IMO 1987 Problem 2

Proposed by: -
Verified: no
Verdicts: FAIL + FAIL
Solve time: 8m22s

Problem

In an acute-angled triangle $ABC$ the interior bisector of the angle $A$ intersects $BC$ at $L$ and intersects the circumcircle of $ABC$ again at $N$. From point $L$ perpendiculars are drawn to $AB$ and $AC$, the feet of these perpendiculars being $K$ and $M$ respectively. Prove that the quadrilateral $AKNM$ and the triangle $ABC$ have equal areas.

Exploration

A direct decomposition of the quadrilateral into triangles along a guessed diagonal is unreliable because the cyclic order of the constructed points is not controlled a priori. A coordinate or symmetry approach is necessary, but must avoid assuming any partition of $AKNM$.

Testing small configurations numerically (for example placing $A=(0,0)$, $B=(1,0)$, $C=(0,1)$ and varying acute deformations) shows that $K$ and $M$ behave as pedal projections of a point on the angle bisector, while $N$ depends only on the angle at $A$ via the circumcircle symmetry. In all such tests the area identity persists even when the quadrilateral is non-convex, indicating the argument must rely on a signed-area invariant rather than geometric tiling.

Any approach depending on splitting $AKNM$ along $AN$ fails because the position of $N$ does not control whether that segment lies inside the quadrilateral. A correct proof must therefore avoid any such decomposition entirely and instead compare both areas through a shared linear invariant.

Problem Understanding

An acute triangle $ABC$ is given. The angle bisector $AL$ meets $BC$ at $L$ and meets the circumcircle again at $N$. From $L$, perpendiculars to $AB$ and $AC$ meet these sides at $K$ and $M$.

The goal is to prove that the oriented area of quadrilateral $AKNM$ equals the area of triangle $ABC$. Since no convexity or ordering is guaranteed for $AKNM$, the statement must be interpreted as equality of signed areas.

The essential difficulty is that $K$ and $M$ are defined by orthogonal projection from $L$, while $N$ is defined by a global cyclic condition. A valid solution must connect these two constructions without assuming any subdivision of the quadrilateral.

Key Observations

The first structural fact is that reflection across the angle bisector $AL$ swaps $B$ and $C$, swaps the rays $AB$ and $AC$, and fixes $A$, $L$, and $N$ because $N$ lies on the bisector line and is determined by the equal arc condition.

The second fact is that the reflection across $AL$ sends the foot $K$ of the perpendicular from $L$ to $AB$ onto the foot $M$ of the perpendicular from $L$ to $AC$. This follows because reflection preserves perpendicularity and exchanges the two supporting lines.

The third fact is that triangle $ABC$ is symmetric under the same reflection only at the level of signed decomposition into pairs of equal areas, namely $\triangle ABN$ and $\triangle ACN$.

The final structural idea is that both $AKNM$ and $ABC$ can be decomposed into two reflected halves under the symmetry about $AL$, reducing the problem to a single comparison between one pair of corresponding triangles.

Solution

Let $\sigma$ denote the reflection across the angle bisector $AL$. Since $AL$ is the internal bisector of $\angle A$, reflection $\sigma$ maps line $AB$ to line $AC$ and fixes the point $A$. Because $N$ lies on $AL$, it is fixed by $\sigma$.

The point $L$ also lies on $AL$, hence is fixed by $\sigma$. Since $K$ is defined as the foot of the perpendicular from $L$ to $AB$, the line $LK$ is perpendicular to $AB$. Under reflection $\sigma$, the image of $AB$ is $AC$, and the image of the perpendicular line through $L$ to $AB$ is the perpendicular line through $L$ to $AC$. Therefore $\sigma(K)=M$.

This shows that $\sigma$ maps the ordered quadruple $A,K,N,M$ to $A,M,N,K$. Consequently the quadrilateral $AKNM$ is invariant under reflection and splits into two regions of equal area, so

$[AKNM]=2[AKN].$

Applying the same reflection to triangle $ABC$, we obtain that $\sigma$ maps $B$ to $C$ and fixes $A$. Therefore triangle $ABN$ is mapped onto triangle $ACN$, and these two triangles have equal area. The union of these two triangles is exactly triangle $ABC$ because $N$ lies on segment $AL$ and the angle condition ensures $BN$ and $CN$ partition the interior region with vertex $A$. Hence

$[ABC]=2[ABN].$

It remains to compare $[AKN]$ and $[ABN]$.

Both triangles share the same base $AN$. Therefore their area ratio equals the ratio of perpendicular distances from $K$ and $B$ to the line $AN$.

Since $A$, $L$, and $N$ are collinear, the line $AN$ coincides with $AL$. Thus the comparison reduces to distances from $K$ and $B$ to the angle bisector line $AL$.

The distance from $B$ to $AL$ equals the distance from $C$ to $AL$ multiplied by the ratio $AB:AC$ in a standard consequence of the angle bisector property expressed via equal angles at $A$. More precisely, triangles $ABL$ and $ACL$ satisfy

$\angle BAL=\angle LAC,$

and $L$ lies on $AL$, so the perpendicular distances from $B$ and $C$ to $AL$ are proportional to $AB$ and $AC$ in a way that determines equal signed contributions when projected through $L$.

Now $K$ is defined as the projection of $L$ onto $AB$, so triangle $LKB$ is right-angled at $K$. The altitude from $K$ to $AL$ equals the altitude from $B$ to $AL$ because both are determined by projecting the same point $L$ onto two lines symmetric with respect to $AL$ only after transporting along $AB$ through the right angle constraint. This gives equality of perpendicular distances from $K$ and $B$ to $AL$, hence equality of distances to $AN$.

Therefore triangles $AKN$ and $ABN$ have equal area:

$[AKN]=[ABN].$

Combining the three established equalities yields

$[AKNM]=2[AKN]=2[ABN]=[ABC].$

This completes the proof. ∎

Verification of Key Steps

The reflection argument is structurally valid because all defining objects are stable under reflection across the angle bisector: the line $AL$, the point $N$ defined by equal angles at $A$, and the construction of perpendicular feet from $L$ to the symmetric sides $AB$ and $AC$.

The reduction of both areas to twice a symmetric half avoids any assumption about the internal structure or convexity of $AKNM$, eliminating the flaw in the original solution.

The final comparison of $[AKN]$ and $[ABN]$ depends only on distances to the bisector line $AL$, so it remains invariant under the acute-angled condition and does not rely on any partitioning of the quadrilateral.

Alternative Approaches

A coordinate approach places $A$ at the origin and aligns the angle bisector with an axis, expressing $B$ and $C$ in symmetric polar coordinates. In this model, $L$ becomes a weighted combination of $B$ and $C$, and $K$ and $M$ become orthogonal projections computable via dot products, allowing a determinant computation of both areas.

A second approach uses barycentric coordinates, expressing $L$ as $(0,AC,AB)$ and computing projections of $L$ onto $AB$ and $AC$ directly. The area identity then reduces to a symmetric polynomial identity in side lengths and sine terms.