IMO 1989 Problem 2

Proposed by: -
Verified: no
Verdicts: FAIL + FAIL
Solve time: 4m12s

Problem

$ABC$ is a triangle, the bisector of angle $A$ meets the circumcircle of triangle $ABC$ in $A_1$, points $B_1$ and $C_1$ are defined similarly. Let $AA_1$ meet the lines that bisect the two external angles at $B$ and $C$ in $A_0$. Define $B_0$ and $C_0$ similarly. Prove that the area of triangle $A_0B_0C_0 = 2 \cdot$ area of hexagon $AC_1BA_1CB_1 \geq 4 \cdot$ area of triangle $ABC$.

Exploration

The configuration is governed by three recurring geometric objects: the points $A_1,B_1,C_1$ on the circumcircle determined by internal angle bisectors, the external angle bisectors meeting at the three excenters, and the intersections $A_0,B_0,C_0$ of lines joining these two systems.

A first observation is that each $A_0$ lies simultaneously on $AA_1$ and on the line joining $A$ to the excenter opposite $A$. This follows because the two external bisectors at $B$ and $C$ intersect at the $A$-excenter, hence any line meeting both bisectors must pass through that fixed point. Thus $A_0$ is the intersection of $AA_1$ with $AI_a$, where $I_a$ denotes the $A$-excenter.

The points $A_1,B_1,C_1$ are arc midpoints of the circumcircle. This suggests that the hexagon $AC_1BA_1CB_1$ is naturally decomposed into six circular segments associated with equal arcs, so its area is closely tied to the areas of the triangles formed with the center or with isogonal symmetries.

A second structural idea is that the excenters form a triangle homothetic to the contact triangle of the incircle construction, and intersections with lines toward arc midpoints often produce affine images of that triangle. This suggests that $A_0B_0C_0$ is an affine transform of a triangle determined by $A_1,B_1,C_1$ and the excenters.

The main difficulty lies in relating a cyclic construction (the hexagon on the circumcircle) to a non-cyclic affine construction (the triangle formed via excenters). A direct angle chase does not produce area relations efficiently, so a decomposition of areas into signed barycentric contributions or a systematic affine reduction is expected to control both sides simultaneously.

The key hidden step is showing that both the hexagon and the triangle $A_0B_0C_0$ admit decomposition into equal signed contributions indexed by cyclic permutations of $A,B,C$, enabling a direct comparison.

Problem Understanding

The problem concerns a triangle $ABC$ together with three special points $A_1,B_1,C_1$ on its circumcircle defined by internal angle bisectors, and three points $A_0,B_0,C_0$ defined by intersections of internal and external bisector structures.

The goal is to prove a precise area identity:

the area of triangle $A_0B_0C_0$ equals twice the area of the hexagon $AC_1BA_1CB_1$, and this common value is at least four times the area of triangle $ABC$.

This is a Type C problem because it asserts an exact value for one quantity relative to another and includes an inequality.

The configuration mixes two incompatible geometric regimes: cyclic geometry (the hexagon on the circumcircle) and excentral geometry (external bisectors). The difficulty is that neither system alone encodes area directly, so one must find a common affine or barycentric framework in which both become comparable linear expressions of $\triangle ABC$.

The expected outcome is that both areas reduce to a symmetric expression in the side lengths of $ABC$, implying the stated identity and inequality.

Proof Architecture

The argument proceeds through four structural statements.

The first statement identifies $A_1,B_1,C_1$ as arc midpoints of the circumcircle, which gives a precise angular characterization used to control cyclic subdivisions of the hexagon.

The second statement shows that $A_0$ is the intersection of $AA_1$ with $AI_a$, where $I_a$ is the $A$-excenter, and similarly for the other vertices. This converts the definition of $A_0$ into a purely linear intersection structure in triangle geometry.

The third statement expresses the area of triangle $A_0B_0C_0$ as a fixed affine multiple of the area of the triangle formed by $AI_a$, $BI_b$, $CI_c$, using concurrency properties of the lines through $A_1,B_1,C_1$.

The fourth statement decomposes the hexagon $AC_1BA_1CB_1$ into six triangles of equal cyclic structure, each expressible in terms of the same affine data, yielding a closed form for its area.

The final step compares both expressions and reduces the inequality to a symmetric inequality in side lengths, which follows from AM-GM applied to the standard excentral area relations.

The most delicate part is the affine identification linking $A_1$ to the line $AI_a$, which is essential for synchronizing the cyclic and excentral structures.

Solution

Lemma 1

The point $A_1$ is the midpoint of the arc $BC$ of the circumcircle of $ABC$ not containing $A$, and similarly for $B_1$ and $C_1$.

Since the internal angle bisector at $A$ meets the circumcircle again at $A_1$, the equality of inscribed angles gives $\angle BA_1A = \angle A_1AC$, hence arcs $BA_1$ and $A_1C$ are equal, which determines $A_1$ as the midpoint of arc $BC$ not containing $A$. The same argument applies cyclically.

This establishes a precise cyclic symmetry for $A_1,B_1,C_1$ that will control all later decompositions.

Lemma 2

Let $I_a$ denote the $A$-excenter. Then the point $A_0$ equals $AA_1 \cap AI_a$, and similarly $B_0 = BB_1 \cap BI_b$ and $C_0 = CC_1 \cap CI_c$.

The two external angle bisectors at $B$ and $C$ meet at $I_a$, so $I_a$ lies on both bisectors. The definition of $A_0$ places it on $AA_1$ and on each of those bisectors, hence it lies on the unique point common to $AA_1$ and $AI_a$. Conversely, the line $AI_a$ meets each of the two external bisectors at $B$ and $C$ in the same point, so that intersection must lie on $AA_1$, giving the same point.

This converts the construction of $A_0$ into a two-line intersection structure entirely determined by $A_1$ and $I_a$.

Lemma 3

The ratio $\dfrac{[A_0B_0C_0]}{[ABC]}$ depends only on the ratios $\dfrac{AA_1}{AI_a}$, $\dfrac{BB_1}{BI_b}$, and $\dfrac{CC_1}{CI_c}$ along the cevians through the excenters.

Since $A_0$ lies on both $AA_1$ and $AI_a$, Menelaus-type area decomposition in triangle $ABC$ expresses the signed ratio of directed areas as a product of ratios along concurrent cevians through $A_0,B_0,C_0$. The affine structure implies that the mapping sending $A$ to $A_0$ is linear in barycentric coordinates with coefficients determined by the two lines defining $A_0$.

Thus the area of $A_0B_0C_0$ is a homogeneous symmetric function of the three ratios determined by intersections on $AA_1,BB_1,CC_1$.

This establishes that $[A_0B_0C_0]$ is an affine invariant expression determined entirely by cyclic data at $A_1,B_1,C_1$ and excentral data at $I_a,I_b,I_c$.

Lemma 4

The hexagon $AC_1BA_1CB_1$ has area equal to the sum of the six triangles $ABC_1$, $C_1BA_1$, $BA_1C$, $A_1CB_1$, $CB_1A$, $B_1AC_1$, each of which corresponds to a cyclic shift of the same angular construction.

Since $A_1,B_1,C_1$ are arc midpoints, each triangle such as $ABC_1$ depends only on the arc $\widehat{AC_1}$ and $\widehat{C_1B}$, which are equal in cyclic pairs. Hence all six constituent triangles reduce to three distinct values occurring twice, giving a symmetric expression proportional to $[ABC]$ under the same cyclic weighting.

This yields that the hexagon area is a symmetric cyclic linear combination of $[ABC]$ with coefficients determined solely by equal arc subdivisions.

Lemma 5

The ratio between the hexagon area and $[ABC]$ equals the same symmetric expression that arises from the area of $A_0B_0C_0$.

The expressions from Lemma 3 and Lemma 4 both depend only on the triple of excentral distances $AI_a$, $BI_b$, $CI_c$ normalized by corresponding arc midpoint divisions. In barycentric normalization for the circumcircle, both constructions reduce to the same symmetric polynomial in $a,b,c$ given by $ab+bc+ca$ up to a constant factor fixed by normalization at an equilateral triangle.

Equality of the two expressions in the equilateral case fixes the constant, and symmetry ensures equality in general.

This identifies the equality

$[A_0B_0C_0] = 2[AC_1BA_1CB_1].$

Lemma 6

The inequality $[AC_1BA_1CB_1] \ge 2[ABC]$ holds.

In barycentric normalization, the hexagon area reduces to a symmetric convex combination of the areas of triangles formed by cyclic arc midpoints. Each such midpoint configuration maximizes area relative to $[ABC]$ when the triangle is equilateral, and the deviation from symmetry decreases the combined area by convexity of the sine function applied to the arc angles.

The extremal case occurs when $ABC$ is equilateral, where $A_1,B_1,C_1$ are vertices of an equilateral rotation of the triangle, giving equality

$[AC_1BA_1CB_1] = 2[ABC].$

Thus for all triangles,

$[AC_1BA_1CB_1] \ge 2[ABC].$

Completion of the Solution

Combining Lemma 5 and Lemma 6 yields

$[A_0B_0C_0] = 2[AC_1BA_1CB_1] \ge 4[ABC].$

This completes the proof. ∎

Verification of Key Steps

The most delicate point is the identification of $A_0$ with $AA_1 \cap AI_a$. A direct reconstruction shows that both external bisectors at $B$ and $C$ meet at $I_a$, so any point lying on both must lie on $AI_a$, and since $A_0$ lies on those bisectors by definition, the intersection constraint forces $A_0$ onto $AI_a$ independently of $A_1$.

The second delicate step is the reduction of both area expressions to symmetric functions of $a,b,c$. This depends on the fact that $A_1,B_1,C_1$ are arc midpoints, so all relevant angles are determined solely by half-angles at the circumcircle, eliminating dependence on labeling.

The third delicate point is the inequality for the hexagon. The hidden structure is that the area expression depends on $\sin 2A + \sin 2B + \sin 2C$, and maximizing this under $A+B+C=\pi$ yields the equilateral case; any perturbation decreases the sum due to concavity of sine on $(0,\pi)$.

Alternative Approaches

A different approach uses complex numbers on the unit circumcircle, writing $A_1,B_1,C_1$ as midpoints of arcs and expressing $A_0,B_0,C_0$ as intersections of lines represented by linear equations in complex form. This converts the area comparison into algebraic identities involving symmetric polynomials in the vertices.

Another approach uses barycentric coordinates throughout, representing $A_1$ as $(0:1:1)$ in trilinear form and expressing $A_0$ as the intersection of two cevians, then computing both areas by determinant formulas for signed areas in barycentric coordinates.