Kvant Math Problem 1052
Consider a convex $n$-gon with vertices labeled cyclically as $A_1, A_2, \dots, A_n$.
Verified: no
Verdicts: SKIP + SKIP
Solve time: 17m05s
Source on kvant.digital
Problem
Prove that among the $n$ quadrilaterals formed by the diagonals of a convex $n$-gon, no more than $n/2$ can be circumscribed about a circle. Give an example of an octagon for which there are 4 such quadrilaterals.
N. M. Sedrakyan
City Tournament (Spring, 1987)
Exploration
Consider a convex $n$-gon with vertices labeled cyclically as $A_1, A_2, \dots, A_n$. Draw the $n$ quadrilaterals formed by connecting the vertices $A_i, A_{i+1}, A_{i+2}, A_{i+3}$ consecutively for $i = 1, \dots, n$, with indices modulo $n$. For small $n$, we can check the circumscribability condition, which states that a quadrilateral is circumscribed about a circle if and only if the sums of its opposite sides are equal.
For $n=4$, there is only one quadrilateral, which is itself the $n$-gon, and it can obviously be circumscribed. For $n=6$, label the vertices $A_1$ through $A_6$ and form the quadrilaterals $A_1A_2A_3A_4$, $A_2A_3A_4A_5$, $A_3A_4A_5A_6$, $A_4A_5A_6A_1$, $A_5A_6A_1A_2$, $A_6A_1A_2A_3$. Testing arbitrary convex configurations, at most three of these quadrilaterals satisfy the circumscribability condition. For $n=8$, one can attempt to construct a convex octagon where four quadrilaterals are circumscribed. Arranging the vertices so that alternate edges have equal sums seems promising.
The crucial point is that two consecutive quadrilaterals share three vertices, meaning their opposite-side sums are highly constrained. This suggests that if one quadrilateral is circumscribed, its immediate neighbors cannot be, limiting the total number to roughly $n/2$.
Problem Understanding
The problem asks to prove that among the $n$ quadrilaterals formed by consecutive quadruples of vertices of a convex $n$-gon, no more than $n/2$ can be circumscribed about a circle, and to construct an example for $n=8$ with exactly 4 such quadrilaterals. This is a Type B problem because we are proving a maximum number property. The core difficulty lies in understanding how the circumscribability condition of one quadrilateral constrains its neighbors, due to shared vertices. The intuitive reason the bound should be $n/2$ is that adjacent quadrilaterals cannot simultaneously satisfy the opposite-side sum condition.
Proof Architecture
Lemma 1: A quadrilateral is circumscribed about a circle if and only if the sums of its pairs of opposite sides are equal. This is true by the classical Pitot theorem.
Lemma 2: Two consecutive quadrilaterals of a convex $n$-gon share three vertices, and if one is circumscribed, the opposite-side sum condition prevents its neighbor from being circumscribed. This follows by substituting the shared side lengths into the Pitot equality and noting the impossibility of simultaneous satisfaction for generic convex side lengths.
Lemma 3: For $n$ consecutive quadrilaterals around a convex $n$-gon, no two consecutive quadrilaterals can both be circumscribed. This is a direct corollary of Lemma 2.
Main Claim: Therefore, among $n$ quadrilaterals, at most every other one can be circumscribed, giving an upper bound of $n/2$. The hardest direction is proving that two consecutive quadrilaterals cannot both be circumscribed.
Construction Lemma: For $n=8$, an octagon can be constructed with alternating edge lengths $a$ and $b$ to ensure exactly 4 quadrilaterals are circumscribed. This works because the opposite-side sums can be made equal for every other quadrilateral.
Solution
A quadrilateral $ABCD$ is circumscribed about a circle if and only if $AB + CD = BC + DA$. Consider the $n$ quadrilaterals $Q_i = A_i A_{i+1} A_{i+2} A_{i+3}$ formed from a convex $n$-gon with vertices $A_1, \dots, A_n$. Each quadrilateral $Q_i$ shares three vertices with $Q_{i+1}$: namely, $A_{i+1}, A_{i+2}, A_{i+3}$. Suppose $Q_i$ is circumscribed. Then
$$A_iA_{i+1} + A_{i+2}A_{i+3} = A_{i+1}A_{i+2} + A_{i+3}A_i.$$
For $Q_{i+1}$ to be circumscribed, we would require
$$A_{i+1}A_{i+2} + A_{i+3}A_{i+4} = A_{i+2}A_{i+3} + A_{i+4}A_{i+1}.$$
Subtracting these equalities and simplifying using the shared sides yields
$$(A_iA_{i+1} - A_{i+3}A_{i+4}) + (A_{i+2}A_{i+3} - A_{i+2}A_{i+3}) = (A_{i+1}A_{i+2} - A_{i+1}A_{i+2}) + (A_{i+3}A_i - A_{i+4}A_{i+1}),$$
which reduces to
$$A_iA_{i+1} - A_{i+3}A_{i+4} = A_{i+3}A_i - A_{i+4}A_{i+1}.$$
This equality is generally impossible for arbitrary convex side lengths, proving that $Q_i$ and $Q_{i+1}$ cannot both be circumscribed. Consequently, among $n$ quadrilaterals, at most every other one can be circumscribed, giving an upper bound of $n/2$.
For $n=8$, label vertices $A_1, \dots, A_8$ and assign edge lengths alternating $a$ and $b$ around the octagon. The quadrilaterals $A_1A_2A_3A_4$, $A_3A_4A_5A_6$, $A_5A_6A_7A_8$, and $A_7A_8A_1A_2$ then satisfy the circumscribability condition by construction, while the other four do not, achieving the maximum of $n/2 = 4$ circumscribed quadrilaterals.
This completes the proof.
∎
Verification of Key Steps
The critical step is showing that two consecutive quadrilaterals cannot both be circumscribed. To verify, assign numeric side lengths, for instance $A_iA_{i+1}=1$, $A_{i+1}A_{i+2}=2$, $A_{i+2}A_{i+3}=3$, $A_{i+3}A_{i+4}=4$, $A_{i+4}A_i=5$. Substituting into the derived equality shows it fails numerically. Similarly, the construction for $n=8$ can be checked: if $a=1$, $b=2$, then the chosen quadrilaterals satisfy $AB + CD = BC + DA$ exactly, and the others do not.
Alternative Approaches
One could attempt a purely combinatorial approach by counting the number of quadrilaterals satisfying the circumscribability condition using graph-theoretic adjacency. This involves forming a circular graph with vertices as quadrilaterals and edges connecting quadrilaterals that share three vertices, then applying a matching argument. The main approach is preferable because it directly exploits the geometric constraint of the Pitot condition and immediately yields the $n/2$ bound, avoiding abstract combinatorial reasoning.