OctoInformBeta

OpenAI preprint, September 2026 · §6. Overlap conventions and dimension of the attractor

Corollary 6.5: Self-similar probability measures with a common positive contraction ratio and arbitrary real third translation

Dimension for three similarities without exact overlaps.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Consider self-similar probability measures for three similarities with a common positive contraction ratio and translations 00, 11 and t∈ℝt\in\mathbb R.

A weighted affine iterated function system SS on a finite alphabet II is a family of real similarities with signed slopes rir_i, real translations tit_i, and weights pip_i satisfying (1): φi(x)=rix+ti,0<|ri|<1,pi>0,∑i∈Ipi=1.\varphi_i(x)=r_i x+t_i, \qquad 0<|r_i|<1, \qquad p_i>0, \qquad \sum_{i\in I}p_i=1.(1)✦ The three-map system Sλ,t,pS_{\lambda,t,p} is the weighted affine system on I={0,1,2}I=\{0,1,2\} given by (2), where 0<λ<10<\lambda<1 and p=(p0,p1,p2)p=(p_0,p_1,p_2) is a vector of strictly positive weights summing to one: φ0(x)=λx,φ1(x)=λx+1,φ2(x)=λx+t.\varphi_0(x)=\lambda x, \qquad \varphi_1(x)=\lambda x+1, \qquad \varphi_2(x)=\lambda x+t.(2)

For a weighted affine system SS, the affine composition of a finite address w=i1⋯inw=i_1\cdots i_n is the map φw\varphi_w in (3), where ⌀\varnothing denotes the empty address; its complete affine map is specified by its signed slope rwr_w and translation twt_w: φw=φi1∘⋯∘φin,φw(x)=rwx+tw,φ⌀(x)=x.\varphi_w=\varphi_{i_1}\circ\cdots\circ\varphi_{i_n}, \qquad \varphi_w(x)=r_wx+t_w, \qquad \varphi_{\varnothing}(x)=x.(3) The system SS has no exact overlaps when equality of the complete affine maps of two finite addresses forces the addresses to coincide, including the empty address.✦

A Borel measure μ\mu on ℝ\mathbb R is self-similar for SS when it satisfies (4), where (φi)*μ(\varphi_i)_*\mu denotes pushforward: μ=∑i∈Ipi(φi)*μ.\mu=\sum_{i\in I}p_i(\varphi_i)_*\mu.(4)✦

The lower Hausdorff dimension dim⁡Hμ\dim_H\mu of a Borel measure μ\mu on ℝ\mathbb R is the extended nonnegative quantity in (5), where dim⁡HE\dim_H E denotes the Hausdorff dimension of a set EE and the infimum runs over Borel sets of positive μ\mu-mass: dim⁡Hμ=inf⁡{dim⁡HE:E⊆ℝ Borel,μ(E)>0}.\dim_H\mu =\inf\bigl\{\dim_H E:E\subseteq\mathbb R\text{ Borel},\ \mu(E)>0\bigr\}.(5)✦

The symbol entropy H(p)H(p) of a strictly positive probability vector p=(p0,p1,p2)p=(p_0,p_1,p_2) is the real number in (6), computed with natural logarithms: H(p)=−∑i=02piln⁡pi.H(p)=-\sum_{i=0}^{2}p_i\ln p_i.(6)

The Claim

Theorem 1. Let 0<λ<10<\lambda<1 and t∈ℝt\in\mathbb R, and let p0,p1,p2p_0,p_1,p_2 be strictly positive weights summing to one. Suppose that Sλ,t,pS_{\lambda,t,p} has no exact overlaps, and let μ\mu be a self-similar probability measure for this system. Then dim⁡Hμ=min⁡{1,H(p)ln⁡(1/λ)}.\dim_H\mu =\min\left\{1,\frac{H(p)}{\ln(1/\lambda)}\right\}.(7)✦

In the paper, the result reads: for 0<λ<10<\lambda<1 and t∈ℝt\in\mathbb R, if the indexed maps in (2) have no exact overlaps, then every strictly positive probability vector p=(p0,p1,p2)p=(p_0,p_1,p_2) gives dim⁡Hμ=min⁡{1,(−∑i=02pilog⁡2pi)/log⁡2(1/λ)}\dim_H\mu=\min\{1,(-\sum_{i=0}^{2}p_i\log_2 p_i)/\log_2(1/\lambda)\}.

What Is Checked

The claim (7) is stated over a self-similar probability measure for the three-map weighted affine system Sλ,t,pS_{\lambda,t,p} under the no-exact-overlap condition on all finite addresses.