0
$\begingroup$

Suppose that $A$ is a finite, nonempty set in an abelian group. If there is a group element with a unique representation as $a-b$ with $a,b\in A$, then none of $A-A$ and $2A$ are small: $$ \min\{|A-A|,|2A|\} \ge 2|A|-1; $$ this is almost immediate for the difference set, while the sumset case is a nontrivial result due to Kemperman and Scherk.

I need an analogue of this result for the additive energy of $A$ defined by $$ E(A):=\{(a_1,a_2,a_3,a_4)\in A^4\colon a_1+a_2=a_3+a_4\}. $$ A straightforward application of the Cauchy-Schwarz inequality gives the well-known estimate $E(A)\ge|A|^4/|2A|$. How small can $E(A)$ be given that there is an element with a unique representation as $a-b$ with $a,b\in A$? Given that there are "many" uniquely representable elements?

What I would need is an explicit estimate that works starting with very small values of $|A|$, not involving excessively large constants and the $O/o$-notation. Even an improvement by a lower-order term can be helpfull.

Improve this question
$\endgroup$
8
  • 2
    $\begingroup$ Additive energy $E(A)$ tends to behave inversely to the sumset size $|2A|$. So wouldn't the analogue of a lower bound on $|2A|$ be an upper bound on $E(A)$ rather than a lower bound? $\endgroup$
    – Terry Tao
    Jan 26, 2022 at 21:09
  • 1
    $\begingroup$ If $G = \mathbb Z/(3n)$ is a finite abelian group, and $A$ consists of all the multiples of $3$ together with $1$, then $1$ has a unique representation as a difference (in fact, every element that's not a multiple of $3$ does) but the additive energy of $A$ is at least $n^3 = (|A| -1 )^3$. So the additive energy can be very large under this condition. $\endgroup$
    – Will Sawin
    Jan 26, 2022 at 21:22
  • $\begingroup$ @TerryTao: I have to think more about the ideology, but anyway, the question is as stated. $\endgroup$
    – Seva
    Jan 26, 2022 at 21:34
  • $\begingroup$ @WillSawin: $(|A|-1)^3$ is about twice larger than $|A|^4/|2A|$, a way better than I need. $\endgroup$
    – Seva
    Jan 26, 2022 at 21:36
  • $\begingroup$ If $A$ is $n$ random elements in $\mathbb R$ then $E(A) = 2n^2- n $, $|2A|$ is $n (n+1)/2$, so $|A|^4/ 2|A|$ is $ 2n^3/ (n+1) = 2n^2 - 2n + 2 + O(1/n)$, which barely differs from $E(A)$. Of course some (in fact, every) element is uniquely represented. Is this OK for you? $\endgroup$
    – Will Sawin
    Jan 26, 2022 at 21:45

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.