A note on the sum-product problem for fractal sets
Adam Cushman, William O’Regan · 2026
Read A note on the sum-product problem for fractal sets by Adam Cushman, William O’Regan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-21949.txt.
Introduction
Cushman and O'Regan show a sum-product bound for fractal sets, using recent advances in incidence geometry for balls and tubes and sum-product theory in the discrete setting. For s between 0 and 1/2, take any set A in the reals with Hausdorff dimension s. Then either the upper-box dimension of AA, or the lower-box dimension of A plus A, must be at least 29s/23. A slightly better bound of 33s/26 is obtained when the sum-set is replaced with the smoother difference-set.
Catalog record
| CREATOR | Adam Cushman, William O’Regan |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46065 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.21949
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A note on the sum-product problem for fractal sets Thanks: WOR is supported in part by an NSERC Alliance grant administered by Pablo Shmerkin and Joshua Zahl Adam Cushman, William O’Regan Abstract . Utilising recent advances in incidence geometry for balls and tubes, and advances in sum-product theory in the discrete setting, we show that for 0 < s ≤ 1 / 2 0<s\leq 1/2 and for any A ⊂ ℝ A\subset\mathbb{R} with Hausdorff dimension s , s, either the upper-box dimension of A A , AA, or the lower-box dimension of A + A A+A must be at least 29 s / 23 . 29s/23. We obtain the slightly better bound of 33 s / 26 33s/26 when we replace the sum-set with the smoother difference-set. 1. Introduction The Erdős–Volkmann ring problem [ EV66 ] asks whether there are any (Borel) subrings of \mathbb{R} which do not have Hausdorff dimension 0 or 1. In [ Fal85 ] Falconer showed (as a corollary of progress on the distance problem) that there are no Borel subrings of Hausdorff dimension strictly between 1/2 and 1. Motivated by the distance problem (and other problems) in [ KT01 ] Katz and Tao suggested the discretised ring problem, which is a more quantitative version of ring problem. They asked that, for a discrete set A, which at scale \delta resembles a set of Hausdorff dimension 1/2, to show that there exists a uniform c_{0}>0 so that N_{\delta}({A+A})+N_{\delta}({AA})>\delta^{-1/2-c_{0}}. This was resolved in [ Bou03 ] by Bourgain. Independently, using a completely different method, Edgar and Miller [ EM03 ] found a clever proof of the (Erdős–Volkmann) ring problem. It remains open as to how large the constant c_{0} ought to be. In terms of fractal dimension, the discretised ring problem may ask for A\subset\mathbb{R} with Hausdorff dimension 0<s<1 how large (1.1) \underline{\dim}_{\mathrm{B}}((A+A)\times AA) or \dim_{\mathrm{H}}((A+A)\times AA) would have to be. When 0<s\leq 2/3 the work of [ RW23 ] combined with the argument of Elekes [ Ele97 ] show that both are bounded below by 5s/2. We are not able to make any improvement of this quantity in this note, however, we remark that by product formulae the quantity ( 1.1 ) may be bounded above by both (1.2) \underline{\dim}_{\mathrm{B}}{(A+A)}+\overline{\dim}_{\mathrm{B}}{AA}, and (1.3) \overline{\dim}_{\mathrm{B}}{(A+A)}+\underline{\dim}_{\mathrm{B}}{AA}. Focussing our attention on ( 1.2 ), the lower bound of 5s/2 tells us that (1.4) \max\{\overline{\dim}_{\mathrm{B}}{(AA)},\underline{\dim}_{\mathrm{B}}{(A+A)}\}\geq 5s/4. The aim of this note is to improve this lower bound to 29s/23 when 0<s\leq 1/2. It was obtained (using a different technique) in [ DO25 ] that (1.5) \max\{\underline{\dim}_{\mathrm{B}}{(A/A)},\overline{\dim}_{\mathrm{B}}{(A+A)}\}\geq 43s/34, for 0<s\leq 1/2. It is unclear what the correct lower bound ought to be. In the discrete case, for A\subset\mathbb{R} it is conjectured that for all \epsilon>0 there exists c>0 so that |A+A|+|AA|>c|A|^{2-\epsilon}. In the fractal case, certainly the improvement cannot be larger than \min\{s,1-s\}. Further, in [ MO25 ] there is an example showing that the increase cannot be larger that (1-s)/2. When s is small, it is unclear whether to expect an improvement of s. For all 0<s<1 there is an example A\subset\mathbb{R} in [ MO25 ] with \underline{\dim}_{\mathrm{B}}(AA)=\underline{\dim}_{\mathrm{B}}(A+A)=s, and so ( 1.4 ) cannot be improved much in the sense of replacing \overline{\dim}_{\mathrm{B}} with a finer definition of dimension. As previously mentioned, the results presented here may be viewed as companion results to [ DO25 , Theorem 1.10 and Theorem 1.11] . There, similar results were obtained with AA replaced with A/A and A-A replaced with A+A, and with 33/26 replaced by the slightly worse 43/34. The reader is invited to view [ DO25 , Subsection 1.4] for the results presented there. Theorem 1.1 . Let 0<s\leq 1/2. Let A\subset\mathbb{R} satisfy \dim_{\mathrm{H}}A=s. Then (1.6) \max\{\overline{\dim}_{\mathrm{B}}{(AA)},\underline{\dim}_{\mathrm{B}}{(A+A)}\}\geq 29s/23. (1.7) \max\{\overline{\dim}_{\mathrm{B}}{(AA)},\underline{\dim}_{\mathrm{B}}{(A-A)}\}\geq 33s/26. Theorem 1.1 will follow from discretised versions below. The deduction is more-or-less identical to the one presented in [ DO25 ] so we omit this here. Theorem 1.2 . Let 0<s\leq 1/2 . There exists C,\delta_{0},\epsilon>0 so that the following holds for all 0<\delta<\delta_{0}. Let A\subset[1/2,1] be a (\delta,s,\delta^{-\epsilon}) -set. Then one of the following must hold. (1.8) N_{{\rho}}({AA})>\rho^{-29s/23}\text{ for some }\delta<\rho<\delta^{C\epsilon}, (1.9) N_{\delta}({A+A})>\delta^{-29s/23+C\epsilon}. Theorem 1.3 . Let 0<s\leq 1/2 . There exists C,\delta_{0},\epsilon>0 so that the following holds for all 0<\delta<\delta_{0}. Let A\subset[1/2,1] be a (\delta,s,\delta^{-\epsilon}) -set. Then one of the following must hold. (1.10) N_{{\rho}}({AA})>\rho^{-33s/26}\text{ for some }\delta<\rho<\delta^{C\epsilon}, (1.11) N_{\delta}({A-A})>\delta^{-33s/26+C\epsilon}. Note that 29/23=1.26086..., and 33/26=1.26923... The main key ingredient is the recent Szémeredi–Trotter strength incidence inequality for balls and tubes. Theorem 1.4 . Let 0<s\leq d\leq 1 . Set \sigma=\min\{s+d,2-s-d\}. Assume \mathcal{T} is a (\delta,s,d,K_{1},K_{2}) -quasi-product set. Consider a shading Y of \mathcal{T} such that Y(T) is a (\delta,\sigma,K_{3}) -KT set for each T\in\mathcal{T} . Then we have \mathcal{I}(\mathcal{T},Y)\lessapprox K_{3}^{\frac{1}{3}}(K_{1}K_{2})^{2/3}(\delta^{-s-d}\#\mathcal{T})^{1/{3}}\#Y(\mathcal{T})^{2/3}. Here a shading Y is an assignment of a distinct collection of \delta -balls to each T\in\mathcal{T}. We write Y(\mathcal{T})=\bigcup_{T\in\mathcal{T}}Y(T). We write \mathcal{I}(\mathcal{T},Y)=\sum_{T\in\mathcal{T}}\sum_{B\in Y(\mathcal{T})}1_{B\cap T\neq\emptyset}1_{B\in Y(T)}. The method to obtain Theorem 1.3 is quite different from the one in [ DO25 ] , although both rely on Theorem 1.4 . Here, we combine ideas present in sum-product literature in the discrete case. We sketch the main ideas below, focusing on the difference-set case. The sum-set case follows using standard modifications. We define the \delta - discretised 3 -energy of two sets A,B by (1.12) \mathrm{E}_{3,\delta}(A,B)=\sum_{z\in\delta\mathbb{Z}}\Big(\sum_{(a,b)\in A\times B}1_{|a-b-z|\leq\delta}\Big)^{3}. If A=B then we write \mathrm{E}_{3,\delta}(A)=\mathrm{E}_{3,\delta}(A,A). We let t\geq s be the ‘discrete Hausdorff dimension’ of A-A. The strategy can be broken into two distinct parts. First, we use Theorem 1.4 to upper bound \mathrm{E}_{3,\delta}(A,A-A) and \mathrm{E}_{3,\delta}(A) in terms of \#A,N_{\delta}({A-A}),N_{\delta}({AA})\text{ and }\delta^{-t}. Second, we find a lower bound for some multiplicative combination of N_{\delta}({A-A}) , N_{\delta}({AA}) , \mathrm{E}_{3,\delta}(A) , and \mathrm{E}_{3,\delta}(A,A-A) using a Cauchy-Schwarz argument. This framework is present in the discrete sum-product literature, where Theorem 1.4 is now playing the role of the Szemeredi-Trotter theorem. See, for example, [ SS11 ] , [ RS22 ] , [ Blo25 ] , and [ Cus25 ] . For the lower bound, we follow [ Cus25 ] quite closely. The idea is to produce many solutions to the equation p_{1}-p_{2}=p_{3}+O(\delta) , where the p_{i} lie in a ‘popular’ separated subset of A-A . However, a difficulty arises in our applications as one can not use the natural definition for the popular set. Indeed, the energy estimates obtained using Theorem 1.4 require the popular set to be more refined. To handle this, we introduce two notions of popular set. We have a refined set D_{\text{pop}} which is suitable for our estimates obtained using Theorem 1.4 , and a weaker set P , which is chosen so that \pi^{-1}(P) captures almost all pairs A\times A , where \pi(x,y)=x-y is the difference map. We note that P is not fit for use in Theorem 1.4 , but since we never apply energy estimates to P , this does not prove to be an obstacle. We produce solutions to the equation p_{1}-p_{2}=p_{3}+O(\delta) where p_{1},p_{2}\in D_{\text{pop}} and p_{3}\in P . Our lower bound, Lemma 3.6 , follows from a few applications of the Hölder inequality. 2. Preliminaries For A\subset\mathbb{R}^{d} let \mathcal{D}_{\delta}(A) denote the collection of \delta -dyadic squares which intersect A. We let N_{\delta}(A)=\#\mathcal{D}_{\delta}(A). We use \mathcal{N}_{\delta}(A) to denote the \delta -neighbourhood of A. Definition 2.1