Let $ABC$ be a triangle with $AB<BC<CA$ and let $D$ be a variable point on $BC$. The point $E$ on the circumcircle of $ABC$ is such that $\angle BAD=\angle BED$. The line through $D$ perpendicular to $AB$ meets $AC$ at $F$. Show that the measure of $\angle BEF$ is constant as $D$ varies.
Problem
Source: CGMO 2024 P4
Tags: geometry, CGMO
12.08.2024 23:31
It suffices to show that $\overline{EF}$ passes through a fixed point $X$ on $(ABC)$. I claim this is the point such that $AX=AC$. We use moving points, animating $D$ linearly, so $F$ is clearly linear as well. $E$ can be found as follows: intersect $(ABC)$ with the reflection of $(ABD)$ about $\overline{BC}$. An inversion (with fixed radius) at $B$ sends $(ABC)$ to a fixed line and $(ABD)$ to a circle through a fixed point and a degree $1$ point, so the image of $E$ has degree $1$. Thus $E$ has degree $2$. Furthermore since $E,X$ lie on the same fixed circle it's well-known that $\overline{EX}$ has degree $1$, so the statement $F \in \overline{EX}$ has degree $2$. Thus we check three cases: When $D=C$, $E,F$ conincide. When $\overline{AD} \perp \overline{AB}$, the projection of $A$ through $D$ onto $(ABC)$ is just the $B$-antipode, so $E=A$. Since $F=A$ as well, $E,F$ again coincide. When $D=\infty_{\overline{BC}}$, we want to show $E$ is the point such that $ACXE$ is an isosceles trapezoid. With a suitable phantom point argument and by reflecting everything across the perpendicular bisector of $\overline{BC}$, it's equivalent to show that if $Y$ is the point on $(ABC)$ with $\angle YBC=90^\circ$ then $Y$ is the midpoint of arc $AE'$, where $E' \in (ABC)$ is defined by having $CA=CE'$. This follows from angle chasing.
12.08.2024 23:33
13.08.2024 01:34
. Claim: $E'D'=E'B=E'F$ $$\angle E'BD'=\angle CAD'=2\angle COD'=2\angle CE'D'$$$$\angle D'F'B=90^{\circ}-\angle BAC+\angle BAD'=90^{\circ}-\angle CAD'=90^{\circ}-\frac{1}{2} \angle CE'D'=\frac{1}{2}\angle DE'B$$ To finish, $$\angle CE'F'=180^{\circ}-\angle F'E'B=180^{\circ}-2\angle F'D'B=180^{\circ}-2(\angle ACB+\angle BAC-90^{\circ})=2\angle ABC$$
Attachments:

13.08.2024 21:04
After $\sqrt{bc}$ invert all we need is: New problem wrote: On $\triangle ABC$ with circumcircle $\omega$ let $H_C$ be a point in $\omega$ such that $CH_C \perp AB$. An arbitrary point $D'$ moves in $\omega$ and now we let $D'H_C \cap AB=F'$. Prove that the circumcenter $E'$ of $\triangle BF'D'$ lies in $BC$. $$\angle E'BD'=90-\angle BF'D'=90-(90-\angle BAC+\angle BAD')=\angle BAC-\angle BAD'=\angle CAD'=\angle CBD'$$Is what gives the desire conclusion, now let $X'$ a point in $BC$ such that $AB=AX'$, note that as $\angle BF'E'=\angle ABC=\angle AXB$ we get that $AF'E'X'$ is cyclic, notice that $D',E',F'$ are inverses of $D,E,F$ in the original problem so inverting back we get $E,F,X$ colinear where $X$ is a fixed point (it's $X'$ inverted) so $\angle BFE=180-\angle BAX$ which is fixed thus we are done .
14.08.2024 03:10
bruh why invert only once Let $H$ be the foot from $D$ to $AB$. $\sqrt{ac}$ invert. $D'$ is a point on $AB$. $H$ is sent to the point on $BC$ such $\angle BD'H'=90^\circ$. $F'$ is simply the intersection of $(BD'H')$ and $(ABC)$. $E'$ is the intersection of the reflection of $CD'$ over $AB$ with $AC$. Our goal is to prove that $\angle BF'E'$ is fixed. We can just subtract/add fixed angles and it is clear that we only want $\angle AF'E'$ fixed. In fact, it suffices to prove that $AF'$ is the bisector of $\angle CF'E'$, as then $\angle AF'E'=\angle CF'A=\angle CBA$ we would be done. However, clearly $AB$ is the bisector of $\angle CD'E'$. Hence we just want to show that $F'$ lies on the apollonian circle determined by $E',A,C$. Now invert again!!! $\sqrt{bc}$ invert. $D''$ is a point on $AC$. $H''$ is the intersection of semicircle $D''A$ with $(ABC)$. $(D''H''C)$ intersects $BC$ at $F''$. We want to prove that $D''F''$ is perpendicular to $AB$. However, this is a clear angle chase as $\angle CF''D''=90^\circ-\angle CH''A=90^\circ-\angle B$ hence the perpendicularity is clear to see.
28.08.2024 15:23
This proof is not by me. I'm sending it just because it's supremely elegant, and no one is talking about it here yet. Let $H$ be the orthocenter of $\triangle ABD$. $\angle BED = \angle BAD = \pi - \angle BHD \Rightarrow BHDE$ is cyclic. Hence $\angle FAE = \angle CBE = \angle DHE \Rightarrow AHEF$ is cyclic. Therefore, $\angle BEF = \angle AEB + \angle AEF = \angle ACB + \angle AHD = \angle ACB + \pi - \angle ABC$ is a constant. $\square$
Attachments:
