5

Let $BCD$ $(BC<BD)$ be a triangle inscribed in circle $(I)$. Let $E\in BC, F\in BD, G\in CD$ be such that $DE\perp BC, CF\perp BD, BG\perp CD$. Let $L$ be the orthocenter of triangle $BCD$. Let $H$ be the intersection point of $EG$ and $CF$, $J$ be the intersection point of $GF$ and $DE$. Let $K$ be the midpoint of $LI$.

Prove that $BK\perp HJ$.

Figure

MvG
  • 42,596
Blind
  • 1,116
  • Well, I don't know how much you have worked on this, but I would recommend 1. drawing an accurate picture and 2. finding all the results related to the things defined in the problem. – user357980 Jun 14 '17 at 03:50
  • I have tried but I cannot find the solution. It is a difficult question. Do you have any hint for me? – Blind Jun 14 '17 at 03:51
  • I spent one day for this problem but i cannot find the way to solve it. – Blind Jun 14 '17 at 03:52
  • I am not really a geometry person, but I would recommend drawing one very big and clean picture and then (maybe by labeling a copied version of that drawing) try to put in angles and see if there are any patterns that jump out. – user357980 Jun 14 '17 at 04:04
  • I would say that it doesn't look like an easy problem either. – user357980 Jun 14 '17 at 04:05
  • your suggestion is not useful for me.... – Blind Jun 14 '17 at 04:08
  • 1
    $K$ is the nine-point centre – CY Aries Jun 14 '17 at 04:17
  • @CYAries Thank you for your hint. How can we continue? – Blind Jun 14 '17 at 04:26
  • @Blind I haven't started working on the problem yet. Just read the question and noticed that you have a nine-point centre. – CY Aries Jun 14 '17 at 04:43
  • @CYAries Thank you very much for your kind hint. I appreciate you if you could continue your work. – Blind Jun 14 '17 at 04:56
  • Is I the center of the circle? If so, how do we know that it is on EF. And, most perplexing, whatever happened to A????????? – marty cohen Jun 14 '17 at 21:40
  • @martycohen: Sorry, I picked a bad configuration. In general $I$ is not on $EF$. Just encountered this by careless accident. Updated the figure to show a more generic situation. As to why the letter $A$ was omitted, I don't know either. Perhaps $I$ started out being called $A$ and got renamed later? Irrelevant, though. – MvG Jun 14 '17 at 21:53
  • 1
    This is a nice, non-obvious problem. Since you've commented that MvG's homogeneous coordinate approach is beyond your experience, please indicate how to better tailor a response to you. You should say something about the context in which the problem arose (found in a textbook, online, etc) and give an idea about the tools that are expected to be used here (simple high school geometry? "contest-level" geometry? coordinates and/or vectors? etc). – Blue Jun 15 '17 at 22:52
  • I tried (see below) but it was not very successful. Maybe there is some other more elegant ways to do it. – Mick Jun 16 '17 at 16:24
  • @Blue I am grateful if you could give a elementary and geometrical proof for this problem. You are right, I am looking for simple high school geometry. – Blind Jun 17 '17 at 03:25
  • Is this problem from a competition? – Ricardo Largaespada Jun 24 '17 at 13:55

3 Answers3

4

This may not be the kind of proof you are looking for, but if you can't come up with some nice theorems to use, sometimes brute force computation can yield a proof.

I'm using homogeneous coordinates. Without loss of generality, the circle is the unit circle, with $B=[1:0:1]$ at the right. Using the tangent half angle formula, there have to exist parameters $c$ and $d$ such that $C=[c^2-1:2c:c^2+1]$ and $D=[d^2-1:2d:d^2+1]$ as any point on the unit circle except for $B$ itself can be expressed like this. We also need the matrix $M:=\operatorname{diag}(1,1,0)$ which turns a line into the point at infinity orthogonal to that. We can use this to construct your perpendiculars. Joining points to form lines, as well as intersecting lines to obtain points, can be expressed using the cross product.

So (with a bit of help from a computer algebra system I'd suggest) one can perform the following computation, canceling common factors wherever they occur:

\begin{align*} I &= [0:0:1] \\ B &= [1:0:1] \\ C &= [c^2-1:2c:c^2+1] \\ D &= [d^2-1:2d:d^2+1] \\ BC &= B\times C = [c:1:-c] \\ CD &= C\times D = [1-cd:-c-d:cd+1] \\ BD &= B\times D = [d:1:-d] \\ DE &= (M\cdot BC)\times D = [d^2+1:-cd^2-c:2cd-d^2+1] \\ CF &= (M\cdot BD)\times C = [c^2+1:-c^2d-d:-c^2+2cd+1] \\ BG &= (M\cdot CD)\times B = [-c-d:cd-1:c+d] \\ E &= DE\times BC = [c^2d^2+c^2-2cd+d^2-1:2c^2d+2c:c^2d^2+c^2+d^2+1] \\ F &= CF\times BD = [c^2d^2+c^2-2cd+d^2-1:2cd^2+2d:c^2d^2+c^2+d^2+1] \\ G &= BG\times CD = [c^2d^2+c^2+2cd+d^2-1:2c+2d:c^2d^2+c^2+d^2+1] \\ L &= DE\times CF \\&= [3c^2d^2+c^2+d^2-1:2c^2d+2cd^2+2c+2d:c^2d^2+c^2+d^2+1] \\ H &= CF\times(G\times E) \\&= [c^4d^3+c^4d+4c^3d^2-4c^2d^3+2c^3-2c^2d-d^3-2c+d\\&\quad:2c^3d^3+4c^3d+2c^2d^2-2cd^3+4c^2-2d^2\\&\quad:c^4d^3+c^4d+2c^3d^2+2c^3+2cd^2-d^3+2c-d] \\ J &= DE\times(G\times F) \\&= [c^3d^4-4c^3d^2+4c^2d^3+cd^4-c^3-2cd^2+2d^3+c-2d\\&\quad:2c^3d^3-2c^3d+2c^2d^2+4cd^3-2c^2+4d^2\\&\quad:c^3d^4+2c^2d^3+cd^4-c^3+2c^2d+2d^3-c+2d] \\ K &= I_3\cdot L + L_3\cdot I \\&= [3c^2d^2+c^2+d^2-1:2c^2d+2cd^2+2c+2d:2c^2d^2+2c^2+2d^2+2] \\ BK &= B\times K \\&= [2c^2d+2cd^2+2c+2d:-c^2d^2+c^2+d^2+3:-2c^2d-2cd^2-2c-2d] \\ HJ &= H\times J \\&= [c^2d^2-c^2-d^2-3:2c^2d+2cd^2+2c+2d:-c^2d^2+c^2-6cd+d^2-3] \\ 0 &= BK^T\cdot M\cdot HJ = \langle BK,M\cdot HJ\rangle = \langle M\cdot BK,HJ\rangle \end{align*}

Let's repeat this in words. $I$ is the origin, $B$ the point $(1,0)$. $C,D$ are two generic points on the circle, excluding $B$ itself. $BC,BD,CD$ are lines joining these points. For $DE$ you take the line $BC$, compute the point at infinity orthogonal to that, then join it with $D$. $E$ is the intersection of that with $BC$. $H$ and $J$ are points of intersection between two lines again, as indicated. $L$ is computed as the intersection of two of the altitudes, $\langle L,BG\rangle=0$ verifies that the third altitude passes through that as well. $K$ is a midpoint, which you can compute by summing homogeneous coordinate vectors after ensuring that the homogenization term (i.e. the last coordinate) is the same. We do this by cross-multiplying with the last coordinate of the respective other term. In the end, you verify orthogonality by computing some bilinear form and checking that it evaluates to zero, or in more geometric terms by ensuring that the orthogonal point at infinity of one line is incident with the other line. q.e.d.

MvG
  • 42,596
  • Thank you for your kind help. Unfortunately, I cannot follow your ideas. I am grateful to you if you could provide a simpler proof for this problem. – Blind Jun 14 '17 at 21:58
  • I am waiting for simple high school geometry. Thank you for all support. – Blind Jun 17 '17 at 03:27
3

This is NOT a solution. I just want to share my finding which is too long to be included in a comment.

enter image description here

As pointed out, K is the center of the nine-point circle. Its properties, together with LK = KI, generate several pairs of parallel lines (marked in blue) but they don’t seem to help.

The standard way to prove $BK \bot HJ$ is to prove $\angle BMH = 90^0$; where M is the intersection of BK and HJ. This is done if we can prove BEBJ is cyclic by showing $\angle MJE = \angle MBE$. Or alternately, if we can show the purple lines are parallel; where EN is drawn to be perpendicular to BK extended.

Mick
  • 17,141
2

I am gonna continue on Mick's writings and image. I hope you know using trigonometry on geometry.

We should prove $\angle MJE = \angle MBE$ , so $\angle HJL = \angle KBE$. $$ $$

We know that $\angle HJL+\angle JHL=B=\angle KBE+\angle KBF$ $$ $$

So it's enough to show that $$\dfrac {\sin(HJL)}{\sin(JHL)}=\dfrac {\sin(KBE)}{\sin(KBF)}$$

$$ $$ (Left as an exercise: if $x+y=z+t\lt180$ and $\dfrac {\sin(x)}{\sin(y)}=\dfrac {\sin(z)}{\sin(t)}$, then $x=z$ and $y=t$)

$$ $$

Let $P$ and $Q$ be the projections of $K$ onto $BC$ and $BD$, respectively. Then,

$KP=KB*\sin(KBE)$ and $KQ=KB*\sin(KBF)$, so

$$\dfrac {\sin(KBE)}{\sin(KBF)}=\dfrac {KP}{KQ}$$

and also (Let $R$ be the radius of circumcircle of $BCD$)$$KP=\dfrac {LE+IT}{2}=\dfrac {2R\cos(B)\cos(C)+R\cos(D)}{2}=\dfrac {R\cos(B-C)}{2}$$, $$KQ=\dfrac {LF+IS}{2}=\dfrac {2R\cos(B)\cos(D)+R\cos(C)}{2}=\dfrac {R\cos(B-D)}{2}$$ $$\Rightarrow \dfrac {\sin(KBE)}{\sin(KBF)}=\dfrac {KP}{KQ}=\dfrac {\cos(B-C)}{\cos(B-D)}\tag {1}$$ $$ $$

By the triangle $HJL$, $$\dfrac {\sin(HJL)}{\sin(JHL)}=\dfrac {HL}{JL}=\dfrac {HL}{EL}*\dfrac {EL}{JL}$$

Because $\angle HEL=90-D$ and $\angle EHL=90+D-B$ $\quad$(easy to see) $$\dfrac {HL}{EL}=\dfrac {\cos(D)}{\cos(B-D)}$$

Because $GL$ is bisecting $\angle EGJ$

and because $\angle GEJ=90-D$ and $\angle GJE=90+B-C$ $\quad$(easy to see) $$\dfrac {EL}{JL}=\dfrac {EG}{GJ}=\dfrac {\cos(B-C)}{\cos(D)}$$ $$\Rightarrow \dfrac {\sin(HJL)}{\sin(JHL)}=\dfrac {HL}{EL}*\dfrac {EL}{JL}=\dfrac {\cos(B-C)}{\cos(B-D)}\tag {2}$$ $$ $$ By (1) and (2), $$\dfrac {\sin(HJL)}{\sin(JHL)}=\dfrac {\cos(B-C)}{\cos(B-D)}=\dfrac {\sin(KBE)}{\sin(KBF)}$$ $$ $$ $$\Rightarrow \angle HJL = \angle KBE $$ $$\Rightarrow \angle MJE = \angle MBE $$

Merdanov
  • 1,222
  • Nice work for introducing P and R, and the finding of $KP=\dfrac {LE+IT}{2}$ (and similarly KR.). Would it be better if we use r for the circum-radius instead? – Mick Jun 22 '17 at 16:22
  • I substituted Q for R, maybe it's better.(I am very much used to use R for circum-radius) – Merdanov Jun 23 '17 at 02:47