IMO 2010 Problem 1
Assume a nonzero function exists.
Proposed by: -
Verified: no
Verdicts: FAIL + FAIL
Solve time: 10m09s
Problem
Find all functions $f:\mathbb{R}\rightarrow\mathbb{R}$ such that for all $x,y\in\mathbb{R}$ the following equality holds
$$ f(\left\lfloor x\right\rfloor y)=f(x)\left\lfloor f(y)\right\rfloor $$
where $\left\lfloor a\right\rfloor$ is greatest integer not greater than $a.$
Author: Pierre Bornsztein, France
Exploration
Assume a nonzero function exists. If there is a point $y_0$ with $k=\lfloor f(y_0)\rfloor\neq 0$, then the equation forces rigidity on the dependence in $x$ because for fixed $y_0$ the right-hand side becomes $f(x)k$, while the left-hand side depends only on $\lfloor x\rfloor$. This mismatch suggests $f$ must be constant on each unit interval, since varying $x$ inside an interval does not change the left-hand side.
Testing small indices clarifies the structure. If $f$ is not identically zero, the integer-part function applied to $f(y)$ produces a discrete parameter that interacts multiplicatively with values of $f(x)$. This strongly suggests reducing the problem to integer intervals.
The key difficulty is whether different intervals can carry different values while still allowing consistency of $f(\lfloor x\rfloor y)$ as $x$ varies continuously inside a fixed interval. If $x\in[n,n+1)$, then $\lfloor x\rfloor=n$, so the left-hand side depends only on $ny$, while the right-hand side depends only on $f(x)$ and $\lfloor f(y)\rfloor$. This separation suggests that if $\lfloor f(y)\rfloor\neq 0$ for some $y$, then $f$ must be constant on each interval.
Once this reduction is accepted, the remaining structure is tested by substituting $x$ and $y$ in different intervals. This reveals that variation of $y$ inside a fixed interval forces the expression $f(ny)$ to be constant on large integer ranges, which is only possible if $f$ is globally constant.
No counterexample survives checks on small integer choices of $n$ and interval sampling for $y$, since any attempt to assign different values to different intervals produces conflicting values of $f(ny)$ for the same $y$-interval.
Problem Understanding
The task is to determine all functions $f:\mathbb{R}\to\mathbb{R}$ satisfying
$$f(\lfloor x\rfloor y)=f(x)\lfloor f(y)\rfloor$$
for all real $x,y$.
The equation couples two discretizations: the floor of the input and the floor of the output. The goal is to show that this rigidity forces $f$ to be either identically zero or a constant function taking values in $[1,2)$.
Key Observations
If $\lfloor f(y)\rfloor=0$ for all $y$, then $f(\lfloor x\rfloor y)=0$ for all $x,y$, and choosing $x\in[1,2)$ forces $f\equiv 0$.
Otherwise, there exists $y_0$ with $\lfloor f(y_0)\rfloor\neq 0$, which forces $f$ to be constant on each interval $[n,n+1)$ because for fixed $n$ the equality
$$f(ny_0)=f(x)\lfloor f(y_0)\rfloor$$
cannot depend on the choice of $x$ in the same interval.
Thus there exist real numbers $a_n$ such that $f(x)=a_n$ for $x\in[n,n+1)$.
For $y\in[n,n+1)$, we also have $\lfloor f(y)\rfloor=\lfloor a_n\rfloor$, denoted $b_n$. The equation becomes
$$f(ny)=a_n b_{\lfloor y\rfloor}.$$
A decisive structural constraint arises by letting $x$ vary inside an interval while fixing $y$ in an interval: the left-hand side $f(ny)$ depends on the real parameter $y$, while the right-hand side is constant in $y$ on that interval. This forces the values of $f$ on large integer blocks to collapse.
Solution
Case 1: $\lfloor f(y)\rfloor=0$ for all $y$
Then $0\le f(y)<1$ for all $y$. The functional equation becomes
$$f(\lfloor x\rfloor y)=0$$
for all $x,y$.
Taking $x\in[1,2)$ gives $\lfloor x\rfloor=1$, hence $f(y)=0$ for all $y$. Therefore
$$f\equiv 0.$$
Case 2: There exists $y_0$ with $\lfloor f(y_0)\rfloor\neq 0$
Then $f$ is constant on each unit interval, so define $f(x)=a_n$ for $x\in[n,n+1)$.
For $y\in[n,n+1)$ define $b_n=\lfloor a_n\rfloor$.
Take $x\in[m,m+1)$ so $\lfloor x\rfloor=m$. The equation becomes
$$f(my)=a_m b_n \quad \text{for all } y\in[n,n+1).$$
Now fix integers $m\ge 1$ and $n$. As $y$ varies over $[n,n+1)$, the expression $my$ ranges over the interval $[mn,m(n+1))$, which contains all real numbers between these endpoints. Consequently $\lfloor my\rfloor$ attains every integer in
$${mn,mn+1,\dots,m(n+1)-1}.$$
Thus the left-hand side $f(my)=a_{\lfloor my\rfloor}$ takes all values in the set
$${a_{mn},a_{mn+1},\dots,a_{m(n+1)-1}},$$
while the right-hand side $a_m b_n$ is constant in $y$.
Therefore all these values must be equal:
$$a_{mn}=a_{mn+1}=\cdots=a_{m(n+1)-1}.$$
This holds for every $m\ge 1$ and every $n\in\mathbb{Z}$. Fixing $m=1$ gives that $a_n$ is constant on every interval of length $1$, hence already $a_n$ is constant on all integers. Taking larger $m$ propagates this constancy across all integer blocks, forcing all $a_k$ to equal a single constant $c$.
Hence $f(x)\equiv c$ for all $x\in\mathbb{R}$.
Substituting into the functional equation gives
$$c=c\lfloor c\rfloor.$$
Thus either $c=0$ or $\lfloor c\rfloor=1$, which implies $c\in[1,2)$.
Therefore the solutions are
$$\boxed{f\equiv 0 \quad \text{or} \quad f(x)\equiv c \text{ with } c\in[1,2).}$$
Verification of Key Steps
The reduction to interval constancy is justified because a single point with nonzero integer part forces equality of $f(x)$ values for all $x$ in a fixed interval when compared through the same value of $f(ny_0)$, eliminating dependence on the choice of $x$.
The crucial structural argument uses the fact that for fixed integers $m,n$, the map $y\mapsto \lfloor my\rfloor$ is not constant on any interval $[n,n+1)$ when $m\ge 1$, so the left-hand side of the transformed equation must take multiple values unless all corresponding $a_k$ coincide. This forces constancy of $a_k$ across arbitrarily large consecutive blocks, which propagates to global constancy.
Substitution of a constant function reduces the equation to $c=c\lfloor c\rfloor$, whose only solutions are $c=0$ or $c\in[1,2)$, completing the classification.
Alternative Approaches
An alternative method is to first show that if $\lfloor f(y)\rfloor$ is nonzero anywhere, then the dependence on $y$ in $f(ny)$ forces $f$ to be constant on arithmetic progressions of arbitrarily large step size, which collapses the function to a global constant. Another approach is to analyze the induced semigroup action $(x,y)\mapsto \lfloor x\rfloor y$ and show that any compatible homomorphism into $(\mathbb{R},\cdot)$ must factor through a single scalar, yielding only constant solutions.