4
$\begingroup$

Motivation. If we consider any bijection $\varphi:\newcommand{\N}{\mathbb{N}} \N \to \N$, we say integers $m\neq n$ are shrinking with respect to $\varphi$ if $|m-n|>|\varphi(m) - \varphi(n)|$, and expanding if $|m-n|<|\varphi(m) - \varphi(n)|$. I wondered whether it is possible that expanding and shrinking pairs are in disbalance. Let's make this precise.

Precise formulation. For any set $X$, let $[X]^2 = \big\{\{x,y\}: x\neq y\in X\big\}$. For any positive integer $n\in\N$, we set $[n]^2 :=[\{1,\ldots,n\}]^2$. For $P\subseteq [\N]^2$, we let $$\newcommand{\mmu}{\mu_{[\N]^2}}\mmu(P) = \lim\inf_{n\to\infty}\frac{\newcommand{\card}{\text{card}}\card\big(P\cap [n]^2\big)}{\card\big([n]^2\big)},$$where of course $\card([n]^2) = n(n-1)/2$.

By $S_\N$ we define the set of all bijections $\varphi:\N\to\N$. For $\varphi\in S_\N$ we let the set of shrinking pairs be $\newcommand{\shr}{\text{shr}}\shr(\varphi) = \big\{\{m,n\}\in [\N]^2: |m-n| > |\varphi(m) - \varphi(n)|\big\}$, and for the set of expanding pairs $\newcommand{\exp}{\text{exp}}\exp(\varphi)$ we replace $>$ by $<$.

Question. What are the values of $\sup\big\{\mmu\big(\shr(\varphi)\big) :\varphi \in S_\N\big\}$ and $\sup\big\{\mmu\big(\exp(\varphi)\big) :\varphi \in S_\N\big\}$?

UPDATE. In the comment section below, Emil Jeřábek constructs a bijection $\varphi\in S_\N$ with $\mmu\big(\exp(\varphi)\big) = 1$, establishing $\sup\big\{\mmu\big(\exp(\varphi)\big) :\varphi \in S_\N\big\} = 1$.

$\endgroup$
6
  • 1
    $\begingroup$ It’s quite trivial to make almost all pairs expanding: e.g., let $\phi(n)=2n$ unless $n$ is a power of $2$, in which case $\phi(n)=2\log_2n+1$. $\endgroup$ Commented Jun 3 at 8:50
  • 1
    $\begingroup$ @EmilJeřábek I dont think this is a bijection. What is the preimage of 4? $\endgroup$ Commented Jun 3 at 9:08
  • 4
    $\begingroup$ Yeah, I’ve just realized the image of $\phi$ misses nontrivial powers of $2$. But that’s easy to fix: e.g., make $\phi(2^{2k})=2^{k+1}$, $\phi(2^{2k+1})=2k+1$. $\endgroup$ Commented Jun 3 at 9:10
  • 1
    $\begingroup$ Shrinking pairs seem more difficult to handle But as a start, for the inverse of the $\phi$ above, about $1/4$ of the pairs are shrinking. $\endgroup$ Commented Jun 3 at 9:12
  • 2
    $\begingroup$ It seems that the minimal number of non-strictly shrinking (or expanding) pairs for a permutation from $S_n$ is $\Theta(n^{3/2}$, so that the numerator in the $\liminf$ can be taken to any power less than $4/3$ without changing the result, but not further. $\endgroup$ Commented Jun 4 at 13:24

1 Answer 1

7
$\begingroup$

One has $\max\big\{\mu\big(\exp(\varphi)\big) :\varphi \in S_{\mathbb{N}}\big\}=1$, as mentioned by Emil Jeřábek in the comments.

We can also create a function $\varphi$ with $\mu\big(\text{shr}(\varphi)\big)=1$ in the following way. Consider a sequence of intervals $I_n=\{a_n,\dots,a_{n}+n-1\}$ in $\mathbb{Z}$, where $a_1=1$ and $a_{n+1}=a_n+n+2$, so that there are gaps $2$ points between consecutive intervals.

Also let $J_n=\{b_n,\dots,b_{n}+n-1\}$, where $b_1=1$ and $b_{n+1}=b_n+n+1$.

Note that the sets $\cup_nI_n$ and $\cup_nJ_n$ have density $1$ in $\mathbb{N}$.

Partially define a map $\varphi:\mathbb{N}\to\mathbb{N}$ by $\varphi(a_n+k)=b_n+k$ for all $k=1,\dots,n-1$. So $\varphi$ maps $\cup_nI_n$ bijectively to $\cup_nJ_n$. Extend $\varphi$ to a bijection in $\mathbb{N}$, it doesn't matter how.

Note that for any $a\in I_n$, $b\in I_m$ with $n\neq m$, we have $|\varphi(a)-\varphi(b)|<|a-b|$. This implies that $\mu_{[\mathbb{N}]^2}(\text{shr}(\varphi))=1$, as we wanted.

$\endgroup$
0

You must log in to answer this question.

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