arxiv: 2605.01824 · v2 · submitted 2026-05-03 · 🧮 math.DS · math.CA

Recognition: unknown

Self-similarity of unions of self-similar sets and their translations

Zhiqiang Wang

Pith reviewed 2026-05-09 16:42 UTC · model grok-4.3

classification 🧮 math.DS math.CA

keywords self-similar setsiterated function systemsunionstranslationsdirected graphscyclesCantor setsfractals

0 comments

The pith

Unions of a self-similar set with its translations remain self-similar precisely when an associated directed graph contains a cycle.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper determines exactly which strictly increasing translation vectors turn the union of copies of a given self-similar set into another self-similar set. This classification applies to sets generated by a fixed family of equal-ratio iterated function systems and matters because it identifies the algebraic conditions that let simple fractal constructions stay closed under translation unions. The test reduces to building a directed graph whose edges come from the possible mappings under the original contraction maps and then checking whether that graph contains a cycle. The work also exhibits two families of self-similar sets for which no choice of translations preserves the self-similar property.

Core claim

For the self-similar set Γ generated by the maps φ_i(x) = βx + i(1-β)/N with i from 0 to N and 0 < β < 1/(N+1), and for translation vectors t = (t_0, …, t_m) satisfying 0 = t_0 < t_1 < ⋯ < t_m, the union ⋃(Γ + t_j) is itself a self-similar set if and only if the directed graph constructed from the t_j’s contains a cycle.

What carries the argument

Directed graphs built from the translation vectors, whose cycles detect the algebraic relations needed for the union to satisfy its own iterated-function-system equation.

If this is right

The union satisfies a self-similarity equation exactly when the associated directed graph has a cycle.
The criterion is finite and checkable by standard graph algorithms for any fixed number of translations.
There exist self-similar sets generated by equal-ratio systems for which adding any translations prevents the union from being self-similar.
The cycle condition fully captures the necessary overlap relations imposed by the contraction maps.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same graph construction may reveal when unions of self-similar sets preserve other geometric invariants such as Hausdorff dimension.
The method supplies a systematic way to generate new self-similar sets by controlled translation unions inside the given contraction class.
Cycle-free translation choices could serve as counter-examples when testing conjectures about closure properties of self-similar sets.

Load-bearing premise

The underlying set comes from an equal-ratio iterated function system with contraction small enough relative to the number of maps, and the translations are strictly ordered and start at zero.

What would settle it

An explicit choice of N, β, and translation vector t where the constructed graph has no cycle yet the union equals the attractor of some new iterated function system, or the opposite mismatch.

read the original abstract

In this paper, we explore the self-similarity of unions of self-similar sets and their translations. For $N \in \mathbb{N}$ and $0< \beta < 1/(N+1)$, let $\Gamma$ be the self-similar set generated by the IFS \[ \Big\{ \phi_i(x)=\beta x + i \frac{1-\beta}{N}: i=0,1,\ldots, N \Big\}. \] We provide a complete characterization of translation vectors $\boldsymbol{t} =(t_0,t_1, \ldots, t_m) \in \mathbb{R}^{m+1}$ with $0=t_0 < t_1 < \cdots < t_m$ for which the union $\bigcup_{j=0}^m (\Gamma+t_j)$ is a self-similar set, by determining the existence of cycles in associated directed graphs. This extends the result of [Derong Kong, Wenxia Li, Zhiqiang Wang, Yuanyuan Yao, Yunxiu Zhang. On the union of homogeneous symmetric Cantor set with its translations. Math. Z., 2024]. Additionally, we present two types of self-similar sets for which the union with their translations cannot be self-similar.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper gives a graph-cycle test that classifies exactly which ordered translates make the union with a fixed equal-ratio Cantor set self-similar, extending the 2024 symmetric case.

read the letter

The central result is a characterization: for the IFS with N+1 equal-ratio maps and β small, the union of Γ with ordered translates t_j is self-similar precisely when the associated directed graph on the indices has a cycle. The graph tracks admissible differences under the IFS maps, and a cycle supplies the finite compositions needed to cover the union by its own images. This matches the abstract claim and the stress-test description of how overlaps are encoded without contradiction under the strict ordering and contractivity assumptions. The paper also exhibits two explicit families of translations that produce no cycle and therefore cannot yield self-similar unions. That concrete negative result is useful for checking boundary cases. The method stays inside the standard graph encoding of IFS overlaps, which fits the equal-ratio setup cleanly and avoids circularity. The extension from the symmetric 2024 paper is substantive within the subfield, though it inherits the same restrictions to equal contractions and sufficiently small β. A minor limitation is that the characterization does not immediately apply to unequal ratios or larger overlaps; the graph rules would need reworking there. The full details of the graph construction and the cycle-to-equation implication are not visible from the abstract alone, but the logic outlined in the stress-test holds together once the admissible edges are defined from the translation differences. Readers who work on overlaps and classification of self-similar sets will get direct value from the explicit criterion and the counter-examples. The work is incremental but grounded and reproducible in principle, so it belongs in a journal that covers fractal geometry and IFS. I would send it to peer review.

Referee Report

2 major / 2 minor

Summary. The paper claims a complete characterization, for the equal-ratio IFS with contractions β < 1/(N+1), of the strictly ordered translation vectors t = (0 = t0 < t1 < ⋯ < tm) such that the finite union ⋃(Γ + tj) is itself self-similar; the criterion is the existence of cycles in a directed graph whose vertices are the translation indices and whose edges encode admissible overlaps under the IFS maps. The work extends the 2024 result of Kong–Li–Wang–Yao–Zhang on homogeneous symmetric Cantor sets and supplies two families of self-similar sets for which no such non-trivial union is self-similar.

Significance. If the graph-theoretic test is rigorously established, the result supplies an explicit, algorithmic criterion for self-similarity of translated unions of a concrete class of Cantor sets. The construction re-uses standard overlap-graph techniques for IFS but specializes them to the equal-ratio, small-β regime, yielding a practical decision procedure that was previously unavailable beyond the symmetric-Cantor case.

major comments (2)

[§3 (graph construction)] The directed-graph construction (presumably §3) is described only at the level of “admissible translation differences”; the precise adjacency rule—i.e., the condition on tj − tk that produces an edge from vertex j to vertex k—must be stated explicitly, together with the verification that every possible covering of the union by the IFS maps corresponds to a path in this graph. Without this, it is impossible to confirm that the cycle condition is necessary and sufficient for the self-similarity equation.
[main theorem (presumably Theorem 1.1 or §4)] The proof that the absence of cycles implies the union fails to be self-similar (the “only if” direction of the main theorem) relies on the strict ordering and the bound β < 1/(N+1) to control overlaps, yet the argument does not exhibit an explicit contradiction with the self-similarity equation when the graph is acyclic; a concrete infinite-descent or measure argument is needed to close the gap.

minor comments (2)

[Abstract] The abstract states that two types of self-similar sets are shown to yield non-self-similar unions, but neither the types nor the relevant IFS are identified; a one-sentence description would improve readability.
[§1] Notation: the integer N (number of IFS maps) and the integer m (number of translations) are independent parameters; their relationship, if any, should be stated once in the introduction.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough review and insightful comments on our manuscript. We address each of the major comments below and indicate the revisions we plan to make.

read point-by-point responses

Referee: [§3 (graph construction)] The directed-graph construction (presumably §3) is described only at the level of “admissible translation differences”; the precise adjacency rule—i.e., the condition on tj − tk that produces an edge from vertex j to vertex k—must be stated explicitly, together with the verification that every possible covering of the union by the IFS maps corresponds to a path in this graph. Without this, it is impossible to confirm that the cycle condition is necessary and sufficient for the self-similarity equation.

Authors: We agree that the adjacency rule needs to be stated more explicitly for clarity. In the revised version, we will add a precise definition: an edge from vertex j to vertex k exists if there exists an IFS index i such that the translation difference t_j - t_k equals the admissible overlap induced by φ_i on the self-similar set, consistent with the equal-ratio structure and the bound on β. We will also include a lemma verifying that every covering of the union by the IFS maps corresponds to a path in this graph, thereby confirming that the cycle condition is necessary and sufficient for self-similarity. These additions will appear in Section 3. revision: yes
Referee: [main theorem (presumably Theorem 1.1 or §4)] The proof that the absence of cycles implies the union fails to be self-similar (the “only if” direction of the main theorem) relies on the strict ordering and the bound β < 1/(N+1) to control overlaps, yet the argument does not exhibit an explicit contradiction with the self-similarity equation when the graph is acyclic; a concrete infinite-descent or measure argument is needed to close the gap.

Authors: We appreciate the referee's suggestion to strengthen the 'only if' direction. The current argument in Section 4 uses the strict ordering of the t_j and the contraction bound β < 1/(N+1) to control overlaps and derive a contradiction with the self-similarity equation. To make this fully explicit, we will revise the proof to include a concrete infinite-descent argument: assuming the graph is acyclic, repeated application of the IFS maps to any point in the union produces an infinite sequence of distinct translation indices, which is impossible for a finite union. This explicit descent will be added to the proof of the main theorem. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper constructs a directed graph on translation indices whose edges encode admissible IFS compositions that could map the union onto itself. The self-similarity of the union is then equivalent to the existence of a cycle in this graph. Because the graph is defined explicitly from the given equal-ratio IFS and the ordered translations, the cycle condition is not presupposed by the target property; it is derived from the contractivity and overlap control. The citation to the 2024 Math. Z. paper (with author overlap) is used only to note the special case being generalized; the present argument supplies an independent graph-theoretic proof for the general ordered case and does not reduce any load-bearing step to an unverified self-citation. No parameters are fitted, no ansatz is imported, and no known result is merely renamed. The derivation is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper rests on standard theorems about attractors of contractive IFS and on the algebraic properties of unions under translation; no free parameters are fitted and no new entities are postulated.

axioms (2)

standard math Every finite contractive IFS on a complete metric space possesses a unique nonempty compact attractor.
Invoked to guarantee that Γ exists and is the unique self-similar set for the given maps.
domain assumption Self-similarity of a union can be decided by checking algebraic relations among the translation vectors and the IFS maps.
Underlies the reduction of the problem to cycle detection in a directed graph.

pith-pipeline@v0.9.0 · 5521 in / 1575 out tokens · 46299 ms · 2026-05-09T16:42:13.568994+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

19 extracted references · 1 canonical work pages

[1]

Baker and D

S. Baker and D. Kong. Unique expansions and intersections of Cantor sets.Nonlinearity, 30(4):1497– 1512, 2017

2017
[2]

Bandt and A

C. Bandt and A. Kravchenko. Differentiability of fractal curves.Nonlinearity, 24(10):2717–2728, 2011

2011
[3]

Deng, X.-G

G.-T. Deng, X.-G. He, and Z.-X. Wen. Self-similar structure on intersections of triadic Cantor sets. J. Math. Anal. Appl., 337(1):617–631, 2008

2008
[4]

Deng and C.-T

G.-T. Deng and C.-T. Liu. Self-similarity of unions of Cantor sets.J. Math. (Wuhan), 31(5):847–852, 2011

2011
[5]

Elekes, T

M. Elekes, T. Keleti, and A. M´ ath´ e. Self-similar and self-affine sets: measure of the intersection of two copies.Ergodic Theory Dynam. Systems, 30(2):399–440, 2010

2010
[6]

D.-J. Feng, S. Hua, and Y. Ji. When is the union of two unit intervals a self-similar set satisfying the open set condition?Monatsh. Math., 152(2):125–134, 2007

2007
[7]

D.-J. Feng, H. Rao, and Y. Wang. Self-similar subsets of the Cantor set.Adv. Math., 281:857–885, 2015

2015
[8]

J. E. Hutchinson. Fractals and self-similarity.Indiana Univ. Math. J., 30(5):713–747, 1981

1981
[9]

D. Kong, W. Li, and F. M. Dekking. Intersections of homogeneous Cantor sets and beta-expansions. Nonlinearity, 23(11):2815–2834, 2010

2010
[10]

D. Kong, W. Li, Z. Wang, Y. Yao, and Y. Zhang. On the union of homogeneous symmetric Cantor set with its translations.Math. Z., 307(2):Paper No. 35, 20, 2024

2024
[11]

W. Li, Y. Yao, and Y. Zhang. Self-similar structure on intersection of homogeneous symmetric Cantor sets.Math. Nachr., 284(2-3):298–316, 2011

2011
[12]

C. G. Moreira, J. Xi, and Y. Zhang. Graphs of continuous but non-affine functions are never self- similar.Proc. Amer. Math. Soc., 153(6):2501–2512, 2025. SELF-SIMILARITY OF UNIONS 23

2025
[13]

Pedersen and J

S. Pedersen and J. D. Phillips. On intersections of Cantor sets: self-similarity.Commun. Math. Anal., 16(1):1–30, 2014

2014
[14]

X. Wang, K. Jiang, and L. Xi. On self-similarity of bounded regions in the plane.Fractals, 33(07):2550053, 2025

2025
[15]

Z. Wang. Self-embeddings of homogeneous self-similar sets generated by three maps. arXiv:2509.18202, 2025

work page arXiv 2025
[16]

Wen and X

Z.-Y. Wen and X. Zhao. A criterion for a finite union of intervals to be a self-similar set satisfying the open set condition.Asian J. Math., 21(1):185–195, 2017

2017
[17]

J. Xu, J. Xi, and K. Jiang. On self-similarity of quadrangle.Fractals, 32(05):2450096, 2024

2024
[18]

Yao and W

Y. Yao and W. Li. Self-similar structure on intersection of Cartesian product of Cantor triadic sets with their translations.Monatsh. Math., 166(3-4):591–600, 2012

2012
[19]

Y. Zou, J. Lu, and W. Li. Self-similar structure on the intersection of middle-(1−2β) Cantor sets withβ∈(1/3,1/2).Nonlinearity, 21(12):2899–2910, 2008. (Zhiqiang Wang)College of Mathematics and Statistics, Center of Mathematics, Chongqing University, Chongqing 401331, People’s Republic of China & Department of Mathemat- ics, University of British Columbia...

2008