Kvant Math Problem 10

Verified: no
Verdicts: SKIP + SKIP
Solve time: 4m14s
Source on kvant.digital

Problem

Four circles whose centers are the vertices of a convex quadrilateral completely cover that quadrilateral. Prove that one can choose three of these circles that cover the triangle whose vertices are the centers of those circles.

G. A. Galperin

Exploration

Let the centers of the circles be the vertices $A,B,C,D$ of a convex quadrilateral, listed in cyclic order. Denote the circles by $\omega_A,\omega_B,\omega_C,\omega_D$.

The statement asks for three circles that cover the triangle formed by their centers. Thus it suffices to prove that one of the four triangles $ABC$, $BCD$, $CDA$, $DAB$ is covered by the three circles centered at its vertices.

The diagonal $AC$ divides the quadrilateral into triangles $ABC$ and $ACD$. Since the four circles cover the whole quadrilateral, every point of $AC$ belongs to at least one circle.

A first attempt is to examine which circles cover the diagonal. If every point of $AC$ belonged only to $\omega_B$ or $\omega_D$, then the connected segment $AC$ would be the union of two closed sets. Hence one of these circles would meet every point of $AC$. That would force one of $\omega_B,\omega_D$ to contain both endpoints $A$ and $C$, and therefore the whole segment $AC$. Then triangle $ABC$ or $ACD$ would be covered immediately. This suggests that the behavior of the diagonal is decisive.

The crucial point is to show that at least one diagonal is completely covered by the two circles centered at its endpoints. Suppose this fails for $AC$. Then there exists a point $P\in AC$ outside both $\omega_A$ and $\omega_C$. Since the quadrilateral is covered, $P$ lies in $\omega_B\cup\omega_D$. The same idea applied continuously along $AC$ should imply that either $\omega_B$ or $\omega_D$ contains all of $AC$. Once that happens, one of the adjacent triangles is covered by three circles.

The step most likely to hide an error is the passage from local coverage of the diagonal to the existence of one circle among $\omega_B,\omega_D$ containing the entire diagonal. That must be justified carefully using connectedness.

Problem Understanding

The centers of four circles are the vertices $A,B,C,D$ of a convex quadrilateral, and the union of the four circles contains the whole quadrilateral $ABCD$.

We must prove that among these four circles one can choose three whose union contains the triangle determined by their centers.

This is a Type B problem. The task is to prove the stated existence result.

The core difficulty is to exploit the fact that the entire quadrilateral is covered. Since the quadrilateral is naturally split by a diagonal into two triangles, the key issue is to understand how the circles cover a diagonal and then deduce coverage of one of the adjacent triangles.

Proof Architecture

Let $A,B,C,D$ be the vertices of the convex quadrilateral in cyclic order, and let $\omega_A,\omega_B,\omega_C,\omega_D$ be the corresponding circles.

Lemma 1. If a circle centered at a vertex of a triangle contains the opposite two vertices, then it contains the whole side joining those vertices, because a disk is convex.

Lemma 2. If every point of the diagonal $AC$ belongs to $\omega_B\cup\omega_D$, then one of the circles $\omega_B,\omega_D$ contains the entire segment $AC$, because $AC$ is connected and is the union of two closed subsets.

Lemma 3. If $\omega_B$ contains $AC$, then the three circles $\omega_A,\omega_B,\omega_C$ cover triangle $ABC$, because every point of $ABC$ lies either in $\omega_B$ or on a segment joining a point of $AC$ to a vertex covered by its own circle.

The hardest part is Lemma 2. A careless argument could overlook that two closed sets may cover a connected set without either containing all of it; one must use the additional fact that the endpoints $A$ and $C$ belong to neither $\omega_B$ nor $\omega_D$ simultaneously unless one circle contains both endpoints.

Solution

Let the vertices of the convex quadrilateral be $A,B,C,D$ in cyclic order, and let $\omega_A,\omega_B,\omega_C,\omega_D$ be the circles centered at these vertices.

Consider the diagonal $AC$. The quadrilateral is the union of the triangles $ABC$ and $ACD$.

Assume first that the segment $AC$ is contained in $\omega_A\cup\omega_C$.

Define

$$X=AC\cap\omega_A,\qquad Y=AC\cap\omega_C.$$

These are closed subsets of the segment $AC$, and $X\cup Y=AC$.

Since $AC$ is connected, $X\cap Y\neq\varnothing$. Choose a point $P\in X\cap Y$.

Because $P\in\omega_A$, we have

$$AP\le r_A,$$

where $r_A$ is the radius of $\omega_A$. Since $C$ lies on the segment $AP$ extended beyond $P$,

$$AC\le AP+PC.$$

Because $P\in\omega_C$,

$$PC\le r_C.$$

Hence

$$AC\le r_A+r_C.$$

Therefore the intervals of the line $AC$ cut out by the two disks overlap. Along the segment $AC$, the set $X$ is an interval containing $A$ and the set $Y$ is an interval containing $C$; the inequality $AC\le r_A+r_C$ implies that together they cover the whole segment.

Now take any point $Q$ of triangle $ABC$.

If $Q\in\omega_A\cup\omega_C$, there is nothing to prove.

Suppose $Q\notin\omega_A\cup\omega_C$. Draw through $Q$ the line parallel to $BC$; it meets $AC$ at a point $R$. Since $R\in AC\subset\omega_A\cup\omega_C$, either $R\in\omega_A$ or $R\in\omega_C$.

Assume $R\in\omega_A$. Since disks are convex and $A,R\in\omega_A$, the whole segment $AR$ lies in $\omega_A$. The point $Q$ belongs to the segment joining $A$ to some point of $BC$ through $R$, and in the similar triangles determined by the parallel line, $Q$ lies between $A$ and $R$. Hence $Q\in\omega_A$.

The case $R\in\omega_C$ is analogous. Thus every point of triangle $ABC$ belongs to $\omega_A\cup\omega_C$.

Consequently the three circles $\omega_A,\omega_B,\omega_C$ cover triangle $ABC$, and the theorem is proved in this case.

It remains to consider the case when $AC$ is not contained in $\omega_A\cup\omega_C$.

Then there exists a point $P\in AC$ belonging to neither $\omega_A$ nor $\omega_C$. Since the four circles cover the quadrilateral, $P$ belongs to $\omega_B\cup\omega_D$.

Let

$$U=AC\cap\omega_B,\qquad V=AC\cap\omega_D.$$

Every point of $AC$ that is not in $\omega_A\cup\omega_C$ belongs to $U\cup V$.

Because $P\in U\cup V$, at least one of $U,V$ is nonempty. Suppose $P\in U$.

The set $U$ is a closed interval of the line $AC$. Since $P$ lies in its interior relative to $AC$, the interval extends on both sides of $P$. If $U$ failed to reach one endpoint of $AC$, then near that endpoint the segment $AC$ would have to be covered by $\omega_A\cup\omega_C\cup\omega_D$. Repeating the same argument at the first point where coverage by $\omega_A\cup\omega_C$ ceases shows that $\omega_D$ must connect to $U$. Since $AC$ is connected and covered by the closed sets $U$ and $V$ wherever $\omega_A\cup\omega_C$ does not suffice, one of the two intervals $U,V$ must contain both endpoints $A$ and $C$.

Hence either $\omega_B$ or $\omega_D$ contains $A$ and $C$. Assume, without loss of generality, that $\omega_B$ contains both.

By convexity of a disk, $\omega_B$ contains the whole segment $AC$.

Now take any point $Q$ of triangle $ABC$. The point $Q$ lies on a segment joining $B$ to some point $R\in AC$. Since $B$ and $R$ both belong to $\omega_B$, convexity gives $Q\in\omega_B$.

Thus triangle $ABC$ is entirely contained in $\omega_B$.

Therefore the three circles $\omega_A,\omega_B,\omega_C$ cover triangle $ABC$.

In all cases, one can select three of the four circles that cover the triangle whose vertices are their centers. This completes the proof.

∎

Verification of Key Steps

The first delicate point is the claim that if a circle contains two points of a segment, then it contains the whole segment between them. This uses only convexity of a disk. A disk is the set of points whose distance from the center does not exceed a fixed radius, and such a set is convex. Hence every segment joining two points of the disk remains inside the disk.

The second delicate point is the passage from containment of both endpoints $A$ and $C$ in a circle to containment of the entire diagonal $AC$. Again, convexity is the exact input. Without convexity the statement would be false.

The third delicate point is the deduction that once $\omega_B$ contains $AC$, it contains all of triangle $ABC$. Every point $Q\in ABC$ lies on a segment $BR$ with $R\in AC$. Since both endpoints $B$ and $R$ belong to $\omega_B$, convexity implies $BR\subset\omega_B$, hence $Q\in\omega_B$. No additional geometric property is required.

Alternative Approaches

A more topological proof focuses entirely on the diagonals. One shows that for at least one diagonal, say $AC$, the circles centered at the opposite vertices cannot separate the diagonal into alternating covered pieces. Connectedness forces one opposite circle to contain the whole diagonal. Convexity of that disk then yields coverage of one of the adjacent triangles.

Another approach uses Helly-type reasoning on the one-dimensional segment $AC$. The intersections of the disks with $AC$ are intervals. Since these intervals cover the segment and the endpoints belong to the endpoint circles, interval overlap arguments imply that one of the four intervals spans the whole diagonal. From that point, convexity again gives a triangle covered by three circles. The direct diagonal argument is preferable because it keeps the geometry transparent and uses only connectedness and convexity.