---
title: Moran-type Attractors Overview
url: https://www.emergentmind.com/topics/moran-type-attractors
type: topic
---

# Moran-type Attractors Overview

A Moran-type attractor is a compact set generated as the invariant set of a non-stationary (often infinite) sequence of iterated function systems, where the contraction ratios and the branching numbers may vary from stage to stage. This generalizes the classical self-similar (or self-affine) attractors of autonomous IFSs to a much broader context encompassing non-homogeneous and non-stationary geometric constructions. Moran-type attractors and their invariant measures unify many classical and modern examples of fractal sets, such as non-homogeneous Cantor sets and rapidly inhomogeneous self-similar constructions. Their dimension theory, measure-theoretic properties, and spectrality theory have been developed to include generalized contraction maps, separation regimes, and multifractal spectra.

## 1. Definition of Moran-type Attractors

Moran-type attractors are associated with *Moran-type iterated function systems* (MIFS), which are defined as follows. Let $(X, \rho)$ denote a compact metric space. For each $n \geq 1$, let $\Phi_n = \{\phi_{n,1}, \dots, \phi_{n,N_n}\}$, with $2 \leq N_n < \infty$, be a finite family of bi-Lipschitz maps $\phi_{n,j}: X \to X$. Require constants $0 < c_{1,n} \leq c_{2,n} < 1$ such that for all $x, y \in X$,
\[
c_{1,n}\, \rho(x, y) \leq \rho(\phi_{n,j}(x), \phi_{n,j}(y)) \leq c_{2,n} \, \rho(x, y),
\]
and
\[
\lim_{n \to \infty} \prod_{i=1}^n c_{2,i} = 0.
\]
The family $\{\Phi_n\}_{n \geq 1}$ forms a MIFS. The *Moran-type attractor* (or *invariant set*) at level $n$ is the unique compact set $K_n \subset X$ such that
\[
K_n = \bigcup_{j=1}^{N_n} \phi_{n, j}(K_{n+1}),
\]
with
\[
K_n = \left\{ \lim_{k \to \infty} \phi_{n, J_k}(a): J_k \in \Sigma_n^k,\, a \in X \right\}, \quad \text{where } \phi_{n, J} := \phi_{n, j_n} \circ \cdots \circ \phi_{n+k-1, j_{n+k-1}}
\]
and $\Sigma_n^k = \prod_{i=0}^{k-1} \{1, \dots, N_{n+i}\}$ is the symbolic coding at stage $n$. In particular, the attractor $K_1$ is called the (primary) Moran-type attractor of the system [2601.11023].

## 2. Construction of Invariant Measures

Corresponding to each attractor is an associated sequence of *Moran-type measures*. For each $n$, select a probability vector $\mathbf{p}_n = (p_{n,1}, \dots, p_{n,N_n})$ with $p_{n,j} > 0$ and $\sum_j p_{n,j} = 1$. The product measure $\nu_n$ on the infinite code space $\Sigma_n^\mathbb{N}$ induces a Borel probability measure $\mu_n$ on $K_n$ via the canonical coding map $\pi_n$. The measure $\mu_n$ is the unique solution to
\[
\mu_n = \sum_{j=1}^{N_n} p_{n,j}\, (\mu_{n+1} \circ \phi_{n,j}^{-1}),
\]
with $\operatorname{supp} \mu_n = K_n$. The entire measure-theoretic structure is recursively determined by the measures at each stage, providing a measure-theoretic refinement of the attractor's geometric construction [2601.11023].

## 3. Separation Regimes and Regularity Properties

The dimension theory and structure of Moran-type attractors critically depend on separation properties:

- **Moran-type Open-Set Condition (MOSC):** There exist open sets $V_n \subset X$ such that images $\{\phi_{n,j}(V_{n+1})\}_j$ are contained in $V_n$ and have pairwise disjoint interiors, with $\inf_n \operatorname{Leb}^d(V_n) > 0$.
- **Weak Separation Condition (MWSC):** For suitable sets $U_n \subset X$, the number of overlapping images of equivalently contracted maps remains uniformly bounded at each stage.
- **Strong Separation Condition (MSSC):** At every level, the cylinder sets $\phi_{1, J}(K_{n+1})$ are pairwise disjoint.

The validity of some separation regime is often assumed in the proofs of exact dimension formulae and the identification of multifractal properties [2601.11023].

## 4. Dimension Theory of Moran-type Attractors

For Moran-type attractors generated by similarities $\phi_{n, j}$ of contraction ratios $r_{n, j}$, several dimension formulae are established:

- **Box-Counting and Packing Dimensions:** When the MWSC and a uniform lower contraction bound hold,
\[
\underline{\dim}_B K_1 = \liminf_{b \to 0} \frac{\ln \# \mathcal{A}_b}{-\ln b}, \qquad \overline{\dim}_B K_1 = \dim_P K_1 = \limsup_{b \to 0} \frac{\ln \# \mathcal{A}_b}{-\ln b}
\]
where $\mathcal{A}_b$ is the set of all finite compositions whose contraction ratio is at most $b$.

- **Hausdorff Dimension via Pressure:** Define $P_n(s) = \sum_{j=1}^{N_n} r_{n, j}^s$, $S_n(s) = \prod_{i=1}^n P_i(s)$. Under MOSC, if
\[
\liminf_{n \to \infty} S_n(s) \in (0, \infty), \quad \text{then} \quad \dim_H K_1 = s.
\]
Otherwise, $\dim_H K_1$ can be estimated via the asymptotics of $S_n(s)$ [2601.11023].

- **Specific Formula for Staggered IFS:** For the so-called “staggered contraction ratio” systems with $n_j = 2p_j$ and contraction bases $b_k > 2p_k > 2$,
\[
\dim_H(A) = \liminf_{k \to \infty} \frac{\sum_{j=1}^k \ln(2p_j)}{\sum_{j=1}^k \ln b_j}.
\]
This formula generalizes the classical Moran equation and applies even in highly inhomogeneous settings [2412.13427].

## 5. Spectrality and Structural Decomposition

A central question concerns when the invariant measure of a Moran-type attractor is *spectral*, i.e., admits an orthonormal basis of complex exponentials in $L^2(\mu)$.

- **Spectrality Criterion**: For the IFS
\[
\phi_{k,d}(x) = (-1)^d b_k^{-1} (x + d), \quad d \in D_{2p_k} = \{0, \dots, 2p_k-1\}
\]
if all $n_k = 2p_k$ are even, $\{b_k\}$ is bounded, and $2p_k \mid b_k$ for all $k \geq 2$, then the corresponding measure $\mu$ is spectral [2412.13427]. The spectrum can be described as
\[
\Lambda = D_{2p_1} + b_1 D_{2p_2} + b_1 b_2 D_{2p_3} + \cdots.
\]
This generalizes the spectrality results for homogeneous Cantor measures.

- **Fourier and Convolutional Structure**: The Fourier transform of $\mu$ admits the infinite product representation
\[
\widehat\mu(t) = \exp(-2\pi i\,b_1^{-1}t)\,\prod_{k=1}^\infty f_k \Bigl( \frac{t}{b_1\cdots b_k} \Bigr ),
\]
where $f_k(u) = \frac{1}{2p_k} \sum_{d=0}^{2p_k-1} (-1)^d e^{2\pi i d u}$. This analytic structure underpins the proofs of spectrality [2412.13427].

- **Hadamard Triple and Random Convolution Arguments:** The proof of spectrality leverages specific “Hadamard triple” constructions at each level, facilitating a stepwise buildup of the exponential basis via convolutional factorization [2412.13427].

## 6. Assouad-type Dimensions and Multifractal Properties

Recent results have extended the dimension theory of Moran-type attractors to the Assouad dimension, lower dimension, and their corresponding spectra:

- **Assouad Dimension** for a homogeneous Moran set $E$ generated by parameters $\{n_k\}$ and $\{c_k\}$ with $\sup n_k < \infty$,
\[
\dim_A E = \limsup_{k \to \infty} \sup_{\ell \geq 1} \frac{\log(n_{k+1} \cdots n_{k+\ell})}{-\log(c_{k+1} \cdots c_{k+\ell})}
\]
[2407.14837]. The lower dimension is bounded above by the corresponding $\liminf$.

- **Spectra for Cantor-like Sets:** For a more general class including inhomogeneous gaps, the Assouad and lower spectra admit explicit sup/inf characterizations in terms of scaling exponents over variable scales [2407.14837].

- **Connection to Doubling Measures and Separation:** The validity of these dimension formulae requires finite branching, uniform contraction, and sufficient separation (typically, disjoint interiors at each level) [2407.14837].

## 7. Illustrative Examples and Special Cases

- **Non-homogeneous Cantor-type Constructions:** Moran-type attractors include classically self-similar sets as special cases but allow for rapidly changing contractions and branching. For example, systems with $\phi_{n,0}(x) = \tfrac{1}{2}x$, $\phi_{n,1}(x) = \frac{1}{2}(x + \frac{1}{n}\rho^n)$ generate attractors with identical box and Hausdorff dimensions yet can have vanishing $\mathcal{H}^s$-measure at the dimension $s$ [2601.11023].

- **Staggered IFS and Spectral Measures:** For the class $\phi_{k,d}(x) = (-1)^d b_k^{-1}(x + d)$ with even $n_k$ and divisibility $2p_k \mid b_k$, the resulting Moran-type measure exhibits spectrality, contrasting with general inhomogeneous self-similar measures, which typically fail to be spectral unless strong arithmetic conditions are imposed [2412.13427].

- **Assouad Spectrum Computations:** Explicit formulae for Assouad and lower spectra are established for homogeneous and Cantor-like sets under finite-branching and separation, extending and refining the classical results for stationary self-similar sets [2407.14837].

A plausible implication is that Moran-type attractors provide a flexible modeling framework for both geometric and spectral phenomena beyond the homogeneous, self-similar setting, enabling systematic investigation of the interplay between geometric inhomogeneity, measure regularity, and harmonic analysis.

Source: https://www.emergentmind.com/topics/moran-type-attractors