IMO 1990 · Problem 4

Let \(\mathbb{Q}^{+}\) be the set of positive rational numbers. Construct a function \(f:\mathbb{Q}^{+}\to\mathbb{Q}^{+}\) such that \[ f(xf(y))=\frac{f(x)}{y} \] for all \(x,y\) in \(\mathbb{Q}^{+}\).

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

Solution

List the primes in increasing order as \(p_1,p_2,p_3,\ldots\), and pair them as \[ (p_1,p_2),\ (p_3,p_4),\ (p_5,p_6),\ldots. \]

By unique prime factorization, every positive rational \(q\) has a unique expression \[ q=\prod_{k\ge1}p_{2k-1}^{a_k}p_{2k}^{b_k}, \] where \(a_k,b_k\) are integers and only finitely many are nonzero. Define \[ \boxed{\displaystyle f(q)=\prod_{k\ge1}p_{2k-1}^{-b_k}p_{2k}^{a_k}.} \] This is a well-defined positive rational number, since its product also has only finitely many nontrivial factors. In particular, \(f(1)=1\).

Multiplying two positive rationals adds their corresponding prime exponents. The rule \((a_k,b_k)\mapsto(-b_k,a_k)\) preserves addition, so \[ f(uv)=f(u)f(v) \qquad\text{for all }u,v\in\mathbb{Q}^{+}. \] Moreover, applying the rule twice gives \[ (a_k,b_k)\longmapsto(-b_k,a_k)\longmapsto(-a_k,-b_k). \] Consequently, \[ f(f(q)) =\prod_{k\ge1}p_{2k-1}^{-a_k}p_{2k}^{-b_k} =\frac1q \qquad\text{for every }q\in\mathbb{Q}^{+}. \]

Thus, for every \(x,y\in\mathbb{Q}^{+}\), \[ f(xf(y)) =f(x)f(f(y)) =\frac{f(x)}{y}, \] as required.

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