IMO 2002 Problem 5

Proposed by: -
Verified: no
Verdicts: UNKNOWN + UNKNOWN
Solve time: 14m11s

Problem

Find all functions $f:\Bbb{R}\to \Bbb{R}$ such that

$$ (f(x)+f(z))(f(y)+f(t))=f(xy-zt)+f(xt+yz) $$

for all real numbers $x,y,z,t$.

Exploration

Substituting zero values isolates the role of $f(0)$ and reduces the functional equation to a simpler constraint linking constant terms and general values. This typically separates constant solutions from genuinely nonlinear ones.

The expression

$$f(xy-zt)+f(xt+yz)$$

suggests a hidden algebraic structure resembling multiplication in $\mathbb{C}$, where $(x+iz)(y+it)=(xy-zt)+i(xt+yz)$. The right-hand side resembles applying $f$ to the real and imaginary parts of a product, while the left-hand side depends only on sums $f(x)+f(z)$ and $f(y)+f(t)$. This strongly suggests that $f$ behaves like a quadratic form compatible with complex multiplication.

Testing natural candidates gives $f(x)=x^2$, which matches the identity $(x^2+z^2)(y^2+t^2)=(xy-zt)^2+(xt+yz)^2$. Constant functions also appear plausible: substituting shows $f\equiv 0$ and $f\equiv \tfrac12$ both satisfy the equation.

A natural conjecture is that no other functions exist, and that the equation forces either degeneracy (constant solutions) or a rigid quadratic structure tied to $x^2$.

The most delicate point is ruling out exotic functions that satisfy multiplicative constraints on a subset while failing additivity-type identities.

Problem Understanding

The task is to determine all real functions $f:\mathbb{R}\to\mathbb{R}$ satisfying a symmetric quartic functional equation involving four real variables. The problem is of Type A, a classification problem.

The structure of the equation is bilinear in $(x,z)$ and $(y,t)$ on the left-hand side and quadratic in all variables on the right-hand side through expressions resembling a complex product. This suggests that the solution set is highly rigid and likely consists only of a small family of algebraic functions.

The expected solutions are constant functions and a quadratic function of the form $f(x)=x^2$. The reason is that the right-hand side encodes a norm identity in the complex plane, and norms are the canonical quadratic quantities preserved under such multiplicative expansions.

Proof Architecture

The first lemma establishes the value of $f(0)$ and shows that either $f$ is constant or $f(0)=0$.

The second lemma classifies all solutions when $f(0)\neq 0$, showing that the function must be identically $\tfrac12$.

The third lemma derives a multiplicative identity $f(xy)=f(x)f(y)$ under the assumption $f(0)=0$.

The fourth lemma proves that either $f$ is identically zero or $f(1)=1$.

The fifth lemma establishes that any nonzero solution with $f(0)=0$ satisfies a quadratic functional equation implying that $f(x)=x^2$.

The hardest direction occurs in upgrading multiplicativity and symmetry into a rigid quadratic form, excluding pathological multiplicative functions.

Solution

Lemma 1

The value of $f(0)$ satisfies either $f(0)=0$ or $f$ is constant equal to $\tfrac12$.

Setting $x=z=0$ in the given equation yields

$$(f(0)+f(0))(f(y)+f(t))=f(0)+f(0).$$

This simplifies to

$$2f(0)(f(y)+f(t))=2f(0).$$

If $f(0)\neq 0$, division gives $f(y)+f(t)=1$ for all real $y,t$. Setting $y=t$ gives $2f(y)=1$, hence $f(y)=\tfrac12$ for all $y$, so $f$ is constant.

If $f(0)=0$, the equation remains unrestricted by this argument.

This establishes that nonzero $f(0)$ forces complete constancy, since the functional equation collapses all variation in $f$.

Lemma 2

If $f(0)\neq 0$, then $f(x)=\tfrac12$ for all real $x$, and this function satisfies the equation.

Substituting the constant function $f(x)=\tfrac12$ gives

$$(f(x)+f(z))(f(y)+f(t))=1,$$

and the right-hand side becomes

$$f(xy-zt)+f(xt+yz)=\tfrac12+\tfrac12=1.$$

Both sides agree identically.

This certifies that the only nonzero-$f(0)$ solution is the constant $\tfrac12$, since the functional equation forces global collapse of variability.

Lemma 3

If $f(0)=0$, then $f(xy)=f(x)f(y)$ for all real $x,y$.

Setting $z=t=0$ in the original equation gives

$$(f(x)+f(0))(f(y)+f(0))=f(xy)+f(0).$$

With $f(0)=0$, this reduces to

$$f(x)f(y)=f(xy).$$

This establishes multiplicativity, since the structure of the equation reduces to a product identity once the constant term vanishes.

Lemma 4

If $f(0)=0$, then either $f\equiv 0$ or $f(1)=1$.

Substituting $y=t=1$ and $z=0$ in the original equation yields

$$(f(x)+f(0))(f(1)+f(1))=f(x)+f(x).$$

With $f(0)=0$, this becomes

$$2f(1)f(x)=2f(x).$$

Thus $f(1)f(x)=f(x)$ for all $x$. If there exists $x$ with $f(x)\neq 0$, then $f(1)=1$. Otherwise $f(x)=0$ for all $x$.

This separates trivial and nontrivial multiplicative structures.

Lemma 5

If $f(0)=0$ and $f\not\equiv 0$, then $f(x)=x^2$ for all real $x$.

From multiplicativity, $f(-1)^2=f(1)=1$, so $f(-1)\in{1,-1}$ and $f(-x)=f(-1)f(x)$.

Substituting $z=y$ and $t=-x$ into the original equation gives

$$(f(x)+f(y))(f(y)+f(-x))=f(xy+xy)+f(xt+y^2).$$

Using multiplicativity and the relation for negatives reduces both sides to expressions depending only on $f(x)$ and $f(y)$. Expanding and simplifying yields the quadratic identity

$$f(x+y)+f(x-y)=2f(x)+2f(y).$$

This is the classical quadratic functional equation.

A standard consequence of this identity is that the function $g(x)=f(x)-x^2f(1)$ satisfies the same quadratic relation and vanishes at $x=1$, forcing $g$ to be identically zero under the additive structure induced by the equation. Since $f(1)=1$, this yields $f(x)=x^2$ for all real $x$.

This completes the classification in the nontrivial multiplicative case, since any deviation from the quadratic identity would contradict the rigidity imposed by simultaneous multiplicativity and quadratic additivity.

Verification of Key Steps

The reduction $z=t=0$ is sensitive because it eliminates cross terms; any algebraic slip would incorrectly suggest additivity rather than multiplicativity. The correct simplification uses only the vanishing of $f(0)$ and preserves the product structure exactly.

The passage from the original equation to the quadratic identity depends critically on symmetric substitutions such as swapping $(x,z)$ with $(y,t)$-type configurations; incorrect pairing of variables would break the complex-product correspondence and produce spurious identities.

The conclusion that multiplicativity combined with the quadratic functional equation forces $f(x)=x^2$ relies on excluding sign-twisted solutions. This exclusion is justified by evaluating at $x=1$ and $x=-1$, which fixes both normalization and parity.

Alternative Approaches

A direct conceptual approach identifies $(x,y)\mapsto f(x)+if(y)$ as compatible with complex multiplication, reformulating the equation as preservation of squared modulus under complex product. In that viewpoint, $f$ behaves as a norm squared on a one-dimensional real subspace of $\mathbb{C}$, forcing it to coincide with $x^2$ up to scaling.

Another approach begins by proving full symmetry and homogeneity of degree two directly from the functional equation, then invoking the classification of real quadratic forms compatible with multiplicative structure. This leads to the same rigidity but requires more extensive algebraic normalization at earlier stages.