IMO 1984 · Problem 4

Let \(ABCD\) be a convex quadrilateral such that the line \(CD\) is a tangent to the circle on \(AB\) as diameter. Prove that the line \(AB\) is a tangent to the circle on \(CD\) as diameter if and only if the lines \(BC\) and \(AD\) are parallel.

Official text of the problem. Official problem paper (PDF)

Solution

Write \([XYZ]\) for the area of triangle \(XYZ\), and \(d(X,\ell)\) for the perpendicular distance from \(X\) to a line \(\ell\). Let \(M\) and \(N\) be the midpoints of \(AB\) and \(CD\), respectively.

We use the following elementary observation: if two points lie on the same side of a line, the distance of their midpoint from that line is the average of their distances. This follows by projecting the points onto a perpendicular to the line.

By convexity, \(A\) and \(B\) lie on the same side of \(CD\). The given tangency means that \[ d(M,CD)=\frac{AB}{2}. \] Consequently, the triangle area formula gives \[ [ACD]+[BCD] =\frac{CD}{2}\bigl(d(A,CD)+d(B,CD)\bigr) =CD\,d(M,CD) =\frac{AB\cdot CD}{2}. \tag{1} \]

Similarly, \(C\) and \(D\) lie on the same side of \(AB\), so \[ [ABC]+[ABD]=AB\,d(N,AB). \] The circle on \(CD\) as diameter has center \(N\) and radius \(CD/2\). Thus \(AB\) is tangent to this circle exactly when \[ [ABC]+[ABD]=\frac{AB\cdot CD}{2}. \] By (1), this is equivalent to \[ [ABC]+[ABD]=[ACD]+[BCD]. \tag{2} \]

The two diagonals give the area identity \[ [ABC]+[ACD]=[ABD]+[BCD]. \] Hence \[ [ABC]+[ABD]-[ACD]-[BCD] =2\bigl([ABC]-[BCD]\bigr), \] so (2) holds exactly when \[ [ABC]=[BCD]. \]

These last two triangles have the same base \(BC\). Their areas are equal exactly when \(A\) and \(D\) have equal perpendicular distances from \(BC\). Convexity places \(A\) and \(D\) on the same side of \(BC\), so this happens exactly when \(AD\parallel BC\).

All steps are equivalences, proving both directions. The argument uses distances to the full lines and therefore also covers \(AB\parallel CD\) and tangency points outside the side segments. \(\square\)

The solution was checked against John Scholes' IMO solution archive (Kalva).