Critical values for intermediate and box dimensions of projections and other images of compact sets

Nicolas E. Angelini, Ursula M. Molter · 2025

Read Critical values for intermediate and box dimensions of projections and other images of compact sets by Nicolas E. Angelini, Ursula M. Molter (2025) 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-2501-16598.txt.

Introduction

Nicolas Angelini and Ursula Molter wrote a 2025 math paper on the size of shadows of compact sets. When a set in high dimensional space is cast onto a lower space, its fractal size can vary with the casting angle. The authors ask for which sizes the shadow matches a target value, or the size of the set itself.

They study two size notions, one that sits between classic types and the box counting type. Their result covers near all casting spaces in the right measure sense. They also widen the claim past plain shadows. Maps such as wavy random motion paths count under the same frame. As a special case, the claims hold for the box size notion as well.

Catalog record

CREATORNicolas E. Angelini, Ursula M. Molter
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE57 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-16598

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Critical values for intermediate and box dimensions of projections and other images of compact sets Nicolas E. Angelini, Ursula M. Molter Abstract . Given a compact set E ⊂ ℝ d E\subset\mathbb{R}^{d} , we investigate for which values of m m the equality dim θ P V ​ ( E ) = m \dim_{\theta}P_{V}(E)=m or dim θ P V ​ ( E ) = dim θ E \dim_{\theta}P_{V}(E)=\dim_{\theta}E holds for γ d , m \gamma_{d,m} -almost all V ∈ G ⁡ ( d , m ) V\in G(d,m) . Our result extends to more general functions, including orthogonal projections and fractional Brownian motion. As a particular case, when θ = 1 \theta=1 , the results apply to the box dimension. 2020 Mathematics Subject Classification 28A80,28A75,60G22 1. Introduction The variation of the dimension of a set with respect to its orthogonal projection onto m -dimensional linear subspaces of \mathbb{R}^{d} is a classical problem in geometric measure theory, and there is a substantial body of results on this topic. In short, if G(d,m) denotes the set of all m -dimensional subspaces of \mathbb{R}^{d} , the question is: How is \dim P_{V}(E) related to \dim E , where E\subset\mathbb{R}^{d} , V\in G(d,m) , and P_{V} denotes the orthogonal projection onto V . When \dim represents the Hausdorff dimension, the problem has already been studied, yielding the result that if E\subset\mathbb{R}^{d} is a Borel set and P_{V} denotes the orthogonal projection onto an m -dimensional linear subspace V with m\leq d , then \dim_{\rm H}P_{V}E=\min\{\dim_{\rm H}E,m\} (1) for \gamma_{d,m} -almost all V\in G(d,m) . This result was first obtained for the case d=2 , m=1 by Marstrand [ 16 ] and later extended to Borel sets E\subset\mathbb{R}^{d} , d\geq 2 by Mattila [ 17 ] . The lower and upper \theta -intermediate dimensions, \underline{\mbox{\rm dim}}_{\,\theta} and \overline{\mbox{\rm dim}}_{\,\theta} respectively, form a continuous family of dimensions that interpolate between the Hausdorff and the box dimensions of bounded sets in \mathbb{R}^{d} . They satisfy, for all \theta\in(0,1] , the inequalities \dim_{H}E\leq\underline{\mbox{\rm dim}}_{\,\theta}E\leq\overline{\mbox{\rm dim}}_{\,\theta}E\leq\overline{\mbox{\rm dim}}_{\rm B}\,E and \underline{\mbox{\rm dim}}_{\,\theta}E\leq\underline{\mbox{\rm dim}}_{\rm B}\,E , with \overline{\dim}_{1}E=\overline{\mbox{\rm dim}}_{\rm B}\,E . For \theta=1 , we have \underline{\dim}_{1}E=\underline{\mbox{\rm dim}}_{\rm B}\,E . These dimensions are continuous functions of \theta for all \theta\in(0,1] ; however, in general, the \theta -intermediate dimension does not converge to the Hausdorff dimension as \theta\to 0 . For box and \theta -intermediate dimensions, general results such as ( 1 ) do not hold. Nevertheless, as shown in [ 6 ] for the box dimension and in [ 4 ] for \theta -intermediate dimensions, for any fixed \theta\in(0,1] , the value of \dim_{\theta}P_{V}(E) remains invariant for \gamma_{d,m} -almost every V\in G(d,m) . Recall that \dim_{1}P_{V}(E)=\dim_{B}P_{V}(E) . These values of the intermediate dimension are referred to by the authors as the “ (Upper or Lower) \theta -intermediate dimension profiles ,” and are denoted by \overline{\dim}^{m}_{\theta} and \underline{\dim}^{m}_{\theta} , respectively. In general, these profiles are difficult to work with. When dealing with projections, the superscript m in the dimension profiles typically denotes an integer. However, in other contexts – as we will see – the superscript may represent a real number. To distinguish between these cases, from now on we will use t for real numbers and m for integers when referring to dimension profiles. The precise result obtained in [ 6 ] and [ 4 ] is: Theorem 1.1 . Let E\subset\mathbb{R}^{d} be bounded. Then for all V\in G(d,m) , \underline{\mbox{\rm dim}}_{\,\theta}P_{V}(E)\leq\underline{\dim}^{m}_{\theta}E and \overline{\mbox{\rm dim}}_{\,\theta}P_{V}(E)\leq\overline{\dim}^{m}_{\theta}E for all \theta\in(0,1] , and for \gamma_{d,m} -almost all V\in G(d,m) , \underline{\mbox{\rm dim}}_{\,\theta}P_{V}(E)=\underline{\dim}^{m}_{\theta}E and \overline{\mbox{\rm dim}}_{\,\theta}P_{V}(E)=\overline{\dim}^{m}_{\theta}E . The reader can also refer to [ 13 ] for additional results on \theta -intermediate dimensions of projections. In the present paper, we study for which values of m , or under which conditions on the set E , results analogous to Marstrand’s theorem can be obtained for intermediate and box dimensions. We introduce the notions of the upper and lower quasi-Hausdorff dimensions of a set E , denoted by \overline{\dim}_{qH}E and \underline{\dim}_{qH}E , respectively, as \overline{\dim}_{qH}E:=\lim_{\theta\to 0}\overline{\dim}_{\theta}E\quad\text{and}\quad\underline{\dim}_{qH}E:=\lim_{\theta\to 0}\underline{\dim}_{\theta}E. This helps us make the statements and proofs of our results clearer and more straightforward. We show that, given a compact set E\subset\mathbb{R}^{d} and a positive integer m\leq d , the almost sure value of the upper and lower intermediate dimensions of the orthogonal projection onto an m -dimensional subspace satisfies \overline{\dim}_{qH}P_{V}E=\min\left\{m,\overline{\dim}_{qH}E\right\}\quad\text{for }\gamma_{d,m}\text{-almost every }V\in G(d,m), and similarly, \underline{\dim}_{qH}P_{V}E=\min\left\{m,\underline{\dim}_{qH}E\right\}\quad\text{for }\gamma_{d,m}\text{-almost every }V\in G(d,m), see Theorem 3.1 . From this, we deduce: \overline{\mbox{\rm dim}}_{\rm B}\,P_{V}(E)=m\text{ for }\gamma_{d,m}\text{-almost all }V\in G(d,m)\iff m\leq\overline{\dim}_{qH}E, and \underline{\mbox{\rm dim}}_{\rm B}\,P_{V}(E)=m\text{ for }\gamma_{d,m}\text{-almost all }V\in G(d,m)\iff m\leq\underline{\dim}_{qH}E. A direct consequence of Theorem 3.1 is that if E satisfies \underline{\mbox{\rm dim}}_{\rm B}\,E=\underline{\dim}_{qH}E , then \underline{\mbox{\rm dim}}_{\rm B}\,(P_{V}E)=\min\{m,\underline{\mbox{\rm dim}}_{\rm B}\,E\} for almost every V\in G(d,m) . Note that in [ 12 , Corollary 1.3] the authors show that if \underline{\dim}_{\mathrm{B}}E=\dim_{qA}E , then \underline{\dim}_{\mathrm{B}}(P_{V}E)=\min\{\underline{\dim}_{\mathrm{B}}E,m\} for almost every V\in G(d,m) . Since by Theorem 3.9 we have that, \underline{\dim}_{\mathrm{B}}E=\dim_{qA}E\quad\text{ implies }\quad\underline{\dim}_{\mathrm{B}}E=\underline{\dim}_{qH}E, and the converse is not necessarily true, the result proven in this paper is strictly more general than the one in [ 12 ] . All the above statements remain valid if one replaces the lower box dimension and the lower quasi-Hausdorff dimension with their upper counterparts. We also investigate for which values of m the intermediate dimension and the intermediate dimension profiles coincide. We obtain a general lower bound for the profiles in terms of the Assouad and Assouad dimension spectrum \dim^{\alpha}_{A} : \overline{\dim}_{\theta}^{t}E\geq\overline{\mbox{\rm dim}}_{\,\theta}E-\max\{0,\dim_{A}^{\alpha}E-t,(\dim_{A}E-t)(1-\alpha)\}, for all \alpha\in(0,1) , which yields that if m\geq\dim_{qA}E , then for all \theta\in(0,1] we have \dim_{\theta}E=\dim_{\theta}P_{V}E . This result was previously known only in the case of the box dimension. Another known bound that we adapt to intermediate dimensions is the inequality: \overline{\dim}_{\theta}^{s}E\geq\frac{\overline{\dim}_{\theta}^{t}E}{1+\left(\frac{1}{s}-\frac{1}{t}\right)\overline{\dim}_{\theta}^{t}E}, and we demonstrate with a simple example that this inequality is sharp, with equality holding for certain sets. Since these profiles are also related to the dimensions of more general families of functions – such as fractional Brownian motion, as shown in [ 3 ] – we can extend all our results to such functions. Theorem 3.1 and its corollaries provide genuinely new insights only in cases where the intermediate dimensions of the set are not continuous at \theta=0 . To illustrate this point, we now present an example of a set for which our projection theorems yield information that is not captured by existing results. Example 1.2 . Let n\geq 2 . For each 1\leq j<n , it is not difficult to construct a set E\subset\mathbb{R}^{n} such that \dim_{H}E<j\leq\underline{\dim}_{qH}E. Indeed, let F_{1}\subset\mathbb{R}^{j} be a set with j-1<\dim_{H}F_{1}=\dim_{B}F_{1}\leq j ; for instance, F_{1} could be a self-similar set. Define F_{2}=\left(\{0\}\cup\left\{\frac{1}{\log(n)}\right\}_{n\in\mathbb{N}}\right)^{n-j}\subset\mathbb{R}^{n-j}. Then F_{2} is a countable set such that \underline{\mbox{\rm dim}}_{\,\theta}F_{2}=\overline{\dim}_{\theta}F_{2}=n-j for all \theta\in(0,1] . Now define E=F_{1}\times F_{2} . Then \dim_{H}E=\dim_{H}F_{1}\leq j<n-j+\dim_{H}F_{1}=\underline{\dim}_{qH}E, where the last equality follows from the product rule for intermediate dimensions (see [ 1 , Theorem 5.4] ). Therefore, by Theorem 3.1 , we conclude that \underline{\dim}_{qH}P_{V}E=\overline{\dim}_{qH}P_{V}E=j\quad\text{for }\gamma_{n,j}\text{-almost every }V\in G(n,j). In contrast, Marstrand’s projection theorem gives \dim_{H}P_{V}E=\dim_{H}E<j\quad\text{for }\gamma_{n,j}\text{-almost every }V\in G(n,j). 2. Preliminaries Throughout the whole document B(x,r) , r>0 will denote the open ball in \mathbb{R}^{d} with center x and radius r . Given a non-empty set E\subset\mathbb{R}^{d} , \mathbb{I}_{E}(x) will represent the indicator

Read the complete text