90^\circ - C) + 90^\circ + \angle ADO = (180^\circ - C) + \angle ADO.\]](http://latex.artofproblemsolving.com/5/4/e/54ee83646d91366ae5cd4bd8c9b0002b849e4522.png)


![\[\angle NOP = 360^\circ - 90^\circ - C - \angle ODC = 270^\circ - C - \angle ODC.\]](http://latex.artofproblemsolving.com/6/4/b/64b64aa0ffb9cb12306685e93280c404f0bb6c81.png)




Now we're feeling confident about this congruence. Maybe the third thing to prove can be that







![\[BE = 2R \sin\angle BAE = 2R \sin (90^\circ - B) = 2R \cos B.\]](http://latex.artofproblemsolving.com/9/b/f/9bfb740165cf6816602ee06eb754f2f7e395c512.png)


That's all we need, then - success!






Here's a formal, short solution like the type I would write on an olympiad:
Quote:
Note that
Therefore,
, that is,
is the foot of the altitude from
to
. Now we claim that triangles
and
are congruent. For one,
. Consider quadrilateral
, in which
. Also, note that since
is right,
. Then we have
Finally, considering
, we have
where
is the radius of
, and considering right triangle
, we have
by the extended law of sines, so
. This is enough to conclude that
, so
, and thus
are collinear. 
![\[\angle BAD = \angle CAO = \dfrac{\pi - \angle AOC}{2} = \dfrac{\pi - 2 \cdot \angle B}{2} = \dfrac{\pi}{2} - \angle B.\]](http://latex.artofproblemsolving.com/d/3/7/d37c3ccee5187e56d254bfe074dc1e17e6c4d866.png)











![\[\begin{aligned} \angle NOP &= \dfrac{3\pi}{2} - \angle C - \angle ODC \\ &= \dfrac{3\pi}{2} - \angle C - \left(\dfrac{\pi}{2} - \angle ADO \right) \\ &= \pi - \angle C + \angle ADO \\ &= \pi - \angle C + \left(\angle MDO - \angle MDB - \dfrac{\pi}{2}\right) \\ &= \dfrac{\pi}{2} - \angle C + \angle MDO - \angle MBD \\ &= \dfrac{\pi}{2} - \angle C + \angle MDO - \angle DAC \\ &= \dfrac{\pi}{2} - \angle C + \angle MDO - \left(\dfrac{\pi}{2} - \angle C\right) \\ &= \angle MDO. \end{aligned}\]](http://latex.artofproblemsolving.com/e/d/5/ed51c3dd9db8079ef1185ba597bd4eea48a9c4a4.png)

![\[OP = AO \cos \angle AOP = R \cos\left(\dfrac{1}{2} \angle AOC\right) = R \cos \left(\dfrac{1}{2} \cdot 2\angle B\right) = R \cos \angle B\]](http://latex.artofproblemsolving.com/5/e/5/5e58e240b3dd2cbca1dc602ac735fc724ffd8373.png)



![\[MD = \dfrac12 BE = \dfrac12 (2R \sin \angle BAE) = R \sin \left(\dfrac{\pi}{2} - \angle B\right) = R \cos \angle B\]](http://latex.artofproblemsolving.com/a/b/2/ab24c5fb42a969dfcec48e5a366e36d7e517798d.png)





This post has been edited 2 times. Last edited by MSTang, Jan 4, 2016, 7:31 AM
Reason: removed "[P1]"
Reason: removed "[P1]"