18
$\begingroup$

For $k \ge 2$, let

$$u = \{\lfloor{(k - \sqrt{k})n}\rfloor : n \ge 1\}$$ $$v = \{\lfloor{(k + \sqrt{k})n}\rfloor : n \ge 1\}.$$

My computer suggests that $u$ and $v$ are disjoint if and only if $k$ is an odd prime. Can someone give a reference, proof, or counterexample?

$\endgroup$
3
  • 1
    $\begingroup$ Are $k$ and $n$ integers? $\endgroup$
    – LSpice
    Commented Feb 4, 2022 at 17:39
  • $\begingroup$ A tiny bit of data: $\lfloor(2 - \sqrt2)6\rfloor = \lfloor(2 + \sqrt2)1\rfloor$; $\lfloor(4 - \sqrt4)3\rfloor = \lfloor(4 + \sqrt4)1\rfloor$; $\lfloor(6 - \sqrt6)12\rfloor = \lfloor(6 + \sqrt6)5\rfloor$; $\lfloor(8 - \sqrt8)2\rfloor = \lfloor(8 + \sqrt8)1\rfloor$; $\lfloor(9 - \sqrt9)2\rfloor = \lfloor(9 + \sqrt9)1\rfloor$; and, if I have not made a mistake, then these are the least coincidences among the two sets. Do you know any special case (e.g., non-squarefree $k$ don't produce disjoint sets)? $\endgroup$
    – LSpice
    Commented Feb 4, 2022 at 17:45
  • $\begingroup$ $u$ and $v$ are disjoint if and only if $k$ is odd and not a perfect square. See my response below. $\endgroup$
    – GH from MO
    Commented Feb 4, 2022 at 18:36

2 Answers 2

32
$\begingroup$

If $k$ is odd and not a perfect square, then the sets are disjoint. In particular, if $\alpha = \frac{k - \sqrt{k}}{\frac{k-1}{2}}$ and $\beta = \frac{k + \sqrt{k}}{\frac{k-1}{2}}$, then $\alpha$ and $\beta$ are irrational and $\frac{1}{\alpha} + \frac{1}{\beta} = 1$. Therefore, by Beatty's theorem, $A = \{ \lfloor \alpha n \rfloor : n \geq 1 \}$ and $B = \{ \lfloor \beta n \rfloor : n \geq 1 \}$ form a partition of the positive integers. The sets $u$ and $v$ are subsets of $A$ and $B$ respectively (in particular those with $\frac{k-1}{2} \mid n$).

In particular, the sets $u$ and $v$ are disjoint for $k = 15$, which is the smallest non-prime odd number that is not a perfect square.

$\endgroup$
1
  • 3
    $\begingroup$ And these are all the good $k$'s, be a theorem of Skolem (1957). See my response below. $\endgroup$
    – GH from MO
    Commented Feb 4, 2022 at 18:30
20
$\begingroup$

This is a supplement to Jeremy Rouse's nice answer. Let $\alpha$ and $\beta$ be positive irrational numbers. Skolem proved in 1957 (see Theorem 8 in On certain distributions of integers in pairs with given differences) that the Beatty sequences $[\alpha n]$ and $[\beta n]$ are disjoint if and only if $a/\alpha+b/\beta=1$ holds for some positive integers $a$ and $b$. It follows that the OP's sets $u$ and $v$ are disjoint if and only if $k$ is odd and not a perfect square.

$\endgroup$
2
  • 1
    $\begingroup$ The last section of the Skolem 1957 paper notes that his results for sets $\lfloor \alpha n \rfloor$ are extended to more general sets of the form $\lfloor (\alpha n + \beta) \rfloor$ to appear in Norske Vid. Selsk. Fohr., Trondheim. Can someone give a URL for that paper? $\endgroup$ Commented Feb 5, 2022 at 15:41
  • 2
    $\begingroup$ @ClarkKimberling The paper is Über einige Eigenschaften der Zahlenmengen $[\alpha n+\beta]$ bei irrationalem $\alpha α$ mit einleitenden Bemerkungen über einige kombinatorische Probleme, Norske Vid. Selsk. Forh., Trondheim 30 (1957), 42-49. It proves that the sequences $[\alpha n+\gamma]$ and $[\beta n+\delta]$ are disjoint if and only if there are integers $a,b>0$ such that $a/\alpha+b/\beta=1$ and $a\gamma/\alpha+b\delta/\beta\in\mathbb{Z}$. I don't know an URL for this paper. $\endgroup$
    – GH from MO
    Commented Feb 5, 2022 at 16:13

You must log in to answer this question.

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