2
$\begingroup$

Let $A$ be a finite subset of $\mathbb{N}$. We denote the set $\{a_1 +a_2: a_1, a_2\in A\}$ as $2A$. We call the quantity $\sigma[A]:= |2A|/|A|$ as the doubling constant of $A$, and this constant can tell us some information about structure of the set $A$ (e.g., if $\sigma[A]= 2-1/|A|$, then $A$ is an arithmetic progression).

Given a number $k$, and a set $A$ with $\sigma[A]= k$, can we say something about the value of the constant $|3A|/|A|$? Intuitively, it seems that if $k$ is "large" then so should be $|3A|/|A|$.

There already exist some bounds for the general quantity $|nA|/|A|$, but what I want to know is the dependence of $|3A|$ on a given doubling constant, and I can't seem to find anything on this in the literature.

$\endgroup$

2 Answers 2

1
$\begingroup$

The following lemma shows that if the doubling constant $\sigma[A]=K$, then $|nA|/|A|$ cannot be too large.

Lemma. Let $A$ be an additive set with doubling constant $\sigma[A]=K$ for some $K\geq 1$. Then $$|nA|\leq \min(K^{Cn},n^{K^C-1})|A|$$ for all $n\geq 1$ and some absolute constant $C>0$.

Proof. Define a set $H:=A-A$ which is symmetric and contains origin, i.e. $H=-H$ and $0\in H$. Applying Ruzsa's covering lemma to the pair of additive sets $A$ and $H+H$ we obtain $$H+H\subseteq \underbrace{A-A}_{=H}+X \ \text{for some additive set $X$ such that} \ |X|\leq \frac{|H+H+A|}{|A|}.$$ Let's estimate the cardinality of $X$ $$|X|\leq \dfrac{|H+H+A|}{|A|}=\dfrac{|3A-2A|}{|A|}\leq \dfrac{(K^2)^{3+2}|A|}{|A|}= K^{10}.$$

It means that $H$ is a $K_1:=K^{10}$ approximate group.

We notice that $A\subseteq x+H$ for any $x\in A$.

Therefore $A\subseteq a_0+H$ for some $a_0\in A$. By induction we obtain that $nA\subseteq na_0+nH$ for any $n\geq 1$. Hence $$|nA|\leq |na_0+nH|=|nH|.$$ We have shown that $H$ is a $K_1$-approximate group and hence we have: $$|nH|\leq \min (K_1^n, n^{K_1-1})|H|.$$

But $|H|=|A-A|\leq K^2|A|$ and the above inequality implies that $$|nA|\leq K^2\cdot \min(K^{10n}, n^{K^{10}-1})|A|,$$ which can be written: $|nA|\leq \min(K^{Cn},n^{K^C-1})|A|$.

$\endgroup$
6
  • $\begingroup$ Can you explain why we can conclude $|3A-2A| \le (K^2)^{3+2} |A|$? Why does that step follow? $\endgroup$
    – D.W.
    Commented May 27 at 17:33
  • $\begingroup$ Can you explain what is the definition of "$K_1$ approximate group"? How does the cardinality of $X$ imply that $H$ is a $K_1$ approximate group? Why does being a $K_1$ approximate group imply an upper bound on $|nH|$? $\endgroup$
    – D.W.
    Commented May 27 at 17:35
  • $\begingroup$ @D.W., The answer to your first question: Generally, if the set $A$ has a small doubling or difference constant, then the size of the set $|nA-mA|$ is relatively small compared to the size of $|A|$. More precisely, if $|A+A| \leq C|A|$ or $|A-A| \leq C|A|$, then for any non-negative integers $n$ and $m$, we have: $|nA-mA| \leq C^{n+m}|A|$. This inequality is known as the Plünnecke-Ruzsa inequality. So in our case $n=3$ and $m=2$. $\endgroup$
    – RFZ
    Commented May 27 at 19:13
  • $\begingroup$ @D.W., Let $K\geq 1$. An additive set $H$ is said to be a $K$-approximate group if it is symmetric (so $H=-H$), contains the origin, and $H+H$ can be covered by at most $K$ translates of $H$. A few remarks: 1) When we say that $A$ is an additive set in an abelian group $G$, it means that $A$ is a finite non-empty subset of $G$. 2) $H+H$ can be covered by at most $K$ translates of $H$ means that $H+H \subseteq H+B$, where $|B| \leq K$. $\endgroup$
    – RFZ
    Commented May 27 at 19:20
  • $\begingroup$ Might you be open to editing your answer to incorporate that information into the proof? On Stack Exchange, instead of putting clarifications and additional information in the comments, we'd prefer that you revise the answer so it reads well for someone who encounters it for the first time, and so that people don't need to read the comments to follow. $\endgroup$
    – D.W.
    Commented May 27 at 20:06
1
$\begingroup$

Here's a more explicit answer for the case of $\lvert 3A\rvert$:

If $K=\lvert 2A\rvert/\lvert A\rvert$ then $\lvert 3A\rvert \leq K^3\lvert A\rvert$.

This is a special case of Plunnecke's inequality, for which Petridis gave a simple and elegant proof (see https://arxiv.org/abs/1101.3507)

This is actually sharp (up to constants), as the following example of Ruzsa shows: choose some parameters $M\ll N\ll M^2$ and consider in $\mathbb{Z}^3$ the set $A$ formed by three lines of length $N$ (take $N$ integer points along the three coordinate axes) and a box of sidelength $M$ (take all integer points $\{0,\ldots,M-1\}^3$).

One can check that $\lvert A\rvert \asymp M^3$ (most points coming from the box) and $\lvert 2A\rvert\asymp NM^2$ (most points coming from the three $N\times M\times M$ rectangles along the axes) and $\lvert 3A\rvert\asymp N^3$ (since $3A$ is basically the full integer box of sidelength $N$).

This means $K=\lvert 2A\rvert/\lvert A\rvert\asymp N/M$ and $\lvert 3A\rvert/\lvert A\rvert\asymp (N/M)^3\asymp K^3$.

(Choosing $N,M$ appropriately this can be achieved up to $K\approx \lvert A\rvert^{1/3}$.)

$\endgroup$

You must log in to answer this question.

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