---
title: Contractive Iterated-Map Codes
url: https://www.emergentmind.com/topics/contractive-iterated-map-codes
type: topic
---

# Contractive Iterated-Map Codes

Searching arXiv for recent and foundational papers relevant to contractive iterated-map codes and related symbolic/iterative contraction frameworks.
“Contractive iterated-map codes” denotes, in the language of the cited works, structured descriptions attached to iterated systems whose contractive behavior is encoded symbolically, eventwise, or operator-theoretically. For piecewise contractions of the interval, the code is the **natural coding** of an orbit relative to a partition; for logically contractive mappings, the relevant data is an increasing sequence of iterate counts at which contraction occurs; for contractive two-step iterated liftings, the corresponding invariant is the **characteristic function** together with its factorization. Across these settings, the common theme is that contraction need not appear in a uniform one-step form in order to impose rigid asymptotic or structural consequences on the associated iterated-map description [1803.01226][2508.07059][2212.07808].

## 1. Contractive codings in interval dynamics

A **piecewise contraction of \(n\) intervals** (an \(n\)-PC) is a map
\[
f:[0,1)\to[0,1)
\]
for which there exist a constant \(0<\lambda<1\) and a partition of \(I=[0,1)\) into non-degenerate intervals
\[
I_1,\ldots,I_n
\]
such that each restriction \(f|_{I_i}\) is \(\lambda\)-Lipschitz; equivalently,
\[
|f(x)-f(y)|\le \lambda |x-y| \quad\text{for every }x,y\in I_i,\ i=1,\dots,n.
\]
An important subclass is formed by **piecewise \(\lambda\)-affine contractions**, where on each interval \(I_i\),
\[
f(x)=\sigma_i\lambda x+b_i,\qquad \sigma_i\in\{-1,1\},\ b_i\in\mathbb R.
\]

Given such a partition, the **natural coding** of a point \(x\in I\) is the infinite word
\[
\theta=\theta_0\theta_1\theta_2\cdots
\]
over the alphabet
\[
\mathcal A=\{1,\ldots,n\}
\]
defined by
\[
\theta_k=i \iff f^k(x)\in I_i.
\]
The coding records, at each time step, which continuity piece contains the current iterate. The paper also uses the usual notions of periodic and ultimately periodic words: \(\theta\) is periodic if \(\theta=vvv\cdots\), and ultimately periodic if \(\theta=uvvv\cdots\) for some finite words \(u,v\). The language \(L(\theta)\) is the collection of all finite subwords appearing in \(\theta\) [1803.01226].

The notion of **isomorphic** words is central. Two infinite words \(\theta\) and \(\omega\) are isomorphic if they use alphabets of the same cardinality and one is obtained from the other by a bijection of symbols. Accordingly, the classification is not a statement of literal word equality, but of equality up to relabeling. This is a precise way to compare codings generated by different partitions or different interval systems.

## 2. Classification by interval exchange transformations

The central structural statement is a dictionary theorem between codings of injective piecewise contractions and codings of interval exchange transformations. If \(f\) is an injective \(n\)-PC and \(\theta\) is a natural coding of \(f\), then there exist
\[
2\le m\le n
\]
and an \(m\)-interval exchange transformation
\[
T:I\to I
\]
with no attractive periodic orbits such that, after possibly moving forward along the orbit by some \(k\ge 0\), the natural coding of \(f^k(x)\) is either periodic or isomorphic to a non ultimately periodic natural coding of \(T\). In the paper’s exact phrasing: if \(\theta\) is a natural coding of an injective \(n\)-PC, then some infinite subword of \(\theta\) is either periodic or isomorphic to a natural coding of a topologically transitive \(m\)-IET, where \(m\le n\) [1803.01226].

A key corollary is the clean symbolic-dynamical dichotomy: some infinite subword of a natural coding of an injective \(n\)-PC is either periodic or isomorphic to a non ultimately periodic natural coding of a topologically transitive \(m\)-IET, \(2\le m\le n\). In effect, after a finite transient, admissible codes are either eventually periodic or interval-exchange in origin.

The paper emphasizes that the admissible symbolic sequences are therefore highly constrained. In the broader language of contractive iterated-map codes, this means that injective contractive interval maps do not generate arbitrary low-complexity sequences. Rather, they fall into a dichotomy: either the orbit eventually enters a periodic regime, producing an ultimately periodic coding, or, after discarding a finite prefix and up to symbol relabeling, the coding is a natural coding of a topologically transitive interval exchange transformation. This gives a classification theorem for what can appear from injective contractive interval systems.

## 3. Realization, semiconjugacy, and invariant structure

The converse direction is equally strong. Every natural coding of a topologically transitive \(n\)-IET is also a natural coding of some injective \(n\)-PC. More precisely, given any topologically transitive \(n\)-IET \(T\), the construction produces an injective piecewise \((-2)\)-affine contraction
\[
f:I\to I
\]
together with a continuous, surjective, non-decreasing map \(h:I\to I\) such that
\[
h\circ f = T\circ h.
\]
In the statement of the paper, this is written as
\[
\theta_f(x)\sim \theta_T(h(x)),
\]
meaning that the natural coding of \(f\) at \(x\) is isomorphic to the natural coding of \(T\) at \(h(x)\) [1803.01226].

The construction is explicit. It uses a dense orbit \(\{p_k\}\) of the IET, defines a family of intervals \(G_k\) whose lengths are dyadic,
\[
|G_k|=2^{-k},
\]
and then defines a monotone factor map \(h\) collapsing the geometry of the IET orbit into the contraction dynamics. The resulting \(f\) is piecewise affine with slope \(\pm \tfrac12\) on each piece, and the semiconjugacy relation \(h\circ f=T\circ h\) is established by matching orbit symbols under \(h\).

The same work also isolates invariant-measure structure behind this correspondence. If \(f\) is injective and a natural coding has infinite orbit complexity, then the \(\omega\)-limit set of the orbit supports a non-atomic invariant probability measure. The argument decomposes the dynamics into attractors \(\Lambda_1,\dots,\Lambda_r\), shows that infinite \(\omega\)-limit sets are Cantor-like rather than finite, and then constructs a monotone map \(h\) from the invariant measure. This is the mechanism through which the interval exchange factor is produced. A plausible implication is that the coding theorem is not merely combinatorial; it reflects a measure-theoretic and topological organization of the contractive dynamics.

## 4. Complexity bounds and Sturmian realizations

For a natural coding \(\theta\) of an injective \(n\)-PC, the factor complexity
\[
p_\theta(k)=\#L_k(\theta)
\]
is eventually affine:
\[
p_\theta(k)=\alpha k+\beta \quad\text{for all large }k,
\]
with
\[
\alpha\in\{0,1,\dots,n-1\},\qquad \beta\ge 1.
\]
In the transitive IET case, the slope is exactly the number of exchanged intervals minus one. More precisely, if \(\theta\) is a natural coding of a topologically transitive \(m\)-IET, then
\[
p_\theta(k)=\alpha k+\beta
\]
eventually, and if the IET is standard and satisfies Keane’s i.d.o.c., then
\[
p_\theta(k)=(m-1)k+1 \quad\text{for all }k\ge 1
\]
[1803.01226].

These formulas give a sharp asymptotic description of subword growth. For periodic codings, the slope is \(0\); for the interval-exchange branch, the slope is positive and bounded above by \(n-1\). The resulting complexity theorem is one of the main precise consequences of the classification theorem.

The paper also gives a concrete family of examples linking piecewise contractions to classical low-complexity words. For every irrational \(0<\alpha<1\), there exists a transcendental \(\delta\) such that the \(2\)-PC
\[
f(x)=\lambda x+\delta\pmod 1
\]
with \(\lambda=\tfrac12\) in the construction is semiconjugate to the irrational rotation
\[
T(x)=x+\alpha\pmod 1,
\]
and every natural coding of \(f\) is Sturmian. In the special case
\[
\alpha=2-\varphi,\qquad \varphi=\frac{1+\sqrt5}{2},
\]
the parameter \(\delta\) is expressed using the rabbit constant, and its transcendence follows from the known transcendence of that constant. This places contractive symbolic dynamics in direct contact with the standard Sturmian framework and shows that interval-exchange-type symbolic languages are realizable inside genuinely contractive systems.

## 5. Event-indexed contraction and fixed-point behavior

A different formalization of contractive iterated-map behavior appears in the notion of a **logically contractive mapping**. Let \((X,d)\) be a complete metric space and \(T:X\to X\). The map is **nonexpansive** if
\[
\Lip(T)\le 1.
\]
It is **logically contractive** if \(T\) is nonexpansive and there exist some \(\lambda\in(0,1)\) and a strictly increasing sequence of natural numbers
\[
n_1<n_2<n_3<\cdots
\]
such that for all \(k\ge 1\),
\[
\Lip\!\big(T^{\,n_k}\big)\le \lambda^k.
\]
Thus contraction need not occur at each step; it is enough that iterates contract more and more strongly along a designated subsequence of event times [2508.07059].

The paper notes an equivalent simplification: logical contractiveness is equivalent to “nonexpansive and at least one iterate \(T^N\) is a strict contraction.” If \(\Lip(T^N)=\mu<1\), then taking \(n_k=kN\) yields
\[
\Lip(T^{n_k})=\Lip\big((T^N)^k\big)\le \mu^k.
\]
From this, the fixed-point theorem follows: if \(T\) is logically contractive, then \(T\) has a unique fixed point \(z\in X\), and for every \(x\in X\), \(T^n x\to z\) as \(n\to\infty\). Moreover, for every \(x\in X\) and every \(k\ge 1\),
\[
d\!\big(T^{\,n_k}x,\;z\big)\le \lambda^k\, d(x,z).
\]

The rate is naturally **event-indexed** rather than iteration-indexed. To convert it into an explicit bound in terms of the raw iteration counter \(n\), one needs information about event density. If
\[
n_{k+1}-n_k\le M \quad \text{for all }k,
\]
then for any \(n\ge n_1\),
\[
d\!\big(T^n x,\;z\big)\le \lambda^{\,1+\left\lfloor \frac{n-n_1}{M}\right\rfloor} d(x,z).
\]
The canonical bounded-gap schedule is obtained by repeating the first strict event: if \(n_1\) is the first strict contraction time, then
\[
\tilde n_m:=m n_1
\]
satisfies
\[
\Lip(T^{m n_1})=\Lip((T^{n_1})^m)\le \big(\Lip(T^{n_1})\big)^m\le \lambda^m.
\]
This yields
\[
d\!\big(T^n x,\;z\big)\le \big(\Lip(T^{n_1})\big)^{\left\lfloor n/n_1\right\rfloor} d(x,z) \le \lambda^{\left\lfloor n/n_1\right\rfloor} d(x,z).
\]

The worked example is the piecewise map \(T:\mathbb{R}\to\mathbb{R}\)
\[
T(x)= \begin{cases}
0, & |x|\le 1,\\
x-\sgn(x), & 1<|x|<2,\\
\sgn(x), & |x|\ge 2.
\end{cases}
\]
It is nonexpansive, but not a strict contraction. In fact,
\[
T^2\equiv 0.
\]
Hence \(T\) is logically contractive with \(n_1=2\), and the unique fixed point is \(0\). This example shows that iterated contraction can capture finite-time stabilization even when one-step contraction fails.

The same paper clarifies that logical contractiveness is not identical to familiar generalized contraction notions. It shows that logical contractiveness and Meir–Keeler are **incomparable**, and it exhibits an asymptotically nonexpansive map that is not logically contractive, namely the identity map. The point is that the framework emphasizes intermittent genuine contraction events rather than merely asymptotic control [2508.07059].

## 6. Variable-factor events and operator-theoretic factorization

The event-indexed framework extends to **variable-factor logically contractive** mappings. A map \(T\) is VLC if it is nonexpansive and there exist increasing event times \(n_1<n_2<\cdots\) and factors \(\lambda_k\in(0,1]\) such that for all \(k\),
\[
d\!\big(T^{\,n_k}x,\;T^{\,n_k}y\big)\le \Big(\prod_{i=1}^{k}\lambda_i\Big)\, d(x,y).
\]
Writing
\[
\Lambda_k:=\prod_{i=1}^k\lambda_i,
\]
this becomes
\[
d\!\big(T^{\,n_k}x,\;T^{\,n_k}y\big)\le \Lambda_k\, d(x,y).
\]
As soon as some \(\Lambda_k<1\) occurs, the corresponding iterate \(T^{n_k}\) is a strict contraction, so the Banach argument gives a unique fixed point \(z\), and
\[
d\!\big(T^{\,n_k}x,\;z\big)\le \Lambda_k\, d(x,z).
\]
The crucial criterion is
\[
\Lambda_k\to 0 \quad\Longleftrightarrow\quad \prod_{k=1}^\infty \lambda_k=0 \quad\Longleftrightarrow\quad \sum_{k=1}^\infty -\ln \lambda_k=\infty.
\]
When this holds,
\[
T^n x\to z \qquad \text{for every }x\in X.
\]
If event gaps are bounded by \(M\), then
\[
d\!\big(T^n x,\;z\big)\le \Lambda_{\,1+\left\lfloor \frac{n-n_1}{M}\right\rfloor} d(x,z).
\]
The paper remarks that \(\Lambda_k\to 0\) is sufficient, but not necessary for a specific map to converge; if, for example,
\[
\lambda_k=1-\frac{1}{k^2},
\]
then \(\sum -\ln\lambda_k<\infty\), so \(\Lambda_k\downarrow \Lambda_\ast>0\), and the abstract theorem gives no convergence guarantee [2508.07059].

An operator-theoretic analogue of contractive iterated-map coding appears in the study of **contractive two-step iterated liftings**. Let
\[
T=(T_1,\dots,T_d)
\]
be a row contraction on a Hilbert space \(H\), meaning
\[
\sum_{i=1}^d T_i T_i^* \le I.
\]
If \(C=(C_1,\dots,C_d)\) is a row contraction on \(H_C\), then a lifting of \(C\) by another row contraction \(A=(A_1,\dots,A_d)\) on \(H_A\) is a row contraction
\[
E=(E_1,\dots,E_d)
\]
on \(H_E = H_C \oplus H_A\) with block form
\[
E_i = \begin{pmatrix} C_i & 0 \\ B_i & A_i \end{pmatrix}, \qquad i=1,\dots,d.
\]
Such a lifting is contractive iff \(C\) and \(A\) are row contractions and there exists a contraction
\[
\gamma:\mathcal D_{*,A}\to \mathcal D_C
\]
such that
\[
B = D_C\,\gamma\,D_{*,A}.
\]
A lifting is called reduced if \(A\) is completely non-coisometric and \(\gamma\) is resolving; the paper notes that reduced liftings coincide with minimal contractive liftings [2212.07808].

For a minimal contractive lifting \(E\) of \(C\), the **characteristic function**
\[
M_{C,E} := P_{\Gamma\otimes\mathcal D_C}\, W \big|_{\Gamma\otimes\mathcal D_E}
\]
is a multi-analytic operator
\[
\Gamma\otimes\mathcal D_E \to \Gamma\otimes\mathcal D_C
\]
satisfying
\[
(L_i\otimes I)M_{C,E} = M_{C,E}(L_i\otimes I),
\]
and it is determined by its symbol
\[
\theta_{C,E} := M_{C,E}\big|_{e_0\otimes \mathcal D_E}.
\]
The operator has a formal Fourier expansion
\[
M_{C,E} \sim \sum_{\alpha\in\Lambda} R_\alpha \,\theta_{C,E}(\alpha).
\]

In the two-step setup,
\[
C \xleftarrow{\;\text{lift by }A\;} E \xleftarrow{\;\text{lift by }A^1\;} E^1,
\]
the characteristic function of the full iterated lifting factors through the characteristic functions of the constituent liftings together with the Julia–Halmos matrix
\[
J_X= \begin{pmatrix} D_{*,X} & X \\ -X^* & D_X \end{pmatrix},
\]
which is unitary. The paper proves unitary identifications of defect spaces and then derives a factorization of \(M_{C,E^1}\) in terms of \(M_A\), \(M_{A^1}\), and the associated defect-space unitaries. It stresses that the characteristic function of the iterated lifting is not just a naive product of two functions; it is a **unitarily corrected product**. For the minimal part \(\widetilde E\), the restriction formula is
\[
M_{C,\widetilde E} = M_{C,E}\,M_{E,E^1}\,(I_\Gamma\otimes \sigma^{-1}_{\widetilde E}),
\]
so the characteristic function of the minimal part is obtained from the product of constituent characteristic functions by restriction to the appropriate defect space [2212.07808].

Taken together, these works suggest that “contractive iterated-map codes” is best understood as a family of rigorous encodings for systems in which contraction emerges through partitioned orbit itineraries, event-indexed iterates, or lifting-theoretic transfer data. In each setting, the iterated structure is the source of the coding, and contraction is the mechanism that forces strong classification, convergence, or factorization properties.

Source: https://www.emergentmind.com/topics/contractive-iterated-map-codes