-1
$\begingroup$

Let $\mathbb{N}$ denote the set of the positive integers. We consider the following function $f:\mathbb{N}\times \mathbb{N}\to \mathbb{Q}$: $$f(a,b)=\frac{a^2+b^2}{1+ab} \text{ for all } a,b\in\mathbb{N}.$$ This function has the amusing property that if $f(a,b)$ is an integer, then it is a square number.

We consider the following sets:

  1. $W_0 = \{n\in\mathbb{N}: f^{-1}(\{n^2\}) = \emptyset\}$,
  2. $W_\infty = \{n\in\mathbb{N}: |f^{-1}(\{n^2\})| = \aleph_0\}$, and
  3. $W_\text{fin} = \mathbb{N} \setminus (W_0\cup W_\infty)$.

What are the cardinalities of $W_0, W_\infty, W_\text{fin}$?

$\endgroup$
2

1 Answer 1

6
$\begingroup$

$W_\infty=\mathbb N\setminus\{1\}$. To see this, note that for $n>1$ we have a solution $(a,b)=(n,n^3)$ with $a<b$, and if we have one such solution, then $(b,n^2b-a)$ is another one (straightforward calculation) with $b<n^2b-a$, from which we easily construct an infinite sequence of distinct solutions.

For $n=1$, we just have to note that $f(a,b)=1$ iff $a^2-ab+b^2=1$ iff $(2a-b)^2+3b^2=4$ which is easily seen to have only one solution $(1,1)$ in $\mathbb N^2$, so $W_\text{fin}=\{1\}$.

Since you explicitly asked for the cardinalities of the respective sets: $|W_0|=0,|W_\text{fin}|=1,|W_\infty|=\aleph_0$.

Edit: proof that the pairs described in the first paragraph are, up to symmetry, all the solutions.

Suppose $(k,l),0<k<l$ were a solution not constructed in the above way. Moreover, let this be such a solution with minimal $k$ possible. Then $(n^2k-l,k)$ is a solution. We easily see $0\leq n^2k-l<k$. If $n^2k-l=0$, then we verify that $(k,l)$ was in fact a solution $(n,n^3)$, contradicting choice of $(k,l)$. If it is positive, then by minimality of $k$ it is one of the solutions $(a,b)$ constructed above. But then $(k,l)=(b,n^2b-a)$, again contradicting the choice of $(k,l)$. So the construction gives us all solutions.

$\endgroup$
5
  • 1
    $\begingroup$ Are these the only solutions? $\endgroup$ Commented Sep 12, 2016 at 3:11
  • 1
    $\begingroup$ @T.Amdeberhan Up to swapping $a$ and $b$ - yes, this can be shown with the Vieta jumping argument. $\endgroup$
    – Wojowu
    Commented Sep 12, 2016 at 6:30
  • 1
    $\begingroup$ Vieta can be used to show $f(a,b)\in\mathbb{N}$ implies $f(a,b)$ is a perfect square. I don't know how it applies to prove the above list $(a,b)$ and $(b,a)$ are the only solutions. $\endgroup$ Commented Sep 12, 2016 at 10:06
  • 1
    $\begingroup$ @T.Amdeberhan Sorry, when I've received your comment I was away from home and just now I have remembered about it. I have added a sketch of a proof that there are no solutions apart from the ones constructed in the list. $\endgroup$
    – Wojowu
    Commented Sep 18, 2016 at 19:20
  • $\begingroup$ I appreciate your write up in the reply. $\endgroup$ Commented Sep 19, 2016 at 11:49

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .