Pretentious Measure-Preserving Multiplicative Actions

• Pretentious measure-preserving multiplicative actions are probability-preserving dynamical systems indexed by multiplicative semigroups, where prime-step operators serve as local coordinates.
• The spectral framework uses pretentious distances from analytic number theory to decompose functions into components modeled by Dirichlet characters and Archimedean eigenfunctions.
• Recent advances demonstrate mean ergodic theorems, joint ergodicity, and recurrence along rational polynomial iterates, deepening our understanding of multiplicative dynamics.

Pretentious measure-preserving multiplicative actions are probability-preserving dynamical systems indexed by the multiplicative semigroup of positive integers, or its extension to $\mathbb Q_+$, in which the relevant structure is governed not by ordinary invariant or eigenfunctions alone, but by a multiplicative analogue of pretentiousness from analytic number theory. In this setting, prime-step operators play the role of local coordinates, and the central question is whether an $L^2$-function behaves, on most primes in a logarithmic sense, like an invariant function, a Dirichlet-character model, or a more general completely multiplicative eigenmodel. Recent work develops this viewpoint into an ergodic theory of multiplicative actions, proving mean ergodic theorems, decomposition results, joint ergodicity criteria, recurrence theorems for rational polynomial iterates, and structural descriptions of associated Furstenberg systems (Charamaras, 2024, Frantzikinakis, 19 Mar 2025, Frantzikinakis et al., 2023, Frantzikinakis et al., 13 Aug 2025).

1. Formal definition and ambient framework

A multiplicative system is a probability measure-preserving system $(X,\mu,S)$, where $S=(S_n)_{n\in\mathbb N}$ is a family of commuting invertible measure-preserving transformations such that

$$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$

If this identity only holds for coprime $n,m$, the system is called weakly multiplicative (Charamaras, 2024). A multiplicative measure-preserving action is similarly described as a quadruple

$(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$

with

$$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$

and it is extended to $\mathbb Q_+$ by

$T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$

(Frantzikinakis, 19 Mar 2025, Frantzikinakis et al., 13 Aug 2025).

The associated Koopman operators act on $L^2$0 by

$L^2$1

or equivalently $L^2$2 in the second notation (Charamaras, 2024). A system is finitely generated if the set of prime-step transformations is finite:

$L^2$3

depending on notation (Charamaras, 2024, Frantzikinakis, 19 Mar 2025, Frantzikinakis et al., 13 Aug 2025). This hypothesis is central in the strongest convergence, decomposition, and inverse-theorem statements.

The spectral formulation is built from completely multiplicative $L^2$4-valued functions. For $L^2$5 and a multiplicative action $L^2$6, the map

$L^2$7

is positive definite, so by Bochner–Herglotz there exists a finite Borel measure $L^2$8 on the compact group $L^2$9 of completely multiplicative $(X,\mu,S)$0-valued functions such that

$(X,\mu,S)$1

(Frantzikinakis et al., 13 Aug 2025). This identity is the basic bridge between multiplicative dynamics and multiplicative number theory.

2. Pretentiousness as multiplicative spectral structure

The organizing notion comes from the pretentious distance. For multiplicative functions $(X,\mu,S)$2 or $(X,\mu,S)$3, the basic distance is

$(X,\mu,S)$4

and one writes $(X,\mu,S)$5 if $(X,\mu,S)$6 (Frantzikinakis et al., 2023). In the generalized interval form used for recurrence problems,

$(X,\mu,S)$7

with $(X,\mu,S)$8 (Frantzikinakis et al., 13 Aug 2025).

A completely multiplicative function is pretentious if there exist a Dirichlet character $(X,\mu,S)$9 and a real $S=(S_n)_{n\in\mathbb N}$0 such that

$S=(S_n)_{n\in\mathbb N}$1

The pretentious class is denoted $S=(S_n)_{n\in\mathbb N}$2 in the recurrence paper (Frantzikinakis et al., 13 Aug 2025). In the Furstenberg-system analysis, the special pretentious models are Dirichlet characters and Archimedean characters $S=(S_n)_{n\in\mathbb N}$3, $S=(S_n)_{n\in\mathbb N}$4 (Frantzikinakis et al., 2023).

For unitary multiplicative actions $S=(S_n)_{n\in\mathbb N}$5 on $S=(S_n)_{n\in\mathbb N}$6 and nonzero $S=(S_n)_{n\in\mathbb N}$7, the dynamical pretentious distance is

$S=(S_n)_{n\in\mathbb N}$8

If $S=(S_n)_{n\in\mathbb N}$9 comes from a completely multiplicative function $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$0 via $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$1, then $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$2 denotes $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$3 (Charamaras, 2024). This measures whether $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$4 behaves like $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$5 for most primes.

A function $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$6 is called pretending to be invariant if

$$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$7

and the corresponding collection is $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$8 (Charamaras, 2024). More generally, $$S_{nm}=S_n\circ S_m \qquad \forall\, n,m\in\mathbb N.$$9 is a pretentious eigenfunction with pretentious eigenvalue $n,m$0 if

$n,m$1

and the collection is denoted $n,m$2 (Charamaras, 2024). For finitely generated systems, these are exactly the functions satisfying

$n,m$3

(Charamaras, 2024).

The global system-level notion appears explicitly in the generalized Pythagorean recurrence paper: a multiplicative action is pretentious if the spectral measure of every $n,m$4 is supported on the set of pretentious multiplicative functions,

$n,m$5

(Frantzikinakis et al., 13 Aug 2025). This makes pretentiousness a spectral support condition rather than merely a property of selected observables.

3. Mean ergodic theory and decomposition theorems

The mean ergodic theorem for finitely generated multiplicative systems is the main theorem of "Mean value theorems in multiplicative systems and joint ergodicity of additive and multiplicative actions" (Charamaras, 2024). For a finitely generated multiplicative system $n,m$6 and any $n,m$7, the additive averages

$n,m$8

always converge in $n,m$9 (Charamaras, 2024). The limit is governed by the pretentiously invariant factor $(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$0 rather than the ordinary invariant factor. The paper identifies pretentious ergodicity by the condition that these averages always converge to the space average,

$(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$1

for every $(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$2 (Charamaras, 2024).

Two orthogonal decomposition theorems form the structural core of the theory.

First, the rational/aperiodic decomposition:

$(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$3

where $(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$4 is the closed span of functions $(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$5 for which there exists a Dirichlet character $(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$6 with $(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$7, and

$(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$8

(Charamaras, 2024).

Second, the pretentious eigen/pretentious weak-mixing decomposition:

$(X,\mathcal X,\mu,(T_n)_{n\in\mathbb N}),$9

where $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$0 is the closed span of functions with $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$1 for some finitely generated completely multiplicative $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$2, and

$$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$3

(Charamaras, 2024). The paper presents this as the multiplicative analogue of the Jacobs–de Leeuw–Glicksberg decomposition.

Further structural statements sharpen the spectral picture. The pretentiously invariant functions coincide with the $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$4-closure of the pretentiously invariant $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$5-algebra, $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$6 is an algebra, if $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$7 then $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$8, and if $$T_1=\mathrm{id},\qquad T_{mn}=T_m\circ T_n\quad \text{for all }m,n\in\mathbb N,$$9 then $\mathbb Q_+$0 (Charamaras, 2024). These properties show that the multiplicative spectral decomposition clusters into equivalence classes under finite pretentious distance.

A weighted form of the mean ergodic theorem extends the analogy with Halász’s theorem. For $\mathbb Q_+$1 and $\mathbb Q_+$2, the weighted averages

$\mathbb Q_+$3

converge in $\mathbb Q_+$4, and the limit is controlled by the projection $\mathbb Q_+$5 onto $\mathbb Q_+$6 (Charamaras, 2024). The proof uses the spectral theorem for unitary multiplicative actions, the identity

$\mathbb Q_+$7

a decomposition $\mathbb Q_+$8, and a convolution/Euler-product argument modeled on Delange’s treatment of multiplicative functions (Charamaras, 2024).

4. Aperiodicity, pretentious weak mixing, and joint ergodicity

The multiplicative setting refines the usual ergodic hierarchy. A multiplicative system is called aperiodic if averages along every arithmetic progression converge to the space average:

$\mathbb Q_+$9

(Charamaras, 2024). It is pretentiously weak-mixing if $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$0 is pretentiously ergodic (Charamaras, 2024).

These notions admit spectral characterizations. The system is aperiodic if and only if $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$1 consists of constants and $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$2 for every non-principal Dirichlet character $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$3; equivalently, it is pretentiously ergodic and has no nontrivial pretentiously rational spectrum (Charamaras, 2024). Likewise, it is pretentiously weak-mixing if and only if $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$4 consists of constants and $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$5 for every $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$6 with $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$7, equivalently if it has no non-constant pretentious eigenfunctions (Charamaras, 2024).

The paper emphasizes strict implication chains:

$T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$8

while the converses fail in general (Charamaras, 2024). The examples mentioned there include a multiplicative rotation that is ergodic but not pretentiously ergodic, and another that is totally ergodic but not aperiodic because of a pretentious rational eigenvalue (Charamaras, 2024). A common misconception is therefore that the multiplicative theory merely renames standard additive ergodic notions; the cited counterexamples show that pretentiousness detects genuinely finer prime-indexed structure.

The same paper studies joint ergodicity of an additive and a finitely generated multiplicative action acting on the same probability space. Theorem B states that such actions are jointly ergodic whenever no local obstructions arise, and it gives a concrete description of these local obstructions (Charamaras, 2024). The data block does not reproduce the obstruction formula, but it explicitly identifies the theorem as a natural expression of the independence principle between additive and multiplicative structures of the integers. This suggests that joint ergodicity here functions as a precise dynamical formalization of additive–multiplicative independence.

5. Pretentious–aperiodic splitting, mixed seminorms, and rational iterates

"Decomposition results for multiplicative actions and applications" develops a structural theory adapted to multiple recurrence and convergence problems for multiplicative actions with rational iterates (Frantzikinakis, 19 Mar 2025). For every multiplicative action and every $T_{m/n}:=T_m\circ T_n^{-1},\qquad m,n\in\mathbb N$9, there exist orthogonal components

$L^2$00

where $L^2$01 is the subspace of functions whose spectral measure is supported on pretentious multiplicative functions and $L^2$02 is the corresponding aperiodic subspace (Frantzikinakis, 19 Mar 2025). In fact,

$L^2$03

For finitely generated actions, this decomposition has stronger consequences: $L^2$04 is the pretentious projection, $L^2$05 is annihilated by the mixed seminorms introduced in the paper, and $L^2$06 has vanishing averages along the rational polynomial iterates considered there (Frantzikinakis, 19 Mar 2025).

The major technical novelty is a family of mixed seminorms combining additive shifts with multiplicative iterates. For $L^2$07, the seminorms are defined by

$L^2$08

These are tailored to patterns such as

$L^2$09

and more generally to rational combinations of linear forms (Frantzikinakis, 19 Mar 2025).

For finitely generated multiplicative actions, the inverse theorem identifies the vanishing factor of these seminorms exactly:

$L^2$10

More precisely, the following are equivalent: $L^2$11; $L^2$12 in $L^2$13 for every $L^2$14, $L^2$15; $L^2$16; and $L^2$17 for every $L^2$18 (Frantzikinakis, 19 Mar 2025). This yields an exact description of the aperiodic factor as the orthogonal complement of the pretentious factor.

The same paper proves convergence and recurrence theorems for rational polynomials factoring into products of pairwise independent linear forms. For finitely generated multiplicative actions and pairwise independent linear forms $L^2$19 with non-negative coefficients, the averages

$L^2$20

converge in $L^2$21 for all $L^2$22 and every $L^2$23-dimensional grid $L^2$24 (Frantzikinakis, 19 Mar 2025). If all actions are aperiodic and $L^2$25, the limit is the product of the integrals (Frantzikinakis, 19 Mar 2025). A second theorem treats two rational polynomials $L^2$26 factoring linearly, with explicit restrictions on $L^2$27 and an excluded power form for $L^2$28, and again proves $L^2$29 convergence (Frantzikinakis, 19 Mar 2025).

The proofs combine four ingredients stated explicitly in the paper: the decomposition $L^2$30, seminorm control that discards the aperiodic part, concentration estimates for the pretentious part, an averaging-over-congruence-classes "$L^2$31-trick," and mean ergodic theorems for multiplicative actions (Frantzikinakis, 19 Mar 2025). The resulting framework is then applied to density regularity of homogeneous quadratic equations such as

$L^2$32

and more generally

$L^2$33

(Frantzikinakis, 19 Mar 2025).

6. Furstenberg systems, recurrence along quadratic equations, and current scope

The Furstenberg-system perspective studies dynamical systems naturally attached to bounded multiplicative functions rather than arbitrary multiplicative actions. For a bounded sequence $L^2$34, weak-* limits of empirical orbit measures under the shift on $L^2$35 define Furstenberg systems using Cesàro or logarithmic averages (Frantzikinakis et al., 2023). For pretentious multiplicative functions, every Furstenberg system has rational discrete spectrum, hence zero entropy (Frantzikinakis et al., 2023). The same paper proves that the only pretentious multiplicative functions whose Furstenberg systems have trivial rational spectrum are the Archimedean characters $L^2$36, and that a pretentious multiplicative function has ergodic Furstenberg systems if and only if it pretends to be a Dirichlet character (Frantzikinakis et al., 2023). It also shows that for a fixed pretentious multiplicative function, all Furstenberg systems are isomorphic (Frantzikinakis et al., 2023).

These Furstenberg-system results are not formulated as theorems about arbitrary pretentious multiplicative actions, but they clarify the model objects that pretentious spectral components emulate. In particular, they show that pretentiousness is dynamically rigid, yet not uniformly ergodic: the Archimedean factor $L^2$37 contributes an identity component and forces non-ergodicity when $L^2$38 (Frantzikinakis et al., 2023). They also expose a distinction between logarithmic and Cesàro averaging. For logarithmic averages of bounded multiplicative functions,

$L^2$39

where strong stationarity means

$L^2$40

for all $L^2$41 and all shifts $L^2$42, but this equivalence fails for Cesàro averages (Frantzikinakis et al., 2023). This is an objective source of subtlety in the subject: averaging mode affects the limiting dynamical structure.

A later development establishes multiple recurrence for pretentious multiplicative actions along generalized Pythagorean triples. For a Rado triple $L^2$43, meaning

$L^2$44

the main theorem states that for every pretentious multiplicative action $L^2$45, every measurable $L^2$46 with $L^2$47, and every $L^2$48, there exist distinct $L^2$49 such that

$L^2$50

and

$L^2$51

(Frantzikinakis et al., 13 Aug 2025). The proof uses explicit parametrizations of the quadratic equation, structured factors built from Archimedean and finitely supported pretentious components, concentration estimates for pretentious functions, multiplicative averaging sets $L^2$52, $L^2$53, $L^2$54, and Chu’s inequality

$L^2$55

for nonnegative $L^2$56 (Frantzikinakis et al., 13 Aug 2025).

The combinatorial consequence is that if $L^2$57 are pretentious completely multiplicative functions, then for every open arc $L^2$58 containing $L^2$59, there exist distinct $L^2$60 such that

$L^2$61

and

$L^2$62

(Frantzikinakis et al., 13 Aug 2025). This confirms the generalized Pythagorean partition-regularity phenomenon in the pretentious setting.

Across these works, several boundaries of the theory are explicit. The finitely generated hypothesis remains crucial for the strongest mean ergodic, inverse-theorem, and decomposition results (Charamaras, 2024, Frantzikinakis, 19 Mar 2025). For general multiplicative actions, some conclusions weaken and averages may need to be restricted to special sets close to $L^2$63 multiplicatively (Frantzikinakis, 19 Mar 2025). Open problems explicitly mentioned include extensions to Pythagorean triples $L^2$64, higher-degree patterns, and extension of the mixed-seminorm inverse theorem to arbitrary multiplicative actions (Frantzikinakis, 19 Mar 2025). A plausible implication is that the long-term structure of the theory will continue to be organized by the same dichotomy already visible in current results: pretentious components contribute concentration and structured recurrence, while aperiodic components contribute vanishing and cancellation.