0

I was looking at the integral $\int_{0}^{1}\lfloor{ax}\rfloor\lfloor{bx}\rfloor dx$. I found this resource for relatively prime $a$ and $b$ and I could follow the solution and simplified it to $\int_{0}^{1}\lfloor{ax}\rfloor\lfloor{bx}\rfloor dx=\frac{ab}{3}-\frac{a}{12b}-\frac{b}{12a}+\color{red}{\frac{1}{12ab}}-\frac{a}{4}-\frac{b}{4}+\frac{1}{4}$. On this MSE question, I saw a solution that held for non-relatively coprime $a$ and $b$: $\int_{0}^{1}\lfloor{ax}\rfloor\lfloor{bx}\rfloor dx=\frac{ab}{3}-\frac{a}{12b}-\frac{b}{12a}+\color{red}{\frac{\gcd(a,b)}{12ab}}-\frac{a}{4}-\frac{b}{4}+\frac{1}{4}$. My intuition told me to replace $a$ with $\frac{a}{c}$ and $b$ with $\frac{b}{c}$ where $c=\gcd(a,b)$ and somehow cancel out the $c$ but it didn't seem to be going anywhere. How can one extend the first solution to get to the second? Why does that single term change and nothing else?

Dylan Levine
  • 2,122

1 Answers1

1

My first answer was very wrong; I've hopefully made amends by changing it to a correct answer, but it's not a complete answer, merely a suggested approach.

The crucial term to evaluate in the first linked resource is $$ \sum_{u=0}^{ab-1} \Bigl\{ \frac ua \Bigr\} \Bigl\{ \frac ub \Bigr\} = \frac1{ab} \sum_{u=0}^{ab-1} (u\mod a)(u\mod b) $$ where $\{y\} = y - \lfloor y\rfloor$ is the fractional part of $y$. In the linked solution they assume that $\gcd(a,b)=1$. Howeever, in general, $(u\mod a)(u\mod b)$ is periodic with period $\mathop{\rm lcm}[a,b] = \frac{ab}{\gcd(ab)}$, and so the sum can be split into $\gcd(a,b)$ equal sums; it's also possible to characterize the pairs $\bigl( u\mod a, u\mod b\bigr)$ that occur, although it's a bit more complicated when $\gcd(a,b)>1$: we only get the ordered pairs where $(u\mod a) \equiv (u\mod b) \pmod{\gcd(a,b)}$.

Greg Martin
  • 92,241