← All problems

Exact overlaps conjecture

Let {φi(x)=rix+ti}i=1m\{\varphi_i(x)=r_i x+t_i\}_{i=1}^m be a finite self-similar iterated-function system on R\mathbb{R}, where 0<∣ri∣<10<|r_i|<1, and let (pi)i=1m(p_i)_{i=1}^m be positive weights with ∑ipi=1\sum_i p_i=1. Let μ\mu be the associated self-similar measure satisfying μ=∑i=1mpi(φi)∗μ\mu=\sum_{i=1}^m p_i(\varphi_i)_\ast\mu. If χ=−∑i=1mpilog⁡∣ri∣\chi=-\sum_{i=1}^m p_i\log|r_i| is the Lyapunov exponent and hRWh_{\mathrm{RW}} is the asymptotic Shannon entropy of the random walk generated by the affine maps φi\varphi_i, then the conjectured dimension formula is dim⁡Hμ=min⁡{1,hRW/χ}\dim_H\mu=\min\{1,h_{\mathrm{RW}}/\chi\}. In the equivalent exact-overlaps formulation, any dimension loss beyond the value predicted without overlaps must be explained by exact overlaps between distinct finite compositions of the maps, meaning that φi1∘⋯∘φin=φj1∘⋯∘φjk\varphi_{i_1}\circ\cdots\circ\varphi_{i_n}=\varphi_{j_1}\circ\cdots\circ\varphi_{j_k} for two distinct words.

References

Source: arXiv

Progress summary

Refreshed
Open · progress reported

A new preprint claims to settle the one-dimensional case, but the higher-dimensional conjecture remains open and the claim has not been independently verified.

The conjecture predicts that dimension loss for self-similar measures is explained by exact overlaps between distinct iterated maps. Earlier work established strong one-dimensional partial results but not the unrestricted conjecture.

Known results

  • Hochman (2014): dimension drop implies superexponentially close cylinders; for algebraic parameters, either exact overlaps occur or the dimension is the generic value min⁡{1,s}\min\{1,s\}.
  • For systems on the real line with algebraic contraction ratios, the conjectured dimension formula is proved, with arbitrary translations.

October 2026 one-dimensional claim

Samuel Kittle and Constantin Kogler report a variance-based entropy inequality that establishes the conjectured formula for one-dimensional self-similar measures. This is a claimed advance in a specialist preprint and has not been independently verified; higher-dimensional cases are outside its scope.

Current status (as of October 2026): The one-dimensional case is claimed solved by a new preprint but remains unverified; higher-dimensional cases remain open.

Sources

Solutions 1

Claimed by OpenAI. For every finite indexed family of contracting real similarities with nonzero ratios and positive probability weights, the manuscript claims dim_H(mu)=min(1,h_RW/chi), where h_RW is the entropy rate of the full random affine maps and chi is the average negative log absolute contraction ratio. Exact overlaps, repeated generators, unequal ratios and reversed orientations are allowed. This addresses the page’s measure-dimension formula; absolute continuity is not claimed.See full solutionHide full solution